Attachment_9.1_Beale_AF_Basin-Wide_Comprehensive_Analysis.pdf

Beale Air Force Base

COMPREHENSIVE HYDROLOGY STUDY

U.S. ARMY CORP OF ENGINEERS SACRAMENTO DISTRICT

Executive Summary

The purpose of this hydrologic investigation was to provide inflow hydrographs at each of the lakes and ponds and locations along the cantonment to evaluate the potential flood risk to Beale Air Force Base (AFB). This was accomplished by developing a rainfall-runoff model and inputting annual chance exceedance (ACE) design storms for the 50-, 20-, 10-, 4-, 2-, 1-, 0.5-, 0.2-, and 0.1% events. The hydrographs produced from the model were routed through a 1 and 2-dimensional hydraulic model of Beale Air Force Base (AFB) to delineate potential flood prone areas (USACE, 2018). The hydraulic analysis is described in the report “Comprehensive Study for Beale Air Force Base Hydraulic Analysis”.

Additionally, Probable Maximum Precipitation (PMP) and Probable Maximum Floods (PMF) were also developed at select impoundment locations with classified hazard ratings and routed through a hydraulic model. The hydrology provided in this study will assist in understanding the hydrologic hazard of the dam following guidance documents FEMA P-94 “Selecting and Accommodating Inflow Design Floods for Dams” and Engineering Regulation (ER) 1110-8-2 (FR) “Inflow Design Floods for Dams and Reservoirs”.

The model was calibrated to the January 2017 storm and validated to the February storm at three USGS discharge stations. Results showed good agreement between the simulated and observed results with a Nash-Sutcliffe value greater than 0.75 and 0.65, respectively. Equations developed from a regression analysis were performed to estimate time of concentration and storage coefficient values for basins outside the calibration area. The inverse relationship between longest flow path and the square root of river slope reasonably predicted time of concentration values with a correlation of determination of

0.76. Time of concentration had the best correlation with predicting storage coefficient with a correlation of determination of 0.6.

Two storms types were created: the first storm type produced 1%, 0.1%, and Probable Maximum Flood inflow hydrographs at each lake with a hazard classification. General and local storm probable maximum precipitation was developed using the MMC Precip Tool v 1.2 (MMC, 2017). The hydrographs produced from this design storm were used to evaluate the hydrologic hazard of the dam. The routings were performed in a separate HEC-RAS model analysis (USACE, 2018). The second type of design storm represents a storm centered above the junction of Hutchinson Creek and its tributary using 50- to 0.1% ACE NOAA Atlas 14 precipitation-frequency depths. This design storm produced hydrographs that were used to evaluate the flood risk downstream at the cantonment. The hydrograph routings were also performed in the HEC-RAS model analysis (USACE, 2018)

Simulated hydrographs peak values were compared to previous study results. Peak flow values were within 16% for drainage areas greater than 4.8 mi2+

, and within 54% for drainage areas less than 3.1 mi2.

The major dissimilarities were due to the differences in parameter estimates from each model. Greater confidence was placed in the model parameters developed in this study given the good agreement in the calibration effort.

Table of Contents Executive Summary

Introduction

General Watershed Description

Climate and Flooding

Hydrologic Modeling

Basin Delineation

HEC-HMS Model Methods

Model Calibration

Storm Event and Meteorological Model Setup

Calibration and Results

Design Storm

Design Storm 1 and Design Storm 2

Rainfall Depths and Zones Computation

Probable Maximum Precipitation

Results

Comparison to Previous Studies

Additional Limitations & Uncertainty

Conclusion

References:

Appendix I:

Introduction Beale Air Force Base (AFB) reached out to U.S Army Corp of Engineers, Sacramento District to evaluate the hydrologic hazard for their dams and potential flood risk to their AFB. Large rainstorms have historically caused flooding along the cantonment, bridges, and roads. Of even greater concern, inflows to lakes have caused dam overtopping. In water year (WY) 2017, a series of rainstorms caused Upper Blackwelder Lake to fill beyond its storage capacity and required sandbagging to prevent further overtopping. On November 28 to December 2, 2012, heavy rainfall created intense flooding along the cantonment causing several million dollars of damage to infrastructure and equipment (Beale AFB, 2016). The Miller Dam report (Mraz, 1996) describes flow overtopping Miller Dam in WY 1995 and 1996.

Although the AFB has impoundments for the purposes of storing water, many of the dams are in poor condition and have not been maintained since their development.

In 2016, 21 dams on the Beale AFB property were inspected and given a hazard classification rating.

These classifications are described in the document “Selecting and Accommodating Inflow Design Foods for Dams” (FEMA, 2013). Table 1 describes the hazard classification for each dam (not all dams are FEMA reportable). Figure 1 shows the locations of FEMA reportable dams. Dams classified as “High” hazards are if dam failures will cause loss of human life; dams classified as “Significant” hazards are if dams failures will not cause loss of human life but will cause significant economic loss and environmental damages; and dams classified as “Low” hazards are if dam failures results in no loss of human life or low economic and environmental losses. Dams with a High hazard classification must pass the Probable Maximum Flood (PMF); significant hazard classifications must pass the 0.1% ACE event;

and low hazard classifications must pass the 1% ACE.

A basin wide hydrology was developed in 1993 (USACE, 1993) and created 1% year and 0.2% year peak flows and flood plains along Reeds Creek, Hutchinson Creek and its tributaries, Dry Creek, and Best Slough. Although it was not specified, the HEC-1 model was likely used to estimate 1% and 0.2% peak flows. Basin specific hydrology were also developed for Miller Lake (Mraz, 1996) and Frisky and Bedsprings Lake (USACE, 2016). The 1996 Miller Dam analysis estimated 10% ACE to 0.1% ACE peak flows using a HEC-1 rainfall-runoff model to assess the hydrologic hazard to Miller Lake. This study concluded Miller Dam was capable of passing only moderate level flows and recommended installing or modifying the spillway to pass the 0.1% ACE. The last study performed on Beale AFB was in 2016 for Frisky and Bedsprings Lake. An HEC-HMS rainfall-runoff model was developed to produce 50% to 0.2% ACE hydrographs. The goal of the 2016 study was to determine whether preemptively breaching Frisky dam would increase the downstream flood threat to the air force base during a storm event. Results suggested a breach at Frisky Dam would not increase the flood risk. None of the models produced from the previous studies had observed flows for calibration. The hydrology produced from this study improves upon the previous analysis by utilizing a calibrated basin-wide HEC-HMS model to compute ACE% hydrographs at all lakes and specified index points.

The hydrographs developed in this study represents an existing conditions hydrology, and assumes the AFB will not make significant changes to its land use such that a separate future conditions model is necessary. The following task summarizes the development of the existing conditions hydrology:

develop rainfall-runoff model in HEC-HMS and calibrate model to WY 2017 event; develop design storms using NOAA Atlas 14 depths; develop PMP depths using HMR-59 guidance; and simulate storms in calibrated model to obtain annual chance exceedance hydrographs at lakes and index locations.

Table 1: Hydrologic Hazard Classif ication

Name Type of

Dam

USACE Rating

1FEMA

"Reportable" Problems At risk

Downstream

Draft Hazard Classification

L / S / H B Street Pond Earth & Rock FAIR to GOOD Yes Goose Lake Low to

Significant

Beale Lake PCC POOR Yes OT in 2016 Farm land / Flood Plane Low

Bedsprings Lake Earth & Rock FAIL Yes OT in 1998

Small Arms Range. East side

Main Base.

Ditched around.

Low

Broskey Lake Earth & Rock FAIL Yes

Right Abutment OT, early 2000s or before

Farm land / Flood Plane Low

Frisky Lake Earth & Rock FAIL Yes Right

Abutment OT in 2010 +/-

Main Base, Small Arms Range High

Goose Lake Earth & Rock POOR Yes

Right Abutment OT in 2000s and

Main Base, then area between

Main Base and Flight Line. Sewer Treatment Plant

Low to Significant

Hospital Pond Earth & Rock POOR Yes Flood Plane Low

Lower Blackwelder

Lake Earth & Rock POOR to FAIR Yes

Area between Main Base and

Flight Line. Sewer Treatment Plant

Significant

Mad Dog Lake Earth & Rock POOR to FAIR Yes Flood Plane, and Sewer Treatment

Plant Low

Miller Lake Earth & Rock FAILED to POOR Yes Flight Line Low to

Significant Parks Lake No Dam N/A No In Flood Plain

PAVEPAWS

Lake Earth & Rock FAIL Yes OT in 2016 Golf Course Low

Pond #1 Earth & Rock N/A No Sewage Treatment Pond

Pond #2 Earth & Rock N/A No Sewage

Pond #3 Earth & Rock N/A No Sewage

Pond #4 Earth & Rock N/A No Sewage

Shingle Lake No Dam N/A No By Golf Course

Small Arms Range Lake Earth & Rock POOR to FAIR Yes

Small Arms Range, Golf

Course

Low to Significant

Upper Bedsprings Earth & Rock FAILED to

POOR NO OT 2016

Bed Springs Dam then Small Arms

Range.

Low

Upper Blackwelder

Lake Earth & Rock POOR Yes

Right Abutment OT

Lower Blackwelder then area between Main Base and

Flight Line. Sewer Treatment Plant

Significant

Vassar Lake Earth & Rock FAIR to GOOD Yes Camp Beale

HW/Vassar Gate, Farm land Flood

Plane

Low to Significant

1FEMA Reportable impoundments that meet reservoir specifications and meet the hydrologic hazard criteria

Figure 1: Map of FEMA Reportable impoundments at Beale AFB. Impoundments with * by its name are non-reportable

General Watershed Description

The AFB is located in Yuba County between the Bear and Yuba River, approximately 10 miles east of Marysville. The basin is mostly unpopulated with few roads and infrastructure outside the AFB areas.

The AFB boundary lines are contained within three watershed basins: Reeds Creek, Hutchinson Creek, and Dry Creek (Figure 2). Reeds Creek lies north of the base and impounded by Miller Dam. The cantonment lies within the Hutchison Creek basin and impounded by Goose Lake and Upper and Lower Blackwelder. An unnamed creek just east of Hutchinson Creek also flows through the cantonment and impounded by 5 reservoirs: Upper Bedsprings, Bedsprings, Frisky, Pave Paws, Small Arms Range Lake, and Mad Dog. The unnamed creeks joins with Hutchinson Creek south of 6th Street and South Erie Street. Dry Creek is located south of the cantonment and has the greatest contributing area and flow potential. Three lakes exist on the Dry Creek watershed: Beale, Broskey, and Vassar Lake. Dry Creek splits off into Best Slough 0.5 miles downstream of Beale Lake. Figure 3 shows the lakes and streams within Beale AFB

The hydrologic soil type for Reeds Creek and Hutchinson Creek contain mostly D type soils (USDA, assessed 2016) with some C type converging along the creek paths. Type D soils are described as having high runoff potential when saturated with restricted water movement through soils, while Type C soils have moderately high runoff potential with a moderately restricted transmissivity when saturated (USDA, 2007). Dry Creek basin also mostly contains Type D soils and an equal mixture of Type C and B soils. Figure 4 shows the soil classification for Beale AFB.

The National Land Cover Database (Homer et al., 2015) shows most of the land cover along Reeds Creek and Hutchinson Creek basins as pasture and herbaceous plants where cattle annually graze along the foothills. Dry Creek has a more diverse land use with higher coverage of evergreen and deciduous forest in the upper foothills. Figure 5 shows the land use for Beale AFB.

Figure 2: Watershed Basins and Beale Air Force Basin

Figure 3: Watershed of Beale AFB for Main Base. Flowlines obtained from NHDPlus Version2 (EPA, assessed 2017)

Figure 4: Beale Air Force Base Hydrologic Soil Classif ication obtained from SSURGO (Soil Survey Staff accessed 2017)

Figure 5: Beale Air Force Base land use representation using NLCD 2011 (Dewitz et al., 2015)

Climate and Flooding

The climate along Yuba and Nevada county region consists of cool, wet winters, and dry hot summers.

Rainfall season typically occurs from October through April and mean annual precipitation is estimated to be 25 inches (PRISM, accessed 2017).

Large storms that cause the vast majority of flooding in northern California are due to atmospheric rivers. Atmospheric rivers, as shown in Figure 6, are narrow bands in the atmosphere with high concentration of water vapor (NOAA, accessed 2017).

Figure 6: Representation of Atmospheric River, https://scripps.ucsd.edu/news/3066 (Scripps Institution of Oceanography, accessed 2017a)

Hydrologic Modeling The rainfall-runoff modeling used in this study was the Hydrologic Engineering Center’s (HEC) Hydrologic Modeling Systems (HMS) version 4.2. HMS was designed to simulate precipitation-runoff processes for a wide range of hydrology application, including the development of flood hydrographs (HEC, 2016). GIS spatial data was utilized to estimate basin parameters, such as soil type and land use. The spatial datasets were obtained from the Natural Resources Conservation Service (NRCS) and U.S. Geological Survey (USGS). Three discharge stations were available from the USGS to calibrate the model to January 2017 event.

The model time step was selected as 15 minutes. This was to ensure the smallest subbasin had at least 3-4 ordinates on the rising limb of the hydrograph (USACE, 1994). For the calibration model, most of the small lakes and ponds were represented as reservoirs in the watershed model. Reservoir storage-elevation-discharge relationships and reservoir geometry were obtained from 1-meter LIDAR dataset and the hydraulic routing model, HEC-RAS.

https://scripps.ucsd.edu/news/3066

Basin Delineation

The watershed was delineated using ArcHydro for ArcGIS 10.3 and a 10-meter DEM from USGS (USGS, accessed 2016a). The stream paths were “burned” into the DEM using the National Hydrography Dataset (NHD) Plus version 2 flowlines (EPA, accessed 2017). The main watershed pour point was delineated along Highway 70 near Plumas Lake. This points captures flow from the Reeds Creek, Hutchinson Creek, and Dry Creek watersheds. Though the delineation captures the entire study area, not all of the delineated area is of interest for this study. Hydrographs produced from subbasins located west and south of the cantonment were not critical for assessing the flood risk to the cantonment or dams. Additional subbasin delineations were requested at bridges, culverts, gage stations, and inflows to lakes and ponds for the HEC-RAS calibration/validation efforts. Pertinent basin parameters such as the longest flow path, basin slope, and channel slope were also obtained using ArcHydro and GeoHMS version 10.3. The basin delineations and attributes were compared to the Watershed Boundary Dataset (WBD) HUC-12 (NRCS-USDA, accessed 2017).

There are known delineation challenges ArcHydro faces along flat topography with little to no variation in elevation. The ArcHydro software is also not able to easily recognize flow splits without additional modifications. However, most of the inaccurate subbasins delineated by ArcHydro were not critical locations to understanding and quantifying the flood risk.

HEC-HMS Model Methods

The following hydrologic modeling methods were specified in HEC-HMS: Clark Unit Hydrograph for unit hydrograph transform; Muskingum-Cunge for hydrologic routing; and Initial and Constant Loss for loss rates. No baseflow was considered. The HMS model schematic is shown in Figure 7 to Figure 11 and highlights the relevant locations for obtaining design hydrographs.

Figure 7: HEC-HMS Schematic of Beale Air Force Base Watersheds

Figure 8: HEC-HMS Schematic of Reeds Creek Watershed shown in Inset 1

Figure 9: HEC-HMS Schematic of the Upper Cantonment Watershed shown in Inset 2

Figure 10: HEC-HMS Schematic of the Lower Cantonment Watershed shown in Inset 3

Figure 11: HEC-HMS Schematic of the Dry Creek Watershed shown in Inset 4

Initial and Constant method is a simple soil loss accounting measure. The initial loss specifies the amount of rainfall stored prior to runoff, and constant loss determines the infiltration rate after the initial loss is satisfied (HEC, 2016). The initial loss does not have a recovery component and the constant loss infiltration rate is applied through the entire simulation period. The average percent impervious for each subbasin were computed from the National Land Cover Database (NLCD) Percent Impervious (Xian, 2011).

Clark Unit Hydrograph (UH) is a synthetic unit hydrograph method that estimates overland and channel runoff, as well as an estimate of delay due to storage effects (Viessman & Lewis, 2003). A time versus area curve translates precipitation to ahydrograph and routed through a linear reservoir to account for storage attenuation (HEC, 2016). Two main parameters the Clark UH method are time of concentration and storage coefficient. Initial time of concentration (𝑇𝑇𝑐𝑐) values were computed using the lag equation relationship with time of concentration. The lag value is transformed to Tc by a factor of 0.6. The follow equation describes the lag equation and transform.

𝐿𝐿𝑔𝑔 = 1560 ∗ 𝑛𝑛 ∗ [𝐿𝐿𝐿𝐿𝑐𝑐

𝑆𝑆0.5]0.33 (1)

Where 𝐿𝐿𝑔𝑔=Lag Time (min), 𝐿𝐿=Length of longest watercourse (mi), 𝐿𝐿𝑐𝑐=centroid flow path (mi), S = overall slope of longest flow path (ft/mile), and n = Basin roughness. Time of concentration and lag relationship (reference) is computed as:

𝐿𝐿 = 0.6𝑇𝑇𝑐𝑐 (2)

Where 𝑇𝑇𝑐𝑐=time of concentration. These computations were initial estimates and were adjusted during calibration.

The storage coefficient, R, was computed using the following relationship:

𝐾𝐾 = 𝑅𝑅

𝑅𝑅+𝑇𝑇𝑐𝑐

(3)

Where K = Constant; R=Storage coefficient. K was computed using the Tc and R values calibrated for the Bear River watershed, shown in Table 10 of the Bear River Hydrology Report (USACE, 2003), where Beale AFB is located in the northwest section of the Bear River watershed. The K value was estimated to 0.63.

Again, these values for storage coefficient are initial values and were adjusted during calibration.

Muskingum-Cunge was selected as the routing method. Representative cross sections were developed in ArcGIS and also obtained from the 1-D HEC-RAS model. Stream length and slope were obtained during the delineation process. Manning’s roughness values for the right, center, and left channels were adopted from the HEC-RAS model. No adjustments were made to these values.

Initial loss rates were obtained from Table 5-1 Chapter 5 of the Sacramento County Drainage Manual (Sacramento County, 1996), while initial constant loss rates were obtained from NRCS SSURGO database (Soil Survey Staff, accessed 2017). Initial losses, however, were shown to have little to no effect on the results. The SSURGO database contains information about soils for most of the United States. The dataset of interest were the hydrologic soil groups. Soil Groups are sectioned into A, B, C, and D type soils.

Skaggs and Khaleel, 1982 estimated infiltration ranges for each soil group. Most of Reeds Creek and Hutchinson Creek contains mostly D and some C type soils. Dry Creek basin has D, C, and B type soils.

Initial estimates for the constant loss rate were based on the average soil group type and infiltration rates provided by Skaggs and Khaleel, 1982. Constant losses were adjusted during calibration.

Model Calibration The USGS placed three stream gages within Hutchinson Creek watershed during the beginning of water year 2017. One station was placed along Hutchinson Creek downstream of Goose Lake, and the other two stations along an unnamed tributary to Hutchinson Creek. Stations recorded in 15-minute time steps, and flows that ranged between 0-8 cfs were not consistently recorded by the station. Table 2 shows the station name, ID, location, and datum. These stations were used to calibrate unit hydrograph and soil loss parameters.

Table 2: USGS Flow Station

Name ID # Latitude Longitude Datum Hutchinson C BL Doolittle DR NR Marysville

11424550 39˚ 07’ 26.36’’ 121˚24’08.75’’ NAD83

Unnamed Trib Ab Gavin Mandry Dr Nr Marysville, CA

11424548 39˚06’04.25’’ 121˚23’23.81’’ NAD83

Unnamed Trib Bl Gavin Mandry Dr NR Marysville, CA

11424546 39˚06’02.54 121˚23’06.95’’ NAD83

USGS stations shown in Table 2 were located downstream of Goose, Lower Blackwelder, Bedsprings, Frisky, Pave Paws, and Small Arms Range Lakes, where the outlets are controlled through ungated spillways and culverts. Elevation-storage-discharge relationships for each impoundment were developed from HEC-RAS using 1- meter LIDAR to represent the lake topography. The LIDAR was developed in August of 2016 when most of the lakes were nearly dry. More information on the LIDAR is described in the Airborne LiDAR Report (Woolpert, 2016). A stage-discharge relationship was developed for the dam (embankment) crest of each of the Beale AFB lakes mentioned in this report. A simplified hydraulic model was developed in HEC-RAS and was used to establish the relationships. The hydraulic model consisted of a dam profile and a flow hydrograph. Computed model results include both stage and discharge values and when combined, define the stage-discharge relationships. Dam crest profiles were based on lidar data received from the Beale AFB. Flow hydrographs were developed individually for each dam starting with a low discharge and ending with a high discharge.

Storm Event and Meteorological Model Setup HEC-HMS was calibrated to the January 2017 event and validated to the February 2017 event. Water year 2017 was the wettest year on record based on the Northern Sierra Precipitation 8-Station Index, shown in Figure 12. WY 2017 had an estimated 45 atmospheric rivers make landfall on the west coast, as shown in Figure 13. This water year was highly useful for calibrating the model since it represented a wet season with high recorded flows.

Figure 12: Water Year 2017 accumulated rainfal l for Northern Sierra Precipitation 8 Station Index Accumulated from CDEC (DWR, accessed 2017)

Figure 13: Description of observed Atmospheric Rivers in WY 2017 (Ralph et. al . , accessed 2017b)

The HEC-HMS model parameters were calibrated to the January 2017 rainfall event from the 6th and 13th. Flows were the result of an atmospheric river categorized as “extreme” (Figure 13) based on Ralph/CW3E Strength Scale. Two of the three discharge stations displayed the highest peak values during this period, shown in Figure 14, with Hutchinson Creek reaching over 1,000 cfs.

Parameters were checked against the February storm event from the 5th to the 10th. Ideally, the model validation would be performed using storms for other water years; however, discharge values were only available for 2017. Future improvements to the model should consider calibrating/validating the model to other storm events when the data becomes available.

Figure 14: WY 2017 period of record for (top) Hutchinson Creek below Doolitt le Drive near Marysvi l le (11424550); (middle) Unnamed Tributary above Gavin Mandry Drive near Marysvi l le (11424548); and (below) Unnamed Tributary below Gavin Mandry Drive near Marysvi l le (11424546).

The “Gage Weights” method was selected as the Meteorological Model in HEC-HMS. The gage weights method works with recording and non-recording precipitation gages to simulate an event. This provides additional flexibility to specify rainfall depths and temporal distribution separately for each subbasin.

The non-recording gage was used to estimate the total storm depths at each subbasin, while recording gages were represented by Bear River at Wheatland (BRW) for the January event and South Honcut Creek near Bangor (SFH) for the February event. Various other rainfall gages were tested, but temporal distribution was best represented by these gages for calibration and validation.

Gridded precipitation was used to estimate the non-recording gage rainfall depths. Gridded precipitation was generated using HEC-GageInterp for the calibration and validation events. GageInterp generates a sequence of HEC-DSS grids from collection of rainfall gage time-series and interpolation algorithm. The generated gridded precipitation can reasonably account for the spatial distribution of rainfall depths; however, there have been cases where gridded datasets have underestimated the precipitation depths (USACE, 2015).

Rainfall gages were collected from California Data Exchange Center (CDEC). The CDEC online site serves as a centralized clearinghouse for data gathered throughout California. Table 3 identifies precipitation stations used to generate the gridded precipitation dataset. The data was cleaned and edited prior to grid generation. Inverse distance squared (ID2W) was selected as the gage interpolation method with a grid cell size of 100 meters. PRISM (PRISM, accessed 2017) dataset was selected as a bias adjustment.

Once gridded values were generated, rainfall depths for each subbasin were summed for the calibration and validation simulation period from January 6th to 13th and February 5th to 10th.

Table 3: Rainfal l Station within surrounding area used for gridded precipitation development

Station Name ID Long. Lat. Elev. (ft. NGVD29)

AUBURN DAM RIDGE ADR -121.045 38.882 1200

ALLEGHANY ALY -120.875 39.470 4957

BANGOR BGR -121.386 39.381 800

BLUE CANYON BLC -120.709 39.280 5280

BALD MOUNTAIN (USFS) BMT -120.683 38.900 4680

BOWMAN LAKE BOL -120.654 39.448 5390

BEAR R AT ROLLINS RESERVOIR BRE -120.953 39.133 1945

BULLARDS BAR BUD -121.144 39.396 2100

BEAR RIVER AT WHEATLAND BRW -121.407 39.000 72

CAMPTONVILLE (DWR) CAM -121.049 39.451 2755

BEAR RIVER AT CAMP FAR WEST DAM CFW -121.317 39.050 260

DRUM POWER HOUSE DPH -120.767 39.264 3400

DEER CK FOREBAY (PG&E) DRC -120.825 39.300 4455

GRASS VALLEY GVY -121.067 39.206 2400

LINCOLN LCN -121.272 38.882 200

LAKE SPAULDING (PG&E) LSP -120.633 39.317 5156

SACTO R AT MOULTON WEIR (CREST 76.8') MLW -122.023 39.338 85

OUR HOUSE DAM OHD -120.996 39.412 1960

PILOT HILL (CDF) PIH -121.009 38.832 1200

PIKE COUNTY PKC -121.202 39.475 3714

READER RANCH RDH -121.117 39.304 2025

SOUTH HONCUT CREEK NEAR BANGOR SFH -121.372 39.368 630

SUGAR PINE SGP -120.750 39.128 3843

SLAB CREEK SLB -120.699 38.773 1850

SEED ORCHARD RAWS NEAR MICHIGAN BLUFF 4N SOM -120.732 39.091 4300

SECRET TOWN SRT -120.883 39.183 2720

WHITE CLOUD WTC -120.838 39.317 4321

Calibration and Results The main parameters adjusted during the calibration process were the time of concentration, storage coefficient, and constant loss rate. No other parameters were changed. The calibration goal was to obtain Nash-Sutcliffe efficiency coefficient (NSE) > 0.75 and a validation > 0.65. Performance ratings using NSE, as shown in Table 4, were developed by Moriasi et al., 2007 and adopted for this study.

Other comparisons, such as time to peak, hydrograph shape, volume, and peak values were also checked. Matching peak flow was of particular interest since the 2016 analysis showed peak flow values were highly correlated to the max lake stages at Frisky (USACE, 2016).

Table 4: Nash-Sutcl iffe Performance Rating developed by Morisasi et al . , 2007.

Performance Rating Nash-Sutcliffe Very Good 0.75 < NSE < 1.00 Good 0.65 < NSE < 0.75 Satisfactory 0.50 < NSE < 0.65 Unsatisfactory NSE < 0.5

Initial simulations using computed parameters were run to ensure model setup was correct. Computed hydrographs showed reasonable comparisons to the observed hydrograph shape and peak timing, but highly underestimated observed volume. Loss rates were incrementally reduced by 0.01 in/hr, but simulated volume was unable to match observed volume, even with loss rates reduced to zero. The best explanation for the discrepancy in volume was the rainfall depth values were likely underestimated.

Rainfall depths were compared to PRISM daily values which showed higher precipitation depths with PRISM. The rainfall depths were scaled approximately 1.6x to the original rainfall depths which better aligned with the hydrograph volumes.

HEC-HMS’s model optimization capabilities were used to match to the observed hydrographs as close as possible (Figure 15). The optimization method selected the Nelder Mead search algorithm with an objective function of maximizing Nash Sutcliffe for the unit hydrograph parameters. Care was taken while using the optimization to ensure optimized values were generally within the initial estimates.

Final unit hydrograph parameters were the result of optimization and user judgment. Values were not exaggerated to obtain a better goodness of fit. Loss rates were reduced to obtain a better match to the hydrograph volume. Subbasins’ constant loss rates were reduced on average by 12% from their initial estimates. A global 12% reduction to the loss rates was applied to the subbasins outside the calibration region.

Calibrated values and initial estimates are shown in Table 5. Simulated hydrograph shape and timing visually matched well with observed values, shown in Figure 16, Figure 17, and Figure 18. The peak flow value was calibrated within 1% of the Hutchinson gage value and 20% and 26% on the Gavin Mandry Dr gage values. The above 25% difference was of initial concern since the peak value is of highest interest, but became less alarming when validation results showed a better fit to the peak value and differences were reduced below 20%. Overall, the simulated hydrograph matched well with the observed hydrograph and passed the Nash-Sutcliffe criteria of > 0.75.

Figure 15: Map showing USGS gage locations and corresponding subbasin that contribute to those gages.

The optimization routine was performed on these subbasins.

Figure 16: Hutchinson Creek below Doolitt le Dr near Marysvil le, CA – January 6 – 14, 2017

Figure 17: Unnamed Tributary above Gavin Mandry Dr near Marysvi l le, CA – January 6 - 14, 2017

Fl ow (c fs

Date

Simulated

Observed

Fl ow

(c fs

Date

Simulated

Observed

Figure 18: Unnamed Tributary below Gavin Mandry Dr near Marysvi l le, CA – January 6-14, 2017

Table 5: Calibration event simulated comparison with observed hydrographs for January 6 to 14, 2017.

Hydrograph

Hutchinson Creek Below Doolittle Drive Above Gavin Mandry Drive Below Gavin Mandry Drive

Computed Observed Difference Computed Observed Difference Computed Observed Difference Peak

Discharge (cfs) 1021 1020 0% 395 530 -26% 228 284 -20%

Volume (acre-ft) 3584 3444 4% 1337 1238 8% 721 704 2%

Date/Time of Peak

Discharge 1/8/17 13:30

1/8/17 13:30 0

1/8/17 9:30

1/8/17 9:45 15 min

1/08/17 8:15

1/8/17 10:00

1 hr 45 min

Nash- Sutcliffe - 0.92 - - 0.91 - - 0.93 -

RMS (cfs) - 74.5 - - 33.2 - - 16.5 -

Validation results to the February storm event met the criteria for success. Both peak flow and shape matched well with the observed data with a NSE coefficients greater than 0.65. Since the validation was done for the same water year, future testing of the model should be considered when other water year flows are available. Table 6 and Figure 19 to Figure 21 show the validation results

Fl ow (c fs

Date

Simulated

Figure 19: Hutchinson Creek below Doolitt le Dr near Marysvi l le, CA – February 5 – 11, 2017

Figure 20: Unnamed Tributary above Gavin Mandry Dr near Marysvi l le, CA – February 5 - 11, 2017

Fl ow

(c fs

Date

Simulated

Observed

Fl ow

(c fs

Date

Simulated

Figure 21: Unnamed Tributary below Gavin Mandry Dr near Marysvi l le, CA – February 5-11, 2017

Table 6: Validation event s imulated comparison with observed hydrographs for February 5 to 10, 2017

Hydrograph Hutchinson Creek Below Doolittle Drive Above Gavin Mandry Drive Below Gavin Mandry Drive

Computed Observed Difference Computed Observed Difference Computed Observed Differen ce Peak

Discharge (cfs) 795 672 18% 376 365 3% 215 258 -17% Volume (acre-

ft) 2268 2183 4% 828 733 13% 428 460 -7% Date/Time of

Peak Discharge

2/9/2017 18:30

2/9/2017 16:45

1 hr 15 min

2/9/2017 16:00

2/9/2017 16:30 minutes

2/9/2017 15:30

2/9/2017 15:15 15 mins

Nash-Sutcliffe - 0.72 - - 0.74 - - 0.83 -

RMS (cfs) - 91.1 - - 35.2 - - 19.8 -

Fl ow (c fs

Date

Simulated