Flow Durations

Flow-duration curves define the percentage of time flows were equaled or exceeded for a given period (Searcy, 1959). Specific intervals of the curve (1-, 2-, 5-, 10-, 20-, 25-, 30-, 40-, 50-, 60-, 70-, 75-, 80-, 85-, 90-, 95-, 99-percent) were computed for both annual and winter seasons using the R-program EGRET (Hirsch and DeCicco, 2015) for all 97 continuous-record streamgages for the entire periods of record, when applicable. The percent change in the flow-duration intervals for the 80 streamgages included in Watson and others (2005) were also computed. Flow-duration curves help provide information about streamflow behavior in the basin. A steeper sloped flow-duration curve may represent a basin that is more reactive to events or is generally more variable than a basin with a flatter flow-duration curve (fig. 3). The added level of analysis focusing on the computation of flow-duration curves for just the winter season attempts to provide finer information about the basin response to winter conditions. Understanding the winter component of the composite annual flow-duration curve provides an opportunity to assess the influence of the winter season on the annual statistics. Changes in computed flow durations are presented in the subsection titled “Percent Change in Streamflow” in the “Results” section of this report.

Base Flow and Runoff

Streamflow can be separated into the following two basic components: base flow and runoff. Runoff is the portion of streamflow resulting directly from precipitation; base flow is the portion of streamflow sustained by groundwater discharge. Separation of these flow components was achieved using the PART base flow separation program (Rutledge, 1998) within the USGS software Groundwater Toolbox (Barlow and others, 2014). The PART program determines base-flow days according to the number of preceding days with continuous recession. The program then linearly interpolates between base-flow days to determine the portion of base flow during runoff events (Rutledge, 1998, p. 34–36; Rutledge, 2007). The mean base flow and runoff was computed for each year of available record, as well as for the entire periods of record for 96 continuous streamgages. The flow components are not able to be separated for canal flow, as measured at Delaware and Raritan Canal at Port Mercer, N.J. (01460440). Analyzing the amounts of each of these pieces of streamflow records aids in the understanding of water availability and variability.

Streamflow Variability

Skew and kurtosis are measures of extremes within a data distribution. Skewness can be considered as the extent of the tails (extreme events) or degree symmetry of the distribution, whereas kurtosis can be thought of as the heaviness of the tails or peakedness of the distribution. Annual values of skew and kurtosis were computed for all 97 continuous-record streamgages for the given periods of record included in the study using R statistical software (R Core Team, 2019).

The analysis originally followed the previous report in evaluating the variability of streamflow by examining the ratios of flow durations, 10/90 and 25/75. However, this approach was determined to be insufficient because some continuous-record streamgage 90th and 75th percentile flow durations were zero, and, therefore, a ratio could not be calculated. The flow-duration ratios are effective at describing the behavior of the central tendencies of the distribution of flows; however, in the interest of flow variability, the shape and extremes of the distribution were of greater utility. Instead, examination of skew and kurtosis were implemented to supplement trend analysis of high and low flows with changes in streamflow variability.

Streamflow variability can be affected by such factors as increased impervious surface, changes in water use, and changes in climate. High and low flows may not change in magnitude alone, but also could have changes in frequency. Understanding both frequency and magnitude of critical or extreme streamflow aids in water management and planning.

Peak-Flow Ratio

The peak-flow ratio is the annual maximum instantaneous flow divided by the 3-day mean flow for the day before, day of, and day after the peak (Watson and others, 2005). To examine the high-flow event variability, the peak-flow ratio for each year of record was computed for all continuous-record streamgages in the study. A watershed’s ability to store and attenuate runoff is measured by the peak-flow ratio (Watson and others, 2005). The higher the ratio, the flashier the stream with quickly rising and receding peaks. A change in the ratio over time can be an indication of changes in land use or vegetation cover in the basin, as well as changes in the intensity and duration of storms.

Frequency Analysis

A frequency curve relates the magnitude of a variable to its frequency of occurrence (Riggs, 1968). The USGS software Surface Water Toolbox (SWToolbox; Kiang and others, 2018) was used to compute mean high and low flows for a given number of consecutive days (n-day) for each year year of record. For this study, the annual highest and lowest 1-, 7-, and 30-day mean-flow time series were determined for both full-year and winter seasons. The annual period used was based on the climatic year (April 1 to March 31) for low-flow frequency analysis (U.S. Interagency Advisory Committee on Water Data, 1982), and the water year (October 1 to September 30) for high-flow frequency analysis. Under natural conditions, the lowest annual flows typically occur in late summer when evapotranspiration is high and groundwater levels are low. Annual low-flow frequency statistics are, therefore, assumed to represent summer conditions. This assumption does not hold true for streamgages below regulated reservoirs where the lowest flow of the year may occur during winter months, depending on operations of the reservoir. Periods were not truncated to determine strictly summer flow frequencies.

The frequency calculation within SWToolbox determines the plotting positions of the annual n-day series and fits a probability curve to the data using a Log-Pearson type III (LPIII) statistical distribution. Flow exceedance and nonexceedances are then estimated using that curve. A conditional probability adjustment is applied to records with zero flows, in which the probabilities are first determined without including the zero flows, then adjusted according to occurrence of zero flows. Details on the approach are described in the published SWToolbox Techniques and Methods (Kiang and others, 2018).

For high-flow frequency analyses, the probability that a streamflow is exceeded for a given period of time is referred to as the exceedance probability; for low-flow frequency analysis, the probability that streamflow is not exceeded is the nonexceedance probability. The specific high- and low-flow probabilities computed from the curves were the 50-, 20-, 10-, 5-, and 4-percent annual exceedance and nonexceedance probabilities, which correspond to the 2-, 5-, 10-, 20- and 25-year recurrence intervals. The percent change in the high- and low-flow frequency flows for the 80 streamgages included in Watson and others (2005) were also computed and the periods through 2001 were compared against the periods through 2017.

Low-flow frequency statistics such as the annual 7-day low flow with a 10-percent annual nonexceedance probability (7Q10) are often used for permitting and regulation of water resources. New Jersey is among the states that utilize the 7Q10 statistic in several regulatory processes including water allocations permitting (Hoffman and others, 2012) High-flow frequency statistics (as opposed to peak-flow frequency statistics) are not typically used in regulation or structural design but could be considered useful for comparing to peak-flow frequencies and changes to storm variability and water availability.

Trend Analysis

The calculation of frequency statistics and various flow components is helpful for permitting and design, as well as documenting the current state of water resources. Determining if and how water resources are changing over time is critical for water resource management planning, especially in a heavily populated state with a changing climate. Trend analysis was used to further evaluate streamflow records; evidence for trends were computed for: base flow, runoff, all n-day series, skew, kurtosis, and peak-flow ratios. The MK test was used for this study as it is appropriate for hydrologic data. It is a nonparametric test which does not require that the data be normally distributed. A trend was considered to be statistically significant when p<0.05.

The MK test uses the ranks of values in the dataset to detect the presence of a trend. In using the ranks of values, the test is not subject to influence from outliers or rare events, which are common in hydrologic data in the form of floods and droughts. However, it also assumes that the data are independent (have no serial correlation). Statistical evidence of a trend can be influenced by the presence of serial correlation, which can be present in hydrologic data series in the form of persistent wet and dry climate cycles.

The presence of serial correlation can be alleviated one of two ways. It can either be removed from the data before applying the trend test, known as “pre-whitening” (Hamed, 2008). The second option is to modify the test to account for the serial correlation. To account for the possibility of serial correlation in the data for this study, the variance of the MK test statistic was adjusted to compensate for an assumed condition of serial correlation.

The MK test was used for three different assumptions of serial correlation: independence (meaning no serial correlation), short-term persistence (STP), and long-term persistence (LTP) using variance correction methods from Hamed and Ramachandra Rao, 1998; Hamed, 2008; and Hodgkins and others, 2017. Short term persistence assumes a significant correlation of data lagged by 1 year with itself. Long term-persistence assumes significant correlations of data lagged at intervals greater than 1 year. Measures of serial correlation were computed using the lag-1 correlation coefficient for STP and the Hurst coefficient for LTP (Hamed and Ramachandra Rao, 1998; and Hamed, 2008).

Index Site Selection

Although the MOVE1 program incorporates regression error and the use of multiple index sites, the selection of index sites used has a direct influence on the estimated flow trends in residuals at the partial-record streamgages. Knowledge of the local hydrology and index site conditions are important factors that can influence streamgage selection beyond the criteria of little to no known regulation. Other criteria considered in the index site selection process include:

Streamgages were then grouped according to physiographic province (table 2 and fig. 4). Partial-record streamgage basins falling within the same physiographic province all used the same list of index sites.

An exception was made to the 25-year requirement for the unglaciated New England Province, as there are not many index sites in this physiographic province. Three index sites were used for more than one physiographic province because of their proximity and mixed basin characteristics.

Another exception was made with the Neshanic River at Reaville, N.J., USGS site 01398000 streamgage. Although relied upon in the previous study, it failed the KS test for the 30-day low, and the annual 1-day low flow with a 10-percent annual nonexceedance probability (1Q10) is now zero. Having a 1Q10 frequency magnitude of zero means this streamgage can no longer be used in the MOVE1 regression analysis because an input variable value of zero would always result in an estimate of zero at the partial-record streamgage.

Variability—Skew and Kurtosis

Examining the annual distribution of streamflow provided a general assessment of possible changes to stream variability without choosing specific streamflow characteristics, such as the annual 1-day high flow. In using skew and kurtosis, the extremes of the distribution are characterized and assessed for changes. Twenty-four streamgages indicated statistically significant trends in the annual skew of the distribution of daily mean streamflow, all of which were increasing except for one (Saddle River at Lodi, N.J., USGS site 01391500). Of the 24 streamgages, 23 also indicated statistically significant trends for kurtosis; all the kurtosis trends were also increasing except for 01391500. One additional streamgage (Crosswicks Creek at Extonville, N.J., USGS site 01464500) indicated an increase in kurtosis but not skew.

The distribution of streamflow within a given year is generally skewed to the right, or positively, with high-flow events causing the shape to be “right-tailed.” Increasing skew over time suggests that the shape of the distribution is changing, possibly because of larger flow events (the tail) becoming greater in relation to the bulk of the streamflow record. Increasing kurtosis indicates that the data within the distribution are occurring more often in the tail or that extreme events are occurring more often as time progresses. The Saddle River at Lodi, N.J., 01391500 streamgage is the only exception to this trend, where the data indicate that the highest flow events are not as large and are occurring less often.

Record length appears to be a factor in whether skew and kurtosis are significantly changing. Of the 25 streamgages indicating a trend in either skew or kurtosis, 18 (>70 percent) had records longer than 80 years. Meanwhile, streamgages with more than 80 years of record account for 28 percent of the number of streamgages included in this study (fig. 2). Even though more streamgages with longer records had significant positive trends, the magnitude of the trend, as computed by Sen’s slope was less (fig. 8) for those longer record streamgages.

Base Flow and Low n-day Flows

The range or magnitude of the low-flow characteristics tested for trends speaks to possible changes in variability of streamflow. The characteristics tested include annual base flow and n-day series of the lowest 1-, 7-, and 30-day flows during annual and winter seasons. Base flow is the largest or most coarse measure of what is considered “low flow” in this study. The n-day series captures finer scale measurements of low flow, with the lowest 1-day flow representing an extreme event for a given year. Though annual lowest n-day flows series are determined using the full climatic year (April 1–March 31) it is assumed these lowest events occur in the summer, following the typical hydrologic cycle of the northern hemisphere. This assumption does not hold true for streamgages that have controlled or regulated flow. Examining trends and their occurrence of either annual (summer) and winter can inform shifts in climate or seasonal weather patterns.

In all, 46 continuous-record streamgages indicated a statistically significant trend for at least 1 or more of the low-flow characteristics computed in this study and are summarized in table 5 and figure 9; more detailed information for the continuous-record streamgages can be found in the associated date release (Williams and others, 2024). Decreasing flows trends were identified at 19 streamgages, and flow is controlled or regulated at 2 of those streamgages. Increasing flow trends were identified at 27 streamgages, 10 of those streamgages are regulated. Recall from the “Methods, Continuous Records Streamgages” section that for regulated streamgages, periods of record used in the calculations do not include earlier unregulated periods. The same proportion of negative trends were observed in the north and south (about 17 percent of streamgages in each region); positive trends appear to be concentrated in northern New Jersey (fig. 10).

Whereas table 5 and figure 10 illustrate the number of trends for each individual low-flow characteristic, as well as the totals for each season, the trends can also be explored in various alternate groupings. At the coarsest measure of low flow considered—base flow—statistically significant trends were identified at 18 streamgages (6 negative, 12 positive). At 6 of those streamgages, base flow was the only low-flow characteristic with evidence of a trend (2 negative, 4 positive). A larger number of streamgages (39) were identified as having a statistically significant trend for at least one of the finer scale 1-, 7-, or 30-day low-flow series for either annual, winter, or both periods. Only 18 of those 39 streamgages indicated trends during both annual (summer) and winter low flow (3 negative, 15 positive).

For any annual (summer) n-day low-flow series, a total of 32 streamgages indicated statistically significant changes in flow (12 negative, 20 positive). Of the total 32 streamgages, more than half (18) indicated a trend for all 3 n-day series (7 negative, 11 positive). For any winter n-day low-flows, 25 streamgages indicated statistically significant changes in flow (6 negative, 19 positive). Of the 25 total streamgages, only 7 indicated a trend for all 3 n-day series (2 negative, 5 positive).

Sen’s slope was then divided by the streamgage drainage area so that the degree of indicated change could be compared among the streamgages for each statistic. Note, however, that this comparison does not necessarily infer that the trend is the result of basin-wide changes but could also be the result of a point-source discharge or withdrawal at any point in the basin, including immediately upstream of the given streamgage.

The largest statistically significant negative trend in cubic feet per second per square mile per year (ft³/s/mi²/yr) in base flow was −0.0429 ft³/s/mi²/yr at Whippany River near Pine Brook, N.J., USGS site 01381800. However, none of the n-day flow series showed evidence of a statistically significant trend at this location. This streamgage has a relatively short period of record of 21 years. Another streamgage located further upstream on the Whippany River (Whippany River near Morristown, N.J., USGS site 01381400) had the second largest statistically significant negative trend in base flow at −0.0358 ft³/s/mi²/yr. The data at this streamgage also did not show evidence for statistically significant trends in any of the n-day low-flow series. The period of record at Morristown is similar at 22 years. The streamgage Whippany River at Morristown (01381500) is located between the streamgage in Pine Brook (01381800) and the streamgage near Morristown (01381400) and has 96 years of record. At this streamgage, all low-flow characteristics except for the winter 30-day low flow indicated statistically significant positive trends. The different trends indicated at each streamgage are likely influenced not only by length of record, but also by local water use upstream of each streamgage.

The largest positive trend in base flow was 0.0136 ft³/s/mi²/yr at the streamgage at Hohokus Brook at Ho-Ho-Kus, N.J., USGS site 01391000. At this location, the n-day series for both annual and winter flows indicated statistically significant positive trends and account for the largest positive trends in the n-day flows, with the highest at 0.0191 for the 30-day low-flow series (table 6). Streamflow at Hohokus Brook at Ho-Ho-Kus, N.J., 01391000 receives contribution from several point-source dischargers upstream of the streamgage.

The largest statistically significant negative trend in ft³/s/mi²/yr for the n-day low-flows was −0.0142 at East Branch Paulins Kill near Lafayette, N.J., 01443280, for the 1-day annual low-flow (table 6). The period of record for this streamgage is relatively short at 25 years, beginning only in 1993. This streamgage did not have any other flow characteristics that showed a statistically significant trend. In examining the daily flow record at this streamgage, the statistically significant decrease in 1-day low flows appears to be attributable to a few extreme occurrences which appear to be outliers and likely anthropogenic in nature. A quarry on an upstream tributary may be contributing to the change in flow.

Examining the magnitude and direction of trends versus record length can inform the influence of cyclical wet and dry periods on the statistics, as opposed to assuming the existence of and adjusting for serial correlation (fig. 11). Generally, streamgages with less than 60 years of data (record beginning after 1956) were associated with the greatest range of low-flow Sen’s slope. The overall range of Sen’s slope per drainage area is listed in table 6 (−0.0429 to 0.0191 ft³/s/mi²/yr). For records between 20 and 40 years in length (beginning after 1976), most trends were negative (10 of 13 streamgages); for records between 40 and 60 years in length (beginning between 1957 and 1976), most trends were positive (8 of 11). Streamgages with greater than 60 years of record (beginning prior to 1956) have smaller ranges of low-flow Sen’s Slopes (−0.0095 to 0.0048), and most trends were positive (17 of 22).

Runoff and High n-day Flows

The high-flow characteristics examined for trends in this report mirror the range and magnitude of low-flow statistics, with runoff (the counterpart to base flow) as the coarsest measure of the upper half of the hydrograph and the highest 1-, 7-, and 30-day high flows informing possible changes in duration and magnitude of large rainfall events. Examination of n-day high flows can inform water managers of more common flooding conditions as well as extreme events without the large uncertainty and variation that can be associated with frequency analysis of annual peak flows. Highest n-day flows typically occur during the winter. Although it may seem redundant, differences in trends of annual compared to winter high flows can still inform shifts in seasonal climate if differing numbers of trends are observed in each season, though not ideal. The use of annual and winter high values in this study were implemented, in part, to replicate Watson and others (2005).

A total of 35 continuous-record streamgages indicated a statistically significant trend for at least 1 or more of the high-flow characteristics computed in this study (table 7 and fig. 12). Most of the trends indicated increasing flows (6 negative, 29 positive). Of the negative trends, flow is regulated or controlled at 1 streamgage; of the positive trends, 6 streamgages are regulated (fig. 13). All but a few of the positive high-flow trends were in northern New Jersey.

At the coarsest measure of high flow—runoff—statistically significant trends were identified at 17 streamgages (5 negative, 12 positive). Four southern New Jersey streamgages indicated a trend only for runoff (all negative). As was the case for low flow, a larger number of streamgages (31) indicated statistically significant trends at the finer-scale 1-, 7-, or 30-day high flow series for either annual, winter, or both seasons. In the case of high-flow characteristics, approximately half of those streamgages (16) indicated changing flow for both annual and winter n-day values. Only one of the streamgages which had both annual and winter n-day high-flow trends indicated decreasing flow.

For any annual n-day high-flow series, a total of 28 streamgages indicated statistically significant changes in flow (2 decreasing, 26 increasing). Of the 28, only 8 indicated a trend for all 3 n-day series (all positive). For any winter n-day high-flow series, fewer (19) streamgages indicated statistically significant changes in flow (1 decreasing, 14 increasing). Of the 19 streamgages, only 4 indicated a trend for all 3 n-day high flow series (all positive). With a greater number of streamgages having evidence for trends in the annual versus winter series, it suggests that highest flows may be shifting to later in the spring (after April 30).

As with the magnitudes of low-flow trends, Sen’s slope was divided by drainage area so comparisons could be made among streamgages. Unlike the low-flow characteristics, high-flow trends are not likely attributable to point-source discharges and withdrawals, which typically account for a much smaller proportion of the highest n-day flow or base flow values. In addition to shifts in climate, regulation and changes in land cover may contribute to changes in n-day high flows.

The largest statistically significant negative trend in runoff was −0.0129 ft³/s/mi²/yr at the Hackensack River at New Milford, N.J., USGS site 01378500 (table 8), a heavily regulated streamgage located immediately downstream from Oradell Reservoir, with 96 years of record. In addition to runoff, the annual 30-day, winter 7-day, and winter 30-day high-flows also indicated negative trends. The largest statistically significant positive trend in runoff was 0.0051 ft³/s/mi²/yr at the Pequannock River at Macopin Intake Dam, N.J., USGS site 01382500 (table 8), also a heavily regulated streamgage with 56 years of record. All n-day high-flow series also indicated positive trends.

The largest statistically significant negative trend for the n-day high-flow series was −0.1167 ft³/s/mi²/yr at the Whippany River near Pine Brook, N.J., 01381800, the annual 30-day high-flow (table 8). No other n-day high-flow statistics, either annual or winter, showed evidence of a trend. Recall however, that base flow at this streamgage also had the largest negative trend in base flow and it only has 21 years of record.

The largest statistically significant positive trend for the n-day high-flow series was 0.3071 ft³/s/mi²/yr at Mantua Creek at East Holly Avenue at Pitman, N.J., USGS site 01475001, for the annual 1-day high-flow (table 8). All the n-day high-flows, as well as runoff, indicated statistically significant increases. The period of record at 01475001 stretches back to 1940; however, there are two gaps in the record, 1976–2003 and 2006–09, for a total of 45 years of record. This streamgage is occasionally regulated from gates at Kressey Lake.

As was observed in the low-flow trends, the largest range of Sen’s slopes for statistically significant high-flow characteristics were associated with streamgages having less than 60 years of record (−0.1167 to 0.3071; fig. 14). Only two streamgages with record length between 20 and 40 years indicated a statistically significant high-flow trend. The Whippany River near Pine Brook, N.J., 01381800 streamgage showed a negative trend in the 30-day high, whereas Pequest River at Huntsville, N.J., USGS site 01445000, showed a positive trend in 1-day high-flows. For streamgages with records greater than 60 years, the range in Sen’s slope for high-flow trends was −0.0197 to 0.1281 ft³/s/mi²/yr, with decreasing runoff only accounting for negative trends at 4 of the 6 streamgages with negative trends.

Peak-Flow Ratio

As a way to assess both a high-flow characteristic and flow variability, the annual peak-flow ratio was examined. An increasing peak-flow ratio over time would indicate that the annual instantaneous peak is becoming larger in relation to the mean daily flow one day prior, the day of, and one day after the peak. This circumstance suggests that more intense rainfall events are occurring, or that the streamflow is becoming “flashier,” or occurring during times of relatively low base flow in the summer. A decreasing peak-flow ratio over time would indicate that the peak is becoming smaller in relation to associated 3-day mean flow. This circumstance suggests that peaks may be prolonged or increasing in duration, or perhaps occurring on already saturated soils with higher prepeak mean daily flow or possibly smaller peaks happening more frequently on saturated soils.

Twenty-five streamgages indicated a statistically significant trend in peak-flow ratio (Williams and others, 2024). More than 75 percent of the trends indicated an increase in the peak-flow ratio over time. These results coincide with the premise that 1-day high flows may be occurring more often in the annual (summer) series than in the winter season.

The magnitude or the number of statistically significant trends does not appear to be influenced by length of record (fig. 15). Streamgage West Branch Middle Brook near Martinsville, N.J., USGS site 01403150, has an exceptionally high increasing Sen’s slope. This streamgage has the smallest drainage area of all those included in this study, 1.99 square miles.

Changes in Trends at Continuous-Record Streamgages

Among the 80 continuous-record streamgages that were included in both the current and previous (Watson and others, 2005) studies, trends in the annual 1-, 7-, and 30-day high and low, annual, and winter flow series were computed using the MK trend test under the assumption of serial independence. Trend test results indicated a considerable increase in the number of statistically significant positive trends, in both high and low flows, for both seasons (fig. 19).

The total number of positive n-day high-flow trends more than doubled from 37, using data through 2001, to 82, using data through 2017. The largest increase in number of positive trends was observed for the annual 30-day high flow. The total number of negative or decreasing n-day high-flow trends dropped from 5 to 3 for the same datasets. A similar change in the number of trends in n-day low flow was observed, but to a lesser degree, with positive trends ranging from 61 to 82 between the two datasets. The total number of negative trends in n-day low flows declined from 37 to 25 (fig. 19).

Percent Change in Streamflow

The determination of trends in streamflow noted in the previous sections evaluates the full periods of record for statistically significant changes. To quantify changes in streamflow since the previously computed statistics through 2001 (Watson and others, 2005), the percent change in individual streamflow statistics were computed. This determination can provide information of possible impacts of the updated statistics on streamflow permits and prior management strategies, regardless of whether the data show a statistically significant trend. Comparisons in continuous-record streamgage-flow statistics include changes in computed durations (1-, 2-, 5-, 10-, 20-, 25-, 30-, 40-, 50-, 60-, 70-, 75-, 80-, 85-, 90-, 95-, 99-percent) and all high and low flow frequencies (1-, 7-, and 30-day high and low flows at 50-, 20-, 10-, 5-, and 4-percent exceedance and nonexceedances, for annual and winter periods). For partial-record streamgage statistics, percent change in the 1-, 7-, 30-day annual, and 7- and 30-day winter low flows at 10-percent annual nonexceedance probability, the 75th and 50th percentile flow, and the annual mean flow were computed.

Continuous-Record Percent Change in Streamflow

Most changes to computed streamflow characteristics for both flow durations and frequencies were generally within ±10 percent using the additional data collected among the 80 continuous-record streamgages that were included in both this and the previous report. The lower flow characteristics typically accounted for changes greater than ±10 percent from the previous report, with a full range of change from −100 to +500 percent.

A greater number of both annual and winter flow duration values indicated decreases in flow durations, though the change was mostly minimal—between 0 and −5 percent. Histograms demonstrating the distribution of the annual and winter flow percent change among the 1,360  individual computed durations (17 durations among the 80 sites) are shown in figure 20. Of those computed annual flow durations, 61 percent (830) indicated decreases and 38 percent (514) indicated increases in flow from 2001 to 2017; 1 percent (16) indicated zero change. For winter flow durations, 57 percent (778) indicated decreases and 39 percent (525) indicated increases; 4 percent (57) indicated zero change. Only 13 and 9 percent of computed annual and winter durations indicated a change in flow greater than 10 percent. The histograms do not include values of zero change because the zeroes would fall into the −5 to 0 percentage bin, making it appear there were more negative changes than were actually computed.

Percent change was not consistent in either the positive or negative direction for all 17 computed durations for a given site, indicating some changes to the distribution of flow in the duration curves. The 50th through 5th percentile flows indicated mostly negative changes. Lowest flows (99th to 95th percentiles) indicated greatest percent change with most outliers greater than ±50 percent, but, overall, leaned more positive. Boxplots of the distribution of percent change of each of the 17 annual and winter flow durations are shown in figure 21 and figure 22.

The greatest decreases in the lower flow durations (99th to 90th) were observed at both regulated and unregulated sites (USGS sites 01378500, 01407500, 01382500, 01401650, 01381000, and 01396800). The outliers of percent change ranged from −47 to −100 percent. The greatest increases in the lower flow durations were at mostly unregulated sites (USGS sites 01396800, 01401000, 01455500, 01403150, 01466500, 01443900, and 01403540). The outlier changes ranged from +22 to +289 percent.

The computed frequency statistics, which are generally associated with the rarest occurring flow conditions, indicated more increases than decreases for both low and high flows from 2001 to 2017. Histograms demonstrating the distribution of the low and high flow computed percent changes among the 1,920 individual computed frequency statistics (12 n-day frequencies for both annual and winter flows among the 80 sites) are shown in figure 23. The histograms do not include values of zero change. Of the low-flow frequency statistics, 7 percent (125) indicated zero change; for high flow, 1 percent (11) indicated zero change. Generally, most changes in low and high frequency statistics were slightly greater than the changes in durations statistics, ranging mostly from −10 to +15 percent. A greater proportion indicated positive changes in both low and high flow frequencies, at 58 and 75 percent, respectively.

A histogram plot showing the percent change in low-flow statistics was generally between −10 to +15 percent, and high-flow percent change was generally between -5 and +15 percent change for the majority of streamgages. The outliers and ranges of percent change do not indicate a clear pattern or bias for a particular n-day or recurrence interval, as shown in the boxplots in fig. 24 and fig. 25. As mentioned in the “Methods” and “Frequency Analysis” sections, some of the changes in frequency flow statistics at regulated streamgages in this study may be a result of the different approaches used in analysis of data that did not fit the LPIII distribution; that is, using a graphical or visual fit compared to computing a percentage of the flows.

The largest increases in low-flow frequency statistics were at the streamgage below Musconetcong River at outlet of Lake Hopatcong, N.J., USGS site 01455500. The site is heavily regulated with a minimum passing flow. The second largest overall increase was observed at Rockaway River below Reservoir at Boonton, N.J.,  01381000, which is also heavily regulated with a minimum passing flow. Increases were between 0 and +500 percent.

The streamgage with the largest decreases in low-flow frequency statistics was Absecon Creek at Absecon, N.J., USGS site 01410500. Flow at this site is unregulated but is affected by tidal fluctuations. The site with the second most decreases in flow frequencies was at Neshanic River at Reaville, N.J., 01398000. This site is not regulated (has no controlled flow upstream). Decreases were between 0 and −100 percent.

For high-flow frequencies, the two streamgages with the highest increase were Pequest River at Huntsville, N.J., 01445000, and Mantua Creek at East Holly Avenue at Pitman, N.J., 01475001. Neither streamgage is regulated. Changes were between 2 and 50 percent for the two streamgages.

The streamgages with the greatest decrease in high-flow frequencies were West Branch Middle Brook near Martinsville, N.J., 01403150, and Cedar Creek at Western Blvd near Lanoka Harbor, N.J., 01408900. The Martinsville streamgage has a very small drainage area (1.99 square miles), and the Lanoka Harbor streamgage is affected by tide. Neither site is regulated. The decreases were relatively minimal, ranging from −2 to −19 percent.

Partial-Record Streamgage Percent Change in Low Flow

Of the 719 partial-record streamgages in the current analysis using data through 2017, 462 of the streamgages were also computed in the 2005 report, which used data through 2001, and were available for comparison. As indicated earlier, partial-record streamgage statistics were computed using a MOVE1 relation with 22 continuous-record streamgages (index sites), 19 of which were also included in the previous report. The percent change in the computed statistics at the index sites from 2001 to 2017 ranged from −15 to +25 percent, but only about 9 percent of those changes were greater than 10 percent (table 12).

Table 12.    

Percent change in computed low-flow statistics at continuous-record streamgages in New Jersey, from 2001 to 2017. The streamgages were used to estimate low-flow statistics at partial-record streamgages in New Jersey, through 2017.

[Data are given in percentages. USGS, U.S. Geological Survey. Blue shading indicates changes greater than 10 percent, red shading indicates changes less than -10 percent]

Table 12.    Percent change in computed low-flow statistics at continuous-record streamgages in New Jersey, from 2001 to 2017. The streamgages were used to estimate low-flow statistics at partial-record streamgages in New Jersey, through 2017.

USGS station number	Station name	1-day, 10-year low flow	7-day, 10-year low flow	30-day, 10-year low flows	7-day, 10-year low flow, winter	30-day, 10-year low flow, winter	75th percentile	50th percentile
01379000	Passaic River near Millington NJ	−8.0	−15.4	−9.1	8.3	10.4	−4.2	0.0
01380450	Rockaway River at Main Street at Boonton NJ	−7.2	−5.9	−1.2	−0.5	−3.7	−1.2	−0.8
01408500	Toms River near Toms River NJ	−6.7	−5.8	−6.0	−2.2	−2.8	−1.9	−0.9
01396660	Mulhockaway Creek at Van Syckel NJ	−3.4	−4.3	0.4	−4.5	−3.7	0.3	−3.4
01411000	Great Egg Harbor River at Folsom NJ	−2.9	−3.1	−2.0	−0.4	−3.2	−0.5	−1.4
01467150	Cooper River at Haddonfield NJ	−2.3	0.0	−2.0	0.3	3.4	−2.2	−5.3
01408000	Manasquan River at Squankum NJ	−2.0	−4.5	−0.5	−4.6	−1.6	−2.8	−3.0
01467000	North Branch Rancocas Creek at Pemberton NJ	0.0	−1.6	−3.1	−2.5	−2.3	−2.9	−2.0
01408120	North Branch Metedeconk River near Lakewood NJ	0.8	−6.0	−4.5	−1.2	1.7	−2.6	−2.2
01409810	West Branch Wading River near Jenkins NJ	1.1	1.4	0.4	−4.3	−8.4	−2.6	−0.3
01411500	Maurice River at Norma NJ	1.7	−1.9	−1.3	0.8	0.6	0.8	0.7
01440000	Flat Brook near Flatbrookville NJ	2.0	2.0	0.5	0.1	0.2	1.1	1.2
01443500	Paulins Kill at Blairstown NJ	3.8	−2.3	3.0	3.4	3.4	2.1	1.2
01445500	Pequest River at Pequest NJ	4.4	2.5	−0.2	0.9	0.1	−0.3	−1.1
01386000	West Brook near Wanaque NJ	8.7	1.7	−1.1	8.1	10.4	1.8	−2.1
01410150	East Branch Bass River near New Gretna NJ	9.1	7.0	7.1	3.3	1.2	−0.1	−5.3
01446000	Beaver Brook near Belvidere NJ	12.5	11.1	13.6	0.6	−0.8	1.2	1.7
01396582	Spruce Run at Main Street at Glen Gardner NJ	16.3	14.9	16.2	5.9	7.8	13.0	4.1
01401000	Stony Brook at Princeton NJ	25.0	0.0	20.7	6.2	8.9	6.9	0.6

Overall, the changes in the low-flow statistics at the partial-record streamgages were greater than those observed at the continuous-record streamgages. Like the continuous-record low-flow frequency statistics, the change skewed slightly toward more increases in flow. Most percent changes in partial-record streamgages low-flow statistics ranged from −15 to +20 percent (fig. 26). Of the possible 3,696 values of change (8 computed statistics for 462 streamgages), about 17 percent (634) of values indicated zero change and are excluded from the histogram. Percentage change could not be computed for 6 percent (217) of values, as the 2001 computed value was a zero.

A histogram plot of percent change in low-flow statistics at 462 partial-record streamgages in New Jersey shows a broad distribution of streamgages between -30 and +50 percent change. The range of percent change among the individual low-flow statistics is consistent, as seen in the boxplots in figure 27. The greatest changes were observed as outliers at the lower flows, the 1-day annual and 1- and 7-day winter, 10-year statistics. The annual median flow indicated the least amount of change among the partial-record streamgage statistics.