Model Scenarios

Fifty-two hydrologic and hydraulic model scenarios were developed for this study (table 4). The scenarios included 24 HEC–HMS-only scenarios used in determining hydrologic effects of land cover and projected precipitation on a basin scale. An additional 28 scenarios using HEC–HMS and HEC–RAS were used to determine hydraulic changes within the SAFB focus area resulting from detention and projected precipitation options. The HEC–HMS-only scenarios were run on the Silver Creek Basin upstream from the Freeburg streamgage with precipitation frequencies of 2-, 10-, 25, 50-, 100-, and 500-year recurrence intervals (50-, 10-, 4-, 2-, 1- and 0.2-percent AEP) over a 24-hour duration using the normal antecedent response condition in HEC–HMS (refer to “Loss Method” section in this report). Scenarios were run with historical (1992), current (2019), and planned land cover in addition to scenarios with planned land cover and projected (2050) precipitation. The hydrologic characteristics of the simulations—consisting of peak streamflow magnitude, timing of peak streamflow, and cumulative volumes—were compared against the current condition scenario.

Simulated flows from 2-, 5-, 10-, 25, 50-, 100-, and 500-year recurrence interval (50-, 20-, 10-, 4-, 2-, 1- and 0.2-percent AEP) precipitation events distributed over a 24-hour duration were used for the HEC-HMS with HEC-RAS scenarios. The objectives of these scenarios were to determine differences in hydraulic characteristics (water-surface profiles and inundation extents) of streams within and adjacent to SAFB resulting from added detention and the projected (2050) precipitation condition. The detention storage scenarios were developed to assess the potential effects of a representation of amended storage on the hydrology at a reference location rather than to serve as a prescriptive design for mitigation of peak streamflows. Detention structure components (dam height, outlet culvert diameter) incorporated into the HEC–RAS geometry were designed to maintain a storage pool within a designated open area in the target subbasin, store a 25-year recurrence interval event, and retain a targeted range of 20 to 100 acre-feet (acre-ft) of storage per square mile of drainage area (table 5).

Elevation Data

The elevation data used in the HEC–HMS model were obtained from several 1-meter DEMs derived from light detection and ranging (lidar) data published between November 14, 2017, and April 6, 2022 (U.S. Geological Survey, 2021c). The data were acquired and processed by Merrick-Surdex Joint Venture, LLP and Aerial Services, Inc. As per USGS quality level 2 standards (version 1.2; Heidemann, 2018), lidar data required a nonvegetated vertical accuracy of a maximum 10-centimeter (cm) root mean square error, and a vegetated vertical accuracy of a maximum of 30 cm at the 95th percentile (actual was 9.7 cm at the 95th percentile).

Land-Cover Change

Changes in land cover were determined using information and projected development data from national databases and selected surrounding counties and municipalities. The historical land-cover condition used in the basin corresponded to the 1992 Enhanced National Land Cover dataset (Nakagaki and others, 2007). Current conditions were represented by the 2019 National Land Cover Data (Dewitz and U.S. Geological Survey, 2021). Maps of future planned developments on SAFB and vicinity (University of Illinois, 2006; The Village of Shiloh, 2018; Saint Clair County, 2023) were georectified in ArcGIS Pro 3.1.2 (Esri, 2023) and mosaiced with the 2019 land cover. The differences in land cover translated into differences in model parameter values used for generating runoff for the historical (1992), current (2019), and planned (to about 2050) land-cover hydrologic models. Differences in land-cover development categories are the primary driver of differences in runoff between historical, current, and future/planned conditions. The change in total development categories (sum of developed open areas, low intensity development, medium intensity development, and high intensity development areas) between current and historical conditions indicated that changes in development within most subbasins were small (less than 10 percent). Subbasins with changes in development of more than 30 percent primarily were in model subbasins adjacent to and within SAFB (refer to fig. 4 inset). Notable changes in development were determined for subbasins in the lower Ash Creek, lower Cardinal Creek, lower South Ditch, and lower Little Silver Creek Basins. Additionally, changes in development of 10 to 20 percent also were determined for several subbasins in the upper Silver Creek Basin near the Troy streamgage (fig. 4). Two subbasins had decreases in development between the two periods of 0 to −3.7 percent, and this can be attributed to differences in land-cover classification between areas defined as “developed open areas” and “pasture” in the 1992 (Nakagaki and others, 2007) and 2019 (Dewitz and U.S. Geological Survey, 2021) land-cover classifications.

For those subbasins not included in available future comprehensive plans, the change in planned development was assumed to be equivalent to that documented between the 1992 (Nakagaki and others, 2007) and 2019 (Dewitz and U.S. Geological Survey, 2021) land-cover classifications. The change in development under planned conditions from existing conditions (fig. 5) indicated a similar spatial development pattern to that for current and historical conditions. There was less than a 10-percent change in developed categories in most upper Silver Creek subbasins, and subbasins with more than a 20-percent change in development were located in the lower one-third of the Silver Creek Basin. The maximum change in planned compared to current development was expected to be higher (as much as 61.4 percent) than that determined for current and historical (40.2 percent) development. The greatest change in development on base corresponded to subbasins in Ash Creek and Cardinal Creek, and in a subbasin in a Silver Creek tributary to the east of SAFB (fig. 5).

Precipitation Data

Accurate precipitation data are needed as inputs to the hydrologic model to simulate the resulting runoff and streamflow. Required current and historical precipitation data were unavailable for stations within the Silver Creek Basin and, therefore, hourly gridded Multi-Radar Multi-Sensor Q3 estimate (Iowa State University, 2021) precipitation data were used for the calibration and validation events. The 1-square-kilometer gridded Multi-Radar Multi-Sensor precipitation data also accounted for differences in the spatial distribution of precipitation within the Silver Creek Basin.

Precipitation values used as input to the hydrologic model for the 52 modeled scenarios (table 4) were developed from precipitation frequency estimates (Angel and others, 2020) for southwest Illinois. Projected climate condition scenarios included adjustment factors to the precipitation frequency estimates that corresponded to a climate model scenario using the representative concentration pathway of 4.5 watts per square meter (vanVuuren and others, 2011) that was projected to 2050. Precipitation corrections for the projected climate scenarios were processed using the Coupled Model Intercomparison Project climate data processing tool version 2.1 (U.S. Department of Transportation, 2022) using the downscaled Coupled Model Intercomparison Project 5 multimodel ensemble, Localized Constructed Analog dataset from Bureau of Reclamation (2022). The precipitation frequency values represent a total magnitude for a specified duration. Time series of precipitation based on these totals were developed by distributing the precipitation in 5-minute increments over the specified 24-hour duration using the Frequency Storm precipitation distribution feature in HEC–HMS and the incorporated TP40 area reduction curve (Hershfield, 1961) adjustment.

Transformation Method

Precipitation exceeding soil infiltration and storage was transformed into runoff in the developed HEC–HMS model using the Clark (1945) unit-hydrograph method. Two parameters needed to define this unit-hydrograph method are time of concentration (Tc) and a storage coefficient (R) (U.S. Army Corps of Engineers, 2022b). The Tc is the time of travel (in hours) it takes for precipitation runoff to travel from the most distant point in a subbasin to the subbasin outlet. R is a storage coefficient (in hours) used to account for storage within the floodplain such as wetlands, reservoirs, and bridges that can produce flood-wave attenuation. Initial estimates of Tc were calculated using the TR–55 methodology (Soil Conservation Service, 1986) for runoff travel time. The computed Tc estimates were used to compute initial estimates of R using R/(Tc+R)=0.4 corresponding to the approximate mean value for this relation in rural basins in west central Illinois (Graf and others, 1982; Straub and others, 2000), and these parameters were adjusted during model calibration.

Loss Method

The infiltration calculations in HEC–HMS are performed by a loss method for each subbasin (U.S. Army Corps of Engineers, 2022b). The Soil Conservation Service runoff curve number (CN) method was used to simulate infiltration losses from precipitation (Soil Conservation Service, 1986; Natural Resources Conservation Service, 2004). The CN values were applied to each subbasin according to the TR–55 methodology (Soil Conservation Service, 1986), and represent hydrologic soil types, land uses and treatments, and antecedent response conditions. Separate composite CNs at the CNII or “normal” antecedent response condition were determined for each subbasin for the historical land-cover condition, the current land-cover condition, and a planned land-cover condition representing a future condition to about 2050. The historical condition used the Enhanced National Land Cover Data 1992 (Nakagaki and others, 2007), the current condition used the 2019 National Land Cover Data (Dewitz and U.S. Geological Survey, 2021), and the planned (to about 2050) condition incorporated proposed development in the basin, where available, with development in remaining subbasins conservatively estimated as being at a level comparable to land-cover changes between 1992 and 2019. The CN grid development uses soil spatial data from the Soil Survey Geographic Database (Natural Resources Conservation Service, 2021), land-cover information from the corresponding time period (historical, current, planned/future), and a common lookup table of CN values corresponding to soil hydrologic classes and land use (Soil Conservation Service, 1986). The CN grids were developed using QGIS ver. 3.18 (QGIS Development Team, 2021) and procedures provided by Merwade (2021).

Base-Flow Method

The recession base-flow method (Chow and others, 1988) was used to simulate base flow within the basin. For this method, the HEC–HMS model requires three parameters: initial streamflow, a recession constant, and a ratio-to-peak constant (U.S. Army Corps of Engineers, 2022b). The initial streamflow was applied to the subbasin as a ratio of cubic feet per second per square mile. The recession constant represents the rate at which base flow recedes after a rainfall event. The ratio-to-peak constant is a threshold that indicates when to begin base flow on the recession limb of a hydrograph. These parameters were estimated and adjusted during the HEC–HMS model calibration.

Streamflow Routing Method

The Muskingum-Cunge method (Cunge, 1969) of routing was selected within HEC–HMS to accurately represent floodplain storage and peak streamflow attenuation conditions in model reaches incorporating multiple subbasin outflows (U.S. Army Corps of Engineers, 2022b). The Muskingum-Cunge routing method was used for the basin-wide assessment of hydrologic change. Individual HEC–HMS subbasin outflows within the two-dimensional HEC–RAS model footprint were used as inflows in the HEC–RAS model. The exception was the cumulative routed flows for Silver Creek and Little Silver Creek that were output from HEC–HMS and used as inflows to the HEC–RAS SAFB focus area model (fig. 3). The HEC–HMS and HEC–RAS models were run consecutively and iteratively for calibration, validation, and scenario development.

Hydrologic Model Calibration and Validation

The basin-scale hydrologic model (entire area upstream from the Freeburg streamgage) was calibrated and validated to observed streamflow data for six high-flow events between 2015 and 2023. The parameters CN, Tc, R, and base-flow recession were the primary parameters used in calibrating the simulated hydrograph peak and shape. Model parameters were adjusted with a target of matching the simulated hydrograph to the corresponding shape, magnitude, and timing of observed hydrographs at the Troy (fig. 6) and Freeburg (fig. 7) streamgages. Final calibration results indicated similarity in the simulated and observed hydrographs (figs. 6, 7; table 6). The Nash-Sutcliffe efficiency (NSE) coefficient (Nash and Sutcliffe, 1970) and percentage bias (PBIAS; Gupta and others, 1999) statistics were used to assess model fit. Values of NSE can vary from negative infinity (∞) to 1. Values of 1 correspond to a perfect match between simulated and observed time series, whereas values of 0 or less indicate that the mean of the observed values is a better predictor than the simulated values. The NSE values of the four calibration events using the Troy streamgage data ranged from 0.95 to 0.98, and NSE values of the four calibration events for the Freeburg streamgage data ranged from 0.81 to 0.98, indicating a good to excellent model fit (Moriasi and others, 2007; table 6). The PBIAS is a measure of the mean tendency of the simulated data to be larger or smaller than observed values. Values of PBIAS can vary from −∞ to ∞ with an optimum value of 0. The calibrated events at the Troy streamgage yielded PBIAS values of −0.46 to 6.8 percent, whereas those at the Freeburg streamgage were −15.8 to 0.75 percent, indicating a good to excellent model fit. A comparison of relevant hydrograph characteristics also indicates similarities, in general, between simulated and observed peak streamflow and time of peak (table 6).

Two additional independent observed high-water events were used to validate the HEC–HMS calibrated models. The NSE values of the validation events at the Silver Creek at Troy streamgage was 0.73 for each event, whereas those at the Silver Creek at Freeburg streamgage were 0.88 and 0.87 indicating that the simulated values were a fair to good predictor of the independent observed hydrographs. The PBIAS values of the validation runs were −14.7 and −13.4 using Troy streamgage data and −8.5 and 4.2 using Freeburg streamgage data, again indicating a fair to good model fit. The simulated and observed peak streamflows and timing were variable but acceptable for the purposes of assessing future/planned land cover and climate effects (table 6).

Hydrologic Model Sensitivity Analyses

The primary hydrologic model parameters that were modified for calibration were Tc, R, and CN (infiltration), and the primary input data consisted of precipitation values. Model sensitivity was analyzed to determine the relative effect of each of these parameters or input datasets on model results. These analyses were conducted by modifying each of these parameters or datasets by +5 percent and −5 percent and determining the resulting change in peak streamflow and cumulative hydrograph volume at the outlet of a selected subbasin (subbasin 1; fig. 2). All sensitivity analyses used a CNII, 10-year, 24-hour duration runoff event. Results of the 5-percent change in sensitivity values indicated that the generated peak streamflow from the HEC–HMS model was most sensitive to changes in the CN parameter (−13.2- and 16.1-percent change), then to precipitation (−7.20- and 9.08-percent change), with the 5-percent changes in all other factors affecting peak streamflow by less than or equal to an absolute 2.72 percent (table 7). The timing of the generated peak streamflows were most sensitive to Tc (−40 and 45 minutes), with all other parameters affecting the peak timing by 10 minutes or less (table 7). The response of cumulative volume to −5 and +5 percent changes in model parameters or precipitation were similar to that of peak streamflow in that the model was most sensitive to CN (−13.1 and 16.0 percent, respectively) and precipitation (−7.18 and 9.04 percent, respectively). The 5-percent changes in the remaining parameters resulted in absolute changes of 0.16 percent or less in cumulative volumes.

Topographic and Bathymetric Data

The elevation data used in the HEC–RAS model were obtained from a 1-meter digital elevation model (DEM) derived from lidar data published in May 2020 (Joseph Weber, Scott Air Force Base, written commun, 2021). The data were acquired and processed by Woolpert, Inc. As per USGS quality level 2 standards (version 1.2; Heidemann, 2018), the lidar data had a nonvegetated vertical accuracy of less than 10 cm (actual was 1.7 cm) root mean square error. The positional horizontal accuracy was 10 cm at a 95-percent confidence level. The accuracy specifications for all lidar datasets met or exceeded the U.S. National Map Accuracy Standards for vertical and horizontal accuracy guidelines for 2-ft contours (American Society for Photogrammetry and Remote Sensing, 2023).

Because the lidar data cannot provide ground elevations below water surfaces, channel cross sections for the model stream reaches were surveyed by a USGS field crew in February 2021. Water depths at the cross-section locations were measured by taping down from bridge crossings and kayaks. Water-surface profiles were interpolated by taping down from known bridge elevations on bridges throughout the stream reaches. The measured water depths were subtracted from the interpolated water surfaces at cross-section locations to collect channel topographic elevation points. The bathymetric data were incorporated into the HEC–RAS model, and the RAS Mapper utility in HEC–RAS was used to generate a DEM of channel bathymetry and to merge the bathymetry DEM with the terrain DEM. Bridge and culvert structures were also surveyed by a USGS field crew in January 2021. A level IV survey (Rydlund and Densmore, 2012) procedure was used in acquiring reference elevations with a resulting vertical accuracy of reference points within 0.32 ft.

Structures

A total of 34 hydraulic structures were represented in the hydraulic model of Silver Creek and its tributaries. These structures have the potential to affect water-surface elevations during flooding along the stream. The 34 structures included thirteen 2-lane or railroad bridges and 21 culverts. All bridge-geometry information is available in the hydraulic models used for analyses and provided in Cigrand and Heimann (2024).

Energy-Loss Factors

Hydraulic analyses require the estimation of energy losses that result from frictional resistance exerted by a channel on the flow. These energy losses are quantified by the Manning’s roughness coefficient (n-value; Chow, 1959). Initial (precalibration) n-values were selected based on field observations and high-resolution aerial photographs that were captured with the May 2020 lidar data (Joseph Weber, Scott Air Force Base, written commun., 2022), and by using tabulated (Chow, 1959) and photographic estimates of n-values (Barnes, 1967).

As part of the calibration process, the initial n-values were varied by flow and adjusted until the differences between simulated and observed water-surface elevations at the continuous water-level monitoring locations (fig. 3) along the study reach from the high-flow calibration events were minimized. The final n-values ranged from 0.03 to 0.046 for the main channel and from 0.035 to 0.12 for the overbank areas simulated in this analysis (Cigrand and Heimann, 2024). The lowest channel coefficients were placed in straight, downstream sections of the model reach, and the highest were placed in coarse-material reaches with vegetated banks and substantial coarse-woody debris. The lowest roughness coefficients on the floodplain were placed in barren lands, and the highest roughness coefficients were placed in densely forested and high-intensity developed areas.

Boundary Conditions and Detention Storage

Hydraulic model input consisted of boundary conditions of inflows obtained from the HEC–HMS model and defined outflow boundary conditions of the estimated normal depth slopes. The input hydrographs for the HEC–RAS model scenarios were generated using a range of precipitation frequency-duration events in HEC–HMS, as detailed in the “Hydrologic Data” section. Subcritical (tranquil) flow regime was assumed for the simulations. Normal depth, based on an estimated average water-surface slope of 0.00015, was used as the downstream boundary condition for Silver Creek and a slope of 0.0007 was used for Ash Creek (Cigrand and Heimann, 2024).

The HEC–RAS model was used to simulate the effects of detention storage and projected climate conditions on water-surface profiles and inundation extents within the model domain. Detention structures were incorporated into the model mesh in the headwaters of Ash Creek and Cardinal Creek and within the South Ditch drainage (fig. 3). Detention structures were placed in areas of the lidar with narrow valleys or “pinch-point” locations that were suitable for substantial volumes of water storage with the placement of a dam. No pinch-point location is suitable for the South Ditch detention, so the existing 3-ft (in diameter) overflow culverts that flow into a field were increased in size to 4 ft, and a berm was incorporated to the field to allow detention storage (table 5). The intent of incorporating detention structures was to determine the generalized potential effectiveness of such structures on flood mitigation rather than provide specific design recommendations. The detention structures incorporated in HEC–RAS were, therefore, designed to optimize detention for a mid-ranged recurrence interval precipitation event of 25 years. The optimization for this sized event was achieved iteratively through culvert sizing and dam height to maintain the inundation area within the constraints of the available open area. The hydrologic response condition for the HEC–RAS scenarios, as determined through the CN values used in HEC–HMS, was selected to correspond to the median value of the three HEC–RAS calibration events. This conservatively high antecedent response condition ensured that the events simulated were of potential consequence and likely to result in flooding.

Hydraulic Model Calibration and Validation

The HEC–RAS model, in conjunction with the HEC–HMS model previously described in the “Hydrologic Model Calibration and Validation” section, was calibrated to two observed high-flow events of July 26–28, 2022, and March 23–31, 2023 (fig. 8; table 8). Following calibration, the model was used to simulate peak conditions for an additional validation event on August 12–15, 2020, to confirm model performance (table 9). Model calibration was completed by adjusting n-values until the results of the hydraulic computations closely agreed with the observed water-surface elevations for given flows at the HWM and pressure transducer locations (fig. 3). The differences between observed peak and simulated water-surface elevations of the two high-flow events at the USGS PT locations ranged from −0.97 to 0.55 ft (mean of about −0.04 ft; table 8). The differences between the observed HWM elevations and simulated water-surface elevations for the validation event ranged from −0.27 to 0.30 ft (mean of about 0.04 ft; table 9). The results demonstrate that the model can accurately simulate water levels throughout the reach.

Hydraulic Model Sensitivity Analyses

The sensitivity of the hydraulic model was assessed using changes in Manning’s n-values and change in detention basin culvert diameter sizing. The Manning’s n-values were modified by +5 and −5 percent, and the culvert diameters used in the detention scenario were modified by +0.5 and −0.5 ft. Sensitivity analyses were used to provide an example of the effects of modification for a basin containing added detention using results from the Ash Creek Basin at Scott Drive (fig. 1). The analyses were run using a 10-year RI, 24-hour duration precipitation and the resulting effects on peak streamflow (determined as the percent difference in peak streamflow) and difference between the maximum water-surface elevation from the existing scenario maximum water-surface elevation were quantified (table 10).

Results indicate that simulated peak streamflows were highly sensitive to the incremental change in culvert size (−12.03- to 29.06-percent change in peak streamflows) and less sensitive to the incremental change in Manning’s n-values (−0.30 to 0.15 percent). The incremental change in culvert size resulted in changes of −0.13 to 0.17 ft in water-surface elevation at the reference location, whereas the small incremental change in Manning’s n still resulted in changes of −0.05 to 0.09 ft.

Ash Creek

The simulated water-surface elevation profiles for the Ash Creek Basin indicated that the simulated detention storage had a greater effect on water levels than did the effects of projected changes in precipitation for all scenarios (figs. 17–20). That is, the differences in water levels between the with detention and without detention scenarios were greater (less than 1 ft to greater than 2 ft) than those of the current and projected precipitation conditions (less than 1 ft). Detention, therefore, could be an effective mitigation practice to reduce peak streamflow conditions during flooding under current and projected climate conditions. The greatest reduction of water-surface elevations as a result of simulated detention occurred at the upstream end of the profile (upstream from Seibert Road; fig. 1) and between stationing of about 2,500 and 10,500 ft (figs. 17–20; included crossings with Enlisted Drive, West Winters Street, Ward Drive, and Scott Drive). The beneficial effects of added detention storage were minimal downstream from station 11,000 ft where Ash Creek flows into a low-lying broad floodplain. However, beneficial results were farther downstream at the Highway 161 crossing (station 12,600). The Highway 161 crossing is overtopped with the 25-year RI condition without added detention (fig. 18) but is not overtopped until the 50-year RI for the with detention scenario (fig. 19).

Ash Creek is channelized through much of the SAFB reaches; therefore, flows are contained within the stream in the 2-, 5-, and 10-year RI scenarios with detention (figs. 21–23). Ash Creek transitions to a low-laying broad floodplain immediately south of SAFB and north of Highway 161. Immediately east of the Ash Creek box culverts on Highway 161 is a low-lying section of road that was not within the extent of inundation until the 25-year RI for the with detention condition (fig. 24) but was overtopped beginning with the 5-year RI condition without the added detention (fig. 22). Some inundation of a residential area along Greenfield Circle was evident in the 5-year event for the without detention scenarios (refer to inset 1 in fig. 22). The inundation in the residential area along Greenfield Circle was more extensive in the 10- and 25-year RI results for the without detention scenarios encroaching along Ash Creek Drive and side streets (refer to inset 1 in figs. 23–24). The first indications of flooding in the residential area adjacent to Ash Creek in the detention scenarios were evident in the 50-year RI with-detention scenario with future precipitation (fig. 25), and the first indications of flooding in the residential area from the with detention scenario with current precipitation is in the 100-year RI event (fig. 26). Flooding magnitude under all scenarios increased in the residential area in the 500-year RI event (fig. 27).

South Ditch

The water-surface profiles developed for South Ditch indicated that the effects of an added detention structure in this basin generally exceeded the anticipated effects of projected changes in precipitation. Differences in water-surface profiles for with detention and without detention scenarios ranged from less than 1 ft to about 2 ft, whereas differences in the profiles resulting from projected precipitation were less than 1 ft (figs. 28–31). The differences resulting from added detention and projected precipitation varied longitudinally along the profile and were greatest upstream from South Ditch tunnel 2 and were greater, in general, upstream from South Ditch tunnel 1 than downstream from this tunnel. Effects of Silver Creek on the downstream end of the water-surface profiles were evident as the water-surface slopes flattened and the effects of detention were no longer evident. The effects of projected precipitation were greater in this area than the effects of detention because of the effect of Silver Creek. The unnamed access road crossing to the landfill was overtopped for events of 10-year RI and greater for the without detention scenarios (fig. 29) and for all scenarios for events of 25-year RI or greater (figs. 29–31). South Ditch tunnel 2 was overtopped for without detention scenarios of RIs of 50 years or greater but only the 500-year RI profile for the scenarios with detention indicated overtopping of tunnel 2 (figs. 30, 31). Of the seven total structure crossings in the profile extent, only tunnel 2 and the unnamed landfill access road crossing were overtopped in the 500-year RI profile (fig. 31).

South Ditch also is channelized through much of the SAFB reaches; therefore, flows were contained within the stream in the 2-, 5-, and 10-year RI scenarios with detention (figs. 21–23). Flood inundation began along South Ditch in the 25-year RI without detention scenarios west of the runways and east of Scott Drive (fig. 24). Flooding is extensive in the residential area adjacent to Ash Creek and along South Ditch and Hangar Road (refer to inset 2 in figs. 21–27) in the 50-year RI event for without detention scenarios (fig. 25). All scenarios for the 100-year RI event indicated that South Ditch and Hangar Road were within the inundation extent, but the inundation extent was substantially smaller for the scenarios with detention and many of the structures near Hangar Road were not affected by flooding (refer to inset 2 in fig. 26). Flooding was extensive across the southwest part of SAFB from Ash Creek and South Ditch in all 500-year RI scenarios (fig. 27). These results demonstrated the substantial protection from flooding afforded by the added detention in addition to the increase of the overflow culvert size (refer to table 5).

Cardinal Creek

Similar to the results from Ash Creek and South Ditch water-surface elevation profile comparisons, the results from the Cardinal Creek profile comparisons also indicated that the differences resulting from detention generally exceeded the differences resulting from projected precipitation (figs. 32–35). The differences in the water-surface profiles in the with and without detention scenarios ranged from less than 1 ft to about 3 ft, whereas the differences as a result of projected precipitation were consistently less than 1 ft. The effects of added detention were greatest in the upper sections of the profiles (station 0–3,000 ft in the 2- and 5-year RI profiles) and were well sustained through much of the longitudinal profile to Pryor Drive (figs. 32, 33). The reduction in the effects of detention and relatively minor slope downstream from Pryor Drive in all RI results were attributable to the effect of inundation from Silver Creek. Scott Lake Road crossing at the downstream end of the profile was inundated in all scenarios beginning with the 5-year RI event (figs. 32–35), which, again, can be attributed to backwater from Silver Creek rather than to headwater flooding from Cardinal Creek. The Pryor Drive crossing was overtopped for the without detention scenarios beginning with the 25-year RI event (figs. 33–35). This crossing was overtopped at the 100-year RI event for the projected precipitation scenario with detention (fig. 34), and for the 500-year RI event under all scenario conditions (fig. 35). Radar Road was overtopped for scenarios without detention beginning with the 50-year RI event and for scenarios with detention only for the 500-year RI event. Results indicate that added detention storage in the Cardinal Creek Basin, optimized for 25-year RI event flows, was still effective at preventing road overtopping for 2 of 3 road crossings for all but the 500-year RI events under existing condition precipitation.

Cardinal Creek also is channelized upon entering SAFB, and flows were essentially contained within the channel in all the 2-year, 5-year, and 10-year RI events (figs. 36–38). Flooding of roads was evident beginning with the 25-year RI event without detention scenarios (fig. 39), and Pryor Drive and surrounding areas were within the inundation extent for the without detention scenarios. The upstream Radar Road intersection was within the inundation extent of the 50-year RI event along with much of Air Guard Way that parallels Gunn Street to the south (fig. 40). Flooding was evident from the detention scenarios beginning with the 100-year RI event (fig. 41) and encroachment of Silver Creek is evident, which exacerbates flooding through the 500-year RI event (fig. 42), covering several blocks on either side of the Pryor Drive and Gunn Street south along Golf Course Road and East Drive.

Silver Creek and East Tributaries

Detention storage was not added in the Silver Creek, Little Silver Creek, and unnamed tributaries 1 and 2 Basins; therefore, differences as result of projected precipitation were greater than differences resulting from added detention (figs. 1.1–1.16). The minor streamflow contributions of Cardinal Creek and South Ditch (two tributaries with added detention) to Silver Creek were not large enough to evoke a response in the Silver Creek water-surface elevation profiles. As a result of projected changes in precipitation, differences in water-surface elevation profiles in Silver Creek, Little Silver, and unnamed tributaries 1 and 2 generally were less than 1 ft except for the 500-year RI profile for Silver Creek in which the differences approached 1 ft (fig. 1.4).

The inundation extent of Silver Creek and eastern tributaries of Little Silver Creek, and unnamed tributaries 1 and 2 covers the extensive bottomland area adjacent to Silver Creek beginning with the 2-year event scenarios (figs. 43–49). The downstream reaches of the eastern tributaries included in the mapping extent are within the influence of Silver Creek backwater effects. The direct effects of Silver Creek flooding of infrastructure on SAFB was primarily associated with the lower Cardinal Creek Basin. It is likely that in local tributaries, including Cardinal Creek, initial flooding associated with headwater sources in the tributaries is followed hours later by flood effects resulting from the Silver Creek main stem. For example, during the July 27, 2022, calibration event, observed water levels on Cardinal Creek, Ash Creek, and South Ditch peaked about 8–9 hours before the water level of Silver Creek at Highway 161 (fig. 8). Flooding of infrastructure by Silver Creek in the lower reaches of the tributaries primarily was associated with the 100-year and 500-year RI events (figs. 48–49).

Uncertainties and Limitations Regarding Use of Hydrologic and Hydraulic Model Results

Although the generated flood profiles and flood-inundation maps represent the water elevations and boundaries of inundated areas with a distinct line, some uncertainty is associated with these maps. The flood boundaries shown were estimated based on water stages at a selected reference location based on the hydrologic response from selected precipitation magnitudes and 24-hour duration. Water-surface elevations along the stream reach were estimated by unsteady-state hydraulic modeling, assumed unobstructed flow, and used streamflows and hydrologic conditions anticipated at the Silver Creek main stem and tributaries. The hydraulic model reflects the land-cover characteristics and any bridge, culvert, levee, or other hydraulic structures existing as of July 2022. Unique meteorological factors (timing and distribution of precipitation) may cause actual streamflows along the modeled reach to vary from those assumed during a flood, which may lead to deviations in the water-surface elevations and inundation boundaries shown. Additional areas may be flooded because of unanticipated conditions such as changes in the streambed elevation or roughness, backwater into major tributaries along a river main stem, or backwater from localized debris or ice jams. The accuracy of the water-elevation profiles and floodwater extent portrayed on these maps also will vary with the accuracy of the DEM used to simulate the land surface.

Additional uncertainties may be inherent or factored into the simulation of flood peaks generated from rainfall-runoff simulations. A hydrologic model was used to simulate flood peaks associated with various probabilistic precipitation amounts. The precipitation was assumed to follow a defined temporal distribution over the duration of the event and an even spatial distribution over the basin. The actual temporal and spatial distribution of precipitation may vary, thereby affecting the timing and magnitude of the predicted flood peak in the main stem or tributary. A single “simple” storm was simulated with a near-base-flow starting condition. For multiple compounding precipitation events, the starting streamflow condition may be considerably higher than base flow and, therefore, the peak condition also will be higher than simulated conditions. Other sources of uncertainty will arise from the selection of the appropriate antecedent runoff condition and the occurrence of atypical precipitation events including rainfall on frozen ground or on a substantial existing snowpack, all of which may affect the timing and magnitude of the predicted flood peak. Although the projected precipitation adjustments for 2050 are based on results from 32 climate models (U.S. Department of Transportation, 2022), these estimates also include uncertainty. The simplified representation of temporal changes in land cover through modification of the HEC–HMS CN loss parameter alone provides uncertainty to the full effect of the changes in the timing of peak streamflows because timing is relatively insensitive to changes in this parameter.