Scientific Investigations Report 2006–5323

**U.S. GEOLOGICAL SURVEY
Scientific Investigations Report 2006–5323**

This study updated the hydrologic analysis for Duck and Jordan Creeks by using streamflow data, a resource not available for the previous FIS. A log-Pearson type III analysis of the streamflow data, weighted with estimates from regional regression equations for southeast Alaska, provided peak streamflow estimates for various recurrence intervals. For the Duck Creek study reach, which is longer than the Jordan Creek study reach, streamflow estimates for locations upstream and downstream from the gaging station reflect changes in drainage area along the study reach.

Hydraulic analysis consisted of construction of a one-dimensional, steady-state hydraulic model using results from the hydrologic analysis, channel geometry and engineered structures field surveys, LIDAR data interpretations, high-tide elevation, and the HEC-RAS modeling software. Modeling produced estimated water-surface profiles for floods with 10-, 50-, 100-, and 500-year recurrence intervals.

Frequency analysis of a series of annual peak-streamflow data produces estimates of peak-streamflow frequency and magnitude, reported as T-year discharges. T is a recurrence interval, or the number of years during which the discharge is expected to be exceeded once, and is the reciprocal of the annual exceedance probability. For example, every year the 50-year peak streamflow, or 50-year flood, has a 1 in 50, or 2 percent, chance of being exceeded. For Duck and Jordan Creeks, analysis of data from the USGS gaging station on each stream provided initial estimates of the peak-streamflow magnitude for the 2-, 5-, 10-, 25-, 50-, 100-, 200-, and 500-year recurrence intervals (50-, 20-, 10-, 4-, 2-, 1-, 0.5-, and 0.2-percent exceedance probabilities, respectively). Weighting these station-data-based estimates with regional equations developed for Alaska (Curran and others, 2003) provided final peak-streamflow estimates.

The USGS computer program PEAKFQ, available at http://water.usgs.gov/software/surface_water.html, automates the log-Pearson Type III frequency distribution analysis recommended in the Interagency Advisory Committee on Water Data’s Bulletin 17B (Interagency Advisory Committee on Water Data, 1982). PEAKFQ facilitated computation of station-data-based peak-streamflow magnitudes for Duck and Jordan Creeks. Systematic peak-streamflow records are available for the USGS gaging station on Duck Creek from 1994-2004 (11 years of record) and on Jordan Creek from 1998-2005 (8 years of record) (table 1). PEAKFQ identified the water year 2000 peak streamflow for Duck Creek as a high outlier, but no historic period could be determined so the frequency distribution was not adjusted. The 1996 historic peak for Jordan Creek was omitted from analysis because it was less than several systematic peaks (table 1). The generalized skew coefficient and standard error of the generalized skew for Streamflow Analysis Region 1 (Curran and others, 2003) were used for the PEAKFQ analysis. Peak-streamflow estimates computed from PEAKFQ are shown with 5- and 95-percent confidence limits in table 2.

Weighting the gaging-station-based peak-streamflow estimate with a regional-equation-based estimate can temper the uncertainties in streamflow estimates from gaging stations with short periods of record (Interagency Advisory Committee on Water Data, 1982). The Duck and Jordan Creek gaging stations have short periods of record that should statistically benefit from incorporating regional hydrology. The regional equations for southeast Alaska correlate peak streamflows to the basin characteristics drainage area, area of lakes and ponds (storage), mean annual precipitation, and mean minimum January temperature (Curran and others, 2003). Drainage area boundaries drawn on paper USGS topographic maps and digitized provided a digital determination of drainage area. Summing the areas of lake and pond polygons on digital hydrography coverages available at http://agdc.usgs.gov/data/usgs/to_geo.html and dividing by drainage area produced the percentage of lakes and ponds. Mean annual precipitation was computed from an Arc/Info AML application applied to the Arc/Info coverage of plate 2 of Jones and Fahl (1994) and mean minimum January temperature was visually estimated from plate 1 of Jones and Fahl (1994). Basin characteristics for Duck and Jordan Creeks (table 3) are similar, except Duck Creek’s drainage area is about one-half as large as Jordan Creek’s drainage area.

The basin characteristics provided input variables for a USGS computer program developed for application of the regional regression equations (available at https://pubs.water.usgs.gov/wri034188). The program computed equation-based estimates of peak-streamflow magnitude and 5- and 95‑percent confidence limits (table 2), as well as the equation’s site-specific equivalent years of record. Following procedures outlined in Curran and others (2003), the gaging-station-based and equation-based estimates were weighted by the station’s years of record and the equation’s equivalent years of record, respectively. The final peak streamflow estimate for the selected recurrence intervals is this weighted estimate (table 2).

The weighted estimates of peak-streamflow magnitude for the
10-, 50-, 100-, and 500-year recurrence intervals at the gaging station provided
inputs for the hydraulic analysis for all of Jordan Creek. Duck Creek is longer
and includes two tributaries, and adjustments to the estimates to account for
location within the watershed improved the simulated water-surface profile,
particularly in the upper study reaches. Although procedures for estimating
peak streamflow at ungaged locations on gaged streams are available (Curran
and others, 2003), these require the use of the regional regression equation.
The regional regression equation is not valid for drainage areas less than 0.72
mi^{2} in this region. The upper reaches of Duck Creek are too small
for the regression equations to be valid, so the estimates for the gaging station
were adjusted by applying a ratio of the drainage areas of the desired location
and the gaging station. Streamflow is not necessarily directly proportional
to drainage area, but comparison to concurrent USGS miscellaneous measurements
of streamflow at various locations along Duck Creek (table
4) shows drainage-area-weighted streamflows are reasonable. Measured values
were available only for streamflows on the order of the 2-year recurrence interval
flood, requiring the assumption that the flow distribution throughout the watershed
would be similar for larger floods. An arbitrary minimum of 10 percent of the
streamflow in the adjacent main stem was assigned to the flow from the ponds
upstream from Nancy Street, and to the unnamed tributary of the stream upstream
from McGinnis Drive, because of the uncertainty inherent in the delineation
of the drainage area boundaries. Final distribution of discharges to locations
in the hydraulic model is shown in table 5.

Estimates provided for the 500-year peak streamflow should be interpreted cautiously. Record lengths for the Duck and Jordan Creek gaging stations, and for the gaging stations used to develop the regional equations, might not be long enough to support extrapolation to such long recurrence intervals (Curran and others, 2003). For critical applications, site-specific data such as paleoflood indicators might be advisable, although the recent post-glacial history of the Mendenhall Valley limits available flood records.

The discharges applied to the limits of the study area are consistently less than those in the previous FIS (table 5, assuming the upstream-most discharge published for the previous FIS is applied to the upstream study area limit). For example, the present estimate of the 100-year flood for Jordan Creek is 36 percent less than the previous value. These differences reflect changes in analysis methods and source data, particularly the introduction of streamflow data, rather than changes in hydrology.

Water-surface profiles were computed for the study reaches using HEC-RAS version 3.1.3 (U.S. Army Corps of Engineers, 2002a, b, c). HEC-RAS is a one-dimensional modeling system that computes water-surface profiles for gradually varied flow by solving the one-dimensional energy equation and for rapidly varied flow (such as flow at hydraulic structures) by solving the momentum equation (U.S. Army Corps of Engineers, 2002a). The HEC-RAS modeling system can simulate unsteady flow; however, this study used only the steady flow capabilities. For this investigation, ponds along Duck and Jordan Creeks were arbitrarily assumed to have minimal storage capacity. Input parameters for steady-flow analysis in HEC-RAS include geometric and elevation data for the channel, culverts, bridges, and roads; roughness coefficients for the channel, overbank areas, and culverts; and flood discharge.

The initial basis for defining the stream network was a GIS-based hydrology coverage provided by the City and Borough of Juneau. The network for this study consists of the main stems of Duck and Jordan Creeks, plus a small unnamed tributary to Duck Creek upstream from McGinnis Drive. A final stream center line coverage for the study area was generated from this hydrology coverage, survey data for this study, and LIDAR data provided by the City and Borough of Juneau (fig. 5). The primary purpose of these stream center lines was to compute data such as reach lengths for the model, necessitating straight paths through the centers of the numerous ponds along both streams. Thus, the stream center lines are not a precise representation of the streams’ location and morphology.

Further simplification of the stream network was useful to simulate flow leaving the main stem of Duck and Jordan Creeks. Inter-basin flow from Jordan Creek to the southeast at Egan Drive and from Duck Creek to Jordan Creek downstream from Egan Drive was simulated as lateral weirs.

The surveying method for cross sections and hydraulic structures included a combination of Global Positioning System (GPS) and conventional survey techniques. Baselines were created from static GPS data collected at local benchmarks and data from National Geodetic Survey (NGS) Continuously Operating Reference Stations (CORS) sites, then processed and used in a network adjustment to obtain latitude, longitude, and ellipsoid heights. Selected benchmarks with published elevations provided constraints on the GEOID99 geoid model, which then was used to obtain elevations relative to Mean Lower Low Water, a commonly used local vertical datum. Using this local geoid model and the constrained coordinates for the benchmarks, temporary benchmarks (appendix A) were established throughout the study area using GPS Real-Time Kinetic (RTK) techniques. These reference marks became hubs and backsights for a conventional survey of most cross sections and miscellaneous points.

Static and RTK GPS data were collected using Trimble 4700 receivers with a microcentered L1/L2 antenna with a ground plane antenna. Selected NGS, Alaska Department of Transportation and Public Facilities (AKDOTPF), and USGS benchmarks provided pre-surveyed locations for base stations. Published horizontal coordinates for four base stations (lev1, ais1, JUN.TIDAL.GPS, and gus2) provided constraints for network adjustment. Horizontal coordinates for this project are referenced to the NAD83 (CORS96) epoch 2003 datum. A 2002 AKDOTPF leveling survey (Tim Reed, Alaska Department of Transportation and Public Facilities, written commun., 2004) and NGS (http://www.ngs.noaa.gov) provided elevations (orthometric heights) for four base stations (JUN.TIDAL.GPS, EDDIE, 95J16, and UW8043) used to constrain the GEOID99 model. Wild T1600 and Wild T1610 total stations and a Trimble datalogger were used for conventional surveying tasks.

Between August 2004 and June 2005, the USGS surveyed 146 cross
sections at 25 culverts and 5 bridges on Duck Creek, and 60 cross sections at
4 culverts and 10 bridges on Jordan Creek (fig. 6*A-F*).
Two cross sections were added at the study margins, one a duplicate of survey
data and one from LIDAR data. Map identification numbers are cross-referenced
to survey identifiers (used as node descriptions in the hydraulic model) in
table 6 and to letter identifiers given to selected
cross sections. Typically, surveys included two cross sections upstream and
two downstream from each hydraulic structure to help model the hydraulic effect
of the structure, and one or two cross sections between hydraulic structures
to characterize the typical channel geometry. Data for culvert dimensions, materials,
and other physical characteristics were collected with measuring tapes and visual
observations. This study does not include culverts replaced since 2004.

The absolute error in surveyed points, or difference from actual elevation, depends on the absolute error in the GPS network and RTK GPS survey and the relative error in the conventional survey. The absolute error in the GPS network is less than 0.03 ft for the study area. The absolute error in the RTK survey was not determined directly for this study, but is estimated to be less than 0.1 ft based on the equipment and procedures used (Trimble, 1999). The relative error for the conventionally surveyed points, or difference from the elevation relative to the hubs, is less than 0.1 ft. The absolute error for the survey is the sum of these errors rounded to one significant figure, or less than 0.2 ft.

GPS and conventional survey data were processed using Trimble Geomatics Office software (Trimble, 1999). Final horizontal and vertical coordinates and elevations (appendix B) were transformed to station and elevation data for cross sections using a customized macro in a Microsoft Excel® spreadsheet. This macro allows selection of two survey points for alignment of the cross section, and then projects all other points onto this alignment to provide a true channel distance along the alignment.

The City and Borough of Juneau provided Light Detection and Ranging (LIDAR) elevation data (fig. 5) that were used to supplement the survey data for selected locations. Discrepancies between the field-surveyed elevations and LIDAR-obtained elevations were on the order of 0.5–1.0 ft, generally showing the LIDAR data at a lower elevation, limiting the use of LIDAR data to applications insensitive to exact elevations. Most commonly, LIDAR data were used to extend cross sections where an obvious, topographic feature was present. LIDAR data were also used to produce an extra cross section upstream from the upstream-most surveyed cross section on Jordan Creek to accommodate a lateral weir simulating overflow from the Jordan Creek network. Review of LIDAR data helped conceptualize inter-basin flow paths near Egan Drive and Glacier Highway.

Flood profile models require an estimate of channel and overbank
roughness, which provide resistance to flow. Hydraulic roughness is characterized
commonly by Manning’s *n*, a coefficient that cannot be measured directly.
Estimates of Manning’s *n* reflect the boundary roughness generated by
bed materials, vegetation, pavement, or other surface material present, as well
as other types of roughness generated by obstacles and variations in channel
parameters over short distances. Empirically derived tables (for example, U.S.
Army Corps of Engineers [2002b]) list Manning’s *n* for surfaces common
to a developed environment, and various publications show photographs of measurement
sites for the natural riverine environment (for example, Barnes [1967]). Initial
estimation of Manning’s *n* for this study relied on comparison of published
values for various surfaces to field observations and photographs of the channel
and hydraulic structures. Final values were obtained from the calibration process,
which involved adjustments of initial values to obtain a reasonable match of
simulated water surfaces with measured water surfaces during a small flood on
November 21, 2005. Final values determined from calibration were used for the
full range of discharges analyzed.

The range of values of Manning’s *n* used in the model
is presented in table 7. In-channel values are relatively
high in many locations to reflect the presence of dense in-channel vegetation.
In some locations, dense vegetation was present in the channel, but channel
sideslopes were lightly vegetated and sloped up to paved sidewalks and roadways.
In these instances, the channel is rougher than the overbank areas.

Flood conditions are conservatively assumed to coincide with a high tide. The starting water-surface elevation in the model for Duck and Jordan Creeks at the downstream end of the study reach is the high-tide elevation of 20 ft, based on analysis from the previous FIS (Federal Emergency Management Agency, 1990). This is the high tide expected during a month when the 10-, 50-, 100-, or 500–year floods are likely to occur.

Model simulation results show backwater effects from a high tide as far as 720 ft upstream from Berners Avenue on Duck Creek and 750 ft upstream from Yandukin Drive on Jordan Creek during a 100-year flood. Water-surface elevations downstream from these locations would be lower than the simulated flood profile if the 100-year flood coincides with a lower tide.

Water-surface elevations were not available to rigorously calibrate the model to the high flows for which it was designed. However, water-surface elevations surveyed for a small flood on November 21, 2005, provided an opportunity to calibrate in-channel roughness values for a lower flow. The poor results of this calibration led to a downward adjustment of final values to within published ranges for the measured conditions.

Model calibration involves adjusting roughness coefficients
to match simulated water-surface elevations as closely as possible to measured
water-surface elevations. Measured discharges during the November 21, 2005,
flood were 40.9 ft^{3}/s at the USGS gaging station on Duck Creek, just
greater than the 2-year flood magnitude, and 138 ft^{3}/s at the gaging
station on Jordan Creek, between the 2- and 5-year flood magnitudes. For both
streams, the model accounted for sediment that partially blocks the culverts
(standard practice assumes sediment is scoured out for high flow, but sediment
was present for the November 2005 flow). Despite the adjustment for partially
blocked culverts, calibrating the model to the measured water-surface elevations
required raising roughness coefficients to 0.09, a value exceeding generally
accepted published ranges of roughness coefficients for the channel conditions
present.

Finalizing roughness coefficients for the desired higher flows required balancing the calibration results from the lower flow with generally accepted published ranges. Roughness coefficients were reduced to within accepted ranges while minimizing the difference in simulated and measured water levels. The maximum final in-channel roughness coefficient was 0.06. Final simulated water-surface elevations for the November 21, 2005, flood were an average of 0.7 ft less than measured water-surface elevations on Duck and Jordan Creeks (fig. 7). The final simulated 100-year water-surface elevation was 0.4 ft lower than the calibrated model’s 100-year water-surface elevation simulation. Although this drop is within a 0.5 ft tolerance for the model, it indicates that the actual 100-year water-surface elevations could be slightly higher than simulated water-surface elevations.