The Stochastic Empirical Loading and Dilution Model (SELDM) was designed to help quantify the risk of adverse effects of runoff on receiving waters, the potential need for mitigation measures, and the potential effectiveness of such management measures for reducing these risks. SELDM is calibrated using representative hydrological and water-quality input statistics. This report by the U.S. Geological Survey, in cooperation with the Federal Highway Administration and the Connecticut, Massachusetts, and Rhode Island Departments of Transportation, documents approaches for assessing flows, concentrations, and loads of highway- and urban-runoff and receiving-stream stormwater in southern New England with SELDM. In this report, the term “urban runoff” is used to identify stormwater flows from developed areas with impervious fractions ranging from 10 to 100 percent without regard to the U.S. Census Bureau designation for any given location. There are more than 48,000 delineated road-stream crossings in southern New England, but because there are relatively few precipitation, streamflow, and water-quality monitoring sites in this area, methods were needed to simulate conditions at unmonitored sites. This report documents simulation methods, methods for interpreting stochastic model results, sensitivity analyses to identify the most critical variables of concern, and examples demonstrating how simulation results can be used to inform scientific decision-making processes. Results of 7,511 SELDM simulations were used to do the sensitivity analyses and provide information decisionmakers can use to address runoff-quality issues in southern New England and other areas of the Nation.

The sensitivity analyses indicate the relatively strong effect of input variables on variations in output results. These analyses indicate that highway and urban runoff quality and upstream water-quality statistics that vary considerably from site to site have the greatest effect on simulated results. Further data are needed to improve available water-quality statistics, and because the number of monitored sites will never approach the number of sites of interest for water-quality management, research is needed to identify methods to select statistics for unmonitored sites and quantify the uncertainties in the selection process. Hydrologically, prestorm streamflows with and without zero flows are the most sensitive and therefore the most important hydrologic variables to quantify. Results of analyses also are sensitive to statistics used for simulating structural best management practices.

Although the focus of the report is on data, statistics, simulation methods, and methods to interpret stochastic simulations, the examples in this report provide results that can be used to inform scientific decision-making processes. The results of 441 simulations that provide regional and site-specific highway and urban runoff yields across southern New England can be used for total maximum daily load analyses. The example stormwater load analysis done for 16 tributaries of the Narragansett Bay demonstrates that highway nitrogen loads are a small fraction of stormwater loads (about 3.6 percent), and a much smaller fraction of all nitrogen loads to the bay, primarily because highways have a small footprint on the land. Examples evaluating the potential effectiveness of end-of-pipe treatment indicate that offsite treatment is warranted in developed areas, and land conservation may be an effective mitigation strategy. The results of these analyses are consistent with conclusions from other simulation and monitoring studies.

Granato, G.E., 2021, Best management practices statistical estimator (BMPSE) version 1.2.0: U.S. Geological Survey software release,

Granato, G.E. and Friesz, P.J., 2021, Model archive for analysis of long-term annual yields of highway and urban runoff in selected areas of California with the Stochastic Empirical Loading Dilution Model (SELDM): U.S. Geological Survey data release,

Granato, G.E., and Jeznach, L.C., 2020, Model archive for analysis of the effects of impervious cover on receiving-water quality with the Stochastic Empirical Loading Dilution Model (SELDM): U.S. Geological Survey data release,

Granato, G.E., Spaetzel, A.B., and Jeznach, L.C., 2022, Model archive for analysis of flows, concentrations, and loads of highway and urban runoff and receiving-stream stormwater in southern New England with the Stochastic Empirical Loading and Dilution Model (SELDM): U.S. Geological Survey data release,

Spaetzel, A.B., Steeves, P.A., and Granato, G.E., 2020, Basin characteristics and point locations of road crossings in Connecticut, Massachusetts, and Rhode Island for highway-runoff mitigation analyses using the Stochastic Empirical Loading and Dilution Model: U.S. Geological Survey data release,

For more information on the USGS—the Federal source for science about the Earth, its natural and living resources, natural hazards, and the environment—visit

For an overview of USGS information products, including maps, imagery, and publications, visit

Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Although this information product, for the most part, is in the public domain, it also may contain copyrighted materials as noted in the text. Permission to reproduce copyrighted items must be secured from the copyright owner.

The authors thank the many people who assisted with this report and the associated digital media. Susan C. Jones of the Federal Highway Administration helped design the study. Adam Fox and Daniel Imig of the Connecticut Department of Transportation, Henry Barbaro and Hung Pham of the Massachusetts Department of Transportation, Mark Nimiroski and Allison Hamel of the Rhode Island Department of Transportation, and Susan C. Jones of the Federal Highway Administration provided oversight and input that improved the content and presentation of information in this report.

Multiply | By | To obtain |

Length | ||
---|---|---|

inch (in.) | 2.54 | centimeter (cm) |

mile (mi) | 1.609 | kilometer (km) |

Area | ||

acre | 4,047 | square meter (m^{2}) |

square mile (mi^{2}) |
2.590 | square kilometer (km^{2}) |

Flow rate | ||

cubic foot per second (ft^{3}/s) |
0.02832 | cubic meter per second (m^{3}/s) |

cubic foot per second (ft^{3}/s) |
28.32 | liter per second (L/s) |

cubic foot per second per square mile ([ft^{3}/s]/mi^{2}) |
0.01093 | cubic meter per second per square kilometer ([m^{3}/s]/km^{2}) |

cubic foot per second per square mile ([ft^{3}/s]/mi^{2}) |
10.93 | liter per second per square kilometer ([L/s]/km^{2}) |

inch per year (in/yr) | 25.4 | millimeter per year (mm/yr) |

Mass | ||

pound, avoirdupois (lb) | 0.4536 | kilogram (kg) |

Yield | ||

pound per acre per year ([lb/acre]/yr) | 1.121 | kilogram per hectare per year ([kg/ha]/yr) |

Temperature in degrees Celsius (°C) may be converted to degrees Fahrenheit (°F) as follows: °F = (1.8 × °C) + 32.

Horizontal coordinate information is referenced to the North American Datum of 1983.

Bacterial concentrations are given in colonies per 100 milliliters or the most probable number per 100 milliliters.

Concentrations of chemical constituents in water are given in milligrams per liter (mg/L), micrograms per liter (µg/L), or nanograms per liter (ng/L).

Specific conductance is given in microsiemens per centimeter at 25 degrees Celsius (µS/cm at 25 °C).

American Association of State Highway and Transportation Officials

annual average daily traffic

basin development factor

basin-lag factor

best management practice

Best Management Practices Statistical Estimator

coefficient of variation

Connecticut Department of Transportation

department of transportation

event mean concentrations

U.S. Environmental Protection Agency

Federal Highway Administration

geographic information system

Highway-Runoff Database

normally distributed random numbers

Kendall-Theil Robust Line

median absolute deviation

Massachusetts Department of Transportation

Municipal Separate Storm Sewer System

National Bridge Inventory

National Land Cover Database

National Oceanic and Atmospheric Administration

National Water Information System

polycyclic aromatic hydrocarbons

Rhode Island Department of Transportation

Stochastic Empirical Loading and Dilution Model

total impervious area

total maximum daily load

U.S. Geological Survey

vehicles per day

wastewater treatment plant

Decisionmakers need information about flows, concentrations, and loads of highway and urban runoff and receiving-stream stormwater to assess potential effects of runoff and the potential to mitigate such risks (

The State highway systems are thin ribbons of land within the surrounding developed and undeveloped areas. Therefore, transportation agencies also need information about the quality and quantity of runoff and BMP discharges from developed areas to assess the risk for water-quality exceedances at highway-stream crossings and to assess the magnitude of runoff loads from State roadways in comparison to developed-area runoff loads in impaired receiving waters. In the national highway-runoff monitoring study by the FHWA (

Stormwater management by State Departments of Transportation (DOTs) is complicated because the DOTs operate extensive linear systems with limited rights of way that cross thousands of receiving waters across each State (

Table 1. Federal Highway Administration definitions of road classes and the associated categories of The National Map and StreamStats from the U.S. Geological Survey

[Road classes are listed in order of increasing functional class. Official road-class definitions are not quantitative. For more extensive definitions, see

FHWA Road class | Definition | The National Map Functional Road Classification category | StreamStats category number | |

Numbers | Names | |||

Local | Local roads provide basic access between residential and commercial properties, connecting them with higher order highways. A route meeting this purpose would connect a home, work, or entertainment trip by connecting the final destination to the roads serving longer trips. These roads commonly have two lanes, low traffic, and low speeds. | 4 | Local road | 4 |

Collector, minor | Minor collectors link local roads with major collectors or arterial roads. These roads provide traffic access and circulation in lower density residential, commercial, or industrial areas; they commonly have two lanes, low traffic, and low speeds. | 3 | Local connecting road | 3 |

Collector, major | Major collectors link local roads and minor collectors to arterial roads. These roads provide direct property access and traffic circulation in higher density residential neighborhoods and commercial and industrial areas. These roads commonly have two or more lanes, moderate traffic, and low to moderate speeds. | 3 | Local connecting road | 3 |

Arterial, minor | Minor arterial roadways provide through-traffic routes in urban areas and travel routes between municipalities in rural areas. These roads provide direct connections to adjacent property and cross streets and commonly have two or more lanes, low to moderate traffic, and moderate speeds. | 2 | Secondary highway or major connecting road | 2 |

Arterial, principal | Principal arterial roadways provide through-traffic routes in urban areas and travel routes between municipalities in rural areas. These roads provide direct connections to adjacent property and cross streets and commonly have more than two lanes, moderate to high traffic, and moderate to high speeds. | 2 | Secondary highway or major connecting road | 2 |

Arterial, freeways and expressways | Freeways and expressways are principal limited-access arterial roadways that are divided limited-access highways. These roads, designed for moderate to high traffic and high speeds, are accessed by traffic ramps, cross streets, railways, and other features through overpasses or underpasses. | 1, 5 | Controlled-access highway (1) or ramp (5) | 1 |

Arterial, interstate | Interstate highways are freeways or expressways that are designed to carry high-speed traffic between States | 1, 5 | Controlled-access highway (1) or ramp (5) | 1 |

Table 2. Road length, ownership, and geometry statistics for Connecticut, Massachusetts, and Rhode Island

[See

FHWA road class | Percentage of total road length | Road length owned (mi) | DOT-owned percentage | Number of lanes | Bridge width (ft/lane) | |||||

National average | State | State DOT | Other | Average, |
Average, |
Median, |
Average | Median | ||

Connecticut | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|

Local | 69.09 | 68.67 | 19.98 | 14,795.33 | 0.13 | 2.0 | 2.0 | 2.0 | 13.6 | 13.3 |

Collector, minor | 6.73 | 3.41 | 32.39 | 703.18 | 4.40 | 2.0 | 2.0 | 2.0 | 13.7 | 13.8 |

Collector, major | 12.90 | 12.40 | 1,103.81 | 1,570.91 | 41.27 | 2.0 | 2.1 | 2.0 | 15.8 | 14.9 |

Arterial, minor | 5.91 | 8.76 | 1,154.38 | 735.61 | 61.08 | 2.2 | 2.5 | 2.0 | 16.4 | 16.0 |

Arterial, principal | 3.76 | 3.87 | 779.18 | 54.81 | 93.43 | 2.6 | 3.0 | 2.0 | 17.1 | 16.4 |

Arterial, freeways and expressways | 0.45 | 1.28 | 279.12 | 0.00 | 100.00 | 4.1 | 4.2 | 4.0 | 19.9 | 19.0 |

Arterial, interstate | 1.16 | 1.61 | 346.34 | 0.00 | 100.00 | 5.4 | 5.1 | 5.0 | 20.2 | 19.0 |

Total | 100.00 | 100.00 | 3,715.20 | 17,859.84 | NA | NA | NA | NA | NA | NA |

Massachusetts | ||||||||||

Local | 69.09 | 67.73 | 54.45 | 24,881.55 | 0.22 | 2.0 | 2.1 | 2.0 | 13.5 | 13.0 |

Collector, minor | 6.73 | 1.67 | 4.46 | 611.97 | 0.72 | 2.0 | 2.0 | 2.0 | 13.3 | 13.0 |

Collector, major | 12.90 | 10.70 | 185.45 | 3,752.12 | 4.71 | 2.0 | 2.1 | 2.0 |
15.0 | 14.8 |

Arterial, minor | 5.91 | 11.81 | 834.18 | 3,513.10 | 19.19 | 2.1 | 2.4 | 2.0 | 16.9 | 16.1 |

Arterial, principal | 3.76 | 5.64 | 1,026.35 | 1,049.66 | 49.44 | 2.5 | 3.0 | 2.0 | 18.2 | 17.5 |

Arterial, freeways and expressways | 0.45 | 0.91 | 324.29 | 9.84 | 97.06 | 4.2 | 4.8 | 4.0 | 19.2 | 19.0 |

Arterial, interstate | 1.16 | 1.54 | 567.37 | 0.46 | 99.92 | 5.6 | 5.8 | 6.0 | 18.0 | 17.3 |

Total | 100.00 | 100.00 | 2,996.55 | 33,818.70 | NA | NA | NA | NA | NA | NA |

Rhode Island | ||||||||||

Local | 69.09 | 68.20 | 23.27 | 4,085.50 | 0.57 | 2.0 | 2.0 | 2.0 | 13.4 | 13.5 |

Collector, minor | 6.73 | 3.10 | 44.52 | 142.36 | 23.82 | 2.0 | 2.2 | 2.0 | 13.4 | 13.0 |

Collector, major | 12.90 | 11.86 | 226.26 | 488.02 | 31.68 | 2.0 | 2.0 | 2.0 | 15.9 | 14.9 |

Arterial, minor | 5.91 | 6.84 | 252.69 | 159.70 | 61.28 | 2.1 | 2.4 | 2.0 | 17.3 | 17.1 |

Arterial, principal | 3.76 | 7.31 | 404.44 | 35.93 | 91.84 | 2.5 | 3.0 | 3.0 | 16.6 | 15.0 |

Arterial, freeways and expressways | 0.45 | 1.53 | 83.49 | 8.63 | 90.63 | 4.0 | 4.1 | 4.0 | 19.1 | 18.5 |

Arterial, interstate | 1.16 | 1.16 | 70.01 | 0.00 | 100.00 | 5.4 | 5.8 | 6.0 | 20.0 | 19.0 |

Total | 100.00 | 100.00 | 1,104.68 | 4,920.14 | NA | NA | NA | NA | NA | NA |

As indicated by the number of road-stream crossings in each State, these DOTs maintain hundreds to thousands of stormwater outfalls and stormwater control measure BMPs. Runoff collected on roadways and structures crossing the stream (bridges or culverts) may be diverted through stormwater conveyances from each roadway approach to the stream and from the structures themselves. Therefore, each road-stream crossing may have multiple outfalls and multiple BMPs. The potential number of BMPs in each State are of concern in part because BMPs are costly to build and maintain with life-cycle costs that can exceed $70,000 per pound per year for some constituents of concern (

The Stochastic Empirical Loading and Dilution Model (SELDM) was developed by the U.S. Geological Survey (USGS) in cooperation with the Federal Highway Administration for simulating stormwater event mean concentrations (EMCs) to indicate the risk for stormwater flows, concentrations, and loads to be above user-selected water-quality goals and to evaluate the potential effectiveness of mitigation measures to reduce such risks (

The purpose of this report is to document approaches for assessing flows, concentrations, and loads of highway and urban runoff and receiving-stream stormwater in southern New England with SELDM. Specifically, this report documents application of statistics for highway and upstream basin properties, hydrologic variables, and stormwater quality that can be used to represent conditions in this area. The report describes methods for interpreting simulation results; documents results of sensitivity analyses designed to guide the selection of input-variables for runoff-quality simulations; and provides example simulations used to illustrate use of simulation results for decision making.

In this report, southern New England is defined as the areas within Connecticut, Massachusetts, and Rhode Island that drain to the ocean or to large rivers that flow into these areas. For example, tributaries to the Connecticut River within these States are included but the main stem and tributaries completely outside these three States are not. For the purpose of calculating basin properties within these States, the southern New England area also includes headwater areas in New Hampshire, New York, and Vermont draining to streams and rivers predominantly located within southern New England. Data from precipitation, streamflow, and water-quality monitoring stations in New Hampshire, New York, and Vermont also were used to supplement data collected within southern New England to improve statistical estimates.

This report is designed to provide information that can be used for robust decision making by highway practitioners, regulators, and decisionmakers. The data, information, and statistics described in this report are intended to facilitate stochastic analysis of the potential effects of stormwater runoff on receiving waters at unmonitored sites (or sites with limited monitoring data). SELDM can be used to simulate long-term conditions at monitoring sites with data, but because there are more than 48,000 delineated road-stream crossings in southern New England stream-basins (

SELDM uses basin characteristics and statistics for storm-event hydrology, stormwater quality, and stormwater treatment to simulate a population of stormflows, concentrations, and loads from runoff-producing events. SELDM uses a stochastic mass-balance approach in which the flows, concentrations, and loads from the upstream basin and a site of interest are used to calculate the combined downstream flows, concentrations, and loads (

Schematic diagram showing the stochastic mass-balance approach for estimating stormflows, concentrations, and loads of water-quality constituents upstream of a highway-runoff outfall, from the highway, and downstream from the outfall (modified from

Figure 1. Schematic diagram showing the stochastic mass-balance approach for estimating stormflows, concentrations, and loads of water-quality constituents upstream of a highway-runoff outfall, from the highway, and downstream from the outfall

In this study, regression equations were used to provide planning-level estimates of selected variables from related variables or to simulate water-quality values by using a transport curve or dependent relation (_{i}_{i} + e_{i}

is the response variable for a given input;

is the intercept of the regression equation;

is the slope of the regression equation;

is the predictor variable input to the equation; and

is the random variation around the line.

In SELDM, the relations between imperviousness and runoff-coefficient statistics are developed in arithmetic space without logarithmic transformations (

In SELDM studies, relations to estimate basin properties, streamflow statistics, concentrations from water-quality transport curves, concentrations from dependent relations, and concentration from other explanatory variables commonly are developed by using the logarithms of data (

If a regression equation is being used to simulate individual values by using the frequency factor method, then the random-variation term _{i}_{n}_{n}_{n}_{n}_{i}

SELDM uses the location (latitude and longitude) of the site of interest, five physical basin characteristics, and the upstream hydrograph recession-ratio statistics to simulate the hydrologic characteristics of the site of interest and the upstream basin (

The drainage area and imperviousness of the site of interest and the upstream basin were used to simulate the volume and timing of stormflows in this study. The drainage area of the site of interest is used to calculate the precipitation volume for that area. The drainage area of the upstream basin is used to calculate the precipitation and prestorm streamflow volumes for that area. The impervious fraction of the site of interest and the upstream basin are used to estimate the runoff coefficient statistics, which are used to transform precipitation volumes into runoff volumes from each area.

The main-channel length and main-channel slope were used to simulate the timing of stormflows in this study. The main-channel drainage length (also known as basin length) is the length of the main channel measured from the point of interest to the basin divide. The main-channel drainage slope (also known as the mean basin slope or 10-85 slope) is the average slope of the main channel upstream from the point of interest. It is calculated by determining the locations and elevations of points at 10 and 85 percent along the main channel from the point of interest to the basin divide and then by dividing the difference in elevations by the channel length between these points. The main-channel length and slope of the drainage network of the site of interest are used to simulate the timing of runoff from the site (

SELDM also has a basin development factor (BDF) variable that can be used with main-channel length and slope to calculate the timing of runoff from the site of interest and the upstream basin (

In SELDM, the site location (latitude and longitude) is used to select regional precipitation, prestorm-streamflow, and upstream-water-quality statistics. The site location also can be used to select precipitation and prestorm-streamflow statistics from nearby monitoring sites. The latitude and longitude coordinates entered can be precise (down to fractions of a second) in order to document the exact location of a particular site of interest and delineate the associated upstream basin, but this precision is not necessary for planning-level regional or statewide analyses. For these analyses, the precision of the coordinates entered can be about one degree of latitude and longitude as long as the selected point falls within the intended region or State. For general or basin-wide analyses, the precision of the selected coordinates can be as coarse as the density of the regional data monitoring networks. For example, the density of National Oceanic and Atmospheric Administration (NOAA) rain-gages included in the SELDM database for southern New England is about 784 square miles per station, so the maximum precision would need to be about 0.23 degrees of latitude and 0.27 degrees of longitude to properly select the nearest rain gage (if they were evenly spaced on a grid). To select the USGS streamgage from within the National SELDM database that is closest to a selected site of interest in southern New England, a precision of about 0.11 degrees of latitude and 0.13 degrees of longitude is needed. This is because the streamgage density in southern New England is about 179 square miles per station.

In this study, representative statistics were needed to do the sensitivity analyses necessary for identifying the effect of different SELDM input variable selections on the results of water-quality simulations. Upstream and highway- (or developed-area-) site characteristics can be determined for a specific site by using the USGS StreamStats application (

The dataset of upstream basin characteristics was developed by delineating stream basins upstream from the intersections between roads and streams in southern New England and analyzing selected basin properties by using GIS software.

The geographic analysis by

Probability plots showing the distribution of various properties of 48,466 drainage basins above roadways and 5,545 basins delineated above arterial roadways, computed by

Figure 2. Probability plots showing the distribution of various properties of 48,466 drainage basins above roadways and 5,545 basins delineated above arterial roadways

The road-crossing basin count may seem large for southern New England, but most delineated basins are nested within larger basins. For example, the Blackstone River Basin above Interstate 95 in Providence, Rhode Island is 475 mi^{2}, has an imperviousness of about 12 percent, and has 2,340 upstream road crossings; the Charles River Basin above Interstate 93 in Boston, Massachusetts is 313 mi^{2}, has an imperviousness of about 23 percent, and has 1,365 upstream road crossings; and the Park River Basin above Interstate 91 in Hartford, Connecticut is 77.2 mi^{2}, has an imperviousness of about 27 percent, and has 539 upstream road crossings. These delineated basins, by definition, do not represent the confluence of tributary streams, but the drainage-area pattern is similar to Giusti’s law, which indicates that the number of upstream basins of any size is about 0.3 times the ratio of the basin area to the selected subbasin area (^{2}, respectively.

The characteristics of southern New England stream basins are shown in

Table 3. Descriptive statistics for basin characteristics of 48,466 stream basins delineated upstream from all road-stream crossings and a subset of 5,545 stream basins delineated upstream from arterial road-stream crossings in southern New England

[See

Variable | Minimum | Maximum | Median | Average | COV | Geometric mean |

48,466 basins above road crossings in southern New England | ||||||
---|---|---|---|---|---|---|

DRNAREA | 0.025 | 1939 | 0.455 | 7.65 | 6.65 | 0.6 |

CSL10_85 | 0.046 | 2186 | 96.5 | 138 | 1.02 | 87 |

Strm_density | 0.001 | 18.7 | 2.32 | 2.52 | 0.556 | 2.1 |

LFPLENGTH | 0.21 | 167 | 1.44 | 3.38 | 2.23 | 1.69 |

LC16IMP | 0 | 85.8 | 3.84 | 9.95 | 1.31 | — |

BLF | 0.008 | 62 | 0.144 | 0.69 | 3.93 | 0.181 |

5,545 basins above arterial-road crossings in southern New England | ||||||

DRNAREA | 0.025 | 1939 | 0.721 | 22 | 5.04 | 1.11 |

CSL10_85 | 0.266 | 1437 | 68.25 | 109 | 1.14 | 63.9 |

Strm_density | 0.005 | 10.8 | 2.3 | 2.45 | 0.51 | 2.1 |

LFPLENGTH | 0.244 | 167 | 1.87 | 6.01 | 2.32 | 2.38 |

LC16IMP | 0 | 79.8 | 10.43 | 14.8 | 0.96 | — |

BLF | 0.012 | 62 | 0.22 | 1.54 | 3.43 | 0.298 |

Information about relations between basin properties is needed to guide the choice of a limited but representative set of values for simulating the potential effect of runoff on receiving waters. To this end, an analysis of correlations between basin properties was done by calculating the nonparametric rank correlation coefficient (Spearman’s rho) and the product-moment correlation coefficient (Pearson's R) for the logarithms of data (

Table 4. Correlation coefficients for basin characteristics of 48,466 stream basins delineated upstream from road crossings in southern New England

[Data from

Basin characteristic variable | Correlation coefficients for basin characteristics | |||||

DRNAREA | CSL10_85 | Strm_density | LFPLENGTH | LC16IMP | BLF | |

Spearman's rho | ||||||
---|---|---|---|---|---|---|

DRNAREA | 1.00 | −0.52 | −0.17 | 0.98 | 0.04 | 0.91 |

CSL10_85 | −0.52 | 1.00 | 0.04 | −0.52 | −0.37 | −0.78 |

Strm_density | −0.17 | 0.04 | 1.00 | −0.11 | −0.11 | −0.10 |

LFPLENGTH | 0.98 | −0.52 | −0.11 | 1.00 | 0.02 | 0.93 |

LC16IMP | 0.04 | −0.37 | −0.11 | 0.02 | 1.00 | 0.16 |

BLF | 0.91 | −0.78 | −0.10 | 0.93 | 0.16 | 1.00 |

Pearson's R on the common logarithms of data | ||||||

DRNAREA | 1.00 | −0.54 | −0.04 | 0.98 | 0.05 | 0.93 |

CSL10_85 | −0.54 | 1.00 | 0.01 | −0.54 | −0.37 | −0.78 |

Strm_density | −0.04 | 0.01 | 1.00 | −0.01 | −0.09 | −0.01 |

LFPLENGTH | 0.98 | −0.54 | −0.01 | 1.00 | 0.05 | 0.95 |

LC16IMP | 0.05 | −0.37 | −0.09 | 0.05 | 1.00 | 0.18 |

BLF | 0.93 | −0.78 | −0.01 | 0.95 | 0.18 | 1.00 |

Correlations for the basin-lag factor (BLF), which is the main-channel length divided by the square root of the main-channel slope, also were calculated (

Regression relations were developed to select representative values of main-channel length and slope from drainage area (

Scatterplots showing relations between drainage area and the main-channel length and slope for 48,466 basins above roadways delineated by

Figure 3. Scatterplots showing relations between drainage area and the main-channel length and slope for 48,466 basins above roadways and regression lines calculated by using a subsample of 6,923 of these basins from the full dataset

Table 5. Regression equation statistics developed by using the Kendall-Theil robust line method for estimating the logarithms of main-channel length and slope from the logarithms of drainage areas of selected stream basins delineated upstream from road crossings in southern New England

[Data from

Variable | KTRLine statistics for logarithms of data | BCF | Retransformed Intercept | ASEE in percent | |||

Intercept | Slope | RMSE | MAD | ||||

CSL10_85 | 1.8784 | −0.29588 | 0.382 | 0.239 | 1.3676 | 75.579 | 108 |

LFPLENGTH | 0.34472 | 0.53421 | 0.079 | 0.052 | 1.0228 | 2.2117 | 18.3 |

The total impervious area (TIA) is an important variable for simulating runoff because it is used to calculate runoff coefficients and basin lagtimes in SELDM (

The stream density, which is the length of all streams in the basin divided by the drainage area, has a smaller range than the other basin characteristics in this study, and the differences between stream density for all the basins and the arterial-upstream basins are relatively minor. One-half of the reciprocal of the stream density can be used to estimate the length of overland flow from drainage divides to tributary stream channels; this estimate is known as the Horton half-distance (

SELDM is nominally a highway-runoff model, but it can be used to simulate runoff for any site of interest by using the characteristics of the site of interest and representative water quality. Because SELDM is a lumped-parameter model, the basin characteristic values chosen for the highway, urban, or other developed areas that are simulated as the site of interest can be literal or interpretive (

If simulations are done to develop annual total maximum daily load (TMDL) yields, then the timing of runoff during individual events is not of concern and the site may be simulated by using an area of 1 acre and a representative TIA value; the remaining basin properties may be specified by using generic values (

In this study, available information about roadway geometry and drainage-system characteristics were used to simulate runoff from hypothetical, but representative sites. Runoff from roadways was simulated by using the paved area rather than the area of the entire right-of-way because roadway-runoff quality data collected in southern New England were collected from paved areas (

The road data incorporated into StreamStats provides information about the lengths of various road classes above any given point on a stream, but information about road widths is needed to estimate the drainage areas of roads within a delineated basin. The

The area of paved roads is needed to calculate the road-runoff flows and loads. Typical road widths may be estimated based on the

Table 6. Pavement areas per mile of roadway by road class, estimated from statistics for the number of lanes by road class and roadway design guidelines for roads in southern New England

[Road classes are defined in

Road types | Number of travel lanes | Commonly used road widths, in feet | Estimated pavement area, in acres per mile |

Local roads and minor collectors without parking | 2 | 20–26 | 2.4–3.2 |

Local roads and minor collectors with parking | 2 | 36–40 | 4.4–4.8 |

Major collectors without parking | 2 | 22–32 | 2.7–3.9 |

Major collectors with parking | 2 | 40–50 | 4.8–6.1 |

Major collectors without parking | 4 | 46–56 | 5.6–6.8 |

Major collectors with parking | 4 | 64–74 | 7.8–9.0 |

Minor and principal full-access arterials | 2 | 28–48 | 3.4–5.8 |

Minor and principal full-access arterials | 4 | 52–72 | 6.3–8.7 |

Limited-access arterials with 2 lanes in each direction | 4 | 40 (80) | 4.8 (9.6) |

Limited-access arterials with 3 lanes in each direction | 6 | 52 (104) | 6.3 (12.6) |

Limited-access arterials with 4 lanes in each direction | 8 | 64 (128) | 7.8 (15.6) |

Because the cost of building and maintaining drainage systems to manage runoff are large, direct drainage to the local land surface is used where possible for infiltration. Highway drainage-design guidelines specify use of grass strips and swales rather than storm sewer systems wherever practical (

Probability plots showing the distribution of pavement drainage areas of highway sites.

Figure 4. Probability plots showing the distribution of pavement drainage areas of highway sites

The NBI (

In SELDM, the length of the drainage flow path is used with its slope to simulate the timing of runoff from the highway or urban site to the point of interest. The site of interest may have two lengths; the physical length for calculating drainage-basin area, and the main-channel drainage length used for calculating the hydrologic basin lagtime for the site. The main-channel drainage length for the site of interest is estimated as the characteristic drainage length that controls the timing of runoff from the drainage divide to the stormwater outfall. For example, when simulating runoff from a bridge with direct-discharge scuppers, the physical length of the bridge may be used to calculate area, and the average distance from the crown of the road to the nearest scupper, which is the average length of the flow path of precipitation on the bridge, may be the hydraulic length. Similarly, on a long stretch of highway with multiple drainages of varying lengths to a parallel stream, the length of that road segment may be used to calculate area, and the average distance from the crown of the road through the drainage system to the nearest outfall may be the hydraulic length. If a highway site stretches across the entire hydrologic stream basin and there is one outfall where it crosses the stream, then the divide-to-divide distance would be used to calculate the roadway area, and the hydraulic length would be the longer distance from the stream to one of the divides. The physical length of highway conveyances can be estimated from the information in

Highway drainage slopes can be estimated by using information from roadway design guidelines and hydraulic design circulars (

Table 7. Highway-drainage slopes estimated from roadway-design guidelines and Federal Highway Administration hydraulic-design circulars

[Design guidelines are AASHTO green, the

Type of slope | Slope estimates | Design guideline(s) | |

Percent | Feet per mile | ||

Pavement cross slopes (across the roadway) | |||
---|---|---|---|

High speed 2 lanes | 1.5–2 | 79.2–106 | AASHTO green HEC-22 |

High speed 3 or more lanes | 1.5–4 | 79.2–211 | AASHTO green HEC-22 |

Intermediate speed | 1.5–3 | 79.2–158 | AASHTO green HEC-22 |

Low speed | 2–6 | 105.6–317 | AASHTO green HEC-22 |

Paved shoulders | 2–6 | 105.6–317 | AASHTO green HEC-22 |

Paved shoulders with curbs | ≥4 | ≥211 | AASHTO green HEC-22 |

Longitudinal-drainage slopes (along the roadway) | |||

Absolute minimum gutter drain | 0.3 | 15.8 | AASHTO green HEC-22 |

Minimum design longitudinal gutter drain | 0.5 | 26.4 | AASHTO green HEC-22 |

Roadside channel, unlined | <2 | <106 | HEC-22 |

Roadside channel, flexible lining | 2–10 | 106–528 | HEC-22 |

Maximum longitudinal road grades (range based on allowable speed) | |||

Local road, level terrain | 5–9 | 264–475 | AASHTO green |

Local road, rolling terrain | 6–12 | 317–634 | AASHTO green |

Local road, mountainous terrain | 10–17 | 528–898 | AASHTO green |

Rural collector, level terrain | 5–7 | 264–370 | AASHTO green |

Rural collector, rolling terrain | 6–10 | 317–528 | AASHTO green |

Rural collector, mountainous terrain | 8–12 | 422–634 | AASHTO green |

Urban collector, level terrain | 6–9 | 317–475 | AASHTO green |

Urban collector, rolling terrain | 7–12 | 370–634 | AASHTO green |

Urban collector, mountainous terrain | 9–14 | 475–739 | AASHTO green |

Rural arterial, level terrain | 3–5 | 158–264 | AASHTO green |

Rural arterial, rolling terrain | 4–6 | 211–317 | AASHTO green |

Rural arterial, mountainous terrain | 5–8 | 264–422 | AASHTO green |

Urban arterial, level terrain | 5–8 | 264–422 | AASHTO green |

Urban arterial, rolling terrain | 6–9 | 317–475 | AASHTO green |

Urban arterial, mountainous terrain | 8–11 | 422–581 | AASHTO green |

Freeway, level terrain | 3–4 | 158–211 | AASHTO green |

Freeway, rolling terrain | 4–5 | 211–264 | AASHTO green |

Freeway, mountainous terrain | 5–6 | 264–317 | AASHTO green |

Minimum storm drain slopes (based on self-cleaning velocity of 3 feet per second) | |||

8 inch pipe, smooth concrete | 0.64 | 33.8 | HEC-22 |

8 inch pipe, ordinary concrete | 0.75 | 39.6 | HEC-22 |

8 inch pipe, corrugated metal pipe | 2.56 | 135 | HEC-22 |

12 inch pipe, smooth concrete | 0.37 | 19.5 | HEC-22 |

12 inch pipe, ordinary concrete | 0.44 | 23.2 | HEC-22 |

12 inch pipe, corrugated metal pipe | 1.49 | 78.7 | HEC-22 |

24 inch pipe, smooth concrete | 0.15 | 7.92 | HEC-22 |

24 inch pipe, ordinary concrete | 0.17 | 8.98 | HEC-22 |

24 inch pipe, corrugated metal pipe | 0.59 | 31.2 | HEC-22 |

36 inch pipe, smooth concrete | 0.09 | 4.75 | HEC-22 |

36 inch pipe, ordinary concrete | 0.1 | 5.28 | HEC-22 |

36 inch pipe, corrugated metal pipe | 0.34 | 18 | HEC-22 |

48 inch pipe, smooth concrete | 0.06 | 3.17 | HEC-22 |

48 inch pipe, ordinary concrete | 0.07 | 3.7 | HEC-22 |

48 inch pipe, corrugated metal pipe | 0.23 | 12.1 | HEC-22 |

Longitudinal bridge drain pipe | 8 | 422 | HEC-21 |

The drainage characteristics, which include drainage area, length, slope, and imperviousness, for other developed land covers also can be estimated from StreamStats, highway design information, and other sources. SELDM can be used to simulate runoff from a particular site or the upstream drainage areas can be aggregated into a site by lumping the areas and using representative hydraulic characteristics (

The annual average daily traffic (AADT) volume, which is a count of the number of vehicles using the roadway per day, is commonly viewed as a basin characteristic of roadway sites that is indicative of runoff quality. AADT data is primarily collected to measure and plan roadway capacity needs, but it has been used, with mixed success, as an explanatory variable for estimating highway-runoff quality (

State DOTs run small numbers of continuous traffic monitoring stations and supplement these stations spatially by using many more short-period counting locations, which are used to estimate AADT values (

The NBI (

Probability plot showing the distribution of annual average daily traffic volumes, in vehicles per day, for all bridges over water and State-maintained bridges over water from the National Bridge Inventory Database (

Figure 5. Probability plot showing the distribution of annual average daily traffic volumes for all bridges over water and State-maintained bridges over water from the National Bridge Inventory Database and highway-runoff monitoring sites from the Highway Runoff Database for locations in southern New England

The population of all southern New England bridges in the NBI (

Box plots showing annual average daily traffic volumes, in vehicles per day per lane, by road class.

Figure 6. Box plots showing annual average daily traffic volumes, in vehicles per day per lane, by road class

SELDM simulates the volume of stormflows from runoff-generating events by using statistics for prestorm streamflows, precipitation, and runoff coefficients (

Regional simulations were done by using prestorm streamflow and precipitation statistics for three U.S. Environmental Protection Agency (EPA) Level III ecoregions that include parts of Massachusetts, Connecticut, or Rhode Island (the Northeastern Highlands, Northeastern Coastal Zone, and Atlantic Coastal Pine Barrens ecoregions;

Map showing U.S. Environmental Protection Agency (

Table 8. U.S. Environmental Protection Agency Level III ecoregions that lie partly within Connecticut, Massachusetts, or Rhode Island

[U.S. Environmental Protection Agency Level III ecoregion numbers, names, and definitions are defined by the ^{2}, square mile]

U.S. Environmental Protection Agency Level III ecoregion definitions | SELDM, area (mi^{2}) |
||

Ecoregion number | Ecoregion name | Definition | |

58 | Northeastern Highlands | The Northeastern Highlands cover most of the northern and mountainous parts of New England as well as the Adirondacks and higher Catskills in New York. It is a relatively sparsely populated region characterized by hills and mountains, a mostly forested land cover, nutrient-poor soils, and numerous high-gradient streams and glacial lakes. Forest vegetation is somewhat transitional between the boreal regions to the north in Canada and the broadleaf deciduous forests to the south. Typical forest types include northern hardwoods (maple-beech-birch), northern hardwoods and spruce, and northeastern spruce-fir forests. Recreation, tourism, and forestry are primary land uses. Farm-to-forest conversion began in the 19th century and continues today. Despite this trend, alluvial valleys, glacial lake basins, and areas of limestone-derived soils are still farmed for dairy products, forage crops, apples, and potatoes. Many of the lakes and streams in this region have been acidified by sulfur depositions originating in industrialized areas upwind from the ecoregion to the west. | 51,371 |

59 | Northeastern Coastal Zone | Similar to the Northeastern Highlands (58), the Northeastern Coastal Zone contains relatively nutrient poor soils and concentrations of continental glacial lakes, some of which are sensitive to acidification; however, this ecoregion contains considerably less surface irregularity and much greater concentrations of human population. Landforms in the region include irregular plains and plains with high hills. Appalachian oak forests and northeastern oak-pine forests are the natural vegetation types. Although attempts were made to farm much of the Northeastern Coastal Zone after the region was settled by Europeans, land use now mainly consists of forests, woodlands, and urban and suburban development, with only some minor areas of pasture and cropland. | 15,882 |

84 | Atlantic Coastal Pine Barrens | This is a transitional ecoregion, distinguished from the coastal ecoregion (63) to the south by its coarser-grained soils, cooler climate, and Northeastern oak-pine potential natural vegetation. The climate is milder than the coastal ecoregion (59) to the north that contains Appalachian oak forests and some northern hardwoods forests. The physiography of this ecoregion is not as flat as that of the Middle Atlantic Coastal Plain (63), but it is not as irregular as that of the Northeastern Coastal Zone (59). The shore characteristics of sandy beaches, grassy dunes, bays, marshes, and scrubby oak-pine forests are more like those to the south, in contrast to the more rocky, jagged, forested coastline found to the north. | 13,369 |

SELDM uses precipitation statistics to stochastically simulate a large series of runoff-generating events. Storm-event precipitation statistics define the characteristics of each simulated storm event and the number of events in the simulation (

Regional simulations were done by using precipitation statistics for three EPA Level III ecoregions that include parts of Massachusetts, Connecticut, or Rhode Island, and statistics for southern New England (

Table 9. Synoptic-precipitation statistics for the southern New England area and selected U.S. Environmental Protection Agency Level III ecoregions that lie in whole or in part within Connecticut, Massachusetts, or Rhode Island

[U.S. Environmental Protection Agency Level III ecoregions are defined in ^{2}, square mile; in/yr, inch per year; in., inch; hr, hour; —, not applicable]

Ecoregion number | Ecoegion name | Number of NOAA stations | Average estimated area per station (mi^{2}) |
Median of long-term average precipitation statistics from measured data | ||||

Runoff-generating events per year | Annual runoff-generating precipitation (in/yr) | Event volume (in.) | Event duration (hr) | Delta (hr) | ||||

58 | Northeastern Highlands | 60 | 856 | 55 | 34.15 | 0.61 | 8.87 | 152 |

59 | Northeastern Coastal Zone | 33 | 481 | 51 | 37.31 | 0.71 | 9.76 | 157 |

84 | Atlantic Coastal Pine Barrens | 15 | 891 | 52 | 35.48 | 0.68 | 8.79 | 159 |

Geographic region (Connecticut, Massachusetts, and Rhode Island) | ||||||||

— | Southern New England | 45 | 784 | 52 | 36.37 | 0.69 | 8.86 | 154 |

Table 10. Synoptic-precipitation statistics from National Oceanic and Atmospheric Administration hourly precipitation-data stations that are in and adjacent to southern New England States

[Synoptic-precipitation statistics were calculated by using the definition of a runoff-generating event with a minimum interevent time of 6 hours and a minimum precipitation volume of 0.1 inch of liquid precipitation; the statistics are the medians of statistics from selected National Oceanic and Atmospheric Administration (NOAA) hourly precipitation-data stations with at least 25 years of data from 1965 to 2006 (

EPA Level III ecoregion | NOAA ID | Precipitation station name | State | Longitude | Latitude | Long-term average statistics from hourly precipitation data | Sensitivity analysis | ||||

Runoff-generating events per year | Annual runoff-generating precipitation (in/yr) | Event volume (in.) | Event duration (hr) | Delta (hr) | |||||||

58 | 065445 | NORFOLK 2 SW | CT | −73.217 | 41.967 | 67 | 46.39 | 0.69 | 10.73 | 126 | Y |

58 | 190666 | BIRCH HILL DAM | MA | −72.117 | 42.633 | 65 | 39.38 | 0.61 | 10.32 | 132 | N |

58 | 193985 | KNIGHTVILLE DAM | MA | −72.867 | 42.283 | 59 | 41.01 | 0.69 | 11.18 | 145 | N |

58 | 194075 | LANESBORO | MA | −73.233 | 42.55 | 50 | 31.65 | 0.63 | 9.12 | 148 | Y |

58 | 194246 | LITTLEVILLE LAKE | MA | −72.883 | 42.267 | 49 | 36.65 | 0.74 | 8.67 | 161 | N |

58 | 196322 | PETERSHAM 3 N | MA | −72.183 | 42.533 | 55 | 36.37 | 0.66 | 8.33 | 148 | N |

58 | 275013 | EDWARD MACDOWELL LAKE | NH | −71.983 | 42.9 | 53 | 37.15 | 0.7 | 8.51 | 153 | Y |

58 | 276550 | OTTER BROOK LAKE | NH | −72.233 | 42.95 | 51 | 30.8 | 0.61 | 7.46 | 159 | N |

58 | 278539 | SURRY MOUNTAIN LAKE | NH | −72.317 | 43 | 57 | 33.44 | 0.59 | 9.98 | 150 | N |

58 | 309670 | YORKTOWN HEIGHTS 1 W | NY | −73.8 | 41.267 | 55 | 40.69 | 0.74 | 7.61 | 151 | N |

58 | 430568 | BENNINGTON 3 N | VT | −73.183 | 42.917 | 57 | 33.08 | 0.58 | 7.18 | 139 | Y |

58 | 437152 | SEARSBURG STATION | VT | −72.917 | 42.867 | 62 | 41.99 | 0.67 | 10.61 | 134 | N |

58 | 438428 | TOWNSHEND LAKE | VT | −72.7 | 43.05 | 56 | 37.74 | 0.67 | 10.63 | 155 | N |

59 | 060806 | BRIDGEPORT SIKORSKY AP | CT | −73.15 | 41.183 | 63 | 41.53 | 0.66 | 10.78 | 140 | Y |

59 | 061488 | COCKAPONSET RS | CT | −72.517 | 41.467 | 42 | 34.38 | 0.82 | 7.44 | 165 | Y |

59 | 063451 | HARTFORD BRAINARD FLD | CT | −72.65 | 41.733 | 48 | 33.71 | 0.7 | 8.86 | 169 | N |

59 | 063456 | HARTFORD BRADLEY FLD | CT | −72.683 | 41.933 | 63 | 43.22 | 0.68 | 11.9 | 140 | N |

59 | 063857 | JEWETT CITY | CT | −71.9 | 41.633 | 52 | 37.31 | 0.71 | 8.08 | 157 | N |

59 | 064488 | MANSFIELD HOLLOW LAKE | CT | −72.183 | 41.75 | 62 | 42.31 | 0.69 | 10.3 | 141 | Y |

59 | 066942 | ROCKVILLE | CT | −72.433 | 41.867 | 45 | 31.39 | 0.7 | 7.92 | 170 | N |

59 | 068138 | STORRS | CT | −72.25 | 41.8 | 46 | 32.83 | 0.71 | 7.32 | 165 | N |

59 | 068330 | THOMASTON DAM | CT | −73.067 | 41.7 | 51 | 37.85 | 0.74 | 8.01 | 157 | Y |

59 | 069388 | WEST THOMPSON LAKE | CT | −71.9 | 41.95 | 45 | 34.27 | 0.76 | 7.53 | 167 | N |

59 | 190120 | AMHERST | MA | −72.533 | 42.383 | 41 | 30.03 | 0.73 | 7.96 | 187 | Y |

59 | 190408 | BARRE FALLS DAM | MA | −72.033 | 42.433 | 51 | 33.01 | 0.65 | 7.66 | 162 | N |

59 | 190575 | BELLINGHAM | MA | −71.483 | 42.1 | 50 | 37.92 | 0.76 | 9.76 | 154 | Y |

59 | 190736 | BLUE HILL OBS | MA | −71.117 | 42.217 | 66 | 48.5 | 0.73 | 12.69 | 133 | Y |

59 | 190770 | BOSTON/LOGAN AP | MA | −71.017 | 42.367 | 61 | 39.99 | 0.65 | 12.03 | 145 | N |

59 | 190840 | BRIDGEWATER | MA | −70.95 | 41.95 | 43 | 30.98 | 0.72 | 7.65 | 179 | Y |

59 | 190998 | BUFFUMVILLE LAKE | MA | −71.9 | 42.117 | 49 | 35.72 | 0.73 | 8.03 | 163 | N |

59 | 192107 | EAST BRIMFIELD LAKE | MA | −72.133 | 42.117 | 64 | 41.11 | 0.65 | 10.7 | 135 | Y |

59 | 195246 | NEW BEDFORD | MA | −70.933 | 41.633 | 62 | 44.78 | 0.72 | 12.36 | 135 | N |

59 | 199093 | WEST BRIMFIELD | MA | −72.267 | 42.167 | 50 | 33.83 | 0.68 | 7.99 | 164 | N |

59 | 199923 | WORCESTER RGNL AP | MA | −71.883 | 42.267 | 65 | 44.35 | 0.68 | 12.59 | 135 | N |

59 | 272174 | DURHAM | NH | −70.95 | 43.15 | 42 | 28.31 | 0.67 | 7.75 | 178 | Y |

59 | 301207 | CARMEL | NY | −73.683 | 41.433 | 41 | 29.52 | 0.72 | 8.86 | 171 | N |

59 | 306674 | PLEASANTVILLE | NY | −73.783 | 41.133 | 46 | 35.34 | 0.76 | 7.55 | 161 | Y |

59 | 307497 | SCARSDALE | NY | −73.8 | 40.983 | 54 | 39.42 | 0.73 | 10.02 | 142 | N |

59 | 309400 | WHITE PLAINS MPL MOOR | NY | −73.733 | 41.017 | 49 | 34 | 0.69 | 10 | 163 | N |

59 | 309576 | WOODLANDS ARDSLEY | NY | −73.85 | 41.017 | 58 | 41.58 | 0.72 | 10.83 | 145 | N |

59 | 375215 | NEWPORT ROSE | RI | −71.35 | 41.5 | 41 | 30.91 | 0.75 | 7.43 | 179 | N |

59 | 376698 | PROVIDENCE/GREEN STATE AP | RI | −71.433 | 41.717 | 62 | 43.6 | 0.7 | 11.56 | 142 | Y |

84 | 193821 | HYANNIS | MA | −70.3 | 41.667 | 52 | 34.62 | 0.67 | 9.43 | 152 | Y |

84 | 196681 | PROVINCETOWN | MA | −70.183 | 42.05 | 49 | 32.29 | 0.66 | 8.79 | 162 | N |

84 | 370896 | BLOCK ISLAND STATE AP | RI | −71.583 | 41.167 | 53 | 34.8 | 0.66 | 10.27 | 157 | N |

In this study, precipitation statistics for individual NOAA precipitation data stations also were used to do TMDL simulations (a level 2 analysis) to provide information about variations in long-term average yields and to do sensitivity analyses on the effect of variations in precipitation statistics on flows, concentrations, and loads of runoff-constituents of concern within the region. Statistics for the 45 precipitation stations within and adjacent to southern New England are shown in

Graph showing the precipitation event statistics for hourly precipitation-data stations in and adjacent to southern New England, the median of sites representing statistics for southern New England, and the medians of statistics for all hourly precipitation data stations within the Northeastern Highlands, Northeastern Coastal Zone, and Atlantic Coastal Pine Barrens U.S. Environmental Agency Level III ecoregions used for annual-yield analyses conducted in Connecticut, Massachusetts, and Rhode Island with the Stochastic Empirical Loading and Dilution Model.

Figure 8. Graph showing the precipitation event statistics for hourly precipitation-data stations in and adjacent to southern New England, the median of sites representing statistics for southern New England, and the medians of statistics for all hourly precipitation data stations within the Northeastern Highlands, Northeastern Coastal Zone, and Atlantic Coastal Pine Barrens U.S. Environmental Agency Level III ecoregions used for annual-yield analyses conducted in Connecticut, Massachusetts, and Rhode Island with the Stochastic Empirical Loading and Dilution Model

SELDM uses streamflow statistics to stochastically simulate a large series of prestorm streamflow volumes from the basin upstream from the point of interest (

Streamflow statistics for three EPA Level III ecoregions that include areas within and outside of Massachusetts, Connecticut, or Rhode Island, (the Northeastern Highlands, Northeastern Coastal Zone, and Atlantic Coastal Pine Barrens ecoregions;

Table 11. Streamflow statistics for the southern New England area and selected U.S. Environmental Protection Agency Level III ecoregions that lie in whole or in part within Connecticut, Massachusetts, or Rhode Island

[Statistics are for the retransformed common logarithms of nonzero flows and the proportion of zero flows; the statistics are the medians and ranges of statistics for each specified region or dataset. Southern New England is defined as the area within Connecticut, Massachusetts, and Rhode Island. The U.S. Environmental Protection Agency Level III ecoregions are defined in ^{2}, square mile; ft^{3}/s/mi^{2}, cubic foot per second per square mile; —, not applicable]

Ecoregion number | Regional dataset name | Number of streamgages | Streamgage density (mi^{2}) |
Long-term average streamflow statistics from measured data | |||||

Statistic | Drainage area (mi^{2}) |
Geometric mean (ft^{3}/s/mi^{2}) |
Geometric standard deviation (unitless) | Skew of logarithms (unitless) | Proportion of zero flow (unitless) | ||||

SELDM streamgages within |
|||||||||
---|---|---|---|---|---|---|---|---|---|

58 | Northeastern Highlands | 60 | 480 | Median | 126 | 1.09 | 2.9 | 0.093 | 0 |

Range | 10.9–491 | 0.0093–2.564 | 1.248–9.147 | −1.479–1.105 | 0–0.00786 | ||||

59 | Northeastern Coastal Zone | 33 | 201 | Median | 63.7 | 1.02 | 2.9 | −0.164 | 0 |

Range | 10.6–497 | 0.3288–1.528 | 1.854–5.717 | −1.437–1.357 | 0–0.02767 | ||||

84 | Atlantic Coastal Pine Barrens | 15 | 393 | Median | 35.2 | 1.04 | 1.96 | 0.108 | 0 |

Range | 10–123 | 0.0608–1.704 | 1.291–6.555 | −1.55–1.735 | 0–0.15288 | ||||

Southern New England region (Connecticut, Massachusetts, and Rhode Island) | |||||||||

— | SELDM | 106 | 163 | Median | 64 | 1.05 | 2.89 | −0.111 | 0 |

Range | 10.6–497 | 0.3288–1.775 | 1.854–5.717 | −1.437–1.357 | 0–0.02767 | ||||

— | 1901–2015 | 385 | 45 | Median | 20.2 | 1.03 | 2.94 | −0.213 | 0 |

Range | 0.35–9660 | 0.1266–6.413 | 1.295–15.195 | −4.738–3.253 | 0–0.26332 | ||||

— | Index | 73 | 483 | Median | 20.1 | 1.01 | 3.38 | −0.255 | 0 |

Range | 0.49–404 | 0.7095–2.695 | 1.448–5.439 | −0.9252–0.3671 | 0–0.01542 | ||||

Southern New England region (Connecticut, Massachusetts, and Rhode Island), streamgages with no zero flows | |||||||||

— | SELDM | 100 | — | Median | 87.9 | 1.06 | 2.86 | −0.104 | 0 |

Range | 12.1–183 | 0.329–1.77 | 1.85–5.37 | −0.948–1.357 | — | ||||

— | 1901–2015 | 330 | — | Median | 24.3 | 1.08 | 2.8 | −0.154 | 0 |

Range | 0.48–9660 | 0.206–6.41 | 1.3–9.76 | −4.738–3.253 | — | ||||

— | Index | 62 | — | Median | 29.7 | 1.05 | 3.25 | −0.217 | 0 |

Range | 0.59–404 | 0.742–2.69 | 1.45–5.44 | −0.925–0.367 | — | ||||

Southern New England region (Connecticut, Massachusetts, and Rhode Island), streamgages with one or more zero flows | |||||||||

— | SELDM | 6 | — | Median | 24.25 | 0.87 | 3.4 | −0.662 | 0.00277 |

Range | 10.6–24.25 | 0.869–0.678 | 3.4–2.71 | −0.662–−1.437 | 0.00277–0.00009 | ||||

— | 1901–2015 | 55 | — | Median | 4.96 | 0.796 | 4.76 | −0.775 | 0.00647 |

Range | 0.35–1544 | 0.127–1.26 | 2.76–15.2 | −1.703–0.675 | 0.00003–0.26332 | ||||

— | Index | 11 | — | Median | 4.96 | 0.809 | 4.47 | −0.527 | 0.00043 |

Range | 0.49–14.4 | 0.7095–1.16 | 4.19–5.12 | −0.874–−0.317 | 0.00005–0.01542 |

Because these three ecoregions also include large areas outside of southern New England, statistics from three other streamflow datasets also were evaluated as alternatives for simulating prestorm streamflows (

Although the drainage-area distributions and periods of record are different, statistics for the three datasets are similar (

Scatterplot showing the distribution of streamflow statistics to the percentage of basins greater than or equal to various values for streamflow from streamgages in the Stochastic Empirical Loading and Dilution Model (SELDM) database (

Figure 9. Scatterplot showing the distribution of streamflow statistics to the percentage of basins greater than or equal to various values for streamflow from streamgages in the Stochastic Empirical Loading and Dilution Model database, southern New England 1901–2015 dataset, and southern New England Index streamgage dataset

Information about relations between the average, standard deviation, and skew of nonzero streamflows is needed to guide the choice of a limited but representative set of values for simulating the potential effect of runoff on receiving waters. To this end, the nonparametric rank correlation coefficient (Spearman’s rho) was calculated among these statistics for each of the three streamflow datasets (

Table 12. Spearman’s rank-correlation coefficients for streamflow statistics for the common logarithms of nonzero flows from streamgages representative of conditions in southern New England

[Datasets (listed in ^{2}, square mile; ft^{3}/s, cubic foot per second; N, number of streamgages in each dataset]

Variable | Drainage area (mi^{2}) |
Geometric mean (ft^{3}/s) |
Geometric mean (ft^{3}/s/mi^{2}) |
Geometric standard deviation (unitless) | Skew of logarithms (unitless) |

SELDM: Stochastic Empirical Loading and Dilution Model database streamgages (N=106) | |||||
---|---|---|---|---|---|

Drainage area (mi^{2}) |
1 | 0.97 | 0.33 | −0.34 | 0.1 |

Geometric mean (ft^{3}/s) |
0.97 | 1 | 0.51 | −0.44 | 0.15 |

Geometric mean (ft^{3}/s/mi^{2}) |
0.33 | 0.51 | 1 | −0.74 | 0.39 |

Geometric standard deviation (unitless) | −0.34 | −0.44 | −0.74 | 1 | −0.67 |

Skew of logarithms (unitless) | 0.1 | 0.15 | 0.39 | −0.67 | 1 |

1901–2015: Southern New England 1901–2015 streamgages (N=385) | |||||

Drainage area (mi^{2}) |
1 | 0.98 | 0.34 | −0.3 | 0.22 |

Geometric mean (ft^{3}/s) |
0.98 | 1 | 0.43 | −0.3 | 0.14 |

Geometric mean (ft^{3}/s/mi) |
0.34 | 0.43 | 1 | −0.75 | 0.34 |

Geometric standard deviation (unitless) | −0.3 | −0.3 | −0.75 | 1 | −0.66 |

Skew of logarithms (unitless) | 0.22 | 0.14 | 0.34 | −0.66 | 1 |

Index: Southern New England index streamgages (N=73) | |||||

Drainage area (mi^{2}) |
1 | 0.99 | 0.25 | −0.51 | 0.4 |

Geometric mean (ft^{3}/s) |
0.99 | 1 | 0.34 | −0.57 | 0.45 |

Geometric mean (ft^{3}/s/mi) |
0.25 | 0.34 | 1 | −0.82 | 0.57 |

Geometric standard deviation (unitless) | −0.51 | −0.57 | −0.82 | 1 | −0.73 |

Skew of logarithms (unitless) | 0.4 | 0.45 | 0.57 | −0.73 | 1 |

Based on these correlations (

Table 13. Regression equation statistics developed by using the Kendall-Theil robust line method for estimating the mean, standard deviation, and skew of the common logarithms of streamflow data and the fraction of zero streamflows in southern New England

[The regression equations were developed by using the logarithms of data (

Dataset | KTRLine statistics (unitless) | Retransformed intercept (unitless) | ASEE (percent) | ||||

Intercept | Slope | RMSE | MAD | BCF | |||

I. Estimate the standard deviation from mean of the logarithms of normalized streamflow | |||||||
---|---|---|---|---|---|---|---|

SELDM | 0.47749 | −0.76816 | 0.07933 | 0.03686 | 1.007 | 3.003 | 18.4 |

1901–2015 | 0.47886 | −0.78480 | 0.09951 | 0.06033 | 1.059 | 3.012 | 23.2 |

Index | 0.53150 | −0.87608 | 0.05561 | 0.03423 | 1.008 | 3.400 | 12.9 |

II. Estimate the skew from standard deviation of the logarithms of normalized streamflow | |||||||

SELDM | 1.30110 | −3.0658 | 0.29736 | 0.15325 | 0.0100 | — | 265 |

1901–2015 | 0.96209 | −2.5050 | 0.49765 | 0.20034 | 0.0130 | — | 203 |

Index | 0.81446 | −2.0241 | 0.21790 | 0.16517 | −0.0340 | — | 79.9 |

III. Estimate the skew from mean of the logarithms of normalized streamflow | |||||||

SELDM | −0.15248 | 1.8916 | 0.43843 | 0.22111 | 0.0160 | — | 391 |

1901–2015 | −0.22941 | 1.3200 | 0.57195 | 0.31934 | −0.0250 | — | 233 |

Index | −0.26097 | 1.6613 | 0.24783 | 0.19030 | −0.0320 | — | 90.9 |

IV. Estimate the skew from mean of the logarithms of normalized streamflow algebraically | |||||||

SELDM | −0.16274 | 2.3550 | 0.45619 | 0.21020 | 0.0200 | — | 407 |

1901–2015 | −0.23746 | 1.9659 | 0.59764 | 0.32401 | −0.0220 | — | 243 |

Index | −0.26136 | 1.7733 | 0.24977 | 0.19619 | −0.0330 | — | 91.6 |

V. Estimate the logarithms of the fraction of zero flows from the mean of the logarithms of streamflow | |||||||

SELDM | 0.12346 | −2.1671 | 1.0737 | 0.51189 | 4.487 | 1.329 | 2120 |

1901–2015 | −1.7239 | −0.82338 | 0.82406 | 0.57027 | 4.426 | 0.0189 | 597 |

Index | −2.3955 | −1.6009 | 0.62519 | 0.34724 | 2.563 | 0.0040 | 263 |

SELDM uses the specified fraction of zero flows to simulate the effect of runoff on ephemeral or intermittent streams by using conditional-probability methods. When the prestorm streamflow value is zero, the runoff or BMP discharge volumes are likely to be a larger fraction of downstream flows than for similar runoff events with nonzero prestorm streamflows. The upstream flow also will depend on the upstream area and lagtime, which are deterministic variables in SELDM, and the upstream runoff coefficients and recession ratios, which are stochastic variables in SELDM. An ephemeral stream has no baseflow; it flows only in response to runoff. Therefore, in theory, the maximum fraction of zero prestorm streamflow for an ephemeral stream is 1. Because the EPA definition of a runoff-generating event has an interevent dry time of 6 hours, multiple runoff events could take place within one day. Given an average number of runoff-generating events per year equal to 52 (

Although the fraction of zero flow is commonly thought to be a function of drainage area, it also depends on physiography, geography, and water use. Streamflow statistics in the datasets selected to represent conditions in southern New England indicate that zero flows occur across a wide range of drainage areas (

Table 14. Percent of streamgages with one or more zero flows by drainage-area category from datasets selected to be representative of conditions in southern New England

[The Stochastic Empirical Loading and Dilution Model (SELDM) regional dataset includes all the streamgages in the southern New England States within SELDM (^{2}, square mile; N, number of streamgages in each dataset; —, not applicable]

Drainage-area range (mi^{2}) |
Count of total streamgages in each area range | Percent of streamgages in each area range with one or more zero flows | Minimum fraction of zero flows | Maximum fraction of zero flows |

SELDM streamgage dataset (N=106) | ||||
---|---|---|---|---|

≤1 | 0 | — | — | — |

>1–2 | 0 | — | — | — |

>2–10 | 0 | — | — | — |

>10–20 | 11 | 18.18 | 0.004908 | 0.027669 |

>20–30 | 18 | 16.67 | 0.000088 | 0.00064 |

>30–50 | 16 | 6.25 | 0.011528 | 0.011528 |

>50 | 61 | 0 | — | — |

1901–2015 streamgage dataset (N=385) | ||||

≤1 | 13 | 46.15 | 0.000342 | 0.066265 |

>1–2 | 17 | 23.53 | 0.003259 | 0.028169 |

>2–10 | 97 | 28.87 | 0.000054 | 0.263315 |

>10–20 | 64 | 9.38 | 0.000249 | 0.027608 |

>20–30 | 40 | 15.00 | 0.000064 | 0.004791 |

>30–50 | 41 | 7.32 | 0.000228 | 0.204473 |

>50 | 113 | 1.77 | 0.000031 | 0.008442 |

Index streamgage dataset (N=73) | ||||

≤1 | 4 | 50 | 0.010567 | 0.015419 |

>1–2 | 1 | 0 | — | — |

>2–10 | 21 | 33 | 0.000048 | 0.006773 |

>10–20 | 10 | 20 | 0.000384 | 0.000624 |

>20–30 | 7 | 0 | — | — |

>30–50 | 6 | 0 | — | — |

>50 | 24 | 0 | — | — |

Regression equations were developed to estimate the fraction of zero flows from the average of the logarithms of flow (

Although SELDM will simulate nonzero prestorm streamflows below the commonly used USGS minimum reported streamflow measurement of 0.01 cubic foot per second (ft^{3}/s; ^{3}/s is substituted into the frequency factor equation and it is rearranged to solve for the lognormal (if the skew is near 0) or log Pearson Type III (if the skew is substantially different from 0) frequency factor, then the resulting equation for the zero-streamflow reporting limit of 0.01 ft^{3}/s (

is the lognormal or log Pearson Type III frequency factor, which is a function of the skew value;

is the logarithm of the minimum streamflow reporting limit, in cubic feet per second (

is the logarithm of drainage area, in square miles;

is the average of the logarithms of streamflow, in cubic feet per second per square mile; and

is the standard deviation of the logarithms of streamflow.

The frequency-factor value (

SELDM simulates runoff from precipitation by using stochastic runoff coefficients, which are the ratio of the volume of runoff in watershed inches to the volume of basin-average precipitation (in inches) during each storm event (

In SELDM, runoff coefficient statistics can be calculated as a function of the total impervious area of the site of interest and the upstream basin by using regression equations (

Table 15. Regression equation statistics developed by using the Kendall-Theil robust line method for estimating the average, standard deviation, and skew of runoff coefficients from the total impervious fraction

[Regression equations were developed for use in the Stochastic Empirical Loading and Dilution Model (SELDM) and other applications by using rainfall and runoff data from 58 highway and 167 nonhighway sites (

Variable | KTRLine statistics for runoff coefficient statistics, using the impervious fraction | ||||||

Segment | Intercept | Slope | RMSE | MAD | MaxX | BCF | |

Highway sites (N=58) | |||||||
---|---|---|---|---|---|---|---|

Average | 1 | 0.03 | 0.755 | 0.169 | 0.14 | 1 | 0.055 |

Standard deviation | 1 | 0.229 | −0.0373 | 0.085 | 0.046 | 1 | −0.018 |

Skew | 1 | 2.13 | −3.32 | 1.46 | 0.748 | 1 | −0.565 |

Nonhighway sites (N=167) | |||||||

Average | 1 | 0.129 | 0.225 | 0.161 | 0.067 | 0.55 | 0.011 |

2 | −0.371 | 1.14 | 0.161 | 0.127 | 1 | 0.011 | |

Standard deviation | 1 | 0.099 | 0.015 | 0.07 | 0.047 | 1 | 0.015 |

Skew | 1 | 1.08 | −0.557 | 1.04 | 0.599 | 0.52 | 0.044 |

2 | 2.22 | −2.73 | 1.04 | 0.595 | 1 | 0.044 |

The timing of runoff from the upstream basin is defined by using the basin lagtime and the hydrograph recession ratio, which is the ratio of the duration of the falling limb to the rising limb (or time to peak) of the hydrograph (

Table 16. Best-fit triangular-hydrograph recession ratios estimated from 20 or more storm-event hydrographs at each listed U.S. Geological Survey streamgage in southern New England

[Basin properties were obtained from source reports (

Streamgage number | Name | Hydrograph-recession ratios | Basin properties (and basin-lag equation variables) | Sensitivity analysis | |||||||||

Min | MPV | Max | Avg | DRNAREA | LENGTH | CSL10_85 | BLF | NLCD Impervious % | Stream density | Reference | |||

01118300 | Pendleton Hill Brook near Clark Falls, CT | 1.35 | 1.35 | 6.05 | 2.92 | 4.02 | 3.92 | 76.5 | 0.45 | 0.41 | 1.58 | Current study | Y |

01115280 | Cork Brook at Rockland Scituate Rd near Clayville, RI | 1.18 | 1.18 | 6.89 | 3.08 | 1.79 | 3.28 | 80.62 | 0.37 | 2.12 | 1.2 | Current study | N |

01117468 | Beaver River near Usquepaug, RI | 1.08 | 1.08 | 4.36 | 2.17 | 8.87 | 8.82 | 33.36 | 1.53 | 1.28 | 1.6 | Current study | N |

01115187 | Ponaganset River at South Foster, RI | 1.35 | 1.35 | 3.91 | 2.2 | 14.4 | 8.09 | 46.73 | 1.18 | 1.07 | 2.59 | Current study | N |

01111300 | Nipmuc River near Harrisville, RI | 1 | 1 | 5.17 | 2.39 | 16 | 7.87 | 34.45 | 1.34 | 1.08 | 2.51 | Current study | N |

01117800 | Wood River near Arcadia, RI | 1.91 | 1.91 | 6.04 | 3.29 | 35.2 | 12.13 | 29.86 | 2.22 | 0.68 | 1.37 | Current study | N |

01115190 | Dolly Cole Brook at Old Danielson Park at S Foster, RI | 2.05 | 2.05 | 6.28 | 3.46 | 4.9 | 4.74 | 78.05 | 0.54 | 1.21 | 2.06 | Current study | Y |

01203510 | Pootatuck River at Sandy Hook, CT | 1.1 | 1.1 | 7.29 | 3.16 | 24.91 | 11.2 | 45.46 | 1.66 | 6.38 | 2.74 | Current study | N |

01188000 | Bunnell Brook near Burlington, CT | 1 | 1 | 2.92 | 1.64 | 4.1 | 4.23 | 52.18 | 0.58 | 2.37 | 2.66 | Current study | Y |

01195100 | Indian River near Clinton, CT | 1.19 | 1.19 | 4.28 | 2.22 | 5.68 | 6.84 | 68.87 | 0.82 | 2.43 | 3.95 | Current study | Y |

01208950 | Sasco Brook near Southport, CT | 1.15 | 1.15 | 3.58 | 1.96 | 7.38 | 6.02 | 54.26 | 0.82 | 5.94 | 3.44 | Current study | Y |

01184100 | Stony Brook near West Suffield, CT | 1.23 | 1.23 | 2.49 | 1.65 | 10.4 | 7.02 | 11.76 | 2.05 | 2.62 | 3.19 | Current study | N |

01208990 | Saugatuck River near Redding, CT | 1.08 | 1.08 | 5.73 | 2.63 | 21 | 12.06 | 29.51 | 2.22 | 1.37 | 2.77 | Current study | N |

01203805 | Weekeepeemee River at Hotchkissville, CT | 1 | 1 | 3.75 | 1.92 | 27.05 | 11.55 | 70.95 | 1.37 | 0.82 | 2.48 | Current study | N |

01123000 | Little River near Hanover, CT | 1 | 1 | 4.36 | 2.12 | 30 | 17.03 | 20.76 | 3.74 | 0.51 | 3.53 | Current study | N |

01187300 | Hubbard River near West Hartland, CT | 1.38 | 1.38 | 5.93 | 2.9 | 19.9 | 10.49 | 74.22 | 1.22 | 0.18 | 1.9 | Current study | N |

01187800 | Nepaug River near Nepaug, CT | 1 | 1 | 3.14 | 1.71 | 23.5 | 11.38 | 34.69 | 1.93 | 1.01 | 2.49 | Current study | N |

01194000 | Eightmile River at North Plain, CT | 1.22 | 1.22 | 3.46 | 1.97 | 20.1 | 10.33 | 48.99 | 1.48 | 0.65 | 2.66 | Current study | N |

01208873 | Rooster River at Fairfield, CT | 1.08 | 1.08 | 2.76 | 1.64 | 10.71 | 9.82 | 51.28 | 1.37 | 36.82 | 1.9 | Current study | N |

01111300 | Nipmuc River near Harrisville, RI | 1 | 2.53 | 5.73 | 3.09 | 16 | 7.79 | 30.4 | 1.41 | 1.08 | 2.51 | N | |

01187300 | Hubbard River near West Hartland, CT | 1.67 | 1.67 | 9.13 | 4.16 | 19.9 | 10.4 | 67.5 | 1.26 | 0.18 | 1.9 | N | |

01094400 | North Nashua River at Fitchburg, MA | 1 | 2.23 | 5 | 2.74 | 63.4 | 17.8 | 40.7 | 2.78 | 5.98 | 1.74 | N | |

01094500 | North Nashua River near Leominster, MA | 1 | 2.83 | 4.27 | 2.7 | 110 | 25.6 | 32.6 | 4.49 | 10.6 | 1.82 | N | |

01095220 | Stillwater River near Sterling, MA | 1.66 | 2.01 | 4.05 | 2.57 | 30.4 | 11.5 | 39.3 | 1.83 | 1.52 | 1.67 | N | |

01096000 | Squannacook River near West Groton, MA | 1.03 | 1.95 | 2.66 | 1.88 | 64.4 | 18.3 | 41.7 | 2.83 | 2.13 | 2.01 | N | |

01097000 | Assabet River at Maynard, MA | 1 | 1.83 | 5.04 | 2.62 | 116 | 28.1 | 4.69 | 13 | 11.1 | 2.5 | Y | |

01097300 | Nashoba Brook near Acton, MA | 1.16 | 1.16 | 3.18 | 1.83 | 12.9 | 5.83 | 8.62 | 1.99 | 7.85 | 2.75 | Y | |

01100600 | Shawsheen River near Wilmington, MA | 1.08 | 1.34 | 3.08 | 1.83 | 36.5 | 16.2 | 8.61 | 5.52 | 25.3 | 1.68 | N | |

01102500 | Aberjona River at Winchester, MA | 1 | 1 | 5.11 | 2.37 | 24.1 | 10.3 | 9.64 | 3.32 | 40.6 | 1.81 | N | |

01103280 | Charles River at Medway, MA | 1.16 | 2.34 | 9.67 | 4.39 | 65.7 | 21.4 | 7.83 | 7.65 | 14.2 | 2.33 | N | |

01105500 | East Branch Neponset River at Canton, MA | 1.25 | 2.2 | 6.23 | 3.23 | 27.2 | 8.32 | 23.4 | 1.72 | 20 | 2.36 | Y | |

01105600 | Old Swamp River near South Weymouth, MA | 1 | 1 | 3.39 | 1.8 | 4.47 | 4.76 | 10.3 | 1.49 | 25.5 | 2 | N | |

01105730 | Indian Head River at Hanover, MA | 1.77 | 1.85 | 4.62 | 2.75 | 30.2 | 13.3 | 9.92 | 4.24 | 14.8 | 2.29 | N | |

01108000 | Taunton River near Bridgewater, MA | 1 | 1.58 | 5.44 | 2.67 | 258 | 33.5 | 3.63 | 17.6 | 9.71 | 2.09 | N | |

01109000 | Wading River near Norton, MA | 1.02 | 2.22 | 3.99 | 2.41 | 43.3 | 19.6 | 7.55 | 7.12 | 9.22 | 2.12 | N | |

01109060 | Threemile River at North Dighton, MA | 1 | 1.22 | 5.56 | 2.59 | 84.3 | 32.5 | 5.91 | 13.3 | 10.8 | 2.37 | N | |

01109070 | Segreganset River near Dighton, MA | 1.41 | 1.46 | 4.75 | 2.54 | 10.6 | 7.36 | 8.67 | 2.5 | 3.94 | 2.41 | N | |

01111200 | West River below West Hill Dam, near Uxbridge, MA | 1 | 2.68 | 4.47 | 2.72 | 27.8 | 13.3 | 13.7 | 3.61 | 2.55 | 2.68 | N | |

01162500 | Priest Brook near Winchendon, MA | 1.2 | 2.51 | 4.56 | 2.76 | 19.2 | 15.2 | 19 | 3.49 | 0.52 | 1.82 | N | |

01163200 | Otter River at Otter River, MA | 1.48 | 2.25 | 3.48 | 2.4 | 34.1 | 12.3 | 16.5 | 3.02 | 9.14 | 1.86 | N | |

01169000 | North River at Shattuckville, MA | 1.18 | 2.87 | 3.24 | 2.43 | 89.9 | 22.6 | 49 | 3.23 | 0.56 | 2.05 | N | |

01169900 | South River near Conway, MA | 1 | 1.38 | 3.81 | 2.06 | 24.1 | 14.6 | 58.1 | 1.91 | 0.89 | 2.05 | Y | |

01170100 | Green River near Colrain, MA | 1 | 2.07 | 4.09 | 2.39 | 41.3 | 19.1 | 59.4 | 2.48 | 0.24 | 2.38 | Y | |

01171500 | Mill River at Northampton, MA | 1.37 | 1.5 | 3.9 | 2.26 | 54 | 18 | 76.1 | 2.06 | 1.94 | 1.97 | N | |

01173500 | Ware River at Gibbs Crossing, MA | 1 | 1 | 4.42 | 2.14 | 197 | 43.7 | 25.1 | 8.72 | 1.22 | 2.04 | N | |

01174565 | West Branch Swift River near Shutesbury, MA | 1 | 2.64 | 6.22 | 3.29 | 12.6 | 7.83 | 61 | 1 | 0.24 | 1.89 | N | |

01174600 | Cadwell Creek near Pelham, MA | 1.21 | 2.02 | 4.02 | 2.42 | 0.6 | 1.91 | 129 | 0.17 | 0.43 | 1.93 | N | |

01174900 | Cadwell Creek near Belchertown, MA | 1 | 2.77 | 4.56 | 2.78 | 2.89 | 3.95 | 135 | 0.34 | 0.17 | 1.99 | N | |

01175670 | Sevenmile River near Spencer, MA | 1.59 | 1.67 | 8.6 | 3.95 | 8.69 | 7.95 | 39.4 | 1.27 | 0.71 | 2.84 | Y | |

01181000 | West Branch Westfield River at Huntington, MA | 1.22 | 2.13 | 6.11 | 3.15 | 93.7 | 23.3 | 44.2 | 3.5 | 0.43 | 1.6 | N | |

01331500 | Hoosic River at Adams, MA | 1.24 | 2.1 | 4.12 | 2.49 | 46.7 | 14.4 | 10.3 | 4.49 | 1.52 | 1.74 | N |

Table 17. Summary statistics for the best-fit triangular-hydrograph recession ratios estimated from 20 or more storm-event hydrographs at each listed U.S. Geological Survey streamgage in southern New England

[Basin properties were obtained from source reports or calculated by using StreamStats. Min, minimum; MPV, most probable value; Max, maximum; Avg, the average of the three recession-ratio statistics; DRNAREA, drainage area in square miles; LENGTH, main-channel length in miles; CSL10_85, main-channel slope in feet per mile; BLF, basin-lag factor, which is the basin length (LENGTH) in miles divided by the square root of the channel slope (CSL10_85) in feet per mile; NLCD, National Land Cover Database; Impervious %, total impervious area in percent; SELDM, Stochastic Empirical Loading and Dilution Model; —, not applicable]

Statistic | Hydrograph-recession ratios | Basin properties (and basin-lag equation variables) | ||||||||

Min | MPV | Max | Avg | DRNAREA | LENGTH | CSL10_85 | BLF | NLCD Impervious % | Stream density | |

Minimum | 1 | 1 | 2.49 | 1.5 | 0.6 | 1.91 | 3.63 | 0.17 | 0.17 | 1.2 |

Median | 1.1 | 1.5 | 4.42 | 2.34 | 24.1 | 11.38 | 34.69 | 1.99 | 1.52 | 2.06 |

Average | 1.2 | 1.67 | 4.8 | 2.56 | 38.43 | 13.28 | 39.99 | 3.14 | 5.96 | 2.23 |

Maximum | 2.05 | 2.87 | 9.67 | 4.86 | 258 | 43.7 | 135 | 17.6 | 40.6 | 3.95 |

SELDM default values | 1 | 1.85 | 4.4 | 2.42 | — | — | — | — | — | — |

Spearman’s rank correlation analyses were done in an attempt to provide guidance on the selection of hydrograph recession-ratio statistics by using basin properties. The rank correlations among recession-ratio statistics were about 0.193 between the minimum and the most probable value, about 0.189 between the minimum and maximum, and about 0.228 between the most probable value and the maximum. Correlations between the three triangular-distribution statistics and the drainage area, main-channel length, main-channel slope, basin-lag factor, and imperviousness were of mixed sign and had absolute values ranging from 0.00027 and 0.292. These low correlations indicate that basin properties cannot be used to quantitatively select recession-ratio statistics. These results are similar to the results of correlation analyses done by

Statistics were calculated for 21 water-quality properties and constituents of concern (

Table 18. Runoff-quality constituents analyzed in this study with counts of the number of highway-runoff sites, urban-runoff sites, and the best management practice treatment analysis method

[Pcode is the water-quality parameter code denoted by the letter p and the five-digit identification number from the

Pcode | Constituent name in NWIS | HN | UN | SQ | BMP |

Water-quality properties | |||||
---|---|---|---|---|---|

p00076 | Turbidity, water, unfiltered, nephelometric turbidity units | 12 | 35 | 21 | M |

Sediment and solids | |||||

p00530 | Solids, suspended, water, milligrams per liter | 19 | 241 | 7 | M |

p80154 | Suspended sediment concentration, milligrams per liter | 18 | 30 | 35 | M |

Nutrient constituents, unfiltered | |||||

p00600 | Total nitrogen, water, unfiltered, milligrams per liter | 18 | 67 | 65 | M |

p62855 | Total nitrogen [nitrate + nitrite + ammonia + organic nitrogen], analytically determined, in milligrams per liter | 16 | 67 | 6 | M |

p00665 | Phosphorus, water, unfiltered, milligrams per liter | 19 | 196 | 69 | M |

Minor and trace inorganics, unfiltered | |||||

p01027 | Cadmium, water, unfiltered, micrograms per liter | 13 | 49 | 0 | M |

p01034 | Chromium, water, unfiltered, recoverable, micrograms per liter | 13 | 32 | 0 | M |

p01042 | Copper, water, unfiltered, recoverable, micrograms per liter | 13 | 146 | 0 | M |

p01051 | Lead, water, unfiltered, recoverable, micrograms per liter | 13 | 97 | 0 | M |

p01067 | Nickel, water, unfiltered, recoverable, micrograms per liter | 13 | 36 | 0 | M |

p01092 | Zinc, water, unfiltered, recoverable, micrograms per liter | 13 | 169 | 0 | M |

p71900 | Mercury, water, unfiltered, recoverable, micrograms per liter | 15 | 6 | 0 | S |

Organic constituents | |||||

pXXX05 | PAHs EPA 8310, water, unfiltered, micrograms per liter, (Sum of 16 PAHs not censored) | 12 | 8 | 0 | M |

Biological constituents | |||||

p31616 | Fecal coliform, M-FC MF (0.45 micron) method, water, colonies per 100 milliliters | 12 | 29 | 19 | M |

p31625 | Fecal coliforms, M-FC MF (0.7 micron) method, water, colony forming units per 100 milliliters | 12e | 29e | 12 | S |

P31649 | Enterococci, m-E MF method, water, colony forming units per 100 milliliters | 4 | 7 | 13 | S |

p31673 | Fecal streptococci, KF streptococcus MF method, water, colony forming units per 100 milliliters | 4eu | 4 | 13 | S |

p50468 | 7 | 11 | 14 | M | |

p50569 | Total coliforms, defined substrate test method (DSTM), water, most probable number per 100 milliliters | 4 | 8e | 7 | S |

Major ionic constituents | |||||

p00940 | Chloride, water, filtered, milligrams per liter | 13 | 58 | 40 | M |

Available data for simulating runoff and receiving water quality are limited in comparison to the number of sites where estimates of water quality may be needed (

Although the nominally dissolved (filtered) fraction of many constituents is of regulatory concern, the whole-water (unfiltered) concentrations were simulated because sediment concentrations and the distribution between the filtered and unfiltered fractions can change as runoff travels from developed and agricultural surfaces through conveyances and stormwater treatment facilities (

SELDM provides three methods for simulating stormflow quality (

All concentrations were simulated by using statistics for the logarithms of data. The logarithms of concentration and stormflow commonly are analyzed and simulated by using the logarithms of the data because these variables vary by orders of magnitude and are bounded by zero (

Highway-runoff quality statistics for commonly measured properties and constituents (

Table 19. Statistics for the common logarithms of data used to simulate highway-runoff quality in southern New England with the Stochastic Empirical Loading and Dilution Model (SELDM)

[Pcode is a water-quality parameter code denoted by the letter p and the five-digit identification number from the

Pcode | Logarithmic statistics—Median | Average estimate—Low | Average estimate—High | |||

Average | Standard deviation | Skew | Percentage not skewed | |||

Water−quality properties | ||||||
---|---|---|---|---|---|---|

p00076 | 1.53 | 0.631 | 0 | 100 | 1.24 | 1.636 |

Sediment and related constituents | ||||||

p00530 | 1.666 | 0.3132 | 0 | 100 | 1.296 | 1.891 |

p80154 | 1.905 | 0.5726 | 0 | 83 | 1.628 | 2.866 |

Nutrient constituents, unfiltered | ||||||

p00600 | 0.054 | 0.256 | 0 | 89 | −0.146 | 0.218 |

p62855 | 0.047 | 0.2583 | 0 | 100 | −0.132 | 0.245 |

p00665 | −0.944 | 0.3627 | 0 | 95 | −1.172 | −0.271 |

Minor and trace inorganics, unfiltered | ||||||

p01027 | −0.772 | 0.4931 | 0 | 85 | −1.023 | −0.508 |

p01034 | 1.07 | 0.389 | 0 | 100 | 0.695 | 1.202 |

p01042 | 1.433 | 0.3783 | 0 | 100 | 1.098 | 1.623 |

p01051 | 0.909 | 0.5265 | 0 | 100 | 0.593 | 1.134 |

p01067 | 0.606 | 0.4792 | 0 | 100 | 0.333 | 0.746 |

p01092 | 2.093 | 0.4057 | 0 | 100 | 1.654 | 2.287 |

P71900 | −2.121 | 0.0915 | 0 | 87 | −2.151 | −2.056 |

Organic constituents | ||||||

pXXX05 | 0.284 | 0.5514 | 0 | 100 | −0.097 | 0.628 |

Biological constituents | ||||||

p31616 | 3.052 | 0.5135 | 0 | 92 | 2.867 | 3.345 |

p31625* | 3.052 | 0.5135 | 0 | 92 | 2.867 | 3.345 |

P31649 | 3.399 | 0.649 | 0 | 100 | 3.27 | 3.428 |

p31673** | 4.07 | 0.449 | 0 | 100 | 3.562 | 4.809 |

p50468 | 3.01 | 0.7118 | 0 | 86 | 2.346 | 3.107 |

p50569*** | 3.861 | 0.5729 | 0 | 75 | 3.511 | 4.119 |

Major ionic constituents | ||||||

p00940 | 1.747 | 1.022 | 0 | 85 | 1.47 | 1.975 |

[Pcode is a water-quality parameter code denoted by the letter p and the five-digit identification number from the

Correlation between the average and standard deviation | Correlation between the average and AADT | Source | ||

Spearman’s rho | 95-percent confidence intervals of the correlation coefficient value | Spearman’s rho | 95−percent confidence intervals of the correlation coefficient value | |

Water-quality properties | ||||
---|---|---|---|---|

−0.17 | −0.72−0.51 | 0.45 | −0.24−0.84 | MA 2009 |

Sediment and related constituents | ||||

−0.108 | −0.57−0.41 | 0.3 | −0.23−0.69 | NC 2011, NH 2015 |

−0.08 | −0.54−0.42 | 0.44 | −0.05−0.76 | MA 2002, 2009, 2017 |

Nutrient constituents, unfiltered | ||||

−0.003 | −0.5−0.49 | 0.28 | −0.25−0.68 | NC 2011, MA 2017 |

−0.29 | −0.71−0.28 | 0.76 | 0.38−0.92 | MA 2009, 2017 |

−0.386 | −0.73−0.12 | 0.51 | 0.04−0.8 | MA 2002, 2009, 2017 |

Minor and trace inorganics, unfiltered | ||||

−0.15 | −0.69−0.49 | 0.91 | 0.69−0.98 | MA 2009, 2010 |

−0.4 | −0.8−0.26 | 0.76 | 0.3−0.93 | MA 2009, 2010 |

−0.324 | −0.77−0.34 | 0.8 | 0.39−0.95 | MA 2009, 2010 |

−0.522 | −0.91−−0.32 | 0.8 | 0.56−0.95 | MA 2009, 2010 |

−0.538 | −0.86−0.09 | 0.9 | 0.65−0.97 | MA 2009, 2010 |

−0.379 | −0.8−0.28 | 0.84 | 0.49−0.96 | MA 2009, 2010 |

0.229 | −0.37−0.69 | 0.004 | −0.55−0.55 | NC 2011 |

Organic constituents | ||||

−0.18 | −0.72−0.5 | 0.93 | 0.73−0.98 | MA 2009 |

Biological constituents | ||||

−0.011 | −0.63−0.62 | 0.19 | −0.49−0.73 | CA 2018, WA 2015 |

−0.011 | −1−1 | 0.19 | −1−0.99 | CA 2018, WA 2015 |

0 | −1−1 | −0.2 | −1−0.99 | CA 2018, SC 2008 |

−0.8 | −1−0.97 | NA | NA | BMPSE Urban runoff |

0.393 | −0.67−0.93 | −0.43 | −0.93−0.64 | CA 2018, OR 2016, SC 2008 |

−0.4 | −1−0.99 | −0.8 | −1−0.97 | CA 2018, SC 2008 |

Major ionic constituents | ||||

−0.58 | −0.87−0.03 | 0.39 | −0.27−0.8 | MA 2009 |

Estimated from p31616.

Estimated from urban-runoff quality data.

Estimated from concentrations for two sites with p50569 data and two sites with p31507 (total coliform, completed test, water, most probable number per 100 milliliters) data.

Rank correlation (rho) analysis using Spearman’s rho was used to evaluate the correlation between the average and standard deviation of the logarithms of concentrations to determine whether the values used for simulation could be selected independently. Rank correlation values for highway-runoff constituents ranged from −0.8 to 0.393 with a median correlation of −0.18 (

Concentrations of highway runoff for all constituents were simulated by using a skew value of 0. This value was selected because the percentage of datasets with skew values that were not significantly different from 0 at the 95-percent confidence limit (

Rank correlation analysis using Spearman’s rho also was used to examine relations between the geometric mean concentrations of constituents and the AADT reported for each highway monitoring site. These rank correlation values ranged from −0.8 to 0.93 with a median correlation of 0.445 (

Table 20. Regression equation statistics developed by using the Kendall-Theil robust line method for estimating the average of the common logarithms of highway-runoff constituents from the common logarithms of average daily traffic volumes

[Regression equations developed for constituents with statistically significant Spearman's rank correlation coefficients greater than 0.5 in

Pcode | Number of pairs | Intercept | Slope | MAD | RMSE | ASEE (percent) |

Nutrient constituents, unfiltered | ||||||
---|---|---|---|---|---|---|

p62855 | 16 | −1.3436 | 0.30245 | 0.05676 | 0.10546 | 24.6 |

p00665 | 19 | −2.0356 | 0.23646 | 0.17960 | 0.39132 | 112 |

Minor and trace inorganics, unfiltered | ||||||

p01027 | 13 | −3.3174 | 0.55352 | 0.06998 | 0.13683 | 32.3 |

p01034 | 13 | −0.64391 | 0.37269 | 0.10397 | 0.18525 | 44.7 |

p01042 | 13 | −0.97216 | 0.52301 | 0.09180 | 0.30056 | 78.4 |

p01051 | 13 | −1.2288 | 0.46496 | 0.09303 | 0.24062 | 59.9 |

p01067 | 13 | −0.95930 | 0.34036 | 0.02890 | 0.32466 | 86.5 |

p01092 | 13 | 0.09490 | 0.43449 | 0.13888 | 0.23461 | 58.2 |

Organic constituents | ||||||

pXXX05 | 12 | −2.2863 | 0.55804 | 0.08692 | 0.16006 | 38.1 |

Although regression equations are provided, the results will not be used for simulations in this study because there are many complications for application of specific AADT estimates. AADT may not be the primary causal variable. Increased AADT is associated with increases in the imperviousness of land covers within a mile radius of highway-runoff monitoring sites (

National urban-runoff quality statistics for commonly measured properties and constituents (

Table 21. Statistics for the common logarithms of national urban-runoff quality data used to simulate developed-area runoff quality in southern New England with the Stochastic Empirical Loading and Dilution Model (SELDM)

[Statistics were calculated by using the Best Management Practices (BMPs) Statistical Estimator version 1.2.0 (

Pcode | Average | Standard deviation | Skew | Percentage not skewed | Spearman’s rho, average standard deviation | ||||

Water-quality properties | |||||||||
---|---|---|---|---|---|---|---|---|---|

p00076 | 1.33 | 0.306 | 0 | 91 | 0.17 | ||||

Sediment and related constituents | |||||||||

p00530 | 1.68 | 0.381 | 0 | 90 | 0.01 | ||||

p80154 | 2.08 | 0.47 | 0 | 90 | 0.19 | ||||

Nutrient constituents, unfiltered | |||||||||

p00600 | 0.158 | 0.258 | 0 | 82 | −0.04 | ||||

p62855 | — | — | — | — | — | ||||

p00665 | −0.760 | 0.295 | 0 | 82 | −0.05 | ||||

Minor and trace inorganics, unfiltered | |||||||||

p01027 | −0.276 | 0.274 | 0 | 88 | −0.33 | ||||

p01034 | 0.647 | 0.218 | 0 | 94 | −0.05 | ||||

p01042 | 1.13 | 0.258 | 0 | 86 | −0.05 | ||||

p01051 | 1 | 0.343 | 0 | 88 | −0.19 | ||||

p01067 | 0.712 | 0.271 | 0 | 97 | −0.08 | ||||

p01092 | 1.82 | 0.265 | 0 | 86 | −0.09 | ||||

p71900 | 0.536 | 0.313 | 0 | 100 | 0.66 | ||||

Organic constituents | |||||||||

pXXX05 | −0.422 | 0.417 | 0 | 100 | 0.07 | ||||

Biological constituents | |||||||||

p31616 | 3.49 | 0.801 | 0 | 90 | 0.04 | ||||

p31625* | 3.49 | 0.801 | 0 | 90 | 0.04 | ||||

P31649 | 3.32 | 0.687 | 0 | 86 | 0.18 | ||||

p31673 | 4.07 | 0.449 | 0 | 100 | — | ||||

p50468 | 3.25 | 0.694 | 0 | 100 | −0.15 | ||||

p50569** | 4.26 | 0.39 | 0 | 50 | −0.4 | ||||

Major ionic constituents | |||||||||

p00940 | 1.12 | 0.522 | 0 | 75 | 0.5 |

Estimated from statistics for p31616, fecal coliform, M-FC MF (0.45 micron) method.

Estimated from statistics for p31507, total coliform, completed test, water, most probable number per 100 milliliters.

Rank correlation analysis using Spearman’s rho was used to evaluate the cross-correlation between the average and standard deviation of the logarithms of concentrations to determine whether the values used for simulation could be selected independently. Rank correlation (rho) values for urban-runoff constituents ranged from −0.4 to 0.66 with a median correlation of −0.040 (

Concentrations of urban runoff for all constituents were simulated by using a skew value of 0. This value was selected because the percentage of datasets with skew values that were not significantly different from 0 at the 95-percent confidence limit (

The risk assessment process is the foundation of the regulatory framework for numeric and narrative water-quality criteria (

Available stormwater data can provide information about the distribution of event mean concentrations (EMCs) to estimate the potential for exceeding a specified concentration standard or an assigned load limit, but statistics calculated from available data must be extrapolated to estimate long-term exceedance probabilities. Version 1.1.0 of the HRDB contains 106,441 concentration values with data for 414 different water-quality constituents (

Table 22. Stream water-quality monitoring stations on minimally developed, developed, and wastewater-affected receiving streams that were used to develop individual and categorical transport-curve statistics for simulating upstream water quality in southern New England with the Stochastic Empirical Loading and Dilution Model (SELDM)

[Streamgage names can be found in the National Water Information System (NWIS, ^{2}, square mile; MA, Massachusetts; CT, Connecticut; RI, Rhode Island; —, data not available]

USGS streamgage number | USGS streamgage name | NWIS drainage area, in mi^{2} |
Number of WWTP | Density of road crossing per mi^{2} |
NLCD 2011 land cover, in percent of drainage area | ||||

Crop/hay | Wetlands | Forest | Developed | Impervious area | |||||

Minimally developed (MD) | |||||||||
---|---|---|---|---|---|---|---|---|---|

01174575 | Dickey Brook Tributary Near Cooleyville, MA | 1.06 | 0 | 3.8 | 0.40 | 6.24 | 92.5 | 0.28 | 0.07 |

01174570 | Dickey Brook Near Cooleyville, MA | 1.19 | 0 | 2.5 | 0 | 13.8 | 84.6 | 0 | 0.14 |

01198122 | Ironworks Brook, On East Rd., At Sheffield, MA | 11.2 | 0 | 2.1 | 4.05 | 11.5 | 76.7 | 4.28 | 0.19 |

01174565 | West Branch Swift River Near Shutesbury, MA | 12.6 | 0 | 1.0 | 0.51 | 2.71 | 92.5 | 3.09 | 0.25 |

01170100 | Green River Near Colrain, MA | 41.4 | 0 | 2.8 | 3.12 | 1.81 | 89.1 | 4.23 | 0.28 |

01187850 | Clear Bk Nr Collinsville, CT | 0.59 | 0 | 1.7 | 0 | 0.44 | 92.9 | 4.31 | 0.29 |

01125415 | Muddy Brook At Childs Hill Rd Nr Woodstock, CT | 20.2 | 0 | 3.3 | 14.2 | 13.9 | 64.0 | 6.06 | 0.67 |

01172680 | Natty P Bk Templeton Rd (Ds) Nr Hubbardston, MA | 1.63 | 0 | 2.5 | 2.71 | 9.02 | 80.1 | 6.69 | 0.73 |

01115110 | Huntinghouse Bk At Elmdale Rd At N Scituate, RI | 6.23 | 0 | 1.6 | 2.56 | 8.95 | 81.5 | 5.78 | 0.74 |

01135300 | Sleepers River (Site W-5) Near St. Johnsbury, VT | 42.9 | 0 | — | 13.5 | 2.81 | 75.3 | 4.85 | 0.87 |

01118356 | Ashway River At Extension 184 Near Ashway, RI | 26.6 | 0 | 2.0 | 5.84 | 12.9 | 74.3 | 4.04 | 0.89 |

01187830 | Phelps Brook At Mill Dam Road Near Collinsville, CT | 2.7 | 0 | 4.8 | 5.61 | 14.9 | 71.3 | 7.27 | 0.91 |

01201020 | Lake Waramaug Bk Nr Warren, CT (Inflow Site 26) | 6.6 | 0 | 4.7 | 9.37 | 11.2 | 72.2 | 6.06 | 0.94 |

04282636 | Little Otter Cr @ Middlebrk Rd, Nr Ferrisburg, VT | 43.4 | 0 | — | 44.7 | 9.95 | 39.8 | 4.74 | 0.94 |

04282634 | Little Otter Cr Ab Middlebrk Rd Nr Ferrisburg, VT | 42.4 | 0 | — | 44.3 | 10.1 | 39.9 | 4.77 | 0.95 |

01201030 | Lk Waramaug Bk Nr New Preston, CT (Inflow Site 2) | 8.59 | 0 | 4.7 | 10.3 | 8.86 | 73.3 | 6.01 | 0.95 |

01169900 | South River Near Conway, MA | 24.1 | 0 | 3.4 | 7.89 | 2.66 | 81.9 | 6.37 | 0.96 |

01201010 | Lake Waramaug Bk At Warren, CT (Inflow Site 7) | 3.37 | 0 | 5.0 | 2.86 | 17.3 | 71.5 | 6.72 | 1.08 |

01142500 | Ayers Brook At Randolph, VT | 30.5 | 0 | — | 17.8 | 2.39 | 72.3 | 6.11 | 1.08 |

01187800 | Nepaug R Nr Nepaug, CT | 23.5 | 0 | 3.2 | 5.98 | 6.90 | 78.2 | 7.43 | 1.14 |

01208990 | Saugatuck River Near Redding, CT | 21 | 0 | 4.9 | 1.31 | 9.26 | 74.9 | 12.4 | 1.36 |

01197802 | Williams River, At Railroad Br, Nr Gt. Barrington, MA | 43.2 | 0 | 3.3 | 6.01 | 9.17 | 74.3 | 7.32 | 1.38 |

01118360 | Ashaway River At Ashaway, RI | 28.6 | 0 | 2.1 | 6.07 | 13.3 | 72.4 | 5.23 | 1.43 |

01172800 | Natty Pond Brook Near Hubbardston, MA | 5.48 | 0 | 1.8 | 7.13 | 19.0 | 63.7 | 7.86 | 1.5 |

01117420 | Usquepaug River Near Usquepaug, RI | 36.1 | 0 | 1.6 | 5.06 | 14.6 | 68.1 | 7.91 | 1.67 |

01125436 | Tributary To Mill Bk At South Woodstock, CT | 0.24 | 0 | 0 | 48.6 | 9.34 | 31.8 | 8.82 | 1.88 |

01117471 | Beaver River Shannock Hill Rd, Near Shannock, RI | 11.2 | 0 | 1.4 | 5.14 | 11.7 | 72.9 | 6.75 | 1.96 |

01125435 | Tributary To Mill Bk At Woodstock, CT | 0.19 | 0 | 0 | 45.6 | 11.0 | 32.9 | 8.84 | 2.07 |

01073554 | Exeter River At Wells Village Rd, Near Sandown, NH | 6.52 | 0 | — | 11.6 | 9.80 | 65.6 | 6.40 | 2.11 |

01073572 | Fordway Brook At Lane Road, Near Raymond, NH | 5.79 | 0 | — | 0.63 | 12.6 | 71.4 | 9.51 | 2.19 |

01118055 | Tomaquag Brook, At Rt. 216, At Bradford, RI | 6.71 | 0 | 2.7 | 6.33 | 18.8 | 64.9 | 7.33 | 2.22 |

01195100 | Indian River Near Clinton, CT | 5.68 | 0 | 6.9 | 1.44 | 9.33 | 73.2 | 14.1 | 2.23 |

01188000 | Bunnell Brook Near Burlington, CT | 4.1 | 0 | 4.4 | 10.2 | 9.80 | 66.7 | 11.5 | 2.36 |

01073562 | Towle Brook At Towle Road, Near Chester, NH | 2.5 | 0 | — | 4.63 | 10.7 | 69.7 | 11.6 | 2.56 |

01118400 | Shunock River Near North Stonington, CT | 17.2 | 0 | 2.7 | 7.99 | 13.1 | 69.1 | 7.73 | 2.56 |

01184100 | Stony Brook Near West Suffield, CT | 10.4 | 0 | 4.0 | 17.1 | 28.1 | 41.3 | 8.70 | 2.76 |

01094340 | Whitman River Near Westminster, MA | 21.6 | 0 | 2.5 | 4.13 | 9.38 | 70.0 | 9.59 | 2.76 |

01192883 | Coginchaug River At Middlefield, CT | 29.8 | 0 | 5.7 | 11.5 | 8.27 | 64.7 | 13.5 | 2.84 |

01115114 | Rush Brook Near Elmdale Rd Near North Scituate, RI | 4.7 | 0 | 2.3 | 4.08 | 16.5 | 69.4 | 9.25 | 3.07 |

01184490 | Broad Brook At Broad Brook, CT | 15.5 | 0 | 3.4 | 29.6 | 6.5 | 45.2 | 17.8 | 3.90 |

01195399 | Farm River At Totoket Road At Totoket, CT | 12.9 | 0 | 4.5 | 13.8 | 6.54 | 58.4 | 15.9 | 4.33 |

01115183 | Quonapaug Bk At Rt 116 Nr North Scituate, RI | 1.96 | 0 | 2.6 | 4.08 | 20.1 | 58.4 | 16.2 | 4.43 |

01104405 | Hobbs Brook At Mill St Nr Lincoln, MA | 2.16 | 0 | 6.0 | 2.23 | 29.8 | 51.1 | 15.9 | 4.56 |

01109070 | Segreganset River Near Dighton, MA | 10.6 | 0 | 2.5 | 3.08 | 23.7 | 57.0 | 14.7 | 4.87 |

01101000 | Parker River At Byfield, MA | 21.3 | 0 | 4.8 | 5.58 | 23.9 | 51.1 | 16.8 | 4.95 |

Developed (D) | |||||||||

01104390 | Stony Brook At Kendal Green, MA | 10.4 | 0 | 3.8 | 2.95 | 19.9 | 48.2 | 24.4 | 5.60 |

01208950 | Sasco Brook Near Southport, CT | 7.38 | 0 | 10 | 0.51 | 10.7 | 49.5 | 38.3 | 5.88 |

01192370 | Porter Brook Near Manchester, CT | 2.2 | 0 | 4.1 | 4.67 | 3.58 | 61.9 | 27.4 | 6.36 |

011277916 | Stony Brook At Rt 1 Near Flanders, CT | 1.86 | 0 | 9.7 | 0.07 | 16.7 | 60.5 | 18.4 | 7.24 |

01098340 | Course Brook At Natick, MA | 3.44 | 0 | 2.0 | 6.68 | 14.2 | 57.6 | 20.7 | 7.51 |

01104475 | Stony Brook Res., Unnamed Trib 1, Near Weston, MA | 0.85 | 0 | 11 | 0.82 | 7.37 | 32.7 | 57.9 | 9.55 |

01163200 | Otter River At Otter River, MA | 34.1 | 0 | 4.0 | 3.26 | 20.1 | 46.5 | 22.2 | 9.67 |

01109000 | Wading River Near Norton, MA | 43.3 | 0 | 3.9 | 1.77 | 21.0 | 42.5 | 30.9 | 11.3 |

01108410 | Mill River At Spring Street At Taunton, MA | 43.5 | 0 | 4.6 | 2.61 | 23.3 | 38.9 | 31.2 | 11.5 |

01117351 | White Horn Brk At Ministerial Rd Nr W Kingston, RI | 3.94 | 0 | 1.3 | 4.93 | 19.1 | 38.5 | 28.4 | 11.8 |

01098450 | Snake Brook At Wayland, MA | 2.1 | 0 | 7.1 | 0.65 | 11.8 | 37.6 | 49.1 | 13.3 |

01190045 | Podunk R At South Windsor, CT | 3.74 | 0 | 3.7 | 9.45 | 15.8 | 21.1 | 52.8 | 14.8 |

01105000 | Neponset River At Norwood, MA | 34.7 | 0 | 4.7 | 1.07 | 14.7 | 32.8 | 46.6 | 17.9 |

01192704 | Mattabesset River At Route 372 At East Berlin, CT | 48.1 | 0 | 5.0 | 3.93 | 7.25 | 36.9 | 48.0 | 18.3 |

01073040 | College Brook At Mill Pond Road, At Durham, NH | 0.88 | 0 | — | 5.60 | 4.52 | 22.7 | 62.6 | 24.2 |

01195490 | Quinnipiac River At Southington, CT | 17.4 | 0 | 5.6 | 0.83 | 5.06 | 25.9 | 65.2 | 24.7 |

01105583 | Monatiquot River At East Braintree, MA | 28.7 | 0 | 4.6 | 0.29 | 12.1 | 23.5 | 59.5 | 29.0 |

01105600 | Old Swamp River Near South Weymouth, MA | 4.5 | 0 | 3.8 | 0.02 | 14.9 | 19.3 | 65.2 | 29.6 |

01102345 | Saugus River At Saugus Ironworks At Saugus, MA | 20.8 | 0 | 7.5 | 0.13 | 10.4 | 17.4 | 66.7 | 30.9 |

01098320 | Beaverdam Brook At Natick, MA | 7.27 | 0 | 5.6 | 0.83 | 10.9 | 16.9 | 68.6 | 35.4 |

01208873 | Rooster River At Fairfield, CT | 10.6 | 0 | 12 | 0.01 | 0.63 | 5.43 | 92.9 | 36.6 |

01098360 | Pegan Brook At Natick, MA | 0.54 | 0 | 14.8 | 0.19 | 0.89 | 10.2 | 88.6 | 46.1 |

01100568 | Shawsheen River At Hanscom Field Near Bedford, MA | 2.13 | 0 | 0.5 | 0 | 0 | 16.4 | 83.6 | 49.7 |

01103025 | Alewife Brook Near Arlington, MA | 8.36 | 0 | 2.3 | 0.13 | 1.26 | 5.16 | 87.1 | 53.1 |

Wastewater affected (WA) | |||||||||

01199050 | Salmon Creek At Lime Rock, CT | 29.4 | 1 | 2.1 | 7.37 | 8.77 | 72.8 | 5.52 | 0.92 |

01192050 | Hockanum R At Rockville, CT | 25.5 | 1 | 3.5 | 11.9 | 11.1 | 50.3 | 22.4 | 6.37 |

01209700 | Norwalk River At South Wilton, CT | 30 | 2 | 7.6 | 0.77 | 8.42 | 65.3 | 24.2 | 5.91 |

01209572 | Norwalk River At Cannondale, CT | 15.2 | 2 | 7.6 | 0.49 | 9.65 | 64.2 | 24.6 | 6.29 |

01209570 | Norwalk R At Georgetown CT | 14.5 | 2 | 7.9 | 0.49 | 10.0 | 63.6 | 24.9 | 6.34 |

01209710 | Norwalk River At Winnipauk,CT | 33 | 2 | 8.1 | 0.70 | 8.03 | 61.5 | 28.6 | 7.53 |

01189000 | Pequabuck R At Forestville, CT | 45.8 | 2 | 6.0 | 2.95 | 4.73 | 50.0 | 39.8 | 13.9 |

01122610 | Shetucket R At South Windham, CT | 408 | 3 | 4.1 | 4.98 | 10.1 | 71.8 | 9.89 | 2.07 |

01125500 | Quinebaug River At Putnam, CT | 328 | 3 | 4.9 | 5.88 | 12.1 | 63.4 | 13.4 | 3.78 |

01208370 | Naugatuck R Below Fulling Mills Bk At Union City, CT | 215 | 3 | 5.4 | 4.86 | 4.40 | 61.6 | 26.2 | 9.53 |

01112900 | Blackstone River At Manville, RI | 430.63 | 3 | 5.0 | 2.95 | 9.43 | 54.8 | 27.7 | 11.5 |

01125520 | Quinebaug River At Cotton Bridge Road Nr Pomfret, CT | 342 | 4 | 4.8 | 6.17 | 12.1 | 63.0 | 13.6 | 3.87 |

01208500 | Naugatuck River At Beacon Falls, CT | 260 | 4 | 5.4 | 4.71 | 4.43 | 61.4 | 26.6 | 9.51 |

Neither the available data nor simulated results match the commonly used regulatory exposure durations (

Many water-quality constituents of potential concern do not have numeric water-quality criteria, but risk-assessment techniques are still relevant to decision making. Risk-assessment techniques can be used to assess changes in risks for various flow, concentration, and load levels upstream from an area of interest, from an area of interest, and downstream from discharge locations. Risk-based analyses also can be used to assess the potential for mitigation measures to reduce risks for adverse effects on receiving waters.

Scatterplots showing the populations of simulated event mean phosphorus concentrations for highway-runoff quality, stormwater best management practice discharge quality, and receiving stream stormflow upstream and downstream from a discharge point.

Figure 10. Scatterplots showing the populations of simulated event mean phosphorus concentrations for highway-runoff quality, stormwater best management practice discharge quality, and receiving stream stormflow upstream and downstream from a discharge point

Table 23. Water-quality transport-curve statistics developed by using the Kendall-Theil robust line method for estimating the common logarithms of constituent concentrations from the common logarithms of area-normalized streamflow. Regression statistics represent the median of statistics for minimally developed, developed, and wastewater-affected receiving streams in southern New England

[Method used is defined in

Category | Number of sites | Method | Segment 1 | Segment 2 | ||||||

Intercept | Slope | MAD | MaxX | Intercept | Slope | MAD | MaxX | |||

p00076 (turbidity, water, unfiltered, nephelometric turbidity units) | ||||||||||
---|---|---|---|---|---|---|---|---|---|---|

MD | 9 | M | 0.03897 | 0.00000 | 0.15359 | 0.08588 | 0.00000 | 0.45378 | 0.16307 | 1.4557 |

D | 4 | M | 0.34245 | 0.00000 | 0.14689 | 0.00000 | 0.34245 | 0.17643 | 0.17329 | 1.4671 |

WA | 8 | M | 0.36537 | 0.00000 | 0.15287 | 0.10080 | 0.34397 | 0.21231 | 0.16375 | 1.3819 |

p00530 (solids, suspended, water, milligrams per liter) | ||||||||||

MD | 3 | M | 0.77249 | 0.00000 | 0.12444 | 1.1100 | — | — | — | — |

D | 1 | S | 1.9763 | −0.08161 | 0.14541 | 1.0278 | — | — | — | — |

WA | 3 | M | 0.76948 | 0.05686 | 0.21489 | 1.3044 | — | — | — | — |

p80154 (suspended sediment concentration, in milligrams per liter) | ||||||||||

MD | 25 | M | 0.57600 | 0.00000 | 0.24304 | 0.22096 | 0.39409 | 0.82328 | 0.23574 | 2.1362 |

D | 5 | S | 0.38908 | 0.00000 | 0.38908 | 0.07595 | 0.28295 | 1.3973 | 0.41979 | 1.6754 |

WA | 5 | M | 0.72911 | 0.00000 | 0.29905 | 0.21552 | 0.70286 | 0.12180 | 0.24419 | 1.2017 |

p00600 (total nitrogen, water, unfiltered, milligrams per liter) | ||||||||||

MD | 36 | M | −0.2745 | 0.00000 | 0.07942 | 2.1362 | — | — | — | — |

D | 22 | M | 0.07122 | −0.04087 | 0.05770 | −0.98785 | 0.05984 | −0.05239 | 0.06405 | 2.4531 |

WA | 7 | M | −0.0782 | −0.08845 | 0.08241 | 1.8777 | — | — | — | — |

p62855 (total nitrogen [nitrate + nitrite + ammonia + organic nitrogen], analytically determined, in milligrams per liter) | ||||||||||

MD | 4 | S | −0.45204 | 0.08054 | 0.09724 | 0.74451 | −0.53917 | 0.19757 | 0.07975 | 2.0233 |

D | 1 | S | 0.26190 | −0.10615 | 0.07128 | 0.34933 | — | — | — | — |

WA | 1 | S | −0.08175 | 0.07390 | 0.07059 | 1.2016 | — | — | — | — |

p00665 (phosphorus, water, unfiltered, milligrams per liter) | ||||||||||

MD | 38 | M | −1.6978 | 0.00000 | 0.14724 | 0.51602 | −1.7904 | 0.17945 | 0.16093 | 2.1362 |

D | 24 | S | −1.6021 | 0.05494 | 0.14585 | 0.38540 | −1.7407 | 0.41477 | 0.21629 | 2.4531 |

WA | 7 | M | −1.1493 | −0.08588 | 0.16553 | 0.00000 | −1.1493 | −0.02593 | 0.16553 | 1.8777 |

p01027 (cadmium, whole water, in micrograms per liter) | ||||||||||

MD | 0 | — | — | — | — | — | — | — | — | — |

D | 0 | — | — | — | — | — | — | — | — | — |

WA | 4 | M | −0.77276 | −0.00549 | 0.09203 | 1.0344 | — | — | — | — |

p01034 (chromium, water, unfiltered, recoverable, micrograms per liter) | ||||||||||

MD | 0 | — | — | — | — | — | — | — | — | — |

D | 0 | — | — | — | — | — | — | — | — | — |

WA | 1 | S | −0.40895 | 0.14228 | 0.11803 | 0.89449 | — | — | — | — |

p01042 (copper, whole water, in micrograms per liter) | ||||||||||

MD | 0 | — | — | — | — | — | — | — | — | — |

D | 0 | — | — | — | — | — | — | — | — | — |

WA | 6 | M | 0.64983 | −0.15124 | 0.07167 | 0.11889 | 0.65978 | −0.23493 | 0.07568 | 1.0344 |

p01051 (lead, whole water, in micrograms per liter) | ||||||||||

MD | 0 | — | — | — | — | — | — | — | — | — |

D | 0 | — | — | — | — | — | — | — | — | — |

WA | 6 | M | 0.04861 | 0.00000 | 0.16249 | 1.0344 | — | — | — | — |

p01067 (nickel, water, unfiltered, recoverable, micrograms per liter) | ||||||||||

MD | 0 | — | — | — | — | — | — | — | — | — |

D | 0 | — | — | — | — | — | — | — | — | — |

WA | 3 | M | 0.04139 | 0.00000 | 0.06427 | 0.86063 | — | — | — | — |

p01092 (zinc, whole water, in micrograms per liter) | ||||||||||

MD | 0 | — | — | — | — | — | — | — | — | — |

D | 0 | — | — | — | — | — | — | — | — | — |

WA | 6 | M | 1.2756 | 0.00061 | 0.08465 | 0.00000 | 1.2756 | 0.10344 | 0.09367 | 1.0223 |

p31616 (fecal coliforms, M-FC MF (0.45 micron) method, water, colony forming units per 100 milliliters) | ||||||||||

MD | 8 | M | 1.3471 | −0.09694 | 0.4178 | 1.4557 | — | — | — | — |

D | 4 | M | 2.1756 | −0.26212 | 0.41659 | 1.4671 | — | — | — | — |

WA | 7 | M | 2.3758 | 0.17540 | 0.46008 | 1.1532 | — | — | — | — |

p31625 (fecal coliforms, M-FC MF (0.7 micron) method, water, colony forming units per 100 milliliters) | ||||||||||

MD | 3 | M | 1.32737 | −0.24402 | 0.29579 | 1.3740 | — | — | — | — |

D | 3 | M | 2.21585 | 0.29300 | 0.48891 | 1.3740 | — | — | — | — |

WA | 6 | M | 2.40109 | 0.02460 | 0.34465 | 1.3740 | — | — | — | — |

P31649 (enterococci, m-E MF method, water, colony forming units per 100 milliliters) | ||||||||||

MD | 6 | M | 1.1693 | −0.70448 | 0.43240 | 1.1506 | — | — | — | — |

D | 2 | M | 1.9989 | 0.10378 | 0.42241 | 1.0753 | — | — | — | — |

WA | 5 | M | 1.6113 | 0.46071 | 0.50162 | 1.2991 | — | — | — | — |

p31673 (fecal streptococci, KF streptococcus MF method, water, colony forming units per 100 milliliters) | ||||||||||

MD | 4 | M | 1.6568 | −0.08343 | 0.64784 | 1.3915 | — | — | — | — |

D | 2 | M | 1.9982 | −0.02228 | 0.52789 | 1.3651 | — | — | — | — |

WA | 7 | M | 2.1890 | −0.10196 | 0.53979 | 1.3190 | — | — | — | — |

p50468 ( |
||||||||||

MD | 6 | M | 1.9340 | 0.05395 | 0.37856 | 1.9678 | — | — | — | — |

D | 3 | M | 2.2189 | −0.15414 | 0.43140 | 2.4531 | — | — | — | — |

WA | 5 | M | 2.0678 | 0.20847 | 0.32180 | 1.8777 | — | — | — | — |

p50569 (total coliforms, defined substrate test method (DSTM), water, most probable number per 100 milliliters) | ||||||||||

MD | 1 | M | 2.6014 | −0.20815 | 0.17622 | 0.45310 | — | — | — | — |

D | 1 | M | 2.1028 | 0.45174 | 0.33282 | 0.67200 | — | — | — | — |

WA | 5 | M | 2.2044 | 0.00000 | 0.31280 | 1.8777 | — | — | — | — |

p00940 (chloride, water, filtered, milligrams per liter) | ||||||||||

MD | 15 | M | 1.1026 | −0.03807 | 0.07188 | 1.9678 | — | — | — | — |

D | 13 | M | 1.6157 | −0.08522 | 0.06600 | 1.6754 | — | — | — | — |

WA | 12 | M | 1.5208 | −0.20826 | 0.09120 | 1.8777 | — | — | — | — |

Risk-based analyses using SELDM also can be used to evaluate numeric water-quality criteria.

SELDM-derived estimates also can be used to quantitatively examine water-quality exceedance risks in the absence of a numeric criteria. With or without a fixed criterion, SELDM can be used to examine the changes in concentrations that may occur between the upstream and downstream concentrations or between upstream concentrations simulated by using different statistics to determine if such changes are large enough to be detectable with monitoring data. For example,

Water-quality transport curves, which are regression relations between streamflow and instream concentrations (

Transport curves were developed for one water-quality property (turbidity), sediment and solids, unfiltered nutrients, unfiltered minor and trace inorganics (metals), biological constituents, and one major ion (chloride). SELDM uses the selected regression model and error term to generate stochastic estimates of water-quality concentrations for the constituent of interest (

Water-quality transport curves were developed using the nonparametric Kendall-Theil robust line method as implemented in the software program KTRLine version 1.0 (

Data from 82 stream water-quality monitoring stations within and adjacent to southern New England (

Candidate sites were reviewed in GIS software to exclude stations located immediately downstream from limited-access highways or within 0.25 miles of a reservoir or pond outlet. Stations located immediately downstream from limited-access highways were excluded from the analysis because the objective was to characterize background water quality above roadway stormwater outfalls. Stations located downstream from impoundments were excluded from the analysis because particle settling, biological action, and flow mixing in impoundments were expected to change or obscure relations between streamflow and constituent concentrations. Examination of transport curves developed from selected stations indicates that downstream proximity to wetlands may have similar effects, but transport curves for these stations were retained because riparian wetlands are a common feature in stream basins of southern New England. Stations that met the data and geographic requirements were used to determine a drainage-area threshold that allowed for sufficient data to be retained while minimizing the effect of large drainage areas. Larger basins are more likely to include impoundments that alter the flow regime and affect water quality (

Among the 82 stations for which transport curves were developed, 36 are located in Connecticut, 28 are located in Massachusetts, and 10 are located in Rhode Island (^{2} and the maximum is 430.63 mi^{2}; however, stations greater than 50 mi^{2} were only used to develop transport curves for concentrations of metals. Stream stations were stratified by imperviousness and the presence of WWTPs to facilitate selection of representative upstream water quality at unmonitored sites. Minimally developed stream basins were defined as basins without WWTPs and TIA values less than or equal to 5 percent above the monitoring location; 45 monitoring sites met these criteria (

Data for available constituents at individual stations were used to develop 347 water-quality transport curves, which are available in the model-archive data release for this project (

Within SELDM, an additional water-quality transport-curve segment was added for flows above the largest value in the water-quality dataset at the individual monitoring station to reduce the risk of simulating unrealistic instream concentrations, which may take place if positive- or negative-slope segments of the transport curve are extended far beyond available data. This step is necessary because SELDM is designed to simulate runoff events that may take place over a long period of time, the number of simulated events can greatly exceed the number of stream samples available for developing the transport curve, and simulated stormflows over long periods may exceed the range of flows at which water-quality samples were collected. To develop the additional transport-curve segment, the maximum flow value (MaxX,

Dependent water-quality relations, which are equations used to estimate concentrations of one constituent from another more readily available constituent (

Table 24. Regression equation statistics developed by using the Kendall-Theil robust line method for estimating the common logarithms of dependent concentrations from the common logarithms of predictor concentrations

[All regression relations developed by using available data from U.S. Geological Survey streamflow-quality values for a given constituent (

Water-quality data category | Type of relation | Number of pairs | Segment 1 | |||

Intercept | Slope | MAD | MaxX | |||

Estimate p00530 (total suspended solids, milligrams per liter) from p80154 (suspended sediment concentration, milligrams per liter) | ||||||
---|---|---|---|---|---|---|

All data | KTRLine | 40 | 0.13945 | 0.66207 | 0.20179 | 1.9085 |

NonWWTP | KTRLine | 21 | 0.08805 | 0.50000 | 0.18399 | 1.7482 |

WWTP | KTRLine | 19 | 0.17569 | 0.79210 | 0.14613 | 1.9085 |

Estimate p62855 (total nitrogen, milligrams per liter) from p00600 (total nitrogen, milligrams per liter) | ||||||

All data | KTRLine | 1432 | −0.01261 | 0.99603 | 0.02112 | 0.66276 |

Estimate p00600 (total nitrogen, milligrams per liter) from p62855 (total nitrogen, milligrams per liter) | ||||||

All data | KTRLine | 1432 | 0.00805 | 0.97298 | 0.02054 | 0.63649 |

Estimate particulate-associated cadmium from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04049; |
SedC | 13 | −3.8938 | 1.0000 | 0.17150 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p34825; |
SedC | 15 | −3.0270 | 1.0000 | 0.25193 | — |

Streambed sediment, smaller than 2 mm, stage 2 recoverable (p53686; |
SedC | 14 | −4.4686 | 1.0000 | 0.30381 | — |

Streambed sediment, smaller than 2 mm, stage 1 recoverable (p01028; |
SedC | 15 | −3.4035 | 1.0000 | 0.22145 | — |

Estimate particulate-associated chromium from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04052; |
SedC | 14 | −1.7366 | 1.0000 | 0.31950 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p34840; |
SedC | 15 | −1.1602 | 1.0000 | 0.21255 | — |

Streambed sediment, smaller than 2 mm, stage 2 recoverable (p53689; |
SedC | 14 | −2.9708 | 1.0000 | 0.48193 | — |

Streambed sediment, smaller than 2 mm, stage 1 recoverable (p01029; |
SedC | 14 | −2.9684 | 1.0000 | 0.48019 | — |

Estimate particulate-associated copper from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04054; |
SedC | 14 | −2.4117 | 1.0000 | 0.37764 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p34850; |
SedC | 15 | −1.3544 | 1.0000 | 0.30824 | — |

Streambed sediment, smaller than 2 mm, stage 2 recoverable (p53691; |
SedC | 14 | −2.6740 | 1.0000 | 0.23776 | — |

Streambed sediment, smaller than 2 mm, stage 1 recoverable (p01043; |
SedC | 14 | −2.8297 | 1.0000 | 0.83926 | — |

Estimate particulate-associated lead from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04130; |
SedC | 14 | −1.7470 | 1.0000 | 0.15080 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p34890; |
SedC | 15 | −1.1094 | 1.0000 | 0.28541 | — |

Streambed sediment, smaller than 2 mm, stage 2 recoverable (p53707; |
SedC | 14 | −2.9709 | 1.0000 | 0.25861 | — |

Streambed sediment, smaller than 2 mm, stage 1 recoverable (p01052; |
SedC | 15 | −2.4314 | 1.0000 | 0.48871 | — |

Estimate particulate-associated nickel from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04132; |
SedC | 14 | −2.1504 | 1.0000 | 0.30958 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p34925; |
SedC | 15 | −1.5018 | 1.0000 | 0.17898 | — |

Streambed sediment, smaller than 2 mm, stage 2 recoverable (p53705; |
SedC | 14 | −2.9720 | 1.0000 | 0.26135 | — |

Streambed sediment, smaller than 2 mm, stage 1 recoverable (p01068; |
SedC | 14 | −3.2816 | 1.0000 | 0.64887 | — |

Estimate particulate-associated zinc from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04131; |
SedC | 14 | −1.4394 | 1.0000 | 0.18526 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p35020; |
SedC | 15 | −0.69867 | 1.0000 | 0.22469 | — |

Streambed sediment, smaller than 2 mm, stage 2 recoverable (p53723; |
SedC | 14 | −2.1926 | 1.0000 | 0.22176 | — |

Streambed sediment, smaller than 2 mm, stage 1 recoverable (p01093; |
SedC | 14 | −2.2344 | 1.0000 | 0.56463 | — |

Estimate particulate-associated mercury from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Streambed sediment smaller than 2 mm, total digestion (p04133; |
SedC | 14 | −4.7598 | 1.0000 | 0.38823 | — |

Streambed sediment smaller than 0.0625 mm, total digestion (p34910; |
SedC | 15 | −3.7780 | 1.0000 | 0.33741 | — |

Estimate particulate-associated PAH, in micrograms per liter from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Background soil ( |
SedC | 18 | −2.2028 | 1.0000 | 0.3253 | — |

Soil near pavement ( |
SedC | 42 | −1.8956 | 1.0000 | 0.452775 | — |

Streambed sediment (TIA < 6%; |
SedC | 14 | −3.2374 | 1.0000 | 0.96119 | — |

Estimate total PAH, in micrograms per liter from p80154 (suspended sediment concentration, milligrams per liter) | ||||||

Background soil times 1.04 ( |
SedC | 18 | −2.1858 | 1.0000 | 0.32530 | — |

Soil near pavement times 1.04 ( |
SedC | 42 | −1.8786 | 1.0000 | 0.45278 | — |

Streambed sediment (TIA<6%) times 1.04 ( |
SedC | 14 | −3.2204 | 1.0000 | 0.96119 | — |

Dependent water-quality relations for total metal concentrations were developed by using a five-step process developed by

First the suspended-sediment transport curves were developed by using concurrent streamflow and concentration data (

In the third step for developing dependent water-quality relations for total metal concentrations, regression relations developed by _{i}_{i}_{i}_{i}

is the logarithm of an individual

is the average of the logarithms of _{i}

is the standard deviation of the logarithms of

is a value of a standard-normal variate for the

Table 25. Dependent water-quality relations calculated by using metal-sediment distribution coefficients and regression relations between results of simulations for suspended sediment and particulate metals developed by using the Kendall-Theil robust line method

[Calculated statistics are for the common logarithms of distribution coefficient and concentration data and are for the logarithms of distribution coefficient and concentration values from

Particulate metal | Kd equations from SSC ( |
Total concentration (CT) from particulate concentration (CP) | Total concentration (CT) from suspended-sediment concentration (SSC) | |||||||

Average Kd | Standard deviation of Kd | |||||||||

Intercept | Slope | Intercept | Slope | Intercept | Slope | MAD | Intercept | Slope | MAD | |

Cadmium | 5.9299 | −0.83 | 0.4378 | 0.02 | 0.19376 | 0.96424 | 0.15980 | −4.0885 | 0.93897 | 0.24832 |

Chromium | 5.6335 | −0.53 | 0.3160 | 0.02 | 0.14095 | 0.92655 | 0.11432 | −2.4798 | 0.80608 | 0.34553 |

Copper | 5.1399 | −0.43 | 0.4362 | −0.02 | 0.11837 | 0.75449 | 0.19577 | −1.8019 | 0.65575 | 0.24101 |

Lead | 5.9289 | −0.45 | 0.3674 | −0.03 | 0.00606 | 0.91709 | 0.07397 | −2.6583 | 0.85751 | 0.18857 |

Nickel | 4.5944 | −0.06 | −0.4318 | 0.74 | −0.2046 | 0.57888 | 0.17579 | −1.7497 | 0.40090 | 0.22893 |

Zinc | 5.6628 | −0.51 | 0.6848 | −0.12 | 0.18566 | 0.88364 | 0.18453 | −1.6708 | 0.80294 | 0.22250 |

Mercury | 5.6884 | −0.58 | 0.4281 | −0.01 | −0.06412 | 0.94007 | 0.07219 | −4.4449 | 0.85690 | 0.27310 |

In the fourth step for developing dependent water-quality relations for total metal concentrations, individual simulated suspended sediment concentrations, particulate metal concentrations, and associated

is the total or whole water concentration, in micrograms per liter for the

is the particulate metal concentration, in micrograms per liter for the

In the fifth step for developing dependent water-quality relations for whole-water metal concentrations, the populations of simulated suspended sediment and particulate metal concentrations were used with the calculated whole-water concentrations to develop dependent relations between the simulated and calculated values by using the Kendall-Theil Robust Line software (

Boxplots showing the distribution of total whole-water metal concentrations in minimally developed basins simulated by using stochastic distribution coefficients, regression relations to simulated particulate-metal concentrations, and regression relations to simulated suspended-sediment concentrations.

Figure 11. Boxplots showing the distribution of total whole-water metal concentrations in minimally developed basins simulated by using stochastic distribution coefficients, regression relations to simulated particulate-metal concentrations, and regression relations to simulated suspended-sediment concentrations

There also was insufficient data to estimate total PAHs in minimally developed basins, so particulate associated concentrations from bed sediment and soil were estimated and these estimates were adjusted to represent whole-water total PAH statistics as a function of simulated suspended-sediment concentrations.

The sediment-associated PAH equations were adjusted to represent whole-water concentrations by using a simple ratio because the partitioning chemistry is complex and is different for individual PAH constituents. Although PAHs are commonly associated with suspended sediment rather than from the dissolved fraction in the water column, many PAH’s are soluble; for example, the solubility of individual PAHs in water ranges from 0.26 µg/L for Benzo(ghi)perylene to 34,400 µg/L for Naphthalene (

Point-source discharges, such as wastewater treatment plants (WWTPs), add to the base-load of many constituents in basins where they are present. Methods are needed to assess potential effects of stormwater runoff from a point of interest on the quality of receiving waters in wastewater-affected basins. Water-quality transport curves were developed by using flow and water-quality data from 13 wastewater-affected water-quality monitoring sites with drainage areas ranging from 14.5 to 431 square miles and impervious percentages ranging from 0.916 to 13.9 percent (

Discharge quality and volume data from 30 to 143 WWTP facilities each located in Connecticut, Massachusetts, and Rhode Island were obtained from the EPA Enforcement and Compliance History website (

Table 26. Logarithmic regression relations between wastewater treatment plant-design flow and the average and standard deviation of monthly average constituent discharge loads for facilities in Connecticut, Massachusetts, and Rhode Island

[Method described in

Parameter code | Parameter description from NWIS | Number of concentration and flow pairs | Average of monthly average load | Standard deviation of monthly average load | ||||||||

Spearman's rho | Intercept | Slope | MAD | Spearman's rho | Standard deviation | |||||||

Rho | Lower 95% confidence interval | Upper 95% confidence interval | Rho | Lower 95% confidence interval | Upper 95% confidence interval | |||||||

p01027 | Total cadmium | 30 | 0.568 | 0.262 | 0.771 | 0.563 | 0.906 | 0.335 | −0.14 | −0.477 | 0.232 | 0.589 |

p01042 | Total copper | 93 | 0.902 | 0.856 | 0.934 | 2.302 | 1.005 | 0.17 | −0.151 | −0.344 | 0.054 | 0.299 |

p01051 | Total lead | 44 | 0.678 | 0.477 | 0.811 | 1.044 | 0.98 | 0.293 | −0.218 | −0.484 | 0.084 | 0.529 |

p00600 | Total nitrogen | 78 | 0.925 | 0.884 | 0.951 | 5.435 | 1.091 | 0.225 | −0.32 | −0.506 | −0.105 | 0.23 |

p00665 | Total phosphorus | 143 | 0.821 | 0.759 | 0.868 | 3.947 | 1.116 | 0.369 | −0.113 | −0.272 | 0.052 | 0.33 |

p01092 | Total zinc | 43 | 0.890 | 0.806 | 0.940 | 2.733 | 1.079 | 0.26 | −0.366 | −0.601 | −0.074 | 0.253 |

Spearman’s rank correlation (rho) values and their 95-percent confidence intervals were calculated for the average and standard deviation of the logarithms of each constituent load and the associated WWTP design flows. The design flows were used to predict the average and standard deviation of actual monthly flows because the design flows are readily available in permits and do not vary without a change in WWTP design (

Developing planning-level estimates of concentrations in the stream upstream from the runoff outfall is a six-step process. The first step is to simulate a population of stormflows and concentrations from a minimally developed or developed basin without a point-source discharge. The example shown in

Scatterplots showing example applications of wastewater-treatment statistics used to develop a water-quality transport curve adjusted to represent the effects of point-source discharges. The steps include using

Figure 12. Scatterplots showing example applications of wastewater-treatment statistics used to develop a water-quality transport curve adjusted to represent the effects of point-source discharges

This method is designed to produce planning-level estimates to assess the risk of water-quality exceedances above and below the point of interest with and without point source contributions, but these estimates are recognized to have considerable uncertainties. The estimates are based on available data, which are the averages and standard deviations of monthly wastewater concentrations and flows. Ideally, estimates calculated by using hourly or daily data would be more accurate, but a properly designed and maintained WWTP should have fairly consistent outflows even with highly variable inflows. As the distance between the WWTP and the simulated point of interest grows, some settling and biological transformation of point-source contributions may be expected; therefore, reductions based solely on dilution may be less than what would occur in a receiving stream. However,

SELDM simulates the potential effect of stormwater control measures by using statistics approximating the net effects of structural and nonstructural BMPs. In this report, structural BMPs, also known as stormwater control measures, are defined as the components of the drainage pathway between the source of runoff and a stormwater discharge location that affect the timing, volume, or quality of runoff. SELDM can be used to explicitly simulate the effects of structural BMPs on the timing, volume, and quality of runoff by using professional judgment or by fitting the trapezoidal distribution to available data (

Hydrograph extension is the practice of slowing the discharge of runoff flows and releasing these flows to the receiving water body over an extended period of time (

Runoff volume reduction is the practice of retaining, detaining, or routing runoff flows to increase the amount of infiltration, evapotranspiration, or diversion between the pavement and the outfall (

Water-quality treatment is the practice of using physical, chemical, and biological processes in an attempt to trap and hold sediment and chemical constituents in runoff (

In this study, 7,511 simulations were done to examine flows, concentrations, and loads of stormwater in southern New England (

Table 27. Summary of Stochastic Empirical Loading and Dilution Model (SELDM) analysis projects used to assess the effect of variations in input values on simulation results and demonstrate results of analyses in southern New England

[All simulation results are documented within project directories by

Project ID number | Project short title | Project title | Sensitivity analysis category | Number of analyses |

01000-Seed | 01 SNE seed analysis | Southern New England random-seed analysis. | Random-seed selection, selection of highway water-quality statistics, and selection of upstream water-quality statistics | 1,000 |

02000-SA-Rain | 02 Sensitivity analysis precipitation | Precipitation-sensitivity analysis using median statistics for 3 ecoregions, southern New England gages, and 17 precipitation gages; using median statistics for other variables. | Selection of hydrologic statistics | 1,008 |

03000-SA-Stream | 03 Sensitivity analysis streamflow | Streamflow-sensitivity analysis (with no zero flows) using median statistics for 3 southern New England datasets and 17 individual values; using median statistics for other variables. | Selection of hydrologic statistics | 528 |

04000-SA-StreamZed | 04 Sensitivity analysis zero streamflow | Streamflow-sensitivity analysis for zero flows using median statistics for 3 southern New England datasets and 17 individual values; using median statistics for other variables. | Selection of hydrologic statistics | 576 |

05000-SA-USTIA | 05 Sensitivity analysis upstream TIA | Sensitivity analysis on upstream imperviousness from 0 to 90 percent, using median statistics for other variables; using median statistics for other variables. | Selection of basin properties with 3 different random seeds | 396 |

06000-SA-USLarge | 06 Sensitivity analysis large upstream basins | Sensitivity analysis on large upstream basins (40, 80, and 100 square miles; 0, 5, 10, and 20 percent TIA); using median statistics for other variables. | Selection of basin properties | 36 |

07000-SA-HwyLS | 07 Sensitivity analysis highway length and slope | Sensitivity analysis on highway-site properties with short and long lengths; steep and shallow slopes; using median statistics for other variables. | Selection of basin properties | 432 |

08000-SA-HwyRv | 08 Sensitivity analysis highway Rv Equations | Sensitivity analysis on highway-site runoff-coefficient equations looking at the highway and nonhighway runoff statistics; using median statistics for other variables. | Selection of hydrologic statistics | 192 |

09000-SA-USRvCor | 09 Sensitivity analysis upstream Rv correlation | Sensitivity-analysis rank correlation of the upstream runoff coefficient statistics to prestorm-flow plotting position; using median statistics for other variables. | Selection of hydrologic statistics | 1,584 |

10000-SA-Ratio | 10 Sensitivity- analysis recession ratios | Sensitivity-analysis recession-ratio statistics; using median statistics for other variables. | Selection of hydrologic statistics | 672 |

11000-HwyYields | 11 Highway- runoff yield analyses | Simulations to develop highway-runoff yields (loads per acre) by using highway-runoff-quality statistics with regional and local precipitation statistics. | Selection of highway-runoff quality and precipitation statistics | 147 |

12000-UrbanYields | 12 Urban- runoff yield analyses | Simulations to develop urban-runoff yields (loads per acre) by using urban runoff-quality statistics with regional and local precipitation statistics. | Selection of urban-runoff quality statistics in comparison to highway-runoff quality statistics and precipitation statistics | 294 |

13000-BMPs | 13 Structural BMP performance | Simulations to examine the long-term average stormwater-control measure performance using statistics from |
Selection of BMP treatment statistics | 576 |

14000-USQW | 14 Upstream water-quality analyses | Simulations to examine various methods to simulate upstream water quality including dependent relations and transport curves formulated by using literature-based, regional, or site-specific statistics. | Selection of upstream water-quality statistics (38 minimally developed total phosphorus transport-curve analyses, 24 developed total phosphorus transport-curve analyses, and 8 metal analyses) | 70 |

[All simulation results are documented within project directories by

Number of water-quality constituents | Water-quality constituent category ( |
|||||||||

Highway | Urban | Upstream | Downstream | Properties | Sediments | Nutrients | Metals | Organics | Biologicals | Majors |

4 | 0 | 4 | 4 | N | Y | Y | N | N | Y | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

3 | 0 | 3 | 3 | N | Y | Y | N | N | N | N |

21 | 0 | 0 | 0 | Y | Y | Y | Y | Y | Y | Y |

0 | 21 | 0 | 0 | Y | Y | Y | Y | Y | Y | Y |

21 | 0 | 5 | 10 | Y | Y | Y | Y | Y | Y | Y |

9 | 0 | 9 | 16 | N | Y | Y | Y | N | N | N |

The number of analyses run in each project (

For most analyses, three representative highway-site configurations were used for simulating highway or urban runoff. These sites were 0.25, 1, and 10 acres of pavement to represent typical contributing drainage areas (

For most analyses, 16 representative upstream basin configurations (

SELDM was developed to be a tool that can be used to transform disparate and complex scientific data into meaningful information about the risk for adverse effects of runoff on receiving waters, the potential need for mitigation measures, and the potential effectiveness of such measures for reducing these risks (

Dilution-factor analysis is useful for assessing potential effects of runoff on receiving waters. The dilution-factor analysis technique was developed by researchers using statistical mass-balance models to quantify potential effects of urban runoff on receiving waters (

Scatterplot of simulated populations of dilution factors, the ratio of highway runoff to downstream flow, for nine combinations of pavement and upstream areas for basins with upstream-impervious fractions equal to zero. The vertical lines show selected exceedance percentiles used in this report to examine the effects of variations in hydrologic variables on receiving-water quality.

Figure 13. Scatterplot of simulated populations of dilution factors, the ratio of highway runoff to downstream flow, for nine combinations of pavement and upstream areas for basins with upstream-impervious fractions equal to zero

Recognizing that runoff from developed areas was a complex mixture of many different water-quality constituents, each with different potential effects on stream biota,

A dilution-factor analysis can provide a quick initial assessment of the risks for water-quality excursions with and without BMP treatment (

A dilution factor analysis has several advantages over an analysis of only a few water-quality constituents. Highway and urban runoff contains a host of different water-quality constituents that can vary by orders of magnitude in concentration (

The dilution factors also can be used to inform risk-based analyses. As previously demonstrated with selected concentrations (

State DOTs and other MS4 permittees need information about potential yields (loads per unit area) of constituents of concern in stormwater runoff and BMP discharges. The National Cooperative Highway Research Program study on TMDLs concluded that SELDM is a useful method for estimating yields and loads of urban and highway runoff and BMP discharges (

SELDM simulates individual event loads and loads for annual load accounting years for constituents, reported in units of mass per unit volume (for example, milligrams per liter;

Application of simulated runoff yields to specific regulatory-TMDL areas in southern New England is beyond the scope of this report, but an example was done to illustrate a simplified method to estimate highway and nonhighway constituent loads from simulated constituent yields. Analysis of simulation results indicates that there may be substantial variations in individual annual yield values (

Boxplot showing the population of 29 annual total-nitrogen yields in pounds per acre per year, from 1,640 individual runoff-generating events, for 100 percent highway and impervious areas simulated by using precipitation statistics for southern New England, highway and urban runoff-coefficient statistics, highway and urban runoff-quality statistics, and treatment by the generic (median) structural best management practice (BMP).

Figure 14. Boxplot showing the population of 29 annual total-nitrogen yields from 1,640 individual runoff-generating events, for 100 percent highway and impervious areas simulated by using precipitation statistics for southern New England, highway and urban runoff-coefficient statistics, highway and urban runoff-quality statistics, and treatment by the generic structural best management practice

Once long-term average constituent yields for different land covers are calculated, information about different land-cover areas obtained from a GIS or from StreamStats can be used to estimate total constituent loads from each land-cover area in a stream basin of interest (

The example constituent load analysis for total nitrogen was done by using 16 basins that are tributaries to Narragansett Bay (

Table 28. StreamStats drainage-basin properties and estimated long-term average total-nitrogen loads from highways and developed areas for selected streams and rivers draining to Narragansett Bay, Rhode Island

[Basin data are from ^{2}); LC11IMP, average percentage of impervious area determined from NLCD 2011 impervious data; LC11DEV, percentage of developed land from the NLCD 2011 classes 21–24; TIA, the total impervious area as a percent of the drainage area; lb/yr, pound per year; —, not applicable]

Basin name | Point of interest used to do the StreamStats delineation | Basin properties | Stormwater runoff constituent loads of total nitrogen, lb/yr | |||||||

Latitude | Longitude | DRNAREA (mi^{2}) |
LC11IMP (percent) | LC11DEV (percent) | Major road TIA (acres) | Other TIA (acres) | Highway loads | Developed-area loads | Percent highway | |

Annaquatucket River | 41.55095 | −71.43849 | 7.48 | 10.8 | 30.3 | 57 | 517 | 544 | 6,150 | 8.13 |

Barrington River | 41.78475 | −71.3302 | 9.57 | 28 | 55.1 | 123 | 1,710 | 1170 | 20,300 | 5.45 |

Blackstone River | 41.87749 | −71.3818 | 475 | 12.1 | 29.5 | 1,586 | 36,800 | 15,100 | 438,000 | 3.33 |

Dead Man Brook | 41.40331 | −71.4599 | 0.83 | 2.07 | 9.63 | 0 | 11 | 0 | 131 | 0 |

Hardig Brook | 41.69767 | −71.4589 | 6.06 | 40 | 76.5 | 67.3 | 1,550 | 643 | 18,400 | 3.38 |

Hunt River | 41.63843 | −71.4471 | 22.7 | 15.2 | 40.8 | 155 | 2,210 | 1480 | 26,300 | 5.33 |

Maskerchugg River | 41.64864 | −71.4563 | 5.78 | 25.2 | 64 | 88.7 | 932 | 847 | 11,100 | 7.09 |

Old Mill Creek | 41.71217 | −71.3754 | 5.38 | 46 | 85 | 8.38 | 1,580 | 80 | 18,800 | 0.42 |

Palmer River | 41.76216 | −71.2844 | 48.8 | 5.21 | 17.6 | 121 | 1,630 | 1160 | 19,400 | 5.64 |

Pawtuxet River | 41.7643 | −71.3907 | 232 | 12.6 | 27.3 | 876 | 18,700 | 8,370 | 223,000 | 3.62 |

Quequechan River | 41.70496 | −71.1611 | 30.2 | 16.8 | 32.3 | 252 | 3,250 | 2,410 | 38,700 | 5.86 |

Sin and Flesh Brook | 41.61814 | −71.2041 | 3.52 | 7.05 | 19.6 | 26.5 | 159 | 253 | 1,890 | 11.8 |

Taunton River | 41.87536 | −71.0943 | 366 | 12 | 32.2 | 1,238 | 28,100 | 11,800 | 334,000 | 3.41 |

Ten Mile River | 41.83868 | −71.3688 | 55.4 | 24.8 | 55 | 364 | 8,790 | 3,480 | 105,000 | 3.21 |

Three Mile River | 41.85408 | −71.1088 | 85.3 | 13.3 | 36.4 | 373 | 7,260 | 3,560 | 86,400 | 3.96 |

Providence River^{1} |
41.82472 | −71.4082 | 74.5 | 27.6 | 50.9 | 601 | 13,200 | 5,740 | 157,000 | 3.53 |

Tributary totals | — | — | 1,429 | 13.8 | 32.8 | 5,937 | 126,399 | 56,637 | 1,504,571 | 3.63 |

Delineated from the confluence of the Woonasquatucket and Moshassuck Rivers.

Results of this planning-level analysis indicate that highways are a small part of the total annual constituent loads. The percent highway constituent load values in

If BMP-discharge nitrogen yields are used for a loading analysis instead of the direct-discharge yields (

The analyses described in this report were designed to produce statistical estimates of stormwater flows, concentrations, and loads from highway and urban land-cover areas with and without BMP treatment and constituent loads from stream basins upstream from stormwater outfalls. SELDM is not calibrated by fitting input values to a historical record; SELDM is calibrated by selecting statistics for hydrologic, runoff-quality, and BMP-treatment variables from robust and representative datasets (

The uncertainty of selected input statistics for a given purpose depends on many factors, and uncertainties in input values are proportional to uncertainties in simulation results. Many uncertainties may affect results from any model. These include uncertainties in individual measurements (

Uncertainties in measured concentrations and loads can be more than 100 percent of the reported values.

If it is assumed that data are fairly precise and representative of conditions of interest, then uncertainties in environmental statistics are a function of sample size (

In this study, the number of water-quality monitoring sites and the number of samples per site varied greatly for the highway, urban, upstream, and BMP data. The number of samples used to develop the 197 water-quality transport curves (summarized in

Uncertainties in the use of statistics from monitored sites to simulate conditions at unmonitored sites may be considerable. The geographic analysis of basin properties in southern New England indicates that there are more than 48,000 road-stream crossings in this area (

The purpose of this study was to examine relations between variables for generic but representative sites rather than an attempt to predict historical or future water quality at a particular location. For this type of analysis, individual uncertainties must be recognized but are not as critical to the interpretation of results from stochastic models such as SELDM as they would be for deterministic or process-based models that use history matching for calibration. In this study, some of the uncertainty in statistics for individual sites and the transfer of statistics from monitored sites for simulations at unmonitored sites is mitigated by use of the median statistics from multiple sites to represent the general characteristics of highway runoff, urban runoff, BMP treatment, and upstream water quality. The median of site statistics represents the central tendency of all site statistics, without the potential influence of extreme outliers that could be caused by monitoring bias or uncharacteristic conditions at a few sites. If the objective of a study is to assess the risk of adverse effects of runoff at a particular stormwater outfall, then operational definitions for selected statistics are needed. In the absence of such definitions, site-specific data may be needed to reach a consensus decision. However, the time and resources needed to obtain sufficient data to substantially reduce uncertainties may preclude most site-specific data collection efforts. Even if monitoring studies were conducted widely, changes in the roadway or the upstream land area would affect water quality at the site of interest, so the monitoring data collected may not represent future conditions at that site (

A model sensitivity analysis is a systematic study to evaluate the effect of perturbations in input values on simulated results (

The focus of the master random-seed (

The focus of the hydrologic sensitivity analyses (

The sensitivities were examined graphically and statistically. The six different hydrologic sensitivity projects included 36 to 1,584 individual analyses (

Because SELDM was designed to help quantify the risk of adverse effects of runoff on receiving waters, the potential need for mitigation measures, and the potential effectiveness of such management measures for reducing these risks, results of the analyses are focused on statistics for selected exceedance risks. The same exceedance risk for one variable or even for seemingly paired variables may come from different individual storm events because the stochastic rather than deterministic approach can reshuffle the values of related variables (

SELDM uses a random-seed management algorithm to ensure that each runoff-quality analysis will be reproducible; details about the Monte Carlo methods used are documented in appendix 1 of the manual (

Two similar SELDM studies, designed to produce long-term annual constituent loads, did multiple-seed sensitivity analyses.

In the current study, a sensitivity analysis with 500 unique, master random seeds was done to reduce the potential for extreme output values and explore the potential variability in event and annual-load accounting year values from simulation to simulation. The results of this analysis are documented in the 01000-Seed project files (

The population of long-term average annual yield values have substantial variability but, in comparison to long-term maximum annual yields, are relatively robust to changes in the random-seed values.

Boxplots showing the distribution of long-term average and maximum annual yields of total phosphorus in pounds per acre per year from highway runoff and structural stormwater best management practice (BMP) discharge for 500 different master random-seed simulations.

Figure 15. Boxplots showing the distribution of long-term average and maximum annual yields of total phosphorus in pounds per acre per year from highway runoff and structural stormwater best management practice discharge for 500 different master random-seed simulations

Stochastic random-seed variability for individual events simulated in SELDM can be substantial. For example,

Boxplots showing the variation in individual-event dilution factors from selected exceedance percentages from 500 master random-seed simulations. These simulations were done by using a 10-acre 100-percent-impervious highway site with a 10-square-mile 10-percent-impervious upstream basin.

Figure 16. Boxplots showing the variation in individual-event dilution factors from selected exceedance percentages from 500 master random-seed simulations

Variability in individual event concentrations also were examined. As an example,

Boxplots showing individual-event concentrations of total phosphorus concentrations (p00665) from selected exceedance percentages from 500 master random-seed simulations for

Figure 17. Boxplots showing individual-event concentrations of total phosphorus concentrations from selected exceedance percentages from 500 master random-seed simulations for highway-runoff quality and upstream-stormflow quality

The variations and extreme values simulated in the master random-seed analyses are consistent with statistical sampling theory (

SELDM uses statistics for three variables; precipitation volume, event duration, and time between event midpoints; to simulate a population of runoff-generating events (

Each of the 21 sets of precipitation statistics were simulated with 48 combinations of 3 highway-site and 16 upstream-basin configurations. Three 100-percent impervious highway-site configurations with areas of 0.25, 1.0, and 10 acres were used for these simulations. The representative lengths (

To examine the sensitivity of the simulated dilution factors to the selected precipitation statistics, variations in dilution factors were compared to variations in precipitation statistics.

Scatterplot of dilution factors for selected exceedance percentiles as a function of precipitation volumes with a 1-acre highway site and 1-square-mile, 0-percent impervious upstream basin. Results are from 21 simulations with precipitation statistics that represent regional medians and the range of individual precipitation-gage statistics.

Figure 18. Scatterplot of dilution factors for selected exceedance percentiles as a function of precipitation volumes with a 1-acre highway site and 1-square-mile, 0-percent impervious upstream basin

As indicated in

The sensitivity of dilution factors to variations in precipitation for all 1,008 simulations is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for the precipitation-statistics sensitivity analyses as a function of the drainage-area ratios.

Figure 19. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for the precipitation-statistics sensitivity analyses as a function of the drainage-area ratios

The standard deviation values shown in

SELDM uses prestorm streamflow statistics to simulate the flow in the receiving stream at the site of interest when the event begins (

Each set of prestorm streamflow statistics were simulated with 48 combinations of highway site and upstream basin configurations. Three 100-percent impervious highway-site configurations with 0.25, 1.0, and 10 acres were used for these simulations. The representative highway site lengths (

To examine the sensitivity of dilution factors to the selected streamflow statistics, variations in dilution factors were compared to variations in the simulated geometric mean streamflows.

Scatterplot of dilution factors for selected exceedance percentiles as a function of geometric mean streamflow with a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin. Results are from 11 simulations with streamflow statistics that represent regional medians and the range of individual streamgage statistics.

Figure 20. Scatterplot of dilution factors for selected exceedance percentiles as a function of geometric mean streamflow with a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin

The sensitivity of dilution factors to variations in simulated streamflow statistics for all 528 simulations is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for the prestorm streamflow-statistics sensitivity analyses as a function of the drainage-area ratios.

Figure 21. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for the prestorm streamflow-statistics sensitivity analyses as a function of the drainage-area ratios

The standard deviation values for the 0.5 percent exceedance risk shown in

This sensitivity analysis indicates that, for estimating the potential effects of runoff on receiving waters, the upstream prestorm streamflow statistics are among the most influential variables. Because the potential effects of runoff are heavily dependent on prestorm streamflow volumes, methods are needed to estimate flow statistics at any site of interest on a stream. The simplest method to refine a regional estimate of a surface-water statistic is to use the selected-station average or selected-station median from the SELDM dataset (

StreamStats can be used to estimate site-specific statistics for any point within each State in southern New England, but the available statistics and available tools are different from State to State. In Connecticut, StreamStats can be used with the Connecticut Streamflow and Sustainable Water Use Estimator (

SELDM uses prestorm streamflow statistics to simulate the flow in the receiving stream at the site of interest when the event begins; in some cases, the streamflow record may contain zero flows (

Ephemeral streams, which flow only in response to runoff are a special case. Precipitation statistics indicate that there are about 41 to 67 runoff-generating events per year in southern New England (

SELDM uses the conditional-probability Monte-Carlo method to simulate the occurrence of zero flows (

To do the zero-flow sensitivity analysis, each set of streamflow statistics were simulated with 48 combinations of highway site and upstream basin configurations, resulting in a total of 576 zero-flow simulations done by using the primary master random seed (number 8,556). Three 100-percent impervious highway-site configurations with areas of 0.25, 1.0, and 10 acres, respectively, were used for these simulations. The representative lengths (

To examine the sensitivity of dilution factors to the selected streamflow statistics, variations in dilution factors were compared to variations in the simulated fraction of zero flows.

Scatterplot of dilution factors for selected exceedance percentiles as a function of the fractions of zero flow with a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin. Results are from 12 simulations with zero-flow statistics that represent regional medians and zero-flow fractions from 0.00001 to 0.25.

Figure 22. Scatterplot of dilution factors for selected exceedance percentiles as a function of the fractions of zero flow with a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin

Below a specified zero-flow fraction of 0.01, the most substantial variations in simulated dilution factors are caused by variations in the statistics for the logarithms of nonzero flows that were used in simulations (

The sensitivity of dilution factors to variations in simulated streamflow statistics for all 576 simulations is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for zero-flow sensitivity analyses as a function of the drainage-area ratios.

Figure 23. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for zero-flow sensitivity analyses as a function of the drainage-area ratios

As expected, the average dilution factors generally increase with increasing drainage-area ratios (

Although dilution factors are invariant to zero-flow ratios less than the exceedance risk of concern (

This sensitivity analysis indicates that, for estimating the potential effects of runoff on receiving waters in SELDM, the basin’s status as a stream with zero flows and the nonzero flow statistics is more important than an exact zero-flow fraction specification, especially if the fraction of zero flows is small.

SELDM uses the rank correlation between prestorm streamflow and the average upstream runoff coefficient to represent potential effects of antecedent wetness on simulated runoff coefficients for individual events (

The effect of rank correlation on simulation results was tested by using rank correlation values of 0.0, 0.5, 0.6, 0.65, 0.7, 0.72, 0.73, 0.75, 0.77, 0.8, and 0.9. These 11 values were selected to cover the range of potential values with a focus on values representing the most robust datasets (

These simulations demonstrated that systematic changes in the dilution factor with changes in prestorm rank correlation were indistinguishable from the effects of the stochastic shuffle that took place because of the change in simulated rank correlations.

Line graph showing dilution factors for the 0.5 exceedance percentiles as a function of the rank correlation coefficient between prestorm streamflow and upstream runoff coefficients for a 1-acre highway site and 1-square-mile, 0-percent impervious upstream basin. Results are from 11 simulations that were repeated by using 3 master random-seed values. SELDM, Stochastic Empirical Loading and Dilution Model.

Figure 24. Line graph showing dilution factors for the 0.5 exceedance percentiles as a function of the rank correlation coefficient between prestorm streamflow and upstream runoff coefficients for a 1-acre highway site and 1-square-mile, 0-percent impervious upstream basin

The sensitivity of dilution factors to variations in simulated prestorm-streamflow correlation coefficient statistics for all 528 simulations using the master random-seed number of 8,556 is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for runoff-coefficient correlation sensitivity analyses as a function of the drainage-area ratios.

Figure 25. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for runoff-coefficient correlation sensitivity analyses as a function of the drainage-area ratios

The standard deviation values shown on

Although variations in dilution factors that are associated with the selected correlation coefficient between prestorm streamflows and runoff coefficients can be substantial, developing a precise site-specific estimate of this correlation coefficient may be difficult (

If dilution factors are used to evaluate potential effects of runoff on receiving waters, then this correlation analysis indicates that a specified prestorm-correlation coefficient would be needed to standardize the dilution-factor response to other hydrologic variables. Because the default SELDM correlation value provides fairly consistent dilution-factor values from seed to seed (

SELDM uses recession ratio statistics to simulate the timing of the upstream stormflow hydrograph limbs and therefore the duration of concurrent flows from the highway site or structural BMP outfall (

Each set of recession-ratio statistics were simulated with 48 combinations of highway site and upstream basin configurations. Three 100-percent impervious highway-site configurations with 0.25, 1.0, and 10 acres were used for these simulations. The representative lengths (

To examine the sensitivity of dilution factors to the selected recession-ratio statistics, variations in dilution factors were compared to variations in the average recession-ratios for different combinations of highway site and upstream basin.

Scatterplot of dilution factors for selected exceedance percentiles as a function of recession-ratio statistics for a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin. Results are from 14 simulations including the average and median of southern New England statistics, and the range of individual streamgage statistics. SELDM, Stochastic Empirical Loading and Dilution Model.

Figure 26. Scatterplot of dilution factors for selected exceedance percentiles as a function of recession-ratio statistics for a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin

The sensitivity of dilution factors to variations in recession-ratio statistics for all 672 simulations is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for recession-ratio sensitivity analyses as a function of the drainage-area ratios.

Figure 27. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for recession-ratio sensitivity analyses as a function of the drainage-area ratios

The standard deviation values shown on

Given the large effort needed to develop and select recession-ratio statistics and the small effect that variations in these statistics have on simulation results, use of the SELDM default recession-ratio statistics is warranted for planning-level analyses in southern New England and other areas of the country. Developing site specific recession-ratio statistics requires one or more years of continuous-record data at a site of interest or a hydrologically similar site and a substantial effort to process and analyze the data (

In the upstream-basin size sensitivity analysis, the effect of upstream basin size is examined by using selected simulations from the upstream-imperviousness simulations (project 05000-SA-USTIA,

The effects of upstream basin size were simulated by using 84 combinations of highway site and upstream basin configurations. The simulations were done with 3 highway-site configurations, 16 upstream-basin configurations from project 05000-SA-USTIA (

The sensitivity of dilution factors to variations in upstream-basin size for all 84 simulations is shown in

Scatterplot of the simulated dilution factors at the 0.5 percent exceedance risk for 84 combinations of highway site area, basin site area, and upstream percent imperviousness over the range of simulated drainage-area ratios. Dilution factors represent highway site areas of 0.25, 1, and 10 acres, and upstream basin-size areas of 0.1, 1.0, 10, 20, 40, and 80 square miles. Simulated upstream impervious areas were 0, 5, 10, and 20 percent. Median statistics were used for precipitation, streamflow without zero flows, and water quality.

Figure 28. Scatterplot of the simulated dilution factors at the 0.5 percent exceedance risk for 84 combinations of highway site area, basin site area, and upstream percent imperviousness over the range of simulated drainage-area ratios

The sensitivity of downstream concentrations to a wide range of basin sizes also is used to estimate an upstream basin area threshold for which an urban or highway discharge would have little effect. The potential sensitivity of downstream concentrations to drainage-area ratios is shown as the ratio of downstream to upstream concentrations in

Scatterplots showing the ratio of simulated downstream to upstream concentrations for 84 combinations of drainage-area ratios.

Figure 29. Scatterplots showing the ratio of simulated downstream to upstream concentrations for 84 combinations of drainage-area ratios

Although there are 84 combinations of highway and upstream basin properties in

Simulation results indicate different drainage-area ratios that would have no effect on the downstream concentration for different water-quality constituents and risk levels (

Simulation results at the 0.1 percent exceedance risk for phosphorus and suspended sediment concentrations are similar in pattern to 0.5 percent exceedance results, but as expected with higher runoff concentrations, a higher proportion of simulated values are greater than the specified concentration ratios (

The results shown in

The upstream basin-size analysis provides a large range of dilution factors (

Scatterplots showing the ratio of downstream to upstream concentrations as a function of the dilution factor.

Figure 30. Scatterplots showing the ratio of downstream to upstream concentrations as a function of the dilution factor

The results shown in

SELDM uses the drainage area, imperviousness, and drainage length and slope to simulate the volume and timing of stormflow from the stream basin upstream of the site of interest (

In this sensitivity analysis, 11 upstream impervious percentages equal to 0, 1, 2.5, 5, 7.5, 10, 20, 30, 45, 60, and 90 percent were simulated. Three 100-percent impervious highway-site configurations with 0.25, 1.0, and 10 acres were used for these simulations. The representative highway lengths (

These simulations demonstrated that the stochastic shuffle that took place because of the change in correlation values between the highway and upstream basin runoff coefficients may be larger than the systematic changes in stormflows caused by small incremental changes in the upstream imperviousness, especially at low impervious values (

Line graph showing dilution factors for the 0.5 exceedance percentiles as a function of the upstream impervious percentage for a 1-acre highway site and 1-square-mile upstream basin. Results are from 11 impervious-percentage simulations that were repeated by using three random master random-seed values.

Figure 31. Line graph showing dilution factors for the 0.5 exceedance percentiles as a function of the upstream impervious percentage for a 1-acre highway-site and 1-square-mile upstream basin

The sensitivity of dilution factors to variations in drainage-area ratios for all 132 simulations with the selected master random-seed number 8,556 is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for upstream-imperviousness sensitivity analyses as a function of the drainage-area ratios.

Figure 32. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for upstream-imperviousness sensitivity analyses as a function of the drainage-area ratios

The standard deviation values shown on

This sensitivity analysis indicates that, for estimating the potential effects of runoff on receiving waters, the upstream impervious percentage is among the most influential hydrologic variables. The upstream impervious percentage, however, is a physical property of the upstream basin at sites of interest that can be easily ascertained by using StreamStats or other tools. Although there can be substantial differences in impervious-percentage estimates for a given drainage basin (

SELDM uses the average, standard deviation, and skew of runoff coefficients to generate runoff volumes from precipitation volumes for each simulated runoff event (

This runoff-coefficient equation sensitivity analysis was designed to quantify potential effects of selecting either the highway runoff-coefficient equations or the nonhighway runoff-coefficient equations for simulating runoff from the site of interest (highway or urban site). This analysis was done to examine the differences in flows and loads that can be attributed to the runoff-coefficient statistics selected for analysis because SELDM is a lumped parameter model that can be used to simulate the quality and quantity of runoff from highways or other land covers. This sensitivity analysis included 192 individual simulations done by using the primary master random-seed variable (number 8,556). Each set of runoff-coefficient equations was simulated by using 96 combinations of highway site and upstream basin configurations. Three 100-percent impervious highway-site configurations with 0.25, 1.0, and 10 acres with representative lengths (

The differences in stormwater flows, concentrations, and loads caused by changes in runoff-coefficient statistics do not have a purely analytic solution because the differences in the average, standard deviation, and skew of runoff coefficients vary with imperviousness, and the imperviousness of the highway site and upstream basin also affects the correlation between the highway and upstream runoff coefficients. The average and standard deviation calculated by using the highway regression relations (

As indicated by the runoff-coefficient statistics, generated dilution-factor values for the 100 percent impervious sites are similar for the highway and nonhighway equations (on average within a 5.1 percent difference), but the dilution-factor values calculated by using the highway runoff-coefficient statistics are on average about 1.6 times the values calculated by using the nonhighway statistics for the 50-percent impervious sites.

Scatterplot comparing dilution factors at the 0.5 percent exceedance risk calculated by using the highway and nonhighway regression relations between the imperviousness of the site of interest and the average, standard deviation, and skew of runoff coefficients for 2 highway site impervious percentages and 48 combinations of highway site and upstream basin configurations.

Figure 33. Scatterplot comparing dilution factors at the 0.5 percent exceedance risk calculated by using the highway and nonhighway regression relations between the imperviousness of the site of interest and the average, standard deviation, and skew of runoff coefficients for 2 highway site impervious percentages and 48 combinations of highway site and upstream basin configurations

The results of these simulations indicate that the effect of the runoff-coefficient equation selection depends on the imperviousness of the simulated site. Because most of the highway sites used to develop the highway runoff-coefficient equations had high impervious percentages and because 26 of the 27 New England highway-runoff monitoring sites with concentration data in the HRDB (

SELDM is a lumped parameter model that uses drainage area, imperviousness, and drainage length and slope to simulate the volume and timing of runoff from the site of interest (highway site) and the upstream basin (

In this series of analyses, the highway-site characteristics were varied to determine the sensitivity of results to a range of highway-site drainage lengths and slopes. The site combination characteristics included three highway drainage areas, 0.25, 1.0, and 10 acres, that were used in simulations with short, medium, and long drainage lengths and shallow, medium, and steep drainage slopes to simulate the timing of runoff from typical highway sites. Drainage lengths were based on bridge, roadway, and upstream basin characteristics (

The simulated highway drainage lengths were selected to be representative of possible values. The short lengths, which are meant to simulate the distance from the centerline crown of the road to the scuppers of a bridge, were 14, 26, and 38 feet for the 0.25-, 1.0-, and 10-acre sites, respectively. These lengths are based on half the width of a 2-, 4-, and 6-lane bridge. The remaining lengths were selected to represent roadways that are approaching the waterway from two directions. The medium length of 300 ft for the 0.25-acre site was based on a 2-lane road width (

The same drainage slopes were used for all three drainage areas. A shallow slope of 10 feet elevation change per mile was selected because it is a commonly used value related to the self-cleaning velocity of flow in a storm-sewer pipe (

To examine the sensitivity of dilution factors to the selected highway-site configurations, variations in dilution factors are compared to variations in the basin-lag factor, which is the drainage length divided by the square root of the drainage slope (

Scatterplot of dilution factors for selected exceedance percentiles as a function of basin-lag-factor values for a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin. Results are from 9 simulations including highway drainage lengths ranging from 14 to 4,000 feet and drainage slopes ranging from 10 to 300 feet per mile. The combination of length and slope is expressed as the basin lag factor, which is the length in miles divided by the square root of slope in feet per mile.

Figure 34. Scatterplot of dilution factors for selected exceedance percentiles as a function of basin-lag-factor values for a 1-acre highway site and 1-square-mile, 0-percent-impervious upstream basin

The sensitivity of dilution factors to variations in the drainage-area ratio for all 432 simulations is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for basin-lag factor sensitivity analyses as a function of the drainage-area ratios.

Figure 35. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for basin-lag factor sensitivity analyses as a function of the drainage-area ratios

The standard deviation values shown on

The low variability in dilution factors for widely varying highway-site configurations indicates that dilution factors are not sensitive to the length and slope of sites of interest, and these do not need to be specified exactly. This lack of sensitivity is good because real highway and urban drainage systems can be complex, as-built plans for existing drainage systems may not be readily available, and detailed drainage plans may not have been drafted at the planning stage for new or upgraded facilities. The lack of sensitivity also indicates that example simulations may be, within the uncertainty of all statistics used to simulate highway or urban runoff, representative of a wide range of similar sites. For example, if a roadway follows a stream, then variations in the roadway drainage pathways from outfall to outfall may not be substantial enough to warrant separate simulations.

SELDM uses highway and urban runoff statistics to simulate concentrations and loads of runoff from the site of interest (

These highway-runoff constituent-yield sensitivity analyses included 147 simulations done by using the primary master random-seed value (number 8,556). In this series of analyses, a 1-acre 100-percent impervious highway site configuration was used to perform the sensitivity analysis and to provide highway runoff yields for potential use in TMDL analyses (for example,

In this series of analyses, the highway runoff constituent concentration statistics were varied to determine the sensitivity of results to a range of geometric-mean highway-runoff concentrations. Statistics for 21 constituents of potential concern (

Boxplots showing simulated populations of long-term average annual constituent yields of

Figure 36. Boxplots showing simulated populations of long-term average annual constituent yields of total nitrogen yields, total phosphorus yields, suspended sediment, and long-term average annual flow-weighted concentrations of

Results of this analysis indicate that selecting representative highway-runoff statistics has a greater influence on simulated runoff-quality results than random variability from master seed to master seed, and the selected precipitation statistics (

This sensitivity analysis focused on highway runoff quality, but similar issues affect urban runoff quality estimates. The results of 294 urban-runoff quality analyses are documented in the 12000-UrbanYields project files (

Because of the large differences in simulated statistics over the range of available data for southern New England and the differences in results for different constituents, an operational definition would be needed to develop a simple decision-support system for assessing potential effects of runoff in southern New England. In North Carolina, the Department of Transportation worked with the State Department of Environmental Quality and the USGS to develop an operational definition to assess risk by using suspended-sediment concentrations as a sentinel water-quality indicator (

SELDM uses instream water-quality statistics to simulate concentrations and loads of stormwater in the receiving stream upstream from the site of interest (

In this upstream stormflow-quality sensitivity analysis, the median total phosphorus (p00665) transport curve (

In this series of analyses, comparison of the stochastic variations with the median transport curve and variations caused by use of different transport curves indicates that water-quality selection uncertainty is much higher than stochastic variability from simulation to simulation.

Boxplots showing populations of simulated event mean total phosphorus (p00665) concentrations in receiving-stream stormflow upstream from a site of interest generated by using a median transport curve and 500 different master random seeds or 38 individual transport curves with the selected master random-seed number 8,556. Simulations are for a 1 square-mile completely pervious upstream basin. The horizontal lines are two example water-quality criteria (0.025 and 0.1 milligram per liter [mg/L]) for total phosphorus that are shown for discussion. Constituents are defined in

Figure 37. Boxplots showing populations of simulated event mean total phosphorus concentrations in receiving-stream stormflow upstream from a site of interest generated by using a median transport curve and 500 different master random seeds or 38 individual transport curves with the selected master random-seed number 8,556

The method used to develop representative regional transport curves may have unanticipated effects on the simulation results. For example, the high-concentration estimates of total phosphorus from the transport curve developed by using the medians of transport-curve coefficients for minimally developed basins, which are shown as the random-seed simulation results, are in the lower quartile of estimates for individual transport curves (

An alternate to regional transport curves or using the median of transport curve coefficients would be to pool data from a selected subset of hydrologically similar stream basins with similar levels of development and develop a representative transport curve from the pooled data. This approach would be suitable for assessing conditions at a particular site of interest without a water-quality dataset. This approach could demand considerable effort, however, because a detailed analysis of different combinations would need to be done and the highly interpretive nature of such an effort may be subject to challenge.

The difference between use of the median transport curve or an individual-station transport curve for a given constituent has two conflicting effects for developing a simple decision method for estimating the potential risks of runoff on receiving waters and the potential for reducing such risks. If the median transport curve underestimates the population of concentrations, then a decision method based on the number of simulated water-quality excursions may lead to underestimation of the risk for upstream and downstream water-quality excursions. This may result in BMP decisions that do not adequately reduce water-quality excursion risks at some sites. Alternately, if the decision method is based on the change in concentrations from upstream to downstream and the highway or urban runoff concentrations are higher than the upstream concentrations, then the decision method based on the median transport curve may lead to an overestimation of the risk for adverse effects of runoff on receiving-water quality. Overestimating risks may result in application of advanced BMPs where they are not needed to protect downstream water quality. The results of the study by

SELDM uses statistics for volume reduction, hydrograph extension, and water-quality treatment to simulate the effects of structural stormwater runoff BMPs on flows, concentrations, and loads of runoff from the site of interest and their effects on flows, concentrations, and loads in stormwater downstream from the discharge outfall (

Table 29. Long-term average best management practice performance for runoff stormflows and constituent loads calculated by using individual storm statistics for 29 annual-load accounting years for southern New England

[The long-term average ratios in this table should not be confused with percent removals; they are the long-term average of individual storm events, each of which may depart substantially from the long-term values (

Code | Name | Long-term average ratio of outflow to inflow, unitless | Performance rank | Adjusted cost rank | ||||||||

Stormflow | SS | TN | TP | Tcu | TPb | TZn | PAH | Bio | ||||

BI | Grass strip (biofilter) | 0.532 | 0.064 | 0.460 | 0.320 | 0.073 | 0.141 | 0.142 | 0.173 | 0.298 | 1 | 1 |

BR | Bioretention | 0.461 | 0.092 | 0.542 | 0.433 | 0.200 | 0.071 | 0.111 | 0.147 | 0.213 | 2 | 7 |

BS | Grass swale (bioswale) | 0.502 | 0.064 | 0.476 | 0.494 | 0.240 | 0.251 | 0.154 | 0.155 | 0.285 | 4 | 4.5 |

DB | Detention basin | 0.605 | 0.074 | 0.518 | 0.408 | 0.330 | 0.219 | 0.224 | 0.169 | 0.369 | 5.5 | 2.5 |

IB | Infiltration basin | 0.903 | 0.061 | 0.648 | 0.620 | 0.217 | 0.180 | 0.211 | 0.265 | 0.514 | 8 | 2.5 |

MD | Manufactured device | 1.000 | 0.144 | 0.735 | 0.533 | 0.639 | 0.523 | 0.537 | 0.289 | 1.580 | 11 | — |

MF | Media filter | 0.838 | 0.079 | 0.606 | 0.107 | 0.390 | 0.165 | 0.161 | 0.257 | 0.510 | 7 | 6 |

PP | Porous pavement | 0.186 | 0.182 | 0.187 | 0.190 | 0.189 | 0.180 | 0.187 | 0.195 | 0.182 | 3 | 8 |

RP | Retention pond | 0.994 | 0.096 | 0.627 | 0.394 | 0.459 | 0.199 | 0.199 | 0.298 | 0.680 | 9 | 9.5 |

WB | Wetland basin | 1.020 | 0.246 | 1.030 | 0.709 | 0.439 | 0.353 | 0.441 | 0.266 | 0.516 | 10 | 9.5 |

WC | Wetland channel | 1.070 | 0.552 | 0.840 | 1.140 | 0.810 | 0.792 | 0.574 | 0.316 | 0.607 | 12 | 4.5 |

Med | Median BMP | 0.764 | 0.104 | 0.549 | 0.423 | 0.295 | 0.133 | 0.160 | 0.232 | 0.434 | 5.5 | — |

Individual treatment statistics were not varied independently and systematically in these analyses because the three treatment variables, volume reduction, hydrograph extension, and water-quality treatment, have interdependent effects on the outflow loads. There are large variations in treatment statistics within and between BMP categories when individual monitoring sites are considered, so performance results for different BMP categories commonly overlap each other.

This BMP performance sensitivity analysis included 576 individual simulations done by using the primary master random-seed value (8,556). Each of the 12 BMPs were simulated by using 48 combinations of highway site and upstream-basin configurations. Three 100-percent impervious highway-site configurations with 0.25, 1.0, and 10 acres were used for these simulations. The representative lengths (

Results of this BMP sensitivity analysis, shown in

The ratios in

The sensitivity of dilution factors to variations in BMP statistics for all 576 simulations is shown in

Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for best management practice (BMP) sensitivity analyses.

Figure 38. Scatterplots of the dilution-factor statistics at the 0.5 percent exceedance risk for best management practice sensitivity analyses

The standard deviation values shown on

Although the large variability in results indicates that BMP selection is an important factor in simulated results, this BMP sensitivity analysis and the comparison to BMP cost data are generalized. Selection of a BMP for a particular site depends on many factors and constraints, some of which may be particular to a linear transportation system with limited rights of way (

For the purpose of TMDL analyses, a median (generic) BMP may be used to represent the effects of many different BMPs in a stream basin of concern (

SELDM was designed to transform complex scientific data into meaningful information about the risk of adverse effects of runoff on receiving waters, the potential need for mitigation measures, and the potential effectiveness of such management measures for reducing these risks (

In this series of analyses, the potential efficacy of onsite BMPs will be discussed in terms of the reductions in downstream concentrations and the reductions in risks for water-quality exceedances that may take place if onsite BMPs are used.

Scatterplot of populations of simulated event mean concentrations of total phosphorus for highway runoff, stormwater best management practice discharges, and receiving-stream stormflow upstream and downstream from a discharge point. Upstream stormflow concentrations were simulated by using water-quality transport curves developed by using data from U.S. Geological Survey (USGS) water-quality monitoring stations 01098360 and 01073562. Simulations are for a 1-acre highway site draining to a 1-square-mile basin. The horizontal lines are two example water-quality criteria (0.025 and 0.1 milligram per liter [mg/L]) for total phosphorus. The approximate one-event-in-3-year risk level (0.5 percent exceedance risk), which is considered protective for aquatic life (

Figure 39. Scatterplot of populations of simulated event mean concentrations of total phosphorus for highway runoff, stormwater best management practice discharges, and receiving-stream stormflow upstream and downstream from a discharge point

As discussed in the section of this report on “Risk-Based Analyses,” numeric water-quality criteria commonly are based on a specified concentration, frequency of occurrence, and exposure duration (

The two examples in

Scatterplot showing the ratio of downstream total phosphorus (p00665) concentration without best management practice (BMP) treatment to the downstream total phosphorus concentration with treatment as a function of the upstream total phosphorus concentration at the 0.5 percent risk level. Upstream concentrations were simulated by using water-quality transport curves for total phosphorus developed by using data from 62 water-quality monitoring stations. Simulations are for a 1-acre highway site draining to a 0-percent impervious, 1-square-mile stream basin. Constituents are defined in

Figure 40. Scatterplot showing the ratio of downstream total phosphorus concentration without best management practice treatment to the downstream total phosphorus concentration with treatment as a function of the upstream total phosphorus concentration at the 0.5 percent risk level

The risk for exceeding the 0.025 milligram per liter and 0.1 milligram per liter total phosphorus concentration criteria in the receiving stream downstream from the runoff discharge point with and without BMP treatment are shown as a function of the risk for exceeding the criterion concentrations upstream from the discharge point in

Scatterplots showing the simulated risk for exceeding the selected total phosphorus criterion concentrations upstream and downstream from a discharge point.

Figure 41. Scatterplots showing the simulated risk for exceeding the selected total phosphorus criterion concentrations upstream and downstream from a discharge point

Comparison of simulation results done by using transport curves for all 62 water-quality monitoring stations indicates that only 3 streams do not exceed the 0.025 mg/L criterion concentration, and 31 do not exceed the 0.1 mg/L criterion concentration at the 0.5 percent risk level during storm events upstream from the simulated stormwater outfall of concern (

For both total phosphorus criterion concentrations in

Decision rules are needed to evaluate conditions at unmonitored sites by using available data because there are about 48,000 road stream crossings in southern New England (

Table 30. Correlation between simulated upstream stormflow statistics and land-cover percentages for 62 water-quality monitoring stations with total phosphorus concentration data in southern New England

[Land-cover characteristics (

NLCD land-cover characteristics | Range of land-cover percentages | Upstream concentration at the 0.5 percent exceedance | Percentage of upstream concentrations that exceed 0.025 mg/L | Percentage of upstream concentrations that exceed 0.1 mg/L | |||

Spearman's rho | 95-percent confidence intervals of the correlation coefficient value | Spearman's rho | 95-percent confidence intervals of the correlation coefficient value | Spearman's rho | 95-percent confidence intervals of the correlation coefficient value | ||

Crop/hay | 0 to 48.6 | 0.25 | −0.02 to 0.46 | ^{a}0.33 |
0.06 to 0.53 | ^{a}0.28 |
0.01 to 0.49 |

Wetlands | 0 to 29.8 | −0.21 | −0.44 to 0.05 | −0.09 | −0.34 to 0.17 | −0.21 | −0.43 to 0.05 |

Forest | 5.16 to 92.9 | ^{a}−0.29 |
−0.5 to −0.02 | ^{a}−0.46 |
−0.6 to −0.17 | ^{a}−0.31 |
−0.51 to −0.04 |

Developed | 4.04 to 92.9 | 0.01 | −0.24 to 0.26 | 0.12 | −0.14 to 0.36 | 0.01 | −0.24 to 0.27 |

Impervious area | 0.28 to 53.1 | 0.05 | −0.21 to 0.3 | 0.18 | −0.08 to 0.41 | 0.05 | −0.21 to 0.3 |

Statistically significant at the 95-percent confidence limit.

Formulation of a stormwater treatment decision rule for unmonitored sites by using relations between land cover and total phosphorus is not simple because the correlations between total phosphorus and land-cover percentages are not strong (

Although relations between land-cover percentages and total phosphorus concentrations are not strong, the findings demonstrated in

Simulation results shown in

The simulation results described in this report and the cited literature indicate that alternatives to onsite BMP stormwater treatment are needed if the objective is to improve receiving-water quality. In a long-term study,

The results of this study and the literature indicate that land-preservation efforts may be the most effective offsite mitigation strategy to maintain instream quality upstream from developed areas. This is because onsite BMPs have the greatest effect for preserving good water quality (

The purpose of this report is to document approaches for assessing flows, concentrations, and loads of highway- and urban-runoff and receiving-stream stormwater in southern New England and demonstrate how results may be used by stormwater practitioners to help inform resource-management decisions. This effort was done to help inform scientific decision-making processes used by the Federal Highway Administration, State Departments of Transportation (DOTs), and their regulatory partners. To this end, data and statistics for basin properties, precipitation, streamflow, and water quality were collected and calculated. Using this information, a total of 7,511 simulations were done using the Stochastic Empirical Loading and Dilution Model (SELDM) to examine flows, concentrations, and loads of stormwater in southern New England and adjacent areas. Results of these simulations were used to demonstrate methods for interpreting stochastic simulation results, identify the most important variables of interest, and demonstrate how simulation results can be used to inform scientific decision-making processes.

In this report, southern New England is defined as the areas within Connecticut, Massachusetts, and Rhode Island that drain to the ocean or to large rivers that flow into these areas. For example, tributaries to the Connecticut River within these States are included but the main stem and tributaries completely outside these three States are not. For the purpose of calculating basin properties within these States, the southern New England area also includes headwater areas in New Hampshire, New York, and Vermont draining to streams and rivers predominantly located within southern New England. Data from precipitation, streamflow, and water-quality monitoring stations in New Hampshire, New York, and Vermont also were used to supplement data collected within southern New England to improve statistical estimates.

The data, information, and statistics described in this report are intended to facilitate stochastic analysis of the potential effects of stormwater runoff on receiving waters at unmonitored sites (or sites with limited monitoring data). SELDM can be used to simulate long-term conditions at monitoring sites with data, but there are more than 48,000 delineated road-stream crossings in southern New England. Therefore, the probability that data will be available at a site of interest is very low. Because most water-quality monitoring sites have less than 1 year of data, much of the data available at monitored sites is not sufficient to characterize long-term stormwater-quality conditions. SELDM can be used to simulate these long-term values by using statistics calculated from the available data. The methods, data, and statistics described in this report and the supporting data releases were designed for use with SELDM but may be used with other methods or models.

Information about basin characteristics, storm-event hydrology, stormwater quality, and stormwater treatment were compiled to document data and statistics needed to facilitate planning-level and detailed simulations in southern New England. Stream basins above 48,466 road crossings in this area were delineated, characterized, and included in StreamStats. Main-channel length is highly correlated to drainage area and to main-channel slope, but other characteristics such as stream density and imperviousness are not correlated to other basin characteristic variables. Regression equations using drainage area as the explanatory variable were developed to calculate associated main-channel length and slope values for planning-level purposes. These three variables, and many other variables, can be obtained from StreamStats to simulate conditions for any given location in southern New England. Basin characteristics for highway sites were developed from highway design guidelines, data on 5,480 bridges over water in southern New England taken from the National Bridge inventory, and information about 2,436 stormwater conveyances supplied by State Departments of Transportation. Precipitation and streamflow statistics were compiled from 45 precipitation and 385 streamgages in and around southern New England to calculate statistics for three U.S. Environmental Protection Agency Level III ecoregions that intersect southern New England, for the southern New England area, and for areas within southern New England. Hydrograph recession ratio statistics were calculated from 51 streamgages in this area. Statistics from 4 to 19 highway-runoff sites, 4 to 196 urban-runoff sites, and 6 to 69 stream sites were used to develop individual and regional water-quality estimates. Even when using National datasets, statistics for some water-quality constituents of interest had to be estimated by using alternative methods. Statistics to estimate potential effects of wastewater treatment plant (WWTP) discharges on stormflow quality at any location downstream from such facilities were calculated by using data in discharge permits from 30 to 143 municipal wastewater treatment plants. Methods used to simulate stormwater treatment by structural best management practices (BMPs) were described in this report. Detailed, cited BMP treatment statistics used in this study were derived from data collected at hundreds of sites. Therefore, detailed statistics for flow reduction, hydrograph extension, and water-quality treatment are cited to a recently completed study and were included in the model-archive data release rather than duplicated within this report.

Many of the simulation results were evaluated in terms of dilution factors. Dilution factors are the ratio of the discharge into a stream at a point of interest divided by the concurrent stormflow immediately below this point, which is composed of the discharge plus the concurrent upstream stormflow. A dilution factor of one indicates that the downstream stormflow is 100 percent urban- or highway-runoff discharge from the site of interest during the period of discharge, and a dilution factor near zero indicates that the runoff discharge from the site of interest is a negligible portion of the concurrent downstream flow. Dilution-factor analyses are useful because there are many different water-quality constituents in highway and urban runoff, each of which may have different ecological effects and regulatory criteria. Furthermore, application of water-quality statistics from monitored to unmonitored sites is more uncertain than application of hydrologic statistics from monitored to unmonitored sites. Although dilution-factor analyses provide good screening-level information, reliance on dilution factors alone may not capture all the information needed to assess risks for adverse effects of runoff on receiving waters or the potential effectiveness of mitigation measures to reduce those risks.

An example planning-level stormwater-loading analysis was done by using highway- and urban-runoff constituent yields calculated by using SELDM with basin properties, road lengths, and land-cover percentages from StreamStats. In this report, the term urban runoff is used to identify stormwater flows from developed land-cover areas with impervious fractions ranging from 10 to 100 percent without regard to the U.S. Census designation for any given location. Yields of total nitrogen from highways and other impervious areas with and without BMP treatment were simulated by using SELDM. To calculate runoff loads from different areas, the highway and urban yields were multiplied by the areas of major roads and other impervious areas, respectively. Loads were calculated for 16 tributaries that drain from areas in Massachusetts and Rhode Island to the Narragansett Bay. Using this method, it was estimated that highway runoff may be about 3.6 percent of the impervious stormwater loads to the basin. If estimated WWTP and onsite wastewater loads to the bay are included in the loading estimate, then the contribution of highway runoff would be less than 0.6 percent of the total annual load. Even the 3.6 percent loading value is much less than uncertainties in stormflow loading estimates to the bay.

Sensitivity analyses were done to identify the variables that have the greatest effects on simulated values. Analysts using SELDM can focus efforts to collect or identify data and calculate statistics for the most sensitive variables. Sensitivity analyses were done by varying each variable of interest in turn while holding other variables constant at representative values for southern New England in SELDM simulations. Example graphs of the simulated dilution factors or concentrations with respect to the variable of interest indicated the sensitivity of results to that variable. Variations caused by the perturbation of each variable over different highway and upstream basin configurations were evaluated by examining the average, standard deviation, and coefficient of variation of all the simulations for each combination of basin properties. The drainage-area ratios, in acres of pavement per square mile of upstream basin, were used to show the populations of average and standard deviations for many of the sensitivity analysis outcomes. However, differences in results for different simulations with same drainage-area ratios indicate that proportionality in drainage areas alone will not precisely define simulation outcomes. For example, hydrologic differences between a 1-acre highway site with a 1-square-mile upstream basin and a 10-acre highway site with a 10-square-mile upstream basin, have the same ratio but lead to different results. Therefore, the drainage-area ratio alone is not sufficient to precisely estimate the proportion of runoff from the site of interest in downstream flows. This result is similar to analyses done by the USGS in cooperation with the North Carolina DOT to develop a decision support system that translates information from StreamStats into screening-level treatment decisions without the need for detailed simulations at every site.

Results of the sensitivity analyses on the hydrologic statistics provided information needed to focus attention and resources on the variables that will provide the needed information. If results are not strongly sensitive to a variable, then the statistics for that variable may be specified with less certainty than variables that cause large changes in output values. In this report, comparisons among sensitivity analysis variables were made by using the average of the coefficients of variation (COV) of dilution factors for each set of simulations. The COVs were calculated by dividing the standard deviation of dilution factors by the average of dilution factors for each combination of highway site and upstream basin. Within southern New England, simulation results were only moderately sensitive to precipitation statistics, which had an average COV of 0.064. This is because the same events are applied to the highway and upstream basin and because precipitation statistics do not vary dramatically in this area. In comparison, selection of BMP statistics, streamflow statistics with zero flows, streamflow statistics without zero flows, and the correlation of upstream runoff coefficients to prestorm streamflow had much larger effects on simulation results with COV values equal to about 0.64, 0.24, 0.22, and 0.12, respectively. Alternatively, selection of the basin-lag factor (length and slope) of the highway drainage system and the upstream hydrograph recession ratio were much less sensitive than precipitation with COV values of about 0.043 and 0.021. Although the highway, urban, BMP, and upstream water-quality results could not be compared to other variables by dilution factor analyses alone, the large variations in simulation results were much greater than those caused by random variations among simulations. Simulation results are sensitive to upstream area and imperviousness, but these variables can be reliably obtained from available data for a given site of interest by using StreamStats or geographic information systems.

The sensitivity analyses for streamflow with and without zero flows indicate that results of simulations are highly sensitive to the statistics selected for upstream prestorm flows. The results seem to indicate that the presence of zero flows has a large effect on dilution factors, but these effects are primarily caused by the nonzero flow statistics; basins with zero flows also have lower and more variable nonzero flow volumes than basins without zero flows. This indicates that regulation of water withdrawals may have substantial effects on downstream stormwater quality. Compared to other variables needed to assess stormwater quality, the data, statistics, analytical tools, and decision support systems for estimating streamflow are plentiful. Southern New England has a dense USGS streamgage network and many streamgages with long periods of record. However, compared to the drainage-area distribution of road stream crossings, which have a median drainage area of 0.455 square miles, available streamgage data are by and large from much larger perennial-stream basins. For example, the minimum drainage areas in the SELDM, 1901–2015, and Index streamgage datasets are 10.6, 0.35 and 0.49 square miles and the median drainage areas are 64.0, 20.2, and 20.1 square miles, respectively.

The sensitivity analyses indicate that simulation results are very sensitive to the water-quality statistics used to characterize runoff and upstream water quality. The number of highway and instream monitoring sites is limited in comparison to the 48,000 road-stream crossings in this area, which limits the ability to quantitatively identify archetypal statistics for different types of unmonitored sites. The number of samples per site for many constituents is limited, which limits the ability to quantitatively identify at-site statistics. The example runoff-quality simulations also indicate that instream effectiveness of runoff treatment also is dependent on the statistics used for simulation. These results indicate that more water-quality data are needed to better characterize highway- and urban-runoff quality data and the quality of receiving waters in southern New England. Unfortunately, the time and resources needed to collect and document high-quality stormwater datasets limit the availability of data in southern New England and across the country. The water-quality sensitivity analysis results also indicate that research is needed to guide selection of water-quality statistics for unmonitored sites and to quantify uncertainties introduced in this selection process.

The BMP sensitivity analysis included comparison of long-term average annual load reductions for selected constituents including sediment, nutrients, trace elements, and bacteria. The results presented as long-term average ratios of outflow to inflow loads should not be confused with a deterministic percent-reduction paradigm. The long-term average constituent loads result from compilation of individual event constituent loads that were simulated by using ratios and correlations between outflow and inflow values. These ratios and correlation coefficient statistics were calculated from data in the international BMP database. In this report, the combined long-term hydrologic and water-quality performances of different BMP categories were ranked and cost estimates from a National Cooperative Highway Research Program study were used to develop an adjusted cost rank for the BMP categories with available cost estimates.

Results of the sensitivity analyses indicate development of a decision support system to simplify the process for assessing potential effects of highway or urban runoff and potential mitigation strategies would depend on operational definitions for the input values used and the decision criteria. In North Carolina, the State DOT worked with the USGS and the State Department of Environmental Quality to create a simple system that translates information from StreamStats into screening-level treatment decisions without the need for detailed simulations at every site. The simple decision support system developed in North Carolina does not preclude additional analyses at sites where more detailed information is needed. The information and results of simulations documented in this report provide the foundation for development of a decision support system within southern New England. Adoption of such a system, however, would require collaboration among decision makers to specify operational definitions and decision rules.

A series of example runoff-quality simulations were done to demonstrate how SELDM can be used to inform water-resource management decisions. Water-quality transport curves, which are relations between stormflow and constituent concentrations, were developed by using total phosphorus data from 38 minimally developed basins and 24 developed basins without upstream wastewater treatment plants. Minimally developed basins and developed basins were operationally defined as having total impervious percentages less than or equal to 5 percent and greater than 5 percent, respectively. Results of these 62 simulations indicate that end-of-pipe structural BMP treatment is ineffective for altering downstream water quality unless upstream concentrations are low. The largest and most statistically significant correlation between selected concentrations and basin properties was percent forest cover. Therefore, the example simulation results indicate that in developed areas offsite stormwater mitigation strategies located within or immediately downstream from the least developed areas of stream basins may have the greatest effects on stormwater quality. Furthermore, alternative strategies such as land conservation, may have the greatest potential to maintain instream water quality. The results of the example analyses are similar to results from other simulation studies, ecological analyses, and long-term field studies in the literature.