1 USDA – Natural Resources Conservation Service 2 3 DRAFT Technical Note 4

5 Hydrologic Analyses of Post-Wildfire

6 Conditions

7 8 April 2015 9

10 Cover photo: Nabours Mountain directly east of Glenwood NM. 11 Courtesy of: USDA Forest Service, Gila National Forest, Silver City NM DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

12 (This page intentionally blank)

DRAFT Technical Note, April 2014 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

13 Contents 14 Introduction ...... 1

15 The role of NRCS in post-fire assessments and modeling ...... 1

16 Important terms and terminology ...... 4

17 Burned area reflectance classification (BARC) ...... 4 18 Burn severity ...... 4 19 Fire intensity ...... 5 20 Fire severity ...... 6 21 Hydrophobicity or soil water repellency ...... 6 22 Landsat differenced normalized burn ratio (dNBR) ...... 7 23 Rapid assessments of vegetation condition after wildfire (RAVG) ...... 7 24 Soil burn severity mapping ...... 7 25 Low soil burn severity ...... 7 26 Moderate soil burn severity ...... 8 27 High soil burn severity ...... 8 28 Wildfire impacts on watershed hydrology...... 8

29 Assessment of post-fire soil and vegetation conditions ...... 9 30 Assessment of post-fire soil hydrophobic conditions ...... 11 31 Hydrologic modeling of burned watersheds ...... 13

32 Methods and models ...... 13 33 Adjustment of event-based runoff modeling components for burned areas ...... 16 34 Runoff curve numbers ...... 16 35 Adjustment of curve numbers for wildfire effects ...... 18 36 Limitations of the CN method ...... 22 37 Time of concentration ...... 23 38 Watershed lag method ...... 23 39 Velocity method ...... 23 40 Adjustment of runoff flow time for wildfire effects ...... 26 41 Infiltration parameters ...... 28 42 Adjustment of infiltration parameters for wildfire effects ...... 31 43 Unit hydrographs ...... 32 44 NRCS dimensionless unit hydrograph ...... 34 45 Adjustment of unit hydrograph for wildfire effects ...... 36 46 Kinematic wave transformation ...... 38

DRAFT Technical Note, April 2015 i DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

47 Adjustment of kinematic wave transformation parameters for wildfire effects ...... 38 48 Time-area histogram synthetic unit hydrograph development ...... 39 49 Sedimentation estimation...... 41

50 Models for estimating sediment transport...... 46 51 Adjustments for modeling sediment transport from burned areas ...... 48 52 Sediment bulking ...... 48 53 Model uncertainties, calibration, validation ...... 52

54 References ...... 53

55 Case Study 1: High Park Fire Colorado ...... CS1-1

56 Background ...... CS1-1 57 Methods ...... CS1-3 58 Runoff Curve Number (CN) Estimation ...... CS1-5 59 Rainfall ...... CS1-7 60 Lag Time ...... CS1-7 61 Flow Routing ...... CS1-8 62 Sediment Bulking ...... CS1-8 63 StreamStats ...... CS1-8 64 Results and Discussion ...... CS1-9 65 Comparison with Regression Predictions ...... CS1-11 66 Accuracy and Limitations ...... CS1-12 67 Conclusions ...... CS1-13 68 Acknowledgements ...... CS1-14 69 References ...... CS1-14 70 Case Study 2: Bitterroot Wildfires, Montana ...... CS2-1

71 Background ...... CS2-1 72 Methods ...... CS2-4 73 Runoff Curve Number (RCN) Estimation for Burn Areas ...... CS2-8 74 Hydrophobic Soils ...... CS2-13

75 Time of Concentration (Tc) and Assumed Watershed Shape...... CS2-14 76 Results and Discussion ...... CS2-15 77 Limitations ...... CS2-18 78 References ...... CS2-19 79 Case Study 3: West Fork Complex Fire, Colorado ...... CS3-1

80 Background ...... CS3-1

DRAFT Technical Note, April 2015 ii DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

81 Methods ...... CS3-4 82 Runoff Curve Number (CN) Estimation ...... CS3-6 83 Rainfall ...... CS3-7 84 Lag Time ...... CS3-9 85 Flow Routing ...... CS3-9 86 Sediment Bulking ...... CS3-10 87 Erosion Modeling ...... CS3-10 88 Results and Discussion ...... CS3-18 89 Runoff Modeling Results ...... CS3-20 90 Sediment Modeling Results ...... CS3-21 91 Modeling Limitations ...... CS3-23 92 Rainfall-Runoff Modeling Limitations ...... CS3-23 93 Sediment Modeling Limitations ...... CS3-25 94 Conclusions ...... CS3-25 95 Acknowledgements ...... CS3-26 96 References ...... CS3-26 97 Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico ...... CS4-1

98 Background ...... CS4-1 99 NRCS Involvement ...... CS4-4 100 Methods and Application ...... CS4-6 101 Choice of hydrologic computer model: HEC-HMS ...... CS4-7 102 Time-area histogram and synthetic UH Estimation ...... CS4-9 103 Roughness in GIS ...... CS4-10 104 Flow slope in GIS ...... CS4-13 105 Velocity relationships used in Whitewater Creek ...... CS4-18 106 Travel time in GIS ...... CS4-19 107 Spreadsheet derivation of time-area histogram and unit hydrograph ...... CS4-20 108 Infiltration Loss Method: Green & Ampt Equation ...... CS4-24 109 Initial Abstraction, Surface Ponding, Canopy Loss ...... CS4-25 110 Baseflow ...... CS4-26 111 Rainfall events examined ...... CS4-26 112 Sediment Bulking ...... CS4-29 113 Results and Discussion ...... CS4-30 114 Modeling observed data ...... CS4-37 115 Conclusions ...... CS4-40

DRAFT Technical Note, April 2015 iii DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

116 References ...... CS4-41 117 Case Study 5: Saratoga Springs, UT ...... CS5-1

118 Abstract ...... CS5-1 119 Background ...... CS5-2 120 Methods ...... CS5-6 121 Pre-fire flow ranges ...... CS5-8 122 Post-fire peaks and volumes of sediment...... CS5-10 123 Results ...... CS5-16 124 AGWA ...... CS5-19 125 Bulking ...... CS5-19 126 Western U.S. regression model...... CS5-19 127 Regional equations ...... CS5-20 128 Conclusions ...... CS5-20 129 Acknowledgements ...... CS5-21 130 References ...... CS5-21 131 132 List of Figures 133 Figure 1. Differentiating between fire intensity and burn severity (from Parsons, et al. 2010)...... 5 134 Figure 2. Soil-water repellency (hydrophobicity) as altered by wildfire (from DeBano 2000b)...... 6 135 Figure 3. The hydrologic cycle (NASA 2012)...... 8 136 Figure 4. Example of burn severity mapping (red=high, orange=med, yellow=low, green=no burn)...... 10 137 Figure 5. Oregon vegetation types, as mapped by LandFire (2010)...... 11 138 Figure 6. Water droplets resisting infiltration into soil due to extreme hydrophobicity. (Doerr et al. 139 2000)...... 12 140 Figure 7. Graphical solution of the CN runoff equation (USDA-NRCS 2004b) ...... 18 141 Figure 8. USDA Forest Service Regions ...... 20 142 Figure 9. Runoff generating processes on a hillslope terminating at a watercourse (Garen and Moore 143 2005) ...... 22 144 Figure 10. Hydrographs from the same hillslope, with different surface conditions (Canfield, et al. 145 2005)...... 27 146 Figure 11. Green-Ampt infiltration into the soil column ...... 28 147 Figure 12. Optimal hillslope hydraulic conductivity, Cerro Grande NM fire (Canfield, et al. 2005) ...... 31 148 Figure 13. HEC-HMS Green and Ampt input interface (USCOE-HEC 2013) ...... 32 149 Figure 14. Runoff hydrographs by convolution and superposition, using 30-minute unit hydrograph...... 33 150 Figure 15. S curve developed from 30-minute unit hydrograph...... 33 151 Figure 16. Dimensionless unit hydrograph, with time of concentration and lag (USDA-NRCS 2010)...... 34

DRAFT Technical Note, April 2015 iv DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

152 Figure 17. Baldy Creek pre-fire and post-fire 30-minute unit hydrographs...... 37 153 Figure 18. Example of an n-value raster, with progressively higher roughness with darker color ...... 40 154 Figure 19. Example of high burn severity raster (red pixels) ...... 40 155 Figure 20. Example of flow slope raster, progressively steeper with darker color ...... 40 156 Figure 21. 2010 post-wildfire debris flow from San Gabriel Mountains, CA ...... 42 157 Figure 22. Alluvial fan, Death Valley National Park ...... 43 158 Figure 23. A stream cross-section with isolines of velocity, maximized over the thalweg...... 43 159 Figure 24. Riffle-pool morphology ...... 44 160 Figure 25. Hypothetical storm using equation 21b (all discharges converted to cfs)...... 50 161 Figure 26. Sediment production and bulking factors for debris production area 7 (LACDPW 2006) ...... 51 162 163 Figure CS1- 1. Sunset through smoke plumes, from Fort Collins on Day 1 of the High Park Fire...... CS1-2 164 Figure CS1- 2. Aerial extent and soil burn severity of the High Park Fire, based on a BARC image. . CS1-2 165 Figure CS1- 3. Modeled catchments and stream channels, with flow computation points with soil burn 166 severity as background...... CS1-4 167 Figure CS1- 4. Example estimated pre- and post-fire hydrographs for the 10-year rain event...... CS1-10 168 Figure CS1- 5. Example map with pre- and post-fire flood prediction estimates for the Poudre Park area. 169 ...... CS1-11 170 Figure CS1- 6. Estimated 25-year peak flow enhancement ratios for the High Park Fire...... CS1-11 171 172 Figure CS2- 1. State of Montana, with inset of Bitterroot River valley ...... CS2-1 173 Figure CS2- 2. Bitterroot River Watershed with inset of area of the wildfires in the year 2000...... CS2-1 174 Figure CS2- 3. Bitterroot fire severity for the fires of 2000 (Mahlum et al. 2011) ...... CS2-2 175 Figure CS2- 4. Lightning strikes in the Bitterroot Valley of Montana, initiating the wildfires of 2000...... 176 ...... CS2-3 177 Figure CS2- 5. One of the most poignant wildfire scenes ever captured, from the Bitterroot in 2000. CS2-3 178 Figure CS2- 6. Fully loaded silt fence placed in the aftermath of the 2000 Bitterroot Valley Complex 179 wildfire...... CS2-7 180 Figure CS2- 7. Laird Creek in 2001 within what was the Bitterroot Fire Valley Complex...... CS2-7 181 Figure CS2- 8. Sheafman Point in October 2008...... CS2-7 182 Figure CS2- 9. Glen Lake trail in the summer of 2012...... CS2-8 183 Figure CS2- 10. The landscape around Glen Lake in the summer of 2012...... CS2-8 184 Figure CS2- 11. Satellite imagery used to detail burn severity of the 2000 Bitterroot Valley Complex 185 wildfire...... CS2-12 186 Figure CS2- 12. Burn severity map and rain gage data for a rain event that happened a year after the 2000 187 Bitterroot Valley Complex wildfire...... CS2-12 188

DRAFT Technical Note, April 2015 v DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

189 Figure CS3- 1. State of Colorado with zoom into the upper Rio Grande watershed ...... CS3-1 190 Figure CS3- 2. Area of the West Fork Complex wildfire, west of South Fork CO...... CS3-1 191 Figure CS3- 3. Aerial extent of three fires known together as the West Fork Complex fire...... CS3-2 192 Figure CS3- 4. Soil burn severity, model pour points, sub-watersheds, and stream channels ...... CS3-5 193 Figure CS3- 5. 25-year cumulative rainfall distributions used in modeling...... CS3-8 194 Figure CS3- 6. Flow direction raster cell example* ...... CS3-12 195 Figure CS3- 7. Conceptual diagram of sediment yield estimation on hillslope zone breaks ...... CS3-17 196 Figure CS3- 8. Example of an area-weighted RUSLE sediement yield raster ...... CS3-18 197 Figure CS3- 9. Example of estimated pre- and post-fire hydrographs for the 25-year event...... CS3-19 198 Figure CS3- 10. Example map providing pre- and post-fire flood predictions for the S. F. of the Rio 199 Grande...... CS3-19 200 Figure CS3- 11. Peak flow magnification ratios for the West Fork Complex wildfire...... CS3-20 201 Figure CS3- 12. Sediment magnification ratios for the West Fork Complex wildfire...... CS3-21 202 Figure CS3- 13. Infiltration excell and saturation excess overland flow (Laboratory of Ecohydrology, 203 EPFL) ...... CS3-24 204 205 Figure CS4- 1. Location of concern: State of New Mexico with county boundaries, USFS Gila National 206 Forest in green, Gila Wilderness area in yellow...... CS4-2 207 Figure CS4- 2. Location of concern, with Whitewater-Baldy Complex Fire location in red...... CS4-2 208 Figure CS4- 3. Location of concern, showing fire location in red, nearby communities and highways...... 209 ...... CS4-3 210 Figure CS4- 4. Whitewater Creek watershed and the community of Glenwood NM...... CS4-3 211 Figure CS4- 5. Gages in the vicinity of Whitewater Creek...... CS4-4 212 Figure CS4- 6. Burned trees near Hummingbird Saddle, Gila National Forest, ...... CS4-5 213 Figure CS4- 7. HEC-HMS input window for modeling options in a single sub-basin...... CS4-8 214 Figure CS4- 8. Whitewater Creek hydrologic subareas...... CS4-8 215 Figure CS4- 9. Upper Whitewater subarea (called “Baldy Fork”) ...... CS4-9 216 Figure CS4- 10. ArcMap Raster Calculator expression creating a pre-fire roughness layer...... CS4-12 217 Figure CS4- 11. Burn raster (red = high, orange = medium, yellow = low, green = no burn)...... CS4-13 218 Figure CS4- 12. Flow direction codes ...... CS4-14 219 Figure CS4- 13. Trapezoidal channel dimensions...... CS4-17 220 Figure CS4- 14. Twenty-one five-minute travel time bands for Baldy Fork, with Attribute Table. .... CS4-20 221 Figure CS4- 15. Pre-fire time-area histogram for Baldy Fork...... CS4-21 222 Figure CS4- 16. Pre-fire spreadsheet. (30-minute unit hydrograph in column H) ...... CS4-22 223 Figure CS4- 17. Unit Hydrographs, 30-minute duration for Baldy Fork...... CS4-22 224 Figure CS4- 18. S-curve from lagged and summed 30-minute unit hydrographs ...... CS4-23 225 Figure CS4- 19. Baldy Fork S-curves for HecHMS, three scenarios ...... CS4-24

DRAFT Technical Note, April 2015 vi DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

226 Figure CS4- 20. Areal reduction factors developed for 6-hour duration storms at Walnut Gulch ...... CS4-28 227 Figure CS4- 21. Two storm centerings for Whitewater Creek and 10km radii...... CS4-29 228 Figure CS4- 22. Whitewater debris flow hazard (Tillery, Matherne, and Verdin (2012)...... CS4-30 229 Figure CS4- 23. Locations of reported HEC-HMS modeling output...... CS4-31 230 Figure CS4- 24. HecHMS results, 100-year storm peaks at Glenwood (no areal reduction)...... CS4-33 231 Figure CS4- 25. HecHMS results, 100-year storm peaks at Glenwood (areal reduction with centering over 232 the upper Whitewood sub-basin) ...... CS4-35 233 4Figure CS4- 26. 100-year post-fire hydrographs at Glenwood between storm centerings in the upper 234 Whitewater sub-basin and the SF Whitewater sub-basin...... CS4-36 235 Figure CS4- 27. 25-year post-fire hydrographs at Glenwood between storm centerings in the upper 236 Whitewater sub-basin and the SF Whitewater sub-basin...... CS4-36 237 Figure CS4- 28. Whitewater Creek NM flood event of 14-15 Sep 2013, gaged rainfall and streamflow ...... 238 ...... CS4-39 239 Figure CS4- 29. Whitewater Creek flood 14-15 Sep 2013, observed and modeled ...... CS4-40 240 241 Figure CS5- 1. State of Utah and fire location, about 40 miles south of Salt Lake City...... CS5-3 242 Figure CS5- 2. Burning watershed from Eagle Mountain, UT, 21 June 2012...... CS5-3 243 Figure CS5- 3. Burning watershed from Eagle Mountain, UT, 22 June 2012...... CS5-5 244 Figure CS5- 4. Burned watershed flow direction into Saratoga Springs, UT from 1 September 2012 event. 245 ...... CS5-5 246 Figure CS5- 5. Sediment laden flow through residential neighborhood...... CS5-6 247 Figure CS5- 6. Typical sediment composition in residential area...... CS5-6 248 Figure CS5- 7. Location of stream gages and study area ...... CS5-9 249 Figure CS5- 8. Stream gage regression output converted to discharge per unit area ...... CS5-9 250 Figure CS5- 9. Burned zone (red) and sub-areas upstream of Saratoga Springs...... CS5-11 251 Figure CS5- 10. Curve number from cover, for hydrologic soil groups (Goodrich et al. 2005) ...... CS5-12 252 Figure CS5- 11. Interdependency of CN and lag time, Cerro Grande Wildfire (McLin et al. 2001) .. CS5-13 253 Figure CS5- 12. Comparison of observed and simulated pre-and post-fire peak discharges per unit drainage 254 basin area in New Mexico. (McLin et al, 2001) ...... CS5-15 255 Figure CS5- 13. Range of annual erosion rates from Bridges (1973)...... CS5-16 256 Figure CS5- 14. Burned watershed pre-fire and post-fire hydrographs from WinTR-20...... CS5-18 257 258 259 List of Tables 260 Table 1. Federal agencies and their roles concerning wildfire events...... 3 261 Table 2. Computer models for rainfall-runoff analysis ...... 14 262 Table 3. Runoff curve numbers for arid and semiarid rangelands (USDA-NRCS 2004a) ...... 17

DRAFT Technical Note, April 2015 vii DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

263 Table 4. Suggested runoff curve numbers for burned areas from Cerrelli (2005)...... 19 264 Table 5. Forest Region 1 curve numbers (Story 2003) ...... 20 265 Table 6. Forest Region 1 curve numbers (Stuart 2000) ...... 20 266 Table 7. Forest Region 3 curve numbers ...... 21 267 Table 8. Forest Region 3 curve numbers ...... 21 268 Table 9. Forest Region 3 pre-fire and post-fire curve numbers...... 21 269 Table 10. Manning’s roughness coefficients for sheet flow (depth generally ≤ 0.1 ft) ...... 24 270 Table 11. Shallow-concentrated flow velocity equations, where S=flow slope in ft/ft ...... 25 271 Table 12. Green Ampt parameters for soil texture classes & horizons (Rawls, et al. 1983, Saxton, et al. 272 1986) ...... 30 273 Table 13. Size classification and potential consequences of debris flows (Santi and Morandi, 2012) ...... 45 274 275 Table CS1- 1. CN assignments implemented in High Park Fire hydrologic modeling. (Highlighted 276 columns indicate values extracted from USDA-NRCS 2004a...... CS1-6 277 Table CS1- 2. Comparison of CN modeling with USGS regression results from StreamStats...... CS1-12 278 279 Table CS2- 1. Runoff curve numbers for high severity burned areas of southwest Montana* ...... CS2-9 280 Table CS2- 2. Recommended runoff curve numbers for moderate severity burned areas ...... CS2-9 281 Table CS2- 3. Recommended RCNs for low severity burned and unburned areas with north and east facing 282 slopes...... CS2-10 283 Table CS2- 4. Recommended RCNs for low severity burned and unburned areas with south and west 284 facing slopes...... CS2-11 285 286 Table CS3- 1. CN assignments implemented in West Fork Complex Fire hydrologic modeling...... CS3-7 287 Table CS3- 2. Rainfall depths (inches) implemented in the modeling...... CS3-8 288 Table CS3- 3. Spatial datasets used for the six RUSLE factors...... CS3-12 289 Table CS3- 4. C factors for common land covers (Theobald et al. 2010) ...... CS3-16 290 291 Table CS4- 1. Pre-fire roughness assumptions...... CS4-10 292 Table CS4- 2. Post-fire (immediately) roughness assumptions...... CS4-10 293 Table CS4- 3. Post-fire (1 year) roughness assumptions...... CS4-11 294 Table CS4- 4. Values of φ, given assumptions of overland flow width and depth...... CS4-15 295 Table CS4- 5. Shallow-concentrated flow parameters...... CS4-16 296 Table CS4- 6. Trapezoidal channel parameters...... CS4-18 297 Table CS4- 7. Whitewater Creek watershed 6-hour duration storm totals...... CS4-27 298 Table CS4- 8. HecHMS output: hydrograph peaks at four locations (no areal reduction) ...... CS4-32

DRAFT Technical Note, April 2015 viii DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

299 Table CS4- 9. HecHMS output: hydrograph peaks at four locations (areal reductions applied, with 300 centerings as shown in figure CS4-22)...... CS4-34 301 Table CS4- 10. Sediment bulked flows for Whitewater Creek...... CS4-37 302 Table CS4- 11. NOAA Precip-Frequency values and Sep 2013 rainfall event from Catwalk gage .... CS4-38 303 304 Table CS5- 1. Six stream gages near Saratoga Springs, UT ...... CS5-8 305 Table CS5- 2. Burned watershed subarea input to WinTr-20...... CS5-11 306 Table CS5- 3. Increase in runoff curve number from pre-fire to post-fire conditions...... CS5-12 307 Table CS5- 4. Output variables available in AGWA...... CS5-14 308 Table CS5- 5. Cubic feet per second per square mile (CSM) from WinTR-20 pre-fire results and selected 309 stream gages...... CS5-17 310 Table CS5- 6. Burned watershed pre-fire and post-fire peak flow output from WinTR-20...... CS5-17 311 Table CS5- 7. Storm totals runoff for various recurrence intervals, input to WinTR-20 and post-fire runoff 312 values for Subarea 1+2...... CS5-18 313 Table CS5- 8. Burned watershed post-fire event total sediment runoff in tons...... CS5-19 314 Table CS5- 9. Saratoga Springs burned watershed pre-fire and post-fire peaks (CSM) from WinTR-20. 315 ...... CS5-20 316 317 318 Acknowledgements 319 Dan Moore, Hydraulic Engineer, Portland, Oregon, USDA-NRCS, West National 320 Technology Support Center (Case Study 4: Whitewater Creek; Gila Wilderness, New 321 Mexico) 322 Nathaniel Todea, Hydraulic Engineer, USDA-NRCS, Salt Lake City, Utah (Case Study 323 5: Saratoga Springs Wildfire Utah) 324 Geoff, Cerrelli, retired, Hydraulic Engineer (Case Study 2: Bitterroot Wildfires, 325 Montana) 326 Steven Yochum, Hydrologist, USDA-NRCS, Lakewood, Colorado (Case Study 1: High 327 Park Fire Colorado, Case Study 3: West Fork Complex Fire, Colorado) 328 John B. Norman, III, Soil Scientist, USDA-MLRA Soil Survey Office, Fort Collins, 329 Colorado (Case Study 3: West Fork Complex Fire, Colorado) 330 Claudia Hoeft, National Hydraulic Engineer, USDA-NRCS, Washington DC. 331 332

DRAFT Technical Note, April 2015 ix DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

333 (This page intentionally blank)

DRAFT Technical Note, April 2015 x DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

334 Introduction 335 Periodic wildfire is a normal occurrence on forests in the Western United States. Wildfire 336 imposes immediate, extensive, and long-term landscape changes, specifically removal of 337 protective vegetation and, sometimes, inhibited soil infiltration, changing the hydrology of 338 landscape and resulting in radically larger flood events following a fire. Often these floods 339 results in excessive streambank erosion, and unusual amounts of sediment and debris. 340 341 To account for these landscape changes, hydrologists use predictive rainfall-runoff models to 342 illustrate and evaluate the range of flood magnitudes and durations expected from given rainfall 343 events in the watershed. These models also assist in planning re-vegetation efforts in burned 344 watersheds, by revealing areas that will provide the greatest positive effect from the application 345 of conservation activities. 346 347 This technical note provides hydrologic guidance for analysis of burned watersheds. It discusses 348 the specific impacts of wildfire on the runoff process; and gives detailed information on several 349 alternative ways to account for wildfire effects in hydrologic computer models. It discusses 350 several hydrologic models and demonstrates specific analysis techniques for obtaining useful 351 estimates of post-wildfire flooding and sedimentation. Five case studies document the use of 352 these techniques for modeling actual wildfire-burned watersheds. 353 354 The role of NRCS in post-fire assessments and modeling 355 Many Federal agencies provide assistance related to wildfire (table 1). In contrast to the public- 356 land focus of many agencies, NRCS helps private landowners with conservation practices that 357 enable good management of their land. As individual landowners seek to take advantage of the 358 benefits of conservation, they rely on NRCS’ technical expertise in agronomy, rangeland 359 management, nutrient and pesticide management, biology, and hydrology, among other 360 disciplines. After a fire, NRCS can provide rapid assessments of expected increased flood peaks, 361 sediment flow in streams, and other hazards on the watershed. As post-fire spatial data becomes 362 available, NRCS can provide downstream communities with more thorough technical analyses of 363 ongoing streamflow and sedimentation potential. NRCS assistance may augment that of the

DRAFT Technical Note, April 2015 1 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

364 USDA Forest Service, as discussed below, or it may serve as the only technical assistance 365 available. 366 367 The USDA Forest Service is authorized to carry out burned area emergency response (BAER) 368 assessments to identify imminent post-wildfire threats to human life and safety, property, and 369 critical natural or cultural resources on National Forest System (NFS) lands and take immediate 370 actions to manage unacceptable risks. The USDA Forest Service typically organizes the BAER 371 process, but other Federal, Tribal, State, and local agencies often participate to cover other 372 Federal and non-Federal lands adjacent to or intermingled with NFS lands in the event of fires 373 exceeding defined thresholds for size, severity, and/or soil resource damage (Safford, et al. 374 2008). 375 376 The intent of the BAER program is to determine whether wildfire-caused changes in soil 377 hydrologic function resulted in hazardous conditions that threaten life, health, property, or 378 critical cultural and natural resources due to flooding, erosion, and debris flows. BAER teams 379 often provide immediate hydrologic assessments. Communities value these rapid assessments 380 precisely because they provide some gauge of the sudden and new vulnerability of the 381 watershed. However, rapid production may also entail a sacrifice in accuracy. 382

DRAFT Technical Note, April 2015 2 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

383 Table 1. Federal agencies and their roles concerning wildfire events. Agency Roles

USDA Forest Service Forest management, information dissemination, Pre-event training, fire tower lookout staffing, status monitoring During-event Infrastructure protection Damage assessment, Emergency Watershed Post-event Protection , Burned Area Emergency Response (BAER Teams)

Nat. Interagency Fire Center* Pre-event Info dissemination, training, agency coordination During-event Fire-fight coordination

US National Weather Service During-event Forecasting fire conditions Post-event Flash flood warnings

US Geological Survey Pre-event Data collection through gage network Short-term gage installation and data collection, Post-event research on hazards caused by wildfire

Fed. Emergency Mgt. Agency Pre-event Disaster preparedness, information dissemination Post-event Financial assistance

US Bureau of Land Management

Pre-event Federal rangeland management 384 *Sponsoring Federal agencies include USDA Forest Service, Bureau of Land Management, 385 NOAA National Weather Service, National Park Service, Bureau of Indian Affairs, US Fish and 386 Wildlife Service, and US Fire Administration—FEMA. 387 388 The congressionally-authorized Emergency Watershed Protection (EWP) program encourages 389 implementation of emergency measures after a natural disaster to help protect against imminent

DRAFT Technical Note, April 2015 3 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

390 hazards to life and property. Through EWP, NRCS provides financial assistance through cost- 391 share arrangements, and technical assistance for such measures as upland mulching, debris and 392 sediments basins, removal of debris and/or excess sediment from streambeds, establishment of 393 cover on critically eroding lands or streambanks, repair of conservation practices, and purchase 394 of floodplain easements. 395 396 Important terms and terminology 397 Because wildfire affects a watershed in more aspects than hydrology, terminology is sometimes 398 confusing. The ecosystem response, for example, may include soil erosion, vegetation 399 regeneration, microbial community structure restoration, faunal recolonization, and invasive 400 species introduction (Keeley 2009). As specifically related to hydrologic response, a short list of 401 terms and definitions follows. The National Wildfire Coordinating Group (NWCG 2006), 402 among others, provides glossaries of wildland fire terminology larger in scope than hydrology. 403 See http://www.nwcg.gov/pms/pubs/glossary/index.htm. Parsons, et al. (2010) contains similar 404 clarifications of wildfire terminology. 405 406 Agencies and personnel involved with wildfires use the following terms extensively. These 407 terms relate to the driving mechanisms of hydrologic modeling and to pertinent GIS spatial data. 408 409 Burned area reflectance classification (BARC) 410 The BARC process is a spatial data analysis technique for creating GIS layers of soil burn 411 severity by comparing the reflectance difference in certain wavelengths between pre- and post- 412 fire aerial images (Safford, et al. 2008). BARC informs BAER decisions about locations 413 requiring field visits. The imagery processing is also called “Landsat differenced Normalized 414 Burn Ratio” or dNBR, which is defined below. Comparison with field-collected data has shown 415 that BARC products tend to be more indicative of post-fire vegetative condition than of soil 416 condition, especially in low to moderately burned areas (Hudak et al. 2004). 417 418 Burn severity 419 Burn severity is a qualitative assessment of the heat pulse directed toward the ground during a 420 fire. It relates to soil heating, large fuel and duff consumption, consumption of the litter and

DRAFT Technical Note, April 2015 4 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

421 organic layer beneath trees and isolated shrubs, and mortality of buried plant parts (NWCG 422 2006). Burn severity is subdivided into two indices, one for soil and one for vegetation. Figure 423 1 illustrates a comparison between burn severity and the term fire intensity. 424 425 Observations of the following changes in soil characteristics (Parsons et al. 2010) form the basis 426 for of soil burn-severity evaluations: 427 • loss of effective ground cover due to consumption of litter and duff 428 • surface color change due to char, ash cover, or soil oxidation 429 • loss of soil structure due to consumption of soil organic matter 430 • consumption of fine roots in the surface soil horizon 431 • formation of water repellent layers that reduce infiltration 432 433 Fire intensity 434 Fire intensity is a description of the physical combustion process of energy release from organic 435 matter; and represents the energy released during various phases of a wildfire (Keeley 2009). 436 The concept applies to various aspects of a currently ongoing wildfire event, such as fire line 437 intensity, which is the rate of heat transfer per unit length of the fire line. By contrast, burn 438 severity is a post-fire condition. (See figure 1.) 439 440 Figure 1. Differentiating between fire intensity and burn severity (from Parsons, et al. 2010).

DRAFT Technical Note, April 2015 5 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

441 442 Fire severity 443 The term fire severity is most commonly used in empirical studies to refer to the immediate 444 impact on soil and vegetation of heat pulses above and below ground, particularly by indexing 445 the degree of organic matter consumed (Keeley 2009). This term is very similar to the term burn 446 severity, which is preferred by BAER specialists. 447 448 Hydrophobicity or soil water repellency 449 A characteristic of soil generally related to organic matter, soil hydrophobicity is the tendency of 450 the soil to resist wetting or infiltration of moisture. A relatively thin hydrophobic layer can form 451 in an unburned forest, due to the leaching of organic matter from the duff into the soil. During 452 wildfire, the hydrophobic layer can shift downward in the soil and increase in thickness. The 453 intense heat of the wildfire produces a marked temperature gradient in the upper soil layer 454 because dry soil is a poor heat conductor (DeBano 1981). That temperature gradient tends to 455 transport vaporized organic matter downward in the soil column until cooler layers can condense 456 it. Subsequent organic coating of soil particles results in a water repellent layer. Refer to figure 457 2 for a schematic between unburned and burned condition. DeBano 2000a and 2000b also 458 provide a review of this topic. 459 460 Figure 2. Soil-water repellency (hydrophobicity) as altered by wildfire (from DeBano 2000b).

DRAFT Technical Note, April 2015 6 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

461 462 Landsat differenced normalized burn ratio (dNBR) 463 By comparing pre-fire and post-fire Landsat imagery, the differenced normalized burn ratio, or 464 dNBR, analysis enables classification of burn severity. The BARC classification for soil burn 465 severity uses this process. A modification of dNBR called RdNBR, where R stands for relative, 466 is used to assign vegetation burn severity. Comparing the absolute difference in images using 467 dNBR tends to assign too much weight to pre-burn vegetation conditions. The relative version, 468 RdNBR, removes this bias (Safford, et al. 2008; Miller and Thode 2007). 469 470 Rapid assessments of vegetation condition after wildfire (RAVG) 471 A process used by USDA Forest Service specialists to prepare spatial mapping of wildfire effects 472 on vegetation, rapid assessments of vegetation condition after wildfire (RAVG) uses Landsat 473 thematic map images and accounts for before and after wildfire conditions. The reduction of 474 vegetation due to wildfire is indexed by basal area loss, where basal area is a forest management 475 term referring to the sum of cross-sectional areas of trees and stems at breast-height for a given 476 section of land. Although the BAER teams generally focus on rapid assessment of soil burn 477 severity using BARC, the RAVG project creates maps of wildfire on vegetation, usually within 478 thirty days of fire containment. 479 480 Soil burn severity mapping 481 The USDA Forest Service BAER teams assess soil burn severity and produce high-resolution 482 GIS-based maps. The process uses remote sensing (BARC) with field verification (Parsons et al. 483 2010). The GIS soil burn severity layers generally use three categories that are defined as 484 follows (Parsons et al. 2010): 485 486 Low soil burn severity 487 Surface organic layers are not completely consumed and roots are generally unchanged, due to 488 minimal heat penetration of the soil. While exposed mineral soil may appear lightly charred, the 489 canopy and understory vegetation generally appears unchanged. 490 491

DRAFT Technical Note, April 2015 7 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

492 Moderate soil burn severity 493 Up to eighty percent of the pre-fire ground cover may be consumed. Roots may be scorched but 494 generally not completely consumed, and soil structure is unchanged. 495 496 High soil burn severity 497 All or nearly all of the pre-fire ground cover is generally consumed, along with roots up to 0.1 498 inches (0.25 cm) in diameter. Charring may be visible on larger roots. Significant bare or ash- 499 covered soil is exposed and soil structure is less stable due to loss of root mass. 500 501 Wildfire impacts on watershed hydrology 502 A diagram of the hydrologic cycle succinctly shows water processes of Earth and transport 503 mechanisms (figure 3, NASA 2012). For example, evaporation changes the water phase from 504 liquid to vapor, while transporting it from the surface to the atmosphere. 505 506 Figure 3. The hydrologic cycle (NASA 2012).

507 508

DRAFT Technical Note, April 2015 8 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

509 For wildfire-impacted watersheds, the most pertinent part of the hydrologic cycle is the phase 510 change from vapor to liquid, accompanied by transport to higher elevations of the land. The 511 rainfall, returning with higher potential energy, must infiltrate or run off as gravity draws it back 512 to the ocean sink. Wildfire tends to speed up this return and proportion more to runoff rather 513 than to infiltration. 514 515 As a highly effective solvent and with considerable momentum and energy, liquid water begins 516 the erosive process upon first contact with the Earth’s surface—as raindrops impact bare soil. 517 Concentrating flows, whether in small rivulets or larger streams, easily entrain silt, sand, and 518 clay particles, leaving erosion scars on land surfaces and stream banks. Vegetation offers some 519 protection, shielding the soil surface from raindrop impact, holding soil in place with root 520 structures, and impeding flow with branches and leaves. Forest litter can also help protect the 521 soil surface. Plants help the soil resist the erosive forces of flowing water both within overland 522 pathways as well as within stream corridors. 523 524 Destruction of vegetation and the infusion of soil hydrophobicity affect runoff by increasing 525 runoff volume and velocity, thus increasing erosivity. Consequently, the healing of a watershed 526 after a fire depends on the return of vegetation to pre-fire conditions and the breakdown of soil 527 hydrophobicity. Over the course of several years, these processes gradually diminish elevated 528 runoff volume, velocity, and erosivity to pre-fire levels. One study found that 10-year rainfall 529 events on a wildfire-impacted landscape resulted in 100-year or 200-year peak floods (Conedera 530 et al. 2003). Another study found that wildfire increased runoff proportionally more for smaller 531 watersheds than for larger ones (Stoof et al. 2011). 532 533 Assessment of post-fire soil and vegetation conditions 534 One of the most convenient ways to assess the post-fire runoff surface condition is to use soil 535 burn severity mapping (figure 4) derived by the BARC process. Using this process, the 536 hydrologist can determine vegetation loss as a percentage of drainage area for any drainage 537 subarea. Another useful spatial data product derived through the RAVG process, produces maps 538 that relate specifically to vegetation rather than soil condition. . Historical information on

DRAFT Technical Note, April 2015 9 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

539 vegetation effects is available through the cooperative project Monitoring Trends in Burn 540 Severity (MTBS) through the website http://www.mtbs.gov/. 541 542 Another useful data product (Figure 5) is from a Federal mapping initiative called LandFire 543 (2010). Also known as Landscape Fire and Resource Management Planning Tools this initiative 544 was established by the US Department of Interior and the USDA Forest Service, with the goal of 545 providing “a comprehensive, consistent, scientifically credible suite of spatial data layers for the 546 entire United States,” (LandFire 2010) including vegetation types, forest canopy, canopy heights, 547 fire regime classes, and more. Subdivided by geographic region, these high-resolution maps are 548 available through the LandFire website at: http://www.landfire.gov/index.php. 549 550 Figure 4. Example of burn severity mapping.

551 (red=high, orange=med, yellow=low, green=no burn) 552

DRAFT Technical Note, April 2015 10 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

553 Figure 5. Oregon vegetation types, as mapped by LandFire (2010).

554 555 Assessment of post-fire soil hydrophobic conditions 556 The development of soil hydrophobicity is multi-factored and does not necessarily correlate with 557 soil burn severity (Parsons et al. 2010). The depth and thickness of wildfire-induced soil 558 hydrophobicity is dependent upon vegetation type, amount of soil organic matter, soil texture, 559 fire residence time, and burn temperature (DeBano 1981, Huffman, et al. 2001). An important 560 pre-condition is that organic matter must exist for the wildfire to vaporize, leading to 561 condensation of the hydrophobic compounds within the soil profile (DeBano et al. 1967). 562 563 Greater water repellency is generally associated with coarse-grained soil texture. Given a limited 564 supply of condensing hydrophobic substances, the smaller surface areas of coarse-grained soils 565 are more completely covered than the greater surface area of silts and clays (Doerr et al. 2000). 566 567 In addition, the susceptibility of soil texture to hydrophobicity may be related to hydrologic soil 568 group classification (MacDonald, et al. 2000). The four hydrologic soil groups (HSGs) are a 569 standard parameter of the NRCS soil survey, fully described in USDA-NRCS 2009b. The soils 570 of HSG A typically have less than 10 percent clay and more than 90 percent sand or gravel. The

DRAFT Technical Note, April 2015 11 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

571 soils of HSG B typically have 50 to 90 percent sand, whereas HSGs C and D have less than 50 572 percent sand content. The coarseness of HSGs A and B gives them a greater tendency toward 573 hydrophobicity, while the higher silt and clay content of HSGs C and D renders them less 574 susceptible to hydrophobicity. However, HSG is specifically assigned based on saturated 575 hydraulic conductivity of the least transmissive soil layer, not to particle size or hydrophobic 576 susceptibility, and should not be considered a definitive precursor for water repellency. (USDA- 577 NRCS 2009b.) 578 579 Hydrophobicity of the upper mineral soil layer can be determined in the field through a simple 580 procedure: 581 1. Scrape away any ash layer. 582 2. Place several individual drops of water on the air-dried surface. 583 3. Soil later is considered hydrophobic if drops remain after one minute. (See figure 6.) 584 585 Figure 6. Water droplets resisting infiltration into soil due to extreme hydrophobicity. (Doerr et al. 2000).

586 587 The depth of the water repellent layer can depend on fire temperature. Less severe burning may 588 produce hydrophobicity of the soil surface. Locations of greater burn severity may result in the 589 top of the hydrophobic layer being at some depth, say 0.5 to 3 inches. Recognizing this 590 possibility, test the lower layers if the top layer is not water repellent. Scrape away 0.5 to 1 inch 591 of soil depth and repeat the water drop test. By continuing this procedure, both the depth of the 592 top and bottom of the hydrophobic layer may be determined for that location. Repeat this field

DRAFT Technical Note, April 2015 12 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

593 procedure over time to investigate hydrophobicity persistence. Expect considerable spatial 594 variability in hydrophobicity. 595 596 More extensive field procedures are available to quantify the degree of hydrophobicity, such as 597 measuring the water drop penetration time (WDPT). For more severe hydrophobicity, ethanol 598 drops are sometimes used in a critical surface tension (CST) test. Huffman et al. 2001, Scott 599 2000, Doerr 1998, and Robichaud et al. 2008 discuss these and other more extensive procedures. 600 601 The persistence of hydrophobicity is dependent upon several factors, including wetting-drying 602 cycles, and both physical and biological action (Huffman, et al. 2001 and DeBano, 1981). 603 604 Contact by moisture is a factor in the breakdown of hydrophobicity. Although first contact with 605 moisture is repelled, when the soil moisture threshold is exceeded, the layer will allow some 606 infiltration. Upon drying, the layer may again become hydrophobic. This wetting and drying 607 process, along with other physical and biological factors, removes hydrophobicity over time. 608 Among the other factors that help break down hydrophobicity are impact of wildlife treading, 609 impact of falling vegetation such as branches, root penetration of the water repellent layer, and 610 the freeze-thaw process. The hydrophobicity may be minimized in as short a time as a few 611 months (Huffman et al. 2001), but more extensive hydrophobic layers can persist for up to six 612 years (Dyrness 1976). 613 614 Hydrologic modeling of burned watersheds 615 Methods and models 616 Numerous methods and computer programs, varying in complexity, are available for modeling 617 rainfall-runoff processes. In general, these models account for the pertinent aspects of the 618 landscape, including land cover, soils, elevation, slope, sub-basin shape, vegetation character, 619 and flow time of concentration. Meteorology is characterized using either actual storm 620 hyetographs of one or more precipitation gages, or theoretical less-frequent event storms, such as 621 the 100-year, 24-hour storm. Precipitation is distributed over time using pre-determined 622 synthetic storm distribution curves that define rainfall intensity throughout the storm. The shape 623 of the runoff hydrograph is often estimated using one of several possible unit-hydrograph

DRAFT Technical Note, April 2015 13 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

624 estimation methods. For larger watersheds flow routing may be needed to transform and 625 attenuate flow through stream systems. 626 627 Several commonly used computer models for rainfall-runoff analysis are listed in table 2. 628 Continuous simulation models, such as Agricultural Non-Points Source Pollution Model 629 (AGNPS, USDA-ARS, 2013a) and the Soil and Water Assessment Tool (SWAT, USDA-ARS, 630 2013b) help analyze runoff, sedimentation, and water quality issues over long time periods, such 631 as decades. For assessment of rainfall-runoff after wildfire, single flood events are usually the 632 concern. To model single events, commonly used computer models are the US Army Corps of 633 Engineers program Hydrologic Engineering Center – Hydrologic Modeling System (HEC-HMS, 634 USACE-HEC, 2013); and the NRCS models, Computer Program for Project Formulation – 635 Hydrology (WinTR-20, USDA-NRCS 2004c) and Small Watershed Hydrology (WinTR-55, 636 USDA-NRCS 2009a). BAER teams also use WILDCAT (Hawkins and Greenberg 1990). The 637 ARS program, KINEROS2 (A Kinematic Runoff and Erosion Model, Smith et al. 2005, Canfield 638 and Goodrich 2005) is also used for wildfire area runoff simulation, through the Automated 639 Geospatial Watershed Assessment (AGWA) GIS-based tool for watershed assessments. The 640 AGWA tool is a GIS interface that automates the parameterization and execution of the SWAT 641 and KINEROS2 models. 642 643 Table 2. Computer models for rainfall-runoff analysis Acronym Name Reference AGNPS Agricultural Non-Point Source Pollution USDA-ARS 2013a Model SWAT Soil and Water Assessment Tool USDA-ARS 2013b HecHMS Hydrologic Engineering Center - Hydrologic USCOE-HEC 2013 Modeling System WinTR-20 Computer Program for Project Formulation - USDA-NRCS Hydrology 2004c WinTR-55 Small Watershed Hydrology USDA-NRCS 2009a

DRAFT Technical Note, April 2015 14 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

Wildcat4 WILDCAT4 flow model Hawkins & Greenberg 1990 KINEROS2 Kinematic Runoff and Erosion Model Smith, et al. 2005 AGWA Automated Geospatial Watershed Assessment Canfield and Tool Goodrich 2005 644 645 In modeling a storm event, given precipitation volume distributed over time (a hyetograph), the 646 hydrologist must determine what fraction infiltrates and/or is stored in surface depressions. The 647 models of table 2 handle these phenomena in one of two general ways, the NRCS curve number 648 (CN) method or physically based methods, such as Green and Ampt (GA). However, runoff 649 analysis is not complete, with either the CN or GA methods. The CN runoff method provides a 650 total runoff volume estimate at a watershed outlet, but nothing about peak flow rate or 651 hydrograph shape. The GA method provides the fraction of precipitation that infiltrates, leaving 652 that which is ready to runoff, but does not route that runoff to the outlet. 653 654 With the NRCS models, WinTR-20 and WinTR-55, runoff volume is computed using the NRCS 655 runoff curve number (CN) method. Runoff hydrographs are developed by applying the runoff 656 volume, estimated using the CN method, and time of concentration to a dimensionless unit 657 hydrograph. Infiltration methods, such as GA, available in models such as HEC-HMS, do not 658 employ runoff curve numbers (although HEC-HMS also has a CN option). 659 660 Several case studies presented in this Technical Note as appendices A through E illustrate the use 661 of these models individually or various combinations of these models. Case studies 1 (appendix 662 A) and 3 (appendix C) illustrate the use of HEC-HMS with the runoff curve number method. 663 Case study 2 (appendix B) illustrates the use of the fire hydrology, or FIRE HYDRO, 664 spreadsheet, an adaptation of the runoff curve number method. Case study 4 (appendix D) 665 illustrates the use of HEC-HMS with the Green and Ampt infiltration method. Finally, case 666 study 5 (appendix E) illustrates the use of AGWA and WinTR-20. 667 668

DRAFT Technical Note, April 2015 15 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

669 Adjustment of event-based runoff modeling components for burned areas 670 The phenomena of rainfall, interception, surface storage, infiltration, and runoff transformation 671 are discussed in hydrology textbooks, such as Bendient, Huber, and Vieux (2012). To use any of 672 the computer models in table 2 for post-wildfire hydrology, the analyst should understand the 673 impact of fire on these various physical processes. Adjustment of hydrologic model features 674 allows comparison of pre-fire versus post-fire estimates of runoff peaks and hydrograph shapes. 675 676 Accounting for changes in runoff hydrology due to wildfire is generally a matter of determining 677 the extent to which runoff has accelerated due to the loss of vegetation or the lack of infiltration 678 due to soil hydrophobicity. Wildfire may alter any or all of several runoff modeling components. 679 Adjustment to CNs, time of concentration estimates, including flow travel times, Manning’s 680 roughness coefficients, soil infiltration characteristics, and unit hydrographs are detailed in the 681 following sections. 682 683 Runoff curve numbers 684 The NRCS National Engineering Handbook, Part 630, Hydrology, Chapter 9, Hydrologic Soil- 685 Cover Complexes (USDA-NRCS 2004a), documents the general process of determining curve 686 numbers, providing tables of curve numbers for agricultural lands and for arid and semi-arid 687 rangelands (table 3). Figure 7 provides a graphical solution of the curve number equation used 688 to compute Q, runoff volume. The CN method depends on the determination of the curve 689 number index for a given watershed and given hydrologic condition. Hydrologic condition is 690 based on combination of vegetation and land surface characteristics as they affect infiltration. 691 The tables also demonstrate the variance of curve number with hydrologic soil group. 692 693

DRAFT Technical Note, April 2015 16 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

694 Table 3. Runoff curve numbers for arid and semiarid rangelands (USDA-NRCS 2004a) Hydrologic Cover type A B C D condition Herbaceous-mixture of grass, weeds, and poor 80 87 93 low-growing brush, with brush the minor fair 71 81 89 element good 62 74 85 Oak-aspen-mountain brush mixture of oak poor 66 74 79 brush, aspen, mountain mahogany, bitter fair 48 57 63 brush, maple, and other brush good 30 41 48 poor 75 85 89 Pinyon-juniper-pinyon, juniper, or both; fair 58 73 80 grass understory good 41 61 71 poor 67 80 86 Sagebrush with grass understory fair 51 63 70 good 35 47 55 Desert shrub-major plants include saltbush poor 63 77 85 88 greasewood, creosotebush, blackbrush, fair 55 72 81 86 bursage, palo verde, mesquite, and cactus good 49 68 79 84 695 Note: Average runoff conditions. (group A only developed for desert shrub) 696 Hydrologic condition defined as follows: 697 poor: < 30% ground cover (litter, grass, and brush overstory) 698 fair: 30 to 70% ground cover 699 good: >70% ground cover

DRAFT Technical Note, April 2015 17 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

700 Figure 7. Graphical solution of the CN runoff equation (USDA-NRCS 2004b)

701 702 Existing methods for adjusting runoff curve numbers to account for wildfire effects are for the 703 most part based heavily upon the hydrologic judgment of practitioners who have studied burned 704 watersheds. A number of publications, including Foltz, et al. (2009) have documented these 705 methods. For a synopsis of the state of the art, research issues related to post-wildfire runoff and 706 erosion processes, and suggested areas needing further work, see Moody, et al. (2013). 707 708 Adjustment of curve numbers for wildfire effects 709 One of the first efforts to establish some methodology for adjustment of runoff curve numbers 710 for the effects of wildfire was Cerrelli (2005). Studying Montana wildfire events in 2000 and 711 2001, Cerrelli created an Excel spreadsheet that computes watershed curve numbers for four 712 conditions: 1) pre-fire, 2) early post-fire with possible hydrophobicity, 3) medium term post-fire 713 with hydrophobicity no longer in effect, but little re-emergent vegetation, and 4) later post fire, 714 after one growing season. For high burn severity areas, the investigation included a table of 715 wildfire affected curve numbers, table 4. The Fire Hydrology, or FIRE HYDRO, spreadsheet 716 and user notes are available at ftp://ftp.wcc.nrcs.usda.gov/wntsc/H&H/wildfire/. 717 718

DRAFT Technical Note, April 2015 18 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

719 Table 4. Suggested runoff curve numbers for burned areas from Cerrelli (2005). Hydrologic soil group High burn severity CN

A 64 B 78 C 85 D 88 720 721 In addition, the guidance for moderate burn severity is to use the published curve number tables, 722 but with cover type in fair condition, and for low burn severity, to adjust CN to the aspect of the 723 fire-burned slopes. For low burn severity, with north and east facing slopes, the guidance is to 724 use cover type good condition, and for south and west facing slopes to use a curve number 725 between cover type fair and good condition. 726 727 Case study 2 of this Technical Note documents Cerrelli’s investigation, method development, 728 and use on the Montana Bitterroot wildfires of 2000. 729 730 The USDA Forest Service provides web-based guidance for determination of wildfire affected 731 runoff curve numbers, which includes Cerrelli (2005). (See: 732 http://forest.moscowfsl.wsu.edu/BAERTOOLS/ROADTRT/Peakflow/CN/. Tables 5 through 9 are from this 733 reference.) Tables 5 and 6 show post-fire runoff curve numbers suggested by two different 734 investigators (Story, 2003, and Stuart, 2000) for Forest Service Region 1. (See figure 8 for a 735 map of the USDA Forest Service Regions.) 736 737

DRAFT Technical Note, April 2015 19 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

738 Table 5. Forest Region 1 curve numbers (Story 2003) Post-fire condition CN range High burn severity, with hydrophobic soils 93-98 High burn severity, without hydrophobic soils 90-95 739 740 Table 6. Forest Region 1 curve numbers (Stuart 2000) Post-fire condition CN range Moderate burn severity 80 Low burn severity 70-72 Unburned 60-64 Moderate burn severity, with BAER treatments 75 Moderate burn severity, with hydrophobic soils 66 741 742 Figure 8. USDA Forest Service Regions 743 744 745 746 747 748 749 750 751 752 753 754 Tables 7 and 8 show CNs suggested by forest service hydrologists for Forest Service Region 3 755 (Arizona and New Mexico). 756 757

DRAFT Technical Note, April 2015 20 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

758 Table 7. Forest Region 3 curve numbers Post-fire condition CN range High burn severity, with hydrophobic soils 95 High burn severity, without hydrophobic soils 92 Moderate burn severity, with hydrophobic soils 89 Moderate burn severity, without hydrophobic soils 87 Low burn severity 80-83 Unburned 55-75 759 760 Table 8. Forest Region 3 curve numbers Post-fire condition CN range High burn severity, with hydrophobic soils 95 High burn severity, without hydrophobic soils 90-91 Moderate burn severity, with hydrophobic soils 90 Moderate burn severity, without hydrophobic soils 85

Low burn severity CNpre-fire + 5 Straw mulch with good coverage 60 Seeding with log erosion barriers, 1-year post-fire 75 Log erosion barriers without hydrophobic soils 85 761 762 Another investigation by the Forest Service produced a table comparing pre-fire to post-fire 763 runoff curve numbers for the 2003 Aspen Fire in Arizona. (See table 9.) Although these Forest 764 Service studies generally use NRCS hydrology models, case study 1 in this technical note 765 demonstrates the use of CNs within the HEC-HMS model. 766 767 Table 9. Forest Region 3 pre-fire and post-fire curve numbers.

CNpost-fire Hydrologic soil CNpre-fire Low burn Moderate burn High burn group severity severity severity B 56 65 -- -- C 67 70-75 80 90 D 77 80-85 90 95 768

DRAFT Technical Note, April 2015 21 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

769 Limitations of the CN method 770 The lack of field data in burned catchments and related research to verify the effect on CN 771 hampers post-fire runoff modeling with curve numbers. There is little research adequate to 772 determine best-fit runoff curve numbers for unburned mountain and forested watersheds. One 773 study of ten Appalachian forested watersheds, however, found that measured CNs varied 774 significantly from published suggested values (Tedela, et al. 2012). The study also found very 775 wide confidence limits for the measured values. One watershed, with a measured mean CN of 776 57, had 95% confidence interval of 32 to 83. Other studies also found inability to achieve a 777 stable CN value (Hawkins 1993, and Springer and Hawkins 2005). 778 779 Hydrologists modeling forested watersheds should understand the difference between two major 780 runoff-generating processes and their likelihood during storms of different magnitudes pre-fire 781 versus post-fire. Forested watersheds in unburned conditions may tend toward saturation-excess 782 overland flow, with runoff generated from relatively small fractions of the subareas, where 783 rainfall depths exceed the soil capacity to retain water. Newly burned watersheds, however, may 784 tend toward more significant infiltration-excess (also called Hortonian) overland flow, with 785 surface runoff due to rainfall intensities exceeding soil infiltration rate. See figure 9. 786 787 Figure 9. Runoff generating processes on a hillslope terminating at a watercourse (Garen and Moore 2005)

788

DRAFT Technical Note, April 2015 22 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

789 Time of concentration

790 Watershed time of concentration Tc is defined as the time required for runoff from the most

791 hydraulically remote point in the watershed to its outlet. In hydrograph analysis, Tc is considered 792 a representation of the time from the end of excess rainfall to the point of inflection on the 793 receding limb. Watershed lag time is defined as the time from the center of excess rainfall to the

794 hydrograph peak. Both lag and Tc are shown on the dimensionless unit hydrograph schematic, 795 Figure 16. These definitions and methods of computation are discussed in NRCS NEH Part 630, 796 Hydrology, Chapter 15, Time of Concentration, USDA-NRCS (2010). 797

798 The estimation of Tc is often accomplished using either the Watershed Lag Method or the 799 Velocity Method. 800 801 Watershed lag method 802 Time of concentration can be estimated using the following equation, called the Watershed Lag 803 Method.

. ( ) . 804 = 0 8 1000 . 0 7 (eq. 1) 𝑙𝑙 𝐶𝐶𝐶𝐶 −9 0 5 805 where: 𝑇𝑇𝑐𝑐 1140𝑌𝑌 806 Tc = time of concentration (hr) 807 l = flow length (ft) 808 CN = curve number (dimensionless) 809 Y = average land slope (percent) 810 811 Velocity method

812 Another way to estimate Tc described in NEH 630 Chapter 15 (USDA-NRCS, 2010) is the

813 velocity method. This method equates Tc to the sum of estimated flow times for three types of 814 flow from the hydraulically most distant point in the watershed, namely, sheet flow, shallow 815 concentrated flow, and channel flow. 816 817 The following equation for sheet flow travel time, provided in NEH 630 Chapter 15, is a 818 derivative of the kinematic wave equation:

DRAFT Technical Note, April 2015 23 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

819 . ( ) . 820 = (eq. 2) . . 0 8 0 007 𝑛𝑛𝑛𝑛 0 5 0 4 821 𝑇𝑇𝑠𝑠ℎ𝑒𝑒𝑒𝑒𝑒𝑒 𝑃𝑃 𝑆𝑆 822 where:

823 Tsheet = travel time (hr) 824 n = Manning’s roughness coefficient for overland flow, obtained from Table 10 825 l = sheet flow length (ft < 100) 826 P = 2-year, 24-hour rainfall (in) 827 S = land slope (ft/ft) 828 829 Table 10. Manning’s roughness coefficients for sheet flow (depth generally ≤ 0.1 ft) Mannings Surface description n value Smooth (concrete, asphalt, gravel, or bare soil) 0.011 Fallow fields (no residue) 0.05 Cultivated soil, with residue <= 20% 0.06 Cultivated soil, with residue > 20% 0.17 Short-grass prairie 0.15 Dense grasses 0.24 Bermuda grass 0.41 Natural rangeland 0.13 Woods with light underbrush 0.40 Woods with dense underbrush 0.80 830 831 For the shallow, concentrated flow and channel flow segments, the following equations may be 832 used. For shallow concentrated flow, velocity may be estimated using table 11. Channel 833 velocity may be estimated with Manning’s equation (equation 4). 834 lshallow lchannel 835 Tshallow = T channel = (eq. 3) 3600Vshallow 3600Vchannel 836

DRAFT Technical Note, April 2015 24 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

837 where:

838 Tshallow = travel time, shallow concentrated flow (hr)

839 Tchannel = travel time, channel flow (hr)

840 lshallow = flow length, shallow concentrated flow (ft, generally less than 1000)

841 lchannel = flow length, channel flow (ft)

842 Vshallow = flow velocity, shallow concentrated flow (ft/sec)

843 Vchannel = flow velocity, channel flow (ft/sec) 844 845 Table 11. Shallow-concentrated flow velocity equations, where S=flow slope in ft/ft Mannings Flow type Velocity equation (ft/sec) n value

Pavement and small upland gullies 0.025

Grassed waterways 0.050

Nearly bare/untilled or alluvial fan flow 0.051

Cultivated straight row crops 0.058

Short-grass pasture 0.073 Minimum tillage cultivation,

contoured or strip-cropped, or 0.101 woodlands

Forest with heavy ground litter or hay 0.202 meadows 846 847 1.486RS2/3 1/2 848 V = (eq. 4) channel n 849 where:

850 Vchannel = average cross-sectional flow velocity (ft/sec)

851 R = hydraulic radius (ft) = A/Pw, where: 852 A = cross-sectional flow area, ft2

853 Pw = wetted perimeter, ft 854 S = flow slope (ft/ft) 855 n = Manning’s n value for channel flow

DRAFT Technical Note, April 2015 25 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

856 857 Adjustment of runoff flow time for wildfire effects 858 For the Watershed Lag Method, post-wildfire effects are entirely accounted for by adjusting the 859 CN. See the section “Adjustment of curve number for wildfire effects.” 860 861 For the sheet flow portion of the Velocity Method, post-wildfire effects are entirely accounted 862 for using Manning’s roughness coefficient. The analyst must decide whether and how much to 863 reduce the n-value, based on reduction of vegetation caused by the fire. Using table 10, for 864 example, a pre-fire condition might justify an n-value of 0.13 or 0.15, associated with natural 865 range or short-grass prairie, reduced to a post-fire n-value of 0.05 or 0.011, associated with 866 fallow ground or smooth surface, (caused by total burning of vegetation). 867 868 Hydrologists should also keep in mind that Manning’s roughness coefficients are of a different 869 scale between overland flow and channel flow, due to the phenomenon of relative roughness, 870 which is a measure of flow depth (or hydraulic radius) over surface roughness height. Given a 871 homogeneous surface, the Manning’s roughness value will not necessarily be the same at all 872 depths. For very shallow flow, at or near the boundary roughness height, Manning’s n-value will 873 be larger. At depths much greater than roughness height, Manning’s n-value will tend to settle to 874 a single value (for a homogeneous surface). 875 876 Overland flow generally has a larger relative roughness than channel flow. The effects of 877 wildfire tend to lower both relative roughness and Manning’s n-value for overland flow, while 878 raising them for collector channels. Early post-fire overland flow roughness values may be set to 879 that of bare soil. Moody and Kinner (2006) suggest that wildfire basically results in overland 880 flow Manning’s roughness values near that of bare soil and that minor variation from this value 881 (0.011) has much less effect on model results than changes in effective rainfall. In addition, the 882 effect of hydrophobicity can be accounted for by lower overland flow roughness values, near 883 those of rough concrete. Canfield, et al. 2005 measured the runoff from a single hillslope with 884 varying roughness conditions and found the roughness effect significant (figure 10). 885 886

DRAFT Technical Note, April 2015 26 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

887 Figure 10. Hydrographs from the same hillslope, with different surface conditions (Canfield, et al. 2005).

888 889 Channels clogged with post-fire debris or sediment have higher Manning’s roughness values 890 than their pre-fire condition. The modeler should consider that, immediately post-fire, channels 891 have elevated roughness values due to transport of significant newly available sediment and 892 debris. With time, these channel roughness values change, although return to pre-fire roughness 893 may take many years. In addition, for several years after the fire, smaller channels, such as 894 collectors, tend to a more pronounced effect of relative roughness. Bed material such as cobbles 895 and debris result in Manning’s n values that decrease as flow depth increases. Riparian 896 vegetation is often extensive in the bed of intermittently flowing channels, and provides 897 additional flow resistance for the first seasonal flood. 898 899 Research has shown that pre-fire steep mountain streams have significantly higher roughness 900 values than lower elevation streams, with Manning’s n values often ranging from 0.1 to 0.2 901 (Yochum et al. 2012). These higher values are due to the presence of large cobbles and boulders, 902 relative to stream cross-section size, as well as step-pools, instream wood, and extreme 903 variability along the longitudinal profile. A single large boulder can sometimes so restrict flow 904 area that significantly greater out of bank flow occurs only for a short section. 905 906 References for selection of roughness values for high gradient streams include both photographic 907 guidance (Barnes 1967, Aldridge and Garrett 1973, Hicks and Mason 1998, Yochum and

DRAFT Technical Note, April 2015 27 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

908 Bledsoe 2010, Yochum et al. in press) and quantitative prediction tools (Limerinos 1970, Hey 909 1979, Jarrett 1984, Bathurst 1985, Yochum et al. 2012). 910 911 Hydrologic modeling of conditions both pre- and post-fire may benefit from a more extensive 912 examination of sub-basin drainage by development of time-area histograms. These are derived 913 using GIS, and can verify whether the estimated longest path in a partially burned watershed is 914 truly the longest, timewise, rather than lengthwise. The time-area histogram also helps 915 determine hydrograph peaking factors for the use of unit hydrograph transformations. (See the 916 Unit Hydrographs section.) 917 918 Infiltration parameters 919 Physically-based infiltration methods, such as GA, do not attempt to index overall watershed 920 conditions, but rather assume that physical processes at a given point on the landscape may be 921 applied at a wider spatial scale. The GA method is a simplification of infiltration equations that 922 incorporate soil characteristics, such as hydraulic conductivity, and physical aspects such as 923 pressure of surface water and matric suction of the soil (due to capillary action). The GA method 924 represents infiltration as a depth of wetted soil, with moisture gradually moving downward in the 925 soil column as a front, and that this movement can be estimated using soil porosity, soil 926 hydraulic conductivity, initial soil water content, and the soil matric suction at the wetted front. 927 Figure 11 shows the Green-Ampt conception of the soil column. 928 929 Figure 11. Green-Ampt infiltration into the soil column 930 931 932 933 934 935 936 937 938

DRAFT Technical Note, April 2015 28 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

939 The GA method is a simplification of Richard’s equation (USCOE-HEC, 2013) for infiltration 940 that has a limited number of parameters that can be reasonably estimated using soils data. As 941 discussed in the HEC-HMS User Manual (USCOE-HEC, 2013) GA infiltration is estimated 942 using equation 5, with parameters correlated with soil texture (Table 13). 943 +−φθ 944 1( if )S (eq. 5) fKt =  945 Ft 946 where:

947 ft = loss during time period, t 948 K = saturated hydraulic conductivity 949 φ = soil porosity (volume air/volume soil)

950 θi = initial soil water content (volume water/volume soil)

951 Sf = wetting front suction

952 Ft = cumulative loss at time t 953 954 The remaining parameter from equation 5 is initial soil moisture content, which ranges between 955 zero (a completely dry soil) and the effective porosity. The user estimates this value using an 956 antecedent soil moisture index and values from table 12 for field capacity and wilting point. 957 958

DRAFT Technical Note, April 2015 29 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

959 Table 12. Green Ampt parameters for soil texture classes & horizons (Rawls, et al. 1983, Saxton, et al. 1986)

texture horizon φeffective Ksat Sf θfieldCap θwiltingPt (cm3/cm3) (cm/hr) (cm) (cm3/cm3) (cm3/cm3) sand combined 0.417 23.56 4.95 0.13 0.06 A 0.431 -- 5.34 -- -- B 0.421 -- 6.38 -- -- C 0.408 -- 2.07 -- -- loamy sand combined 0.401 5.98 6.13 0.15 0.07 A 0.424 -- 6.01 -- -- B 0.412 -- 4.21 -- -- C 0.385 -- 5.16 -- -- sandy loam combined 0.412 2.18 11.01 0.20 0.09 A 0.469 -- 15.24 -- -- B 0.428 -- 8.89 -- -- C 0.389 -- 6.79 -- -- loam combined 0.434 0.68 8.89 0.26 0.12 A 0.476 -- 10.01 -- -- B 0.489 -- 6.40 -- -- C 0.382 -- 9.27 -- -- silt loam combined 0.486 1.30 16.68 0.29 0.11 A 0.514 -- 10.91 -- -- B 0.515 -- 7.21 -- -- C 0.460 -- 12.62 -- -- sandy clay loam combined 0.330 0.30 21.85 0.26 0.16 B 0.330 -- 26.10 -- -- C 0.332 -- 23.90 -- -- clay loam combined 0.309 0.20 20.88 0.33 0.19 A 0.430 -- 27.00 -- -- B 0.397 -- 18.52 -- -- C 0.400 -- 15.21 -- -- silty clay loam combined 0.432 0.20 27.30 0.37 0.19 A 0.477 -- 13.97 -- -- B 0.441 -- 18.56 -- -- C 0.451 -- 21.54 -- -- sandy clay combined 0.321 0.12 23.90 0.35 0.25 B 0.335 -- 36.74 -- -- silty clay combined 0.423 0.10 29.22 0.42 0.26 B 0.424 -- 30.66 -- -- C 0.416 -- 45.65 -- -- clay combined 0.385 0.06 31.63 0.48 0.35 B 0.412 -- 27.72 -- -- C 0.419 -- 54.65 -- -- 960 961

DRAFT Technical Note, April 2015 30 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

962 Adjustment of infiltration parameters for wildfire effects 963 When wildfire results in hydrophobic soil, the spatial extent is generally not over the entire 964 drainage area. The extent to which the hydrophobicity has reduced infiltration also varies from 965 place to place. Along with soil burn severity mapping, field investigation may give the 966 hydrologist considerable insight as to the extent of infiltration reduction caused by 967 hydrophobicity. However, since it is impossible to visit and test every square foot of ground, the 968 hydrologist must estimate the widespread effect of hydrophobicity, and the rate that infiltration 969 returns to normal as water repellency breaks down over time. 970 971 Two modeling options are available to account for hydrophobicity. The hydraulic conductivity 972 may be reduced (GA method) or the percentage of impervious area may be increased. Canfield, 973 et al. (2005), in their field studies of post-wildfire conditions, examined of the effect on saturated

974 hydraulic conductivity. The increase over time of Ksat (Figure 12) may be an indication of the 975 breakdown of hydrophobicity and return of vegetation. 976 977 Figure 12. Optimal hillslope hydraulic conductivity, Cerro Grande NM fire (Canfield, et al. 2005)

978 979 In the HEC-HMS model, both options are available in the same interface, as shown in Figure 13. 980

DRAFT Technical Note, April 2015 31 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

981 Figure 13. HEC-HMS Green and Ampt input interface (USCOE-HEC 2013)

982 983 Unit hydrographs 984 The process of converting a runoff volume to a hydrograph shape is called transformation. 985 Several of the computer models in table 2 perform the transformation of spatially distributed 986 runoff to an outlet using some variation of unit hydrograph (UH) theory. UH theory, advanced 987 by L.K. Sherman in the 1930s, suggests that the runoff behavior of a watershed can be typed, 988 given its physical characteristics, so that whenever a unit of direct runoff is produced in a 989 watershed over a given duration, the hydrograph peak and shape at the outlet will be that of the 990 unit hydrograph. If a rainfall event produces more than one unit of direct runoff, then the outlet 991 hydrograph can be computed by multiplying each ordinate of the UH by the amount of runoff in 992 that duration. 993 994 The following list details a number of assumptions inherent in the theory: 995 1. A given sub-basin has a different unit hydrograph for each possible duration of direct 996 runoff. 997 2. The effective rainfall, defined as the rainfall that becomes direct runoff occurs uniformly 998 over the sub-basin. 999 3. The shape and duration of the UH are independent of the effective rainfall intensity, 1000 which only effects the magnitude of flow in each UH time interval. 1001 4. If a given rainfall event is divided into subsequent 30-minute intervals of effective 1002 rainfall, a 30-minute UH is applied to each interval of effective rainfall, to obtain an 1003 outflow for that time interval. The total outflow is the sum of the runoff hydrographs 1004 produced from each time interval. This is called the superposition principle. (See figure 1005 14.)

DRAFT Technical Note, April 2015 32 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1006 5. The S curve, derived by summing a series of UHs of a given duration, each shifted by that 1007 duration, allows for the computation of runoff hydrographs from a single curve regardless 1008 of effective rainfall duration. (See figure 15.) 1009 1010 Figure 14. Runoff hydrographs by convolution and superposition, using 30-minute unit hydrograph. 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 Figure 15. S curve developed from 30-minute unit hydrograph. 1025 1026 1027 1028 1029 1030 1031 1032 1033

DRAFT Technical Note, April 2015 33 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1034 1035 NRCS dimensionless unit hydrograph 1036 As documented in the NRCS National Engineering Handbook, Part 630, Hydrology, Chapter 16, 1037 Hydrographs (USDA-NRCS 2007), the NRCS approach is to use a standard dimensionless 1038 curvilinear hydrograph estimation (figure 16). 1039 1040 Figure 16. Dimensionless unit hydrograph, with time of concentration and lag (USDA-NRCS 2010). 1041

1042 484AQ 1043 TT = 0.667 (eq. 6) (eq. 7) pc qp = 1044 0.667Tc 1045 Lag = Lag (hr)

1046 Tc = time of concentration (hr)

1047 Tp = time to peak (hr)

1048 Tr = time of recession (hr)

1049 Tb = time base of triangular approximation (hr) 1050 ΔD = duration of excess rainfall (hr)

DRAFT Technical Note, April 2015 34 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1051 t/Tp = ratio between time interval and time to peak (dimensionless) 1052 q = discharge rate at time t (ft3/sec) 3 1053 qp = peak discharge rate at time Tp, (ft /sec)

1054 Qa = runoff volume up to t (in) 1055 A = watershed drainage area (mi2) 1056 Q = total runoff volume (in) 1057 1058 The coefficient 484 in equation 7 is called the peak rate factor (PRF). The PRF is a ratio of the 1059 runoff under the rising limb of the unit hydrograph to the total time base. For the default PRF of 1060 484, the rising side of the dimensionless unit hydrograph represents 37.5 percent of the total 1061 runoff volume. 1062 1063 The PRF may be estimated using drainage times (equation 8) or the proportion of the watershed 1064 drained before the peak (equation 9). 1065 . . 1066 = (eq. 8) or = (eq. 9) ( ) 1290 67 𝑇𝑇𝑝𝑝 1290 67𝐴𝐴𝑟𝑟𝑟𝑟𝑟𝑟𝑟𝑟 1067 𝑃𝑃𝑃𝑃𝑃𝑃 𝑇𝑇𝑝𝑝+𝑇𝑇𝑟𝑟 𝑃𝑃𝑃𝑃𝑃𝑃 𝐴𝐴𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡 1068 where:

1069 Tp = time to peak = 1 hour for the dimensionless unit hydrograph

1070 Tr = time of recession

1071 Arise = basin area drained before peak

1072 Atotal = total basin drainage area 1073 1074 If the PRF of actual watersheds is determined using equation 8 or 9, the effect of overall basin 1075 slope becomes apparent. Flatter sloped ones with more storage effects tend to have lower PRFs, 1076 whereas steeper watersheds with fewer storage effects tend to have higher PRFs. For example, 1077 one Southwest Florida watershed had peak rate factors ranging from between 188 to 257 and 1078 another ranging from between 302 to 390 (Dendy 1987). On the other hand, Whitewater Creek 1079 in New Mexico (Case Study 4 in this Technical Note) was found to have PRFs over 700. 1080

DRAFT Technical Note, April 2015 35 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1081 Note that the curvilinear UH of figure 16 was developed graphically and not an equation, but is a 1082 smooth curve fitted to the triangular estimation, achieving a natural hydrograph shape. If a 1083 different PRF is used, the curvilinear UH has a different shape. The reference Appendix 16B 1084 (USDA-NRCS 2007) provides tables of dimensionless UHs for several other peak rate factors. 1085 1086 The HEC-HMS model gives the user several options for runoff transformation, including the 1087 NRCS dimensionless UH. However, the user should bear in mind that HEC-HMS has encoded 1088 only the default NRCS dimensionless UH, PRF 484, with no option to use a different PRF. 1089 1090 Adjustment of unit hydrograph for wildfire effects 1091 Generally, the hydrologist does not have the benefit data from studies and instrumentation from 1092 burned watersheds. The best option for estimation of timing is hydrologic judgement of changes 1093 in Manning’s roughness, changes in watershed storage, and changes in peak factor. Of these, 1094 general watershed storage effects may be the least important. Steeper mountainous watersheds 1095 do not tend to have significant wetland or ponding areas. If they do, then the effect of wildfire 1096 may be minimal, except that sedimentation would tend to somewhat reduce storage effects. The 1097 HEC-HMS user manual states, “…many studies have found that the storage coefficient, divided 1098 by the sum of the time of concentration and storage coefficient, is reasonably constant over a 1099 region.” However, the hydrologist working with a burned watershed will generally discover that 1100 no such regionally derived storage coefficients exist. 1101 1102 Changes in PRF may be more. Figure 17 shows the change in the 30-minute unit hydrograph at 1103 Baldy Creek (part of case study 4 of this Technical Note). The pre-fire PRF was already above 1104 600, but the wildire effect was to raise it further from 720 to 769. While the PRF can be 1105 estimated using times of concentration (equation 8) the availability of GIS layers may facilitate a 1106 better estimate using time-area histograms. 1107 1108

DRAFT Technical Note, April 2015 36 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1109 Figure 17. Baldy Creek pre-fire and post-fire 30-minute unit hydrographs. 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 The HEC-HMS model offers several synthetic UH transformation methods, which, similarly to 1123 the NRCS dimensionless unit hydrograph, require some estimate of the magnitude and timing of 1124 the UH peak. For the Snyder UH, for example, HEC-HMS requires input of standard lag and 1125 peaking coefficient. For the Clark UH, the model requires input of time of concentration and 1126 storage coefficient. For the NRCS dimensionless UH, time of concentration is required, along 1127 with peak discharge computed from equation 7 (with the option to use a different peak factor 1128 than the standard). 1129 1130 Similarly, to the NRCS peak factor, the HEC-HMS Snyder synthetic unit hydrograph peaking 1131 constant has no small effect on the hydrograph. The HEC-HMS user manual (USACE 2013) 1132 states that “it ranges from 0.4 to 0.8, with lower values associated with steep-rising 1133 hydrographs.” The user manual also states that the peaking constant “is estimated using the best 1134 judgment of the user, or possibly from locally-developed relationships to watershed physical 1135 features.” 1136 1137 While the NRCS hydrology program WinTR-20 facilitates the use of unit hydrographs up to 600 1138 in peak factor (see Appendix 16B of USDA-NRCS 2007), the HEC-HMS model accepts input of 1139 user-derived unit hydrographs and S-curves. See Case Study 4.

DRAFT Technical Note, April 2015 37 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1140 1141 Kinematic wave transformation 1142 The models KINEROS2 and AGWA perform runoff transformation using kinematic wave 1143 transform. HEC-HMS also offers the kinematic wave transform as an option. The kinematic 1144 wave model refers to an approximation of the full unsteady flow routing equations of Saint 1145 Venant. A number of variations on these equations exist based on ignoring terms that, in certain 1146 applications, are considered negligible. (See Ponce and Simons 1977.) The kinematic wave 1147 model retains only one of four terms, and thus cannot accomodate diffusion of hydrograph peaks. 1148 For watershed hydrology applications, the kinematic wave transform is considered perfectly 1149 adequate. For further discussion of applicablity, see the HEC-HMS Technical Reference Manual 1150 (USACE 2013). 1151 1152 The kinematic wave technique performs time-step calculations using differential equations for 1153 flow on various land surface types (overland, collector channels, and stream channels). Ponce 1154 (1991) suggested that the kinematic wave transform not be applied to sub-areas larger than one 1155 square mile because larger areas subject the differential equation solutions to artificial numerical 1156 effects. However, Woolhiser (1992) and Goodrich (1992) point out that difficulties with 1157 numerical stability do not render a model inapplicable to larger watersheds. Goodrich had, at the 1158 time, successfully modeled sub-basins up to 2.5 square miles. These discussions alert the user to 1159 be aware of numerical stability issues related to drainage area. 1160 1161 As discussed in the HEC-HMS Technical Reference Manual (USACE 2013), the equations used 1162 in the kinematic wave transform reduce to various approximations of Manning’s equation. The 1163 user, then, must provide estimations of n-value for the various surfaces (for which tables 10 and 1164 11 may be referenced). 1165 1166 Adjustment of kinematic wave transformation parameters for wildfire effects 1167 Whether direct runoff is transformed into an outlet hydrograph using unit hydrographs or 1168 kinematic wave equations, the hydrologist must assess the post-fire change in runoff travel time. 1169 The alteration of Mannings roughness coefficient is one consideration. Hydrophobicity may also 1170 speed up overland flow runoff. By limiting infiltration, water repellency deepens the flow on the

DRAFT Technical Note, April 2015 38 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1171 overland surface, while also tending to smooth it. Transport of loose soil above the hydrophobic 1172 layer, however, tends to raise roughness coefficient. 1173 1174 Time-area histogram synthetic unit hydrograph development 1175 Unit hydrographs may be developed using streamflow and precipitation records, but gages for 1176 the pre-fire condition may not exist. Even if they do, the records will not facilitate development 1177 of a post-fire UH. Another option is to develop a synthetic unit hydrograph using estimated 1178 time-area histograms of runoff from the burned watersheds. The procedure can also be used to 1179 estimate peak rate factors or to develop dimensionless unit hydrographs for use in NRCS 1180 computer models. 1181 1182 To develop this terrain-based synthetic unit hydrograph, the user creates GIS raster information 1183 that ultimately allows the estimation of time-bands of raster cells, each band defining the area of 1184 a sub-basin that drains to the outlet in the same time interval. To obtain flow velocity, the user 1185 develops a Manning’s roughness raster, based on whether the cell contains overland, shallow- 1186 concentrated, or channel flow (figure 18). Other factors taken into account in the derivation of 1187 the roughness layer are soil types, vegetation types, and burn severity. Figure 19 shows a sub- 1188 basin raster of high burn severity locations. A flow slope raster can also be developed from the 1189 DEM layer (figure 20). Details on the creation of these raster layers with ArcMap (ESRI 2010) 1190 is provided with Case Study 4, Whitewater Creek, Gila Wilderness New Mexico. 1191

DRAFT Technical Note, April 2015 39 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1192 Figure 18. Example of an n-value raster, with progressively higher roughness with darker color 1193

1194 Figure 19. Example of high burn severity raster (red pixels) 1195 1196 Figure 20. Example of flow slope raster, progressively steeper with darker color

1197 1198 Flow velocity relationships may be derived from equations 2, 3, and 4 above. Case Study 4 1199 shows the use of similar flow velocity equations developed with reference to kinematic wave 1200 routing theory. Note that Manning’s roughness values vary significantly between overland flow 1201 and channel values, as shown in Tables 4 and 5. Further guidance on selection of post-wildfire

DRAFT Technical Note, April 2015 40 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1202 roughness coefficients is provided in the section on adjustment of runoff flow time for wildfire 1203 effects. 1204 1205 In addition, the Manning’s equation for stream velocity estimation (equation 4) is dependent on 1206 flow cross-sectional area (within the hydraulic radius term). One way to estimate this parameter 1207 for any location in a watershed is to use hydraulic geometry relations of, say, bankfull flow area 1208 as a function of drainage area. Case study 4 provides further details on these procedures. Since 1209 a hydrograph runs through a range of flow depths and velocities, the use of bankfull parameters 1210 may result in low velocity estimates if the runoff being modeled is a rare event. Recall, however, 1211 that UH theory assumes the timing is identical no matter the intensity of effective rainfall. 1212 1213 From the velocity raster a drainage time raster can be created to determine the number of cells 1214 that drain to the outlet in successive time bands of, say, five minutes. This velocity raster enables 1215 the derivation of a time-area histogram of the watershed. Standard hydrology textbooks such as 1216 Bedient, Huber, and Vieux (2012) show how a unit hydrograph can be easily created in a 1217 spreadsheet, using a time-area histogram. Case Study 4 demonstrates the derivation of synthetic 1218 unit hydrographs by this methodology. 1219 1220 Sedimentation estimation 1221 Estimation of flow rates and velocities of clear streamflow (free of sediment and debris) benefits 1222 from the fact that liquid water has very stable properties, such as density, viscosity, 1223 incompressibility, and homogeneity. Predicting the movement of woody debris is difficult 1224 because the size, shape, and availability of any given mass is random. With sediment, unit 1225 weight is not highly variable, but the range of particle sizes and cohesiveness is extremely wide, 1226 from micron sized cohesive clays, through the range of silts and sands, to enormous boulders. 1227 The susceptibility of gravels and cobbles to the lift and drag forces of streamflow partly depend 1228 on imbrication, or shielding, provided by other larger particles. Once in motion, the distance a 1229 particle will travel depends on the energy of the stream current, which varies longitudinally with 1230 drainage slope and transversely in a flow cross-section both horizontally and vertically. 1231

DRAFT Technical Note, April 2015 41 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1232 Increased sediment transport by floodwater after watershed burn is common and often damaging, 1233 as shown in figure 21. The steepness of mountain streams provides energy to move sediment. 1234 Whenever some physical attribute of the stream lessens its ability to carry sediment, the stream 1235 responds by depositing whatever it can no longer transport. Figure 22 shows the result of a steep 1236 stream emerging from a canyon onto a plain. No longer confined by canyon walls, the stream 1237 may break out in various directions across an alluvial fan. In any of its chosen paths cross the 1238 alluvial fan, the sediment-transport capability is greatly reduced, compared to that in the canyon 1239 from which it emerged. Resulting deposition creates the fan itself, but the location of flood 1240 channels from any given event varies significantly over time. 1241 1242 Figure 21. 2010 post-wildfire debris flow from San Gabriel Mountains, CA

1243 (photo by Robert Looper, USGS) 1244

DRAFT Technical Note, April 2015 42 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1245 Figure 22. Alluvial fan, Death Valley National Park

1246 (photo by Marli Bryant Miller, marlimillerphoto.com) 1247 1248 Rivers in unburned watersheds transport sediment via natural channel processes, depending on 1249 stream type. A number of stream classification systems exist, which evaluate river 1250 geomorphology based on channel size, shape, slope, and bed material. While the velocity and 1251 sediment transport capacity of a stream can vary significantly along its profile, it can also vary 1252 considerably within a single cross-section. The fastest flow is not usually on the surface, but at 1253 about three-quarters of the maximum depth and over the thalweg. (See figure 23.) Velocity is 1254 reduced toward the flow boundaries , especially in the presence of vegetation. 1255 1256 Figure 23. A stream cross-section with isolines of velocity, maximized over the thalweg.

1257 1258 Velocity currents flow not only longitudinally, downstream, but also circulate sideways, in what 1259 are termed transverse or secondary currents. This phenomenon is responsible for the meandering 1260 behavior of rivers (figure 24), as downwelling tends to scour out pools on the outside of bends, 1261 while gravel and sand deposition forms bars on the inside of bends.

DRAFT Technical Note, April 2015 43 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1262 1263 Figure 24. Riffle-pool morphology

1264 1265 Wildfire destroys the erosion protection of vegetation, leaving watersheds vulnerable to not only 1266 elevated sediment-transport levels in otherwise clear streamflow, but also to debris flows and 1267 mudflows. Vegetation removal leaves overland flow slopes highly vulnerable to erosion; swales 1268 and gullies clog with sediment and debris; and hydrophobic soil at a depth of as little as two 1269 inches leaves the unconsolidated soil above it vulnerable to erosive flow. With wildfire, not only 1270 is sediment more vulnerable to erosion, but debris such as logs and partially burned branches are 1271 often washed into the streams. Sometimes the flowing water is so clogged with sediment, that 1272 the flowing mass is termed mudflow or debris flow. Having a higher density and viscosity than 1273 clear water flow, debris flows can be extremely dangerous. 1274 1275 Santi, et al. 2006 define debris-flow as “the rapid flow of saturated material consisting of more 1276 than 20% gravel and coarse material through a steep channel or over steep hillsides.” Because 1277 the volume of debris flows can often be correlated with property damage and other hazards, a 1278 number of studies have examined size classification (Jakob, 2005) and prediction (Santi and 1279 Morandi, 2012; Prochaska, et al., 2008). See table 13. As relates to post-wildfire recovery in the 1280 Western US, Santi and Morandi (2012) state: “...there is a clear progression in decreasing 1281 volume of debris flows as basins recover from the wildfire: it takes approximately 1 year, or at a

DRAFT Technical Note, April 2015 44 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1282 few locations, as much as 3 years, for debris production to return to pre-fire rates.” Note that 1283 accelerated sediment transport rates, as opposed to debris-flows, may persist much longer. 1284 1285 Table 13. Size classification and potential consequences of debris flows (Santi and Morandi, 2012) Size Volume Potential consequences class (103 m3) 1 < 0.1 Localized damage, known to have killed foresters in small gullies, damage small buildings 2 0.1 - 1 Could bury cars, destroy small wooden buildings, break trees, block culverts, derail trains 3 1 - 10 Could destroy larger buildings, damage concrete bridge piers, damage highways or pipelines 4 10 - 100 Could block creeks, destroy parts of villages, destroy bridges, damage infrastructure corridors 5 100 - Ccould destroy larger urban areas, dam up creeks and small rivers 1000 6 - 10 > 5000 Generally restricted to volcanic events, capable of destroying entire cities, damming large rivers 1286 1287 Another finding from Santi, et al. (2006) is that “the majority of material in post-fire debris flows 1288 is eroded from the channels—only a small percentage of the total volume is contributed from 1289 hillslope rilling and sheetwash.” Again, the analyst must distinguish between the much more 1290 sediment-bulked debris flow events and rainfall-runoff events with elevated sediment transport. 1291 The latter may well originate on hillslopes subjected to vegetation removal and hydrophobic 1292 soils associated with wildfire. 1293 1294 Concerning sediment, the design or evaluation of engineering structures in wildfire-impacted 1295 areas may take a number of approaches. Hydrographs modeled as clear flow are often increased 1296 by considering the bulking effect of entrained sediment. A specific sediment-transport computer 1297 model may be used to estimate locations and magnitudes of sediment aggradation or degradation. 1298 Debris flow or mud flow may be estimated or modeled. HEC-RAS has sediment transport 1299 modeling capability, but the available empirical sediment transport equations come with

DRAFT Technical Note, April 2015 45 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1300 significant uncertainties and generally require calibration. Considerable field measurements of 1301 sediment supply and channel geometry are required. 1302 1303 Although the two-dimensional flood-routing model FLO-2D (O’Brien, et al. 1993) is used to 1304 model debris flow (Elliott, et al. 2005) rapid post-wildfire assessments generally preclude its use 1305 due to extensive input requirements and model complexity. For estimation of debris flow 1306 impacts, the analyst should consider storm event frequency, source material availability, and 1307 location of concern relative to location of source material. Modeling debris flow generally 1308 requires a less frequent (higher rainfall peak) storm, such as a ten-year event, although 1309 immediately after the wildfire a 2-year event may suffice. Availability of source material can be 1310 assessed by field observations of channel conditions, primarily, but also steep hillslope 1311 conditions. The further the location of concern from the source material the less likely the effect 1312 of debris flow (as differentiated from sediment-bulked flow). Cannon, et al. (2003), for example, 1313 found that relative degree of debris flow hazard could be assessed by examining the abundance 1314 of colluvium available in stream channels, the degree of channel confinement, and channel 1315 gradient. The study concluded that further research is needed. 1316 1317 Models for estimating sediment transport 1318 For estimating elevated sediment transport, as opposed to debris-flow, three models are 1319 discussed GeoWEPP, AGWA, and RUSLE2. 1320 1321 GeoWEPP is a geo-spatial interface for the Water Erosion Prediction Project (WEPP) model 1322 (Flanagan, et al. 2001). GeoWEPP for ArcGIS is available from: 1323 https://lesami.geog.buffalo.edu/projects/active/geowepp/. The WEPP model itself is available c: 1324 http://www.ars.usda.gov/Research/docs.htm?docid=10621. WEPP is a continuous simulation process- 1325 based model, applicable to hillslope erosion processes (sheet and rill erosion). It uses 1326 stochastically derived climate data from the CLIGEN model (Zhang and Garbrecht, 2003). 1327 WEPP has two distinct disadvantages. First, WEPP is designed for continuous simulation. As a 1328 result, it is less applicable to storm events, which tend to drive post-fire runoff and sediment 1329 transport. Second, existing climate generator models such as CLIGEN and GEM6 are not

DRAFT Technical Note, April 2015 46 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1330 designed to produce less frequent storm event peaks (Theurer, et al. 2010) but rather the day-to- 1331 day kinds of rainfall events evaluated by continuous simulation models. 1332 1333 KINEROS2 (A Kinematic Runoff and Erosion Model, Canfield, Goodrich, and Burns 2005) is 1334 an event oriented, physically based model describing the processes of interception, infiltration, 1335 surface runoff and erosion from small agricultural watersheds. KINEROS2 may be run 1336 independently, or as an option within the Automated Geospatial Watershed Assessment 1337 (AGWA) modeling tool (Goodrich, et al. 2005). AGWA is a GIS-based hydrologic modeling 1338 tool using standardized datasets to develop input parameter files for KINEROS2 and SWAT. 1339 The KINEROS model is applicable to storm events. and AGWA has had recent improvements 1340 for post-wildfire use (Canfield and Goodrich, and Burns, 2005). The KINEROS2 model is 1341 available as a separate download from the USDA-Agricultural Research Service (ARS), 1342 Southwest Watershed Research Center (SWRC) at: http://www.tucson.ars.ag.gov/kineros/. 1343 AGWA modeling tool and documentation are available from USDA-ARS-SWRC at: 1344 http://www.tucson.ars.ag.gov/agwa/. See Case study 5 for an example of the implementation of 1345 AGWA. 1346 1347 Another model sometimes used for post-fire sedimentation estimation is RUSLE2. This 1348 planning-level erosion model, according to the ARS documentation (USDA-NRCS 2013d), is 1349 used to estimate rill and interrill erosion where mineral soil is exposed to the erosive forces of 1350 impacting raindrops, water falling from vegetation, and surface runoff produced by Hortonian 1351 overland flow. The document states that RUSLE2 is applicable to forestlands; however, the 1352 model is intended to estimate annual average erosion from the above sources rather than event- 1353 based erosion. In fact, long-term datasets of stochastically generated climate variables used in 1354 the model tend to represent day-to-day conditions rather than the rarer storm events such as that 1355 of a 100-year recurrence (Theurer et al. 2010). The simplicity of the model and its 1356 conduciveness to spatial estimation using GIS, make RUSLE2 a possible tool for evaluating 1357 sedimentation due to wildfire. See case study 3 of this technical note for example. Theobald et 1358 al. (2010) documents the GIS manipulations involved and Litschert et al. (2014) applies 1359 RUSLE2 and GIS to wildfire in the Southern Rockies Ecoregion. 1360

DRAFT Technical Note, April 2015 47 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1361 Adjustments for modeling sediment transport from burned areas 1362 Regardless of which estimation methods or models are applied, the modeler is advised to pay 1363 attention to burn severity. One study (Benavides-Solorio and MacDonald, 2001) in Colorado 1364 found relatively small differences in runoff rates as a function of burn severity, but a much 1365 higher correlation for sediment yield. That study stated, “...percent ground cover accounted for 1366 81% of the observed variability in sediment yields” and “...large differences in sediment yields 1367 with burn severity should be attributed primarily to the differences in ground cover rather than 1368 the differences in runoff, water repellency, or antecedent soil moisture.” These results imply that 1369 watershed recovery and return of sediment yield to unburned rates should be a function of 1370 vegetation and ground cover, and the study indicates that full recovery may take from three to 1371 nine years. 1372 1373 Sediment bulking 1374 The post-fire hydrographs produced by the models previously discussed do not account for the 1375 phenomenon of sediment bulking. As defined above, debris flow is considered that which 1376 entrains at least 20% gravel and coarse material. Elliott, et al. (2005) delineates 1377 hyperconcentrated flows as ranging from 20 to 47 percent sediment concentration and mudflows 1378 as having a greater than 47% sediment concentration. 1379 1380 A flood of rare recurrence interval, such as the 50-year event, occurring shortly after wildfire, 1381 may cause hyperconcentrated sediment flows or mudflows. One study (Cannon, et al. 2003) 1382 found significant post-wildfire debris flows occurring with storms as frequent as the two-year 1383 event. The study suggests that the generation of high sediment concentrated flows is possible by 1384 progressive sediment bulking over time. Another study (Giraud and McDonald, 2007) found 1385 that debris flows tend to result from channel incision rather than hillslope erosion, but agree with 1386 Cannon, et al. (2003) that smaller events, merely bulked rather than hyperconcentrated. 1387 1388 Post-wildfire flows that do not reach the 20% threshold nevertheless transport considerable 1389 sediment loads. This entrained sediment raises (or bulks) the overall flow rate and, at any 1390 particular location, raises the flood depth. For a hydrograph, the total bulked flow rate is the sum 1391 of the water discharge and the sediment discharge.

DRAFT Technical Note, April 2015 48 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1392 1393 = + (eq. 9) 1394 𝑄𝑄𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏𝑏 𝑄𝑄𝑤𝑤 𝑄𝑄𝑠𝑠𝑠𝑠𝑠𝑠 1395 One way to account for sediment bulking in post-wildfire hydrograph analyses is to apply a 1396 bulking factor to the flood peak (O’Brien and Fullerton 1989, Elliott et al. 2005): 1397

1398 = = (eq. 10) 𝑄𝑄𝑤𝑤+𝑄𝑄𝑠𝑠𝑠𝑠𝑠𝑠 1 1399 𝐵𝐵𝐵𝐵 𝑄𝑄𝑠𝑠 1−𝐶𝐶𝑣𝑣

1400 where Cv is the maximum sediment concentration by volume in percent. 1401 1402 As discussed in Elliott, et al. (2005), if the event contains a sediment concentration just under 1403 hyperconcentrated (20 percent) then the resulting bulking factor is 1.25: 1404

1405 = = 1.25 (eq. 11) . 1 1406 𝐵𝐵𝐵𝐵 1−0 2 1407 Sediment concentration varies during a storm and generally peaks before that of the water 1408 discharge. One way researchers estimate sediment discharge over time is to measure data and 1409 derive simple regression equations. For example, Moody and Martin (2001) found, for two 1410 recently burned Colorado watersheds, 1411 . 1412 = 4.4 (eq. 12a) 1. 5 1413 𝑄𝑄𝑠𝑠𝑠𝑠𝑠𝑠 = 23 𝑄𝑄𝑤𝑤 (eq. 12b) 1 3 1414 where: 𝑄𝑄𝑠𝑠𝑠𝑠𝑠𝑠 𝑄𝑄𝑤𝑤 3 1415 Qw = water flow (m /s)

1416 Qsed = sediment flow (kg/s) 1417 1418 The study reported statistical correlations of r2=0.89 for equation 21a and r2=0.96 for equation 1419 21b. 1420

DRAFT Technical Note, April 2015 49 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1421 Using a hypothetical storm hydrograph, with a peak of about 580 cfs in the second watershed, 1422 and translating both discharges into cfs, a bulk factor for the peak of 1.06 is derived (figure 25). 1423 1424 Figure 25. Hypothetical storm using equation 21b (all discharges converted to cfs).

1425 1426 For post-fire hydrologic analyses, if previous studies for the watershed of concern exist, then 1427 maximum sediment concentrations and bulk factors are better estimated. The bulk factor can be 1428 much higher than the hypothetical one from above. For example, LACDPW (2006) has 1429 compiled curves for debris production and bulk factors for the Los Angeles River, Santa Clara 1430 River, and Antelope Valley watersheds, with wide variation from place to place. Figure 26 1431 shows the data for one of their lower debris production areas with bulking factors in the range of 1432 1.5. 1433 1434 A factsheet from the USGS (Pierson, 2005) provides guidance on field investigations to 1435 distinguish between debris flows and normal flow sediment deposits. The reference suggests 1436 that normal suspended sediment concentrations are five to ten percent by volume, but that even 1437 with hyperconcentrated flow (volumes between 20 and 60 percent) the flow behavior is

DRAFT Technical Note, April 2015 50 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1438 controlled by water, whereas mud flows (even higher sediment concentrations) have behavior 1439 controlled by the entrained sediment. 1440 1441 Among the field indicators listed are, for water flow deposits, most grains are rounded, beds are 1442 stratified with good sorting both horizontally and vertically, with loose consistency when dry. 1443 Debris flow deposits, on the other hand, tend to be angular sand and fine gravel (indicating a 1444 hillslope source), non-stratified, and extremely poorly sorted, with coherent consistency rather 1445 than loose. 1446 1447 Note that when using post-wildfire streamflow records to calibrate modeled hydrographs these 1448 measurements are already bulked. 1449 1450 Figure 26. Sediment production and bulking factors for debris production area 7 (LACDPW 2006)

1451

DRAFT Technical Note, April 2015 51 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1452 Model uncertainties, calibration, validation 1453 The ideal watershed for which to model the rainfall-runoff process is one with an extensive 1454 spatially-distributed data collection network. If instrumented, the instrumentation would produce 1455 an historic record of rainfall amounts in time and space, soil moisture levels, baseflow levels, and 1456 sedimentation rates. However, even a spatially dense network of instruments; say every half 1457 mile, leaves considerable uncertainty, since conditions at points other than the station locations 1458 would have to be assumed. As discussed in Beven (2001), for example, the use of distributed 1459 hydrologic models for prediction of future states suffers a number of limitations, namely 1460 problems of non-linearity, scale, uniqueness, equifinality, and uncertainty. Other researchers 1461 suggest, more generally, that verification and validation of numerical models of natural systems 1462 is impossible (Oreskes, Shrader-Frechette, and Belitz, 1994). 1463 1464 In the real world, the number of data collection stations in a watershed never comes close to true 1465 spatial adequacy. The watershed model is inherently an open system, subject to equifinality— 1466 that a given end state (say, a certain magnitude storm runoff peak at the outlet) can be caused by 1467 many potential combinations of circumstances (say, storms of similar intensity but different areal 1468 extent, centered over different points in the basin). Even if such an extensive data station 1469 network existed, the radical and sudden change wrought by wildfire renders the historical dataset 1470 inapplicable. 1471 1472 Although calibration and validation are be possible with post wildfire hydrologic modeling, the 1473 hydrologist should at least report sources of known uncertainty. If hydrophobicity exists, the 1474 hydrologist should report the evidence of spatial extent or persistence of significant water 1475 repellent layering, and by what means they accounted for it in the model. If data stations are 1476 installed after a fire, this recorded evidence of post-fire events should be used to show the 1477 reasonability of the hydrologic model. Such data does not generally support claims of 1478 calibration or validation, since scarcity of stations leaves open the possibility of more than one 1479 spatial layout of storms having the same impact at the stations. However, one or more 1480 hypothetical storm distributions, which reproduce the measured data could be modeled, to show 1481 whether the model responds reasonably. Case study 4 demonstrates this approach. 1482

DRAFT Technical Note, April 2015 52 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1483 References 1484 Aldridge, B. N., and J. M. Garrett, 1973. Roughness coefficients for stream channels in Arizona. 1485 U.S. Geological Survey Open-File Report, Tucson, Arizona. 1486 1487 Bathurst, J.C. 1985. Flow resistance estimation in mountain rivers. Journal of Hydraulic 1488 Engineering 111(4), 625-643. 1489 1490 Bedient, P.B., Huber, W.C., and Vieux, B.E. 2012. Hydrology and Floodplain Analysis, Prentice 1491 Hall, NJ, 5th Edition, 801 pages. 1492 1493 Benavides-Solario, J.D., and MacDonald, L.H. 2001. Post-fire runoff and erosion from simulated 1494 rainfall on small plots, Colorado Front Range. Hydrological Processes, 15, 2931-2952. 1495 1496 Beven, K.J. 1989. Changing ideas in hydrology -- the case of physically-based models. Journal 1497 of Hydrology 105, 157-172. 1498 1499 Beven, K.J. 2001. How far can we go in distributed hydological modelling? Hydrology and Earth 1500 System Sciences 5, 1-12. 1501 1502 Canfield, H.E., Goodrich, D.C. 2005. Suggested changes to AGWA to account for fire (V. 2.1). 1503 Southwest Watershed Research Center, USDA-ARS, Tucson, Arizona. 1504 1505 Canfield, H.E., Goodrich, D.C., and Burns, I.S. 2005. Selection of parameters values to model 1506 post-fire runoff and sediment transport at the watershed scale in Southwestern forests. 1507 1508 Cannon, S.H., Gartner, J.E., Parrett, C., and Parise, M. 2003. Wildfire-related debris-flow 1509 generation through episodic progressive sediment-bulking processes, western USA. In 1510 Proceedings: Debris-Flow Hazards Mitigation: Mechanics, Prediction, and Assessment. 1511

DRAFT Technical Note, April 2015 53 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1512 Cerrelli, G.A. 2005. FIRE HYDRO, a simplified method for predicting peak discharges to assist 1513 in the design of flood protection measures western wildfires. In Proceedings of the ASCE 2005 1514 Watershed Management Conference, July 2005, Williamsburg VA. pp. 935-941. 1515 1516 CFR (Code of Federal Regulations), 2005. Emergency watershed protection program. 7 CFR 1517 Part 624.4. 1518 1519 Chen, L., Berli, M., Karletta, C. 2013. Examining modeling approaches for the rainfall-runoff 1520 process in wildfire-affected watersheds: using San Dimas Experimental Forest. Journal of the 1521 American Water Resources Association 1-16. 1522 1523 Conedera, M., Peter, L., Marxer, P., Forster, F., Rickenmann, D., Re, L. 2003. Consequences of 1524 forest fires on the hydrogeological response of mountain catchments: a case study of the Riale 1525 Buffaga, Ticino, Switzerland. Earth Surface Processes and Landforms 28, 117-129. 1526 1527 Debano, L.F., Osborn, J.F., Krammes, J.S., Letey, J. 1967. Soil wettability and wetting agents: 1528 our current knowledge of the problem. USDA Forest Service Research Paper PSW-43: 13. 1529 1530 DeBano, L.F. 2000a. Water repellency in soils: a historical overview. Journal of Hydrology, 1531 231-232, 4-32. 1532 1533 DeBano, L.F. 2000b. The role of fire and soil heating on water repellency in wildland 1534 environments: a review. Journal of Hydrology, 231-232, 195-206. 1535 1536 DeBano, L.F. 1981. Water repellent soils: a state-of-the art. USDA Forest Service, General 1537 Technical Report PSW-46. 1538 1539 Dendy, G.S. 1987. A 24-hour rainfall distribution and peak rate factors for use in Southwest 1540 Florida. Masters Thesis, University of Central Florida, Orlando Florida, 172 pages. 1541

DRAFT Technical Note, April 2015 54 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1542 Doerr, S.H. 1998. On standardizing the ‘Water Drop Penetration Time’ and the ‘Molarity of an 1543 Ethanol Droplet’ techniques to classify soil hydrophobicity: A case study using medium textured 1544 soils. Earth Surface Processes and Landforms 23, 663-668. 1545 1546 Doerr. S.H., Shakesby, R.A., Walsh, R.P.D. 2000. Soil water repellency: its causes, 1547 characteristics and hydro-geomorphological significance. Earth Science Reviews, 51, 33-65. 1548 1549 Dyrness, C.T. 1976. Effect of wildfire on soil wettability in the high cascades of Oregon. USDA 1550 Forest Service, Pacific Northwest Research Station, Research Paper PNW-202, 19 p. 1551 1552 Elliott, J.G., Smith, M.E., Friedel, M.J., Stevens, M.R., Bossong, C.R., Litke, D.W., Parker, R.S., 1553 Costello, C., Wagner, J., Char, S.J., Bauer, M.A., and Wilds, S.R. 2005. Analysis and mapping 1554 of post-fire hydrologic hazards for the 2002 Hayman, Coal Seam, and Missionary Ridge 1555 wildfires, Colorado. USGS Scientific Investigations Report 2004-5300, 80 pages. 1556 1557 Engman, E.T. 1986. Roughness coefficients for routing surface runoff. Journal of Irrigation and 1558 Drainage Engineering 112 (1). Amer. Soc. of Civil Eng., New York, NY. pp. 39-53. 1559 1560 ESRI (Environmental Systems Resource Institute). 2010. ArcMap 10.0 ESRI, Redlands, 1561 California. 1562 1563 Fang, X., etal. 2005. Revisit of NRCS unit hydrograph procedures. In Proceedings of the ASCE 1564 Texas Section Spring Meeting, Austin TX, April 2005, 21 pages. 1565 1566 Flanagan, D.C., etal. 2001. Chapter 7: The Water Erosion Prediction Project (WEPP) model. In 1567 Landscape Erosion and Evolution Modeling. Kluwer Academic Publishers, Norwell MA. 51 1568 pages. 1569 1570 Folmar, N.D., and Miller, A.C. 2008. Development of an empirical lag time equation. ASCE 1571 Journal of Irrigation and Drainage Engineering 112(1), pp. 39-53. 1572

DRAFT Technical Note, April 2015 55 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1573 Foltz, R.B., Robichaud, P.R., and Rhee, H. 2009. A synthesis of post-fire road treatments for 1574 BAER teams: methods, treatment effectiveness, and decisionmaking tools for rehabilitation. 1575 USDA Forest Service General Technical Report RMRS-GTR-228. 1576 1577 Garen, D.C., Moore, D.S. 2005 Curve number hydrology in water quality modeling: uses, abuses 1578 and future directions. Journal of the American Water Resources Association 42(2), 377-388. 1579 1580 Giraud, R.E., and McDonald, G.N. 2007. The 2000-2004 fire-related debris flows in northern 1581 Utah. In: Proceedings 1st North American Landslide Conference, AEG Special Publication 23, 1582 1522-1531. 1583 1584 Goodrich, D.C. 1992. Discussion of “The kinematic wave controversy” by Victor M. Ponce 1585 (April, 1991, Vol. 117, No. 4). ASCE Journal of Hydraulic Engineering 118(9): 1334-1335. 1586 1587 Goodrich, D.C., etal. 2005. Rapid post-fire hydrologic watershed assessment using the AGUA 1588 GIS-based hydrologic modeling tool. In Proceedings of the ASCE 2005 Watershed Management 1589 Conference, July 2005, Williamsburg VA. pp. 1-12. 1590 1591 Hawkins, R.H., Woodward, D.E., Hjelmfelt, A.T., Van Mullem, J.A., and Quan, Q.D. 2002. 1592 Runoff curve number method: examination of the initial abstraction ratio. Proceedings Federal 1593 Interagency Hydrologic Modeling Conference, Las Vegas NV. 12pp. 1594 1595 Hawkins, R.H. 1993 Asymptotic determination of runoff curve numbers from data. Journal of 1596 Irrigation and Drainage Engineering 119, 334-345. 1597 1598 Hawkins, R.H., Greenberg, R.J. 1990. WILDCAT4 flow model. School of Renewable Natural 1599 Resources, University of Arizona, Tucson, AZ. BLM contact: Dan Muller, Denver, CO. 1600 1601 Henkle, J.E., Wohl, E., Beckman, N. 2011. Locations of channel heads in the semiarid Colorado 1602 Front Range, USA Geomorphology 129, 309-319. 1603

DRAFT Technical Note, April 2015 56 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1604 Hey, R.D. 1979. Flow resistance in gravel bed rivers, ASCE Journal of the Hydraulics Division, 1605 105(4), 365-379. 1606 1607 Hicks, D.M., Mason, P.D. 1998. Roughness characteristics of New Zealand rivers, 2nd Edition, 1608 Water Resource Publications. 1609 1610 Hudak, A.T., Robichaud, P.R., Evans, J.S., Clark, J., Lannom, K., Morgan, P., Stone, C. 2004. 1611 Field validation of burned area reflectance classification (BARC) products for post fire 1612 assessment. Remote Sensing for Field Users: Proceedings of the Tenth Forest Service Remote 1613 Sensing Applications Conference, Salt Lake City, Utah, April 5-9 2004. 1614 1615 Huffman, E.L., MacDonald, L.H., and Stednick, J.D. 2001. Strength and persistence of fire- 1616 induced soil hydrophobicity under ponderosa and lodgepole pine, Colorado Front Range. 1617 Hydrologic Processes, 15, 2877-2892. 1618 1619 Jakob, M. 2005. A size classification for debris flows. Engineering Geology, 79, 151-161. 1620 1621 Jarrett, R.D. 1984. Hydraulics of high-gradient streams. Journal of Hydraulic Engineering 110 1622 (11), 1519–1539. 1623 1624 Keeley, J.E. 2009. Fire intensity, fire severity and burn severity: a brief review and suggested 1625 usage. International Journal of Wildland Fire, 18, 116-126. 1626 1627 LACDPW (Los Angeles County Department of Public Works), 2006. Sedimentation Manual, 2nd 1628 Edition. 1629 1630 LandFire 2010. LandFire Fact Sheet, http://www.landfire.gov/. 1631 1632 Lawrence, D. S. L. 1997. Macroscale surface roughness and frictional resistance in overland 1633 flow. Earth Surface Processes and Landforms 22: 365-382. 1634

DRAFT Technical Note, April 2015 57 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1635 Limerinos J.T. 1970. Determination of the Manning coefficient from measured bed roughness in 1636 natural channels. U.S. Geological Survey Water Supply Paper 1898-B, 52 pages. 1637 1638 Litschert, S.E., Theobald, D.M., and Brown, T.C. 2014. Effects of climate change and wildfire 1639 on soil loss in Southern Rockies Ecoregion. Catena 118, 206-219. 1640 1641 Livingston, R.K., Earles, T.A., and Wright, K.R. 2005. Los Alamos post-fire watershed 1642 recovery: a curve-number-based evaluation. In Proceedings of the ASCE 2005 Watershed 1643 Management Conference, July 2005, Williamsburg VA. pp. 471-481. 1644 1645 MacDonald, L.H., Sampson, R.N., Brady, D., Juarros, L., and Martin, D. Predicting Post-Fire 1646 Erosion and Sedimentation Risk on a Landscape Scale: A Case Study from Colorado, pp. 57-87. 1647 Chapter 4 in Mapping Wildfire Hazards and Risks, edited by Sampson, R.N., Atkinson, R.D., 1648 and Lewis, J.W. Journal of Sustainable Forestry, Vol. 11, 1/2, 2000. 1649 1650 Maidment, D.R. 1992. Handbook of Hydrology McGraw-Hill, Inc. 1651 1652 McCuen, R.H. and Bondelid, T.R. 1983. Estimating Unit Hydrograph Peak Rate Factors. ASCE 1653 Journal of Irrigation and Drainage Engineering 109(2), 238-250. 1654 1655 Miller, J.D., and Thode, A.E. 2007. Quantifying burn severity in a heterogeneous landscape with 1656 a relative version of the delta Normalized Burn Ratio (dNBR). Remote Sensing of Environment 1657 109, 66-80. 1658 1659 Moody, J.A., etal. 2013. Current research issues related to post-wildfire runoff and erosion 1660 processes. Earth Science Reviews, (in publication). 1661 1662 Moody, J.A., and Martin, D.A. 2009. Synthesis of sediment yields after wildfire in different 1663 rainfall regimes in the western United States. International Journal of Wildland Fire, 18, 96-115. 1664

DRAFT Technical Note, April 2015 58 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1665 Moody, J. A., and Kinner, D.A. 2006. Spatial structure of stream and hillslope drainage networks 1666 following gully erosion after wildfire. Earth Surface Processes and Landforms 31(3): 319-337. 1667 1668 Moody, J.A., and Martin, D.A. 2001. Hydrologic and sedimentologic response of two burned 1669 watersheds in Colorado, USGS Water-Resources Investigations Report 01-4122, 146 pages. 1670 1671 NASA 2012. NOAA National Aeronautics and Space Administration , Water and Energy Cycle 1672 website: http://science.nasa.gov/earth-science/oceanography/ocean-earth-system/ocean-water- 1673 cycle/ 1674 1675 NWCG 2006. Glossary of wildland fire terminology. National Wildfire Coordinating Group, 1676 Incident Operations Standards Working Team. 1677 1678 O'Brien, J. S., Julien, P.Y., and Fullerton, W.T. (1993). Two-dimensional water flood and 1679 mudflow simulation. ASCE Journal of Hydraulic Engineering 119(2): 244-261. 1680 1681 O’Brien, J.S., and Fullerton, W.T. 1989. Hydraulic and sediment transport study of Crown 1682 Pointe flood wall: Breckenridge, Colo., Lenzotti and Fullerton Consulting Engineers, Inc., 34 p. 1683 1684 Oreskes, N., Shrader-Frechette, K., and Belitz, K. (1994) Verification, validation, and 1685 confirmation of numerical models in the earth sciences. Science 263, 641-646. 1686 1687 Parsons, A., Lewis, S.A., Napper, C., and Clark, J.T. 2010. Field guide for mapping post-fire soil 1688 burn severity. USDA Forest Service General Technical Report RMRS-GTR-243, 49 pages. 1689 1690 Pierson, T.C. 2005. Distinguishing between debris flows and floods from field evidence in small 1691 watersheds. USGS factsheet 2004-3142. 1692 1693 Ponce, V.M. 1991. The kinematic wave controversy. ASCE Journal of Hydraulic Engineering 1694 117(4): 511-525. 1695

DRAFT Technical Note, April 2015 59 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1696 Ponce, V.M., and Simons, D.B. 1977. Shallow wave propogation in open channel flow. ASCE 1697 Journal of Hydraulic Engineering 103: 1461-1476. 1698 1699 Prochaska, A.B., Santi, P.M., Higgins, J.D., and Cannon, S.H. 2008. Debris-flow runout 1700 predictions based on the average channel slope (ACS). Engineering Geology, 98, 29-40. 1701 1702 Rallison, R.E. 1980. Origin and evolution of the SCS runoff equation. Proceedings of the 1703 Symposium on Watershed Management; American Society of Civil Engineers, Boise, ID. 1704 1705 Rallison, R.E., and Miller, N. 1982. Past, present, and future SCS runoff procedure. In Rainfall- 1706 Runoff Relationships, V.P. Singh, ed., Water Resources Publ., Littleton, CO, pp. 353-364. 1707 1708 Rawls, W.J., Brakensiek, D.L., and Miller, N. 1983. Green-Ampt infiltration parameters from 1709 soils data. ASCE Journal of Hydraulic Engineering. 109: 62-70. 1710 1711 Robichaud, P.R., Lewis, S.A., Ashmum, L.E. 2008. New procedure for sampling infiltration to 1712 assess post-fire soil water repellency. USDA Forest Service, Rocky Mountain Research Station, 1713 Research Note RMRS-RN-33. 1714 1715 Safford, H.D., etal. 2008. BAER Soil Burn Severity Maps Do Not Measure Fire Effects to 1716 Vegetation: A Comment on Odion and Hanson (2006). Ecosystems, 11:1, 1-11. 1717 1718 Santi, P. and Morandi, L. 2012. Comparison of debris-flow volumes from burned and un-burned 1719 areas. Landslides, 9, 1-13. 1720 1721 Santi, P. etal. 2006. Evaluation of post-wildfire debris flow mitigation methods and development 1722 of decision-support tools. Report to USGS and the Joint Fire Science Program, 78 pages. 1723 1724 Saxton, K.E., Rawls, W.J., Romberger, J.S., and Papendick, R.I. (1986) Estimating generalized 1725 soil-water characteristics from texture. Transactions of the American Society of Agricultural 1726 Engineers 50(4): 1031-1035.

DRAFT Technical Note, April 2015 60 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1727 1728 Scott, D.F. 2000. Soil wettability in forested catchments of South Africa; as measured by 1729 different methods and as affected by vegetation cover and soil characteristics. Journal of 1730 Hydrology 231-232, 87-104. 1731 1732 Smith, R.E., Goodrich, D.C., Woolhiser, D.A., Unkrich, C.L. 2005. KINEROS – A kinematic 1733 runoff and erosion model. In: Computer Models of Watershed Hydrology, V.J. Singh (Editor). 1734 Water Resource Publications, Highland Ranch, Colorado, pp. 597-632. 1735 1736 Springer, E.P., Hawkins, R.H. 2005. Curve number and peakflow responses following the Cerro 1737 Grande Fire on a small watershed. In: Moglen, G.E., ed. Proceedings of Managing Watersheds 1738 for Human and Natural Impacts: Engineering, Ecological, and Economic Challenges, July 19-22, 1739 2005, Williamsburg, VA, USA. American Society of Civil Engineers, Alexandria, VA. 1740 1741 Stoof, C.R., Vervoort, R.W., Iwema, J., van den Elsen, E., Ferreira, A.J.D., Ritsema, C.J. 2011. 1742 Hydrological response of a small catchment burned by an experimental fire. Hydrology and 1743 Earth System Sciences Discussions 8, 4053-4098. 1744 1745 Story, M. 2003. Cited by USDA Forest Service BAER Tools web page: 1746 http://forest.moscowfsl.wsu.edu/BAERTOOLS/ROADTRT/Peakflow/CN/, Accessed June 2013. 1747 1748 Stuart, B. 2000. Maudlow Fire, Burned Area Emergency Rehabilitation (BAER) plan. Townend 1749 MT: USDA Forest Service, Northern Region, Helena National Forest.Cited by USDA Forest 1750 Service BAER Tools web page: 1751 http://forest.moscowfsl.wsu.edu/BAERTOOLS/ROADTRT/Peakflow/CN/, Accessed June 2013. 1752 1753 Tedela, N.H., McCutcheon, S.C., Rasmussen, T.C., Hawkins, R.H., Swank, W.T., Campbell, J.L, 1754 Adams, M.B., Jackson, C.R., and Tollner, E.W. 2012. Runoff curve numbers for 10 small 1755 forested watersheds in the mountains of the Eastern United States. ASCE Journal of Hydrologic 1756 Engineering, 17, 1188-1198. 1757

DRAFT Technical Note, April 2015 61 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1758 Theobald, D.M., Merritt, D.M., and Norman, J.B., III. 2010. Assessment of threats to riparian 1759 ecosystems in the Western U.S. A report presented to The Western Environmental Threats 1760 Assessment Center, Prineville, OR, by the USDA Stream Systems Technology Center and 1761 Colorado State University, Fort Collins, 61p. 1762 1763 Theurer, F.D., Moore, D.S., Merkel, W.H., Quan, Q.D., Moody, H.F., Bingner, R.L., Bonta, J., 1764 and Flanagan, D. 2010. Comparison of historical versus synthetic weather inputs to watershed 1765 models and their effect on pollutant loads. In Proceedings of the 2nd Joint Federal Interagency 1766 Conference, Las Vegas NV, 12pp. 1767 1768 USCOE-HEC (U.S. Army Corps of Engineers, Hydrologic Engineering Center), 2013. Hec- 1769 HMS. Available at http://www.hec.usace.army.mil/software/hec-hms/. Accessed in April, 2013. 1770 1771 USDA-ARS (U.S. Department of Agriculture-Agricultural Research Service), 2013a. AGNPS. 1772 National Sedimentation Laboratory, ARS, USDA, Oxford MS. Available at http://go.usa.gov/KFO. 1773 Accessed in April, 2013. 1774 1775 USDA-ARS (U.S. Department of Agriculture-Agricultural Research Service), 2013b. Soil and 1776 Water Assessment Tool: SWAT. Grassland Soil and Water Research Laboratory, ARS, USDA, 1777 Temple TX. Available at http://swat.tamu.edu/. Accessed in April, 2013. 1778 1779 USDA-ARS (U.S. Department of Agriculture-Agricultural Research Service), 2013c. The Soil 1780 Infiltration Model in KINEROS2. Southwest Watershed Research Center, Tucson AZ. 1781 Available at http://www.tucson.ars.ag.gov/kineros/. Accessed in April, 2013. 1782 1783 USDA-ARS (U.S. Department of Agriculture-Agricultural Research Service), 2013d. Science 1784 Documentation Revised Universal Soil Loss Equation Version 2 (RUSLE2) ARS Washington 1785 DC. 355pp. 1786

DRAFT Technical Note, April 2015 62 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1787 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 2004a. 1788 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 9, Hydrologic soil cover 1789 complexes. Available at http://go.usa.gov/T3eV/. Accessed in April 2013. 1790 1791 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 1792 2004b. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 10, Estimation of 1793 direct runoff from storm rainfall. Available at http://go.usa.gov/T3eV/. Accessed in April 2013. 1794 1795 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 2004c. 1796 WinTR-20 User Guide. Available at http://go.usa.gov/KoZ/. Accessed in April 2013. 1797 1798 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 2009a. 1799 Small Watershed Hydrology WinTR-55 User Guide. Available at http://go.usa.gov/KoZ/. Accessed 1800 in April 2013. 1801 1802 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 1803 2009b. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 7, Hydrologic Soil 1804 Groups. Available at http://go.usa.gov/T3eV/. Accessed in April 2013. 1805 1806 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 2010. 1807 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 15, Time of 1808 Concentration. Available at http://go.usa.gov/T3eV/. Accessed in April 2013. 1809 1810 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation Service), 2007. 1811 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 16, Hydrographs. 1812 Available at http://go.usa.gov/T3eV/. Accessed in April 2013. 1813 1814 USDA-SCS (U.S. Department of Agriculture-Soil Conservation Service), 1991. Engineering 1815 Field Handbook, Chapter 2, Estimating Runoff. Available at 1816 ftp://ftp-fc.sc.egov.usda.gov/NHQ/pub/outgoing/jbernard/CED-Directives/efh/EFH-Ch02.pdf/. Accessed in 1817 April 2013.

DRAFT Technical Note, April 2015 63 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions

1818 1819 Woolhiser, D.A. 1992. Discussion of “The kinematic wave controversy” by Victor M. Ponce 1820 (April, 1991, Vol. 117, No. 4). ASCE Journal of Hydraulic Engineering 118(9): 1337-1339. 1821 1822 Yochum, S., Bledsoe, B. 2010. Flow resistance estimation in high-gradient streams. 2nd Joint 1823 Federal Interagency Conference, Las Vegas, NV, June 27- July 1. 1824 1825 Yochum, S.E., Bledsoe, B.P., David, G.C.L., Wohl, E. 2012. Velocity prediction in high-gradient 1826 channels. Journal of Hydrology. 424-425, 84-98, doi:10.1016/j.jhydrol.2011.12.031. 1827 1828 Yochum, S.E., Comiti, F., Wohl, E., David, G.C.L., Mao, L. In Press. Photographic guidance for 1829 selecting flow resistance coefficients in high-gradient channels. U.S. Department of Agriculture, 1830 Forest Service, Rocky Mountain Research Station, General Technical Report. 1831 1832 Zhang, X.C. and Garbrecht, J.D. 2003. Evaluation of CLIGEN precipitation parameters and their 1833 implication on WEPP runoff and erosion prediction. In Transactions of the American Society of 1834 Agricultural Engineers, 46, 311-320. 1835

DRAFT Technical Note, April 2015 64 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1836 Case Study 1: High Park Fire Colorado 1837 1838 Background 1839 The High Park Fire, in the foothills west of Fort Collins, Colorado (figures CS1-1 and 1840 CS1-2), burned within an area of 88,600 acres between June 9th and July 1st, 2012. The 1841 fire burned primarily in steep, forested terrain, with average slopes ranging from 14 to 49 1842 percent and elevations ranging from 5,300 to 10,200 feet. It was a fast burning fire, with 1843 about 37,000 acres burned in the first three days. It was also a “dirty burn,” with a 1844 substantial amount of unburned area (21,100 acres, 23%) distributed as a patchy mosaic 1845 throughout the fire extent. Of the impacted area, 9,800 acres burned at high soil burn 1846 severity (11%), 36,300 acres burned at moderate severity (40%), and 21,300 acres burned 1847 at low severity (23%). Just as wildfire containment was achieved, the summer monsoon 1848 season started. Increased flooding and debris flows were observed in streams draining 1849 numerous portions of the fire, with local residents noting that some of these floods were 1850 the most severe since the Big Thomson Flood of 1976. Ash mobilized due to the 1851 enhanced runoff flowed into the Cache la Poudre River, a primary source of drinking and 1852 irrigation water for the northern Colorado Front Range. Since part of the Horsetooth 1853 Reservoir watershed was also burned, water supply storage in this reservoir was 1854 threatened with contamination. As expected, the most severe flooding was in catchments 1855 with the highest percentages of high soil burn severity. In these areas the vegetation cover 1856 and soil litter were consumed, leaving surfaces dominated by a bare mineral condition, 1857 reduced surface roughness, and possible hydrophobicity. These high severity burn areas 1858 also likely had substantial destruction of seed banks, forcing longer vegetative recovery 1859 rates and resulting in longer periods of enhanced flooding. Increased flood peaks, flow 1860 volumes, sediment transport, nutrient enrichment, and stream channel destabilization are 1861 expected for a number of years in many streams draining the fire area, threatening life, 1862 property, infrastructure, and water quality. 1863 1864 An initial hydrologic analysis was provided by a Burned Area Emergency Response 1865 (BAER) report for the large catchments draining this fire. Local and state officials, 1866 however, concerned with increased flood hazard potential and noting the preliminary

DRAFT Technical Note, April 2015 C1-1 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1867 nature of the BAER report, requested assistance from the Natural Resources 1868 Conservation Service (NRCS). Of particular interest were flow frequency estimates for 1869 post fire conditions. This case study provides an 1870 overview of the analyses performed to provide these estimates. 1871 1872 Figure CS1- 1. Sunset through smoke plumes, from Fort Collins on Day 1 of the High Park Fire.

1873 1874 Figure CS1- 2. Aerial extent and soil burn severity of the High Park Fire, based on a BARC image.

DRAFT Technical Note, April 2015 C1-2 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1875 1876 Methods 1877 A rainfall-runoff model was developed to simulate the expected runoff response for both 1878 pre- and post-fire conditions, for the 2-year, 10-, 25-, 50-, and 100-year 1-hour rainfall 1879 events. Hydrologic modeling was performed using the program HEC-HMS (version 3.5), 1880 developed by the U.S. Army Corps of Engineers Hydrologic Engineering Center. The 1881 NRCS curve number (CN) technique for estimating direct runoff from rain events was 1882 used in this analysis. As quality control, peak flows estimated using U.S. Geological 1883 Survey (USGS) regression equations (Capesius and Stephens 2009), embedded in USGS 1884 Streamstats, were compared to the CN runoff results for unburned conditions. 1885 1886 The BAER assessment led by the U.S. Forest Service (USFS) provided initial peak flow 1887 estimates used for post-fire flood mitigation planning activities (BAER 2012). This 1888 hydrologic analysis used the WILDCAT model (Hawkins and Greenberg 1990), which 1889 also relies on the CN method. The analysis was constrained to a deadline of one week 1890 after fire containment, per USFS requirements. Hence, the modeling was relatively coarse 1891 and neglected relevant hydrologic mechanisms such as variability in lag time between 1892 pre- and post-fire conditions, vegetation type, differences in runoff between moderate and 1893 high soil burn severity areas, as well as flood attenuation in larger catchments. 1894 Additionally, the BAER modeling used relatively large catchments in some areas, which, 1895 given the large burn area of the High Park Fire, added to the courseness of the BAER 1896 results. For example, the modeling was less able to account for flood wave attenuation. 1897 NRCS initiated a more detailed hydrologic analysis of streams draining the High Park 1898 Fire area, likely enhancing the accuracy of predictions for flood mitigation efforts. Initial 1899 results of this modeling were obtained for key catchments just before the monsoon season 1900 started in early July, which provide preliminary flooding estimates to local and state 1901 officials. These results were refined throughout the summer and the analysis expanded to 1902 provide flood estimates for most streams draining the fire extent (Yochum 2012a, 1903 Yochum 2012b). While the results provide reasonable estimates of the increased flooding 1904 expected from the burned watershed, further monitoring and research underway in some 1905 areas may enable future modeling refinements. These could employ, for example, more

DRAFT Technical Note, April 2015 C1-3 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1906 mechanistic rainfall-runoff simulations than the CN method. Hence, post-wildfire flood 1907 simulation modeling, especially large events involving substantial private land 1908 ownership, more dense settlement, and higher risks to life, property, and infrastructure, 1909 should be viewed as an iterative process with models of varying complexity developed 1910 over time to satisfy specific needs. 1911 1912 The NRCS model estimated runoff using the CN procedure documented in Estimation of 1913 Direct Runoff from Storm Rainfall (USDA-NRCS 2004b). The CN is a simple 1914 catchment-scale method that gives estimates of flood peaks and volumes at watershed 1915 outlets, with more accurate results expected for larger, higher-intensity rain events. The 1916 method is also discussed in Hydrologic Soil Cover Complexes (USDA-NRCS 2004b), 1917 Rallison (1980), and numerous other publications. However, little quantitative 1918 information has been published of the database on which it was developed (Maidment 1919 1992) and many of the curves used in the development have been lost (Woodward 2005). 1920 In general, the method was developed for rural watersheds in various parts of the United 1921 States, within 24 states, was developed for single storms, not continuous or partial storm 1922 simulation, and was not intended to recreate a specific response from an actual storm 1923 (Rallison, 1980). 1924 1925 Figure CS1- 3. Modeled catchments and stream channels, with flow computation points with soil 1926 burn severity as background.

DRAFT Technical Note, April 2015 C1-4 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1927 1928 Catchments and modeled stream channels implemented in the analyses are presented in 1929 figure CS1-3. Overall, 105 delineated catchments were utilized in the modeling. The 1930 average catchment area was 1.6 square miles, with the individual sizes determined by 1931 required computation points (to provide needed flood predictions at roadway crossing, 1932 homes at risk, etc.), catchment morphologic characteristics, and the need for simulating 1933 flood routing and attenuation in the modeled stream channels. 1934 1935 Runoff Curve Number (CN) Estimation 1936 Curve numbers were assigned to the modeled catchments according to hydrologic soil 1937 group, vegetative type, soil burn severity, and ground cover condition (percent cover). 1938 The average catchment CN was computed using an aerial averaging methodology in GIS, 1939 with more than 51,000 polygons computed for the entire modeled extent. 1940 1941 Soil burn severity (figure CS1-3) is the principle driver for increasing flow in runoff 1942 predictions. For this modeling, soil burn severity was measured using the Burned Area 1943 Reflectance Classification (BARC) process from satellite data collected on 7/20/2012, by 1944 researchers at Colorado State University (CSU). BARC uses reflectance recorded in 1945 satellite images to quantify soil burn severity. For defining soil burn severity, BARC 1946 images have the advantage of being comprehensive and relatively-rapidly developable. 1947 However, comparison with field-collected data has indicated that this remotely-sensed 1948 product can be more indicative of post-fire vegetative condition than soil condition, 1949 especially in low to moderately burned areas (Hudak et al. 2004). Qualitative field 1950 assessment of the High Park Fire BARC image indicated a reasonable prediction of burn 1951 severity in high and moderate areas. The soil burn severity imagery depicted in figure 1952 CS1-3 was developed by the USGS Earth Resources Observation and Science Center 1953 (from the same 7/20 satellite data), rather than the CSU interpretation implemented in the 1954 hydrologic modeling. This BARC interpretation was not available at the time of the 1955 analysis, though the burn severity estimates are relatively comparable and substantial 1956 variations in flood peaks are not expected. 1957

DRAFT Technical Note, April 2015 C1-5 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1958 Hydrologic soil group (HSG) classification was selected using soils data published in the 1959 NRCS SSURGO (Soil Survey Geographic) database. Two soil surveys cover the fire 1960 extent: NRCS Larimer County survey (CO644), published in 1980; and USFS Arapahoe- 1961 Roosevelt survey (CO 645), published in 2001. The USFS survey covers the western 1/3 1962 of the fire area. 1963 1964 Vegetation type, from SWReGAP (Southwest Regional Gap Analysis Project) land cover 1965 mapping, was included in the CN assignments used for the modeling. The dominant 1966 vegetation types within the fire boundary were lodgepole, mixed conifer, ponderosa, 1967 shrubs and grass. 1968 1969 The assigned CN values are provided in table CS1-1. Using a fair ground cover 1970 condition, NRCS recommended values (USDA-NRCS 2004a) were applied by 1971 hydrologic soil group for unburned conditions. The CN values for low, moderate and 1972 high severity burn areas, at the four hydrologic conductivity classifications, were 1973 estimated based on CN values developed from post-fire runoff measurements (Livingston 1974 et al. 2005), with additional guidance from Wright et al. (2005) and Goodrich et al. 1975 (2005). 1976 1977 Table CS1- 1. CN assignments implemented in High Park Fire hydrologic modeling. (Highlighted 1978 columns indicate values extracted from USDA-NRCS 2004a. 1979

Description A HSG B HSG C HSG D HSG Cover Cover Unburned Unburned Unburned Unburned Moderate Moderate Moderate Moderate High High High High Low Low Low Low

Herbaceous, Pasture, Alpine Meadow, Park 49 55 67 77 61* 68* 80 86 74* 81* 88 89 82* 86* 92* 95 Oak-aspen—mountain brush mixture of oak brush, aspen, mountain mahogany, bitter brush, maple, and 45* 52* 65 77 48 55 65 86 57 70* 80* 89 63 70 80 92 other brush Ponderosa pine-juniper (grass understory) 49* 57* 65 77 58 65 75 86 73 78* 83* 89 80 85 90 92 Sagebrush (grass understory) 46* 54* 65 77 51 60 75 86 63 70 80* 89 70 75 85 92 Lodgepole Pine Forest 49* 57* 65 77 60 65 70 86 73 78* 83* 89 79 83* 87* 92 Bare soil 77 77 77 77 86 86 86 86 91 91 91 91 94 94 94 94 Wetland 98 98 98 98 98 98 98 98 98 98 98 98 98 98 98 98 * Updated value based on Black Forest and West Fork Complex Wildfires. Not as implimented in High Park analysis.

DRAFT Technical Note, April 2015 C1-6 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 1980 Rainfall 1981 Since the High Park Fire area is most susceptible to flooding from relatively short 1982 duration monsoonal rain events, a 1-hour storm duration was implemented. Rainfall 1983 depths were extracted from NOAA Atlas 2, Volume 3 (Miller et al. 1973) for the 2-, 10-, 1984 25-, 50- and 100-year rainfall events. The rainfall duration and distribution was identical 1985 to that used in the BAER modeling (BAER 2012), and is provided in the project report 1986 (Yochum 2012b). For catchments with drainages areas ≥ 6 square miles, an areal 1987 reduction factor was applied as detailed in Miller et al. (1973). Reduction varied from 1988 0.95 to 0.78. When applied, this area reduction was implemented in all catchments; flow 1989 may be underpredicted in the smaller, upper catchments of such drainages. 1990 1991 Lag Time 1992 Lag time (L) was computed using the watershed lag method (USDA-NRCS 2010). This

1993 equation is: 0.7 lS0.8 ( +1) 1994 L = (eq. CS1-1) 1900Y 0.5 1995 1996 where l is flow length (ft), Y is average watershed land slope (%), and S is maximum 1997 potential retention (in). S may be calculated from equation CS1-2: 1998 1000 1999 S = −10 (eq. CS1-2) cn' 2000 2001 where cn’ is the retardance factor and is approximately equal to the CN. This method 2002 allows the computation of differing lag times for pre- and post-fire simulations, reflecting 2003 the actual physical mechanism of more rapid flow response during post-fire conditions. 2004 The method was developed under a wide range of conditions, including steep, heavily 2005 forested watersheds (USDA-NRCS 2010). Note that equation CS1-1 is equivalent to the 2006 time of concenration equation in the Technical Note main body (equation 1) since lag is 2007 considered to be 0.6 of the time of concentration. 2008

DRAFT Technical Note, April 2015 C1-7 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2009 Flow Routing 2010 A Muskingum-Cunge procedure was used to route flow from upper catchments to stream 2011 outlets, along the modeled stream reaches (figure CS1-3). This 1-dimensional method 2012 simulates flow attenuation but does not provide a numerical solution of the full unsteady 2013 flow routing equations, as provided in such computational models as HEC-RAS. Instead, 2014 in each reach flow routing was estimated using a single simplified cross section, channel 2015 slope, and Manning’s n estimate. Photographic guidance (Yochum and Bledsoe 2010) 2016 was used to help select flow resistance coefficients, and these values were checked by 2017 inspecting the model solutions to verify that the selected Manning’s n values resulted in 2018 subcritical or approximately critical flow velocity. Hence, it was assumed that bedform 2019 development prevents reach-average supercritical flow in these alluvial channels. 2020 2021 Sediment Bulking 2022 A simple multiplication factor was applied to the post-fire flood predictions to account 2023 for sediment bulking in the debris flows. For burned catchments, this multiplication 2024 factor was assumed to be 1.25 if the (severe + moderate) soil burn severity aerial extent 2025 was greater than 50%, and 1.1 for catchments with between 10 and 50 % (severe + 2026 moderate) soil burn severity. 2027 2028 StreamStats 2029 The regional USGS regression equations for peak flow prediction (Capesius and 2030 Stephens 2009), embedded in Streamstats, were used to assess the reasonableness of pre- 2031 fire peak flow predictions. The predictions use input of drainage area and the 6-hour, 2032 100-year precipitation depth to provide expected runoff from rain events, reflecting the 2033 expectation that floods result from summer monsoons. The error bars associated with 2034 these predictions are substantial – typically about 140 percent. 2035 2036 These predictions are based on streamgage data and, hence, provide a level of ground 2037 truthing, but this method accounts for only drainage area and precipitation regime. Other 2038 physical characteristics and processes that are relevant in runoff processes, such as 2039 infiltration capacity, vegetative type, ground cover condition, watershed shape, and flow

DRAFT Technical Note, April 2015 C1-8 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2040 attenuation, are not accounted for. However, due to its foundation in field-collected data 2041 within similar watersheds, this method is valuable for assessing the general 2042 reasonableness of the model predictions for pre-fire conditions. 2043 2044 Results and Discussion 2045 Hydrologic modeling was performed to develop estimates of increased flood hazard and 2046 potential threats to life and property along streams draining the High Park Fire. For 2047 several key catchments, at locations of threats to residences, example hydrographs (figure 2048 CS1-4) show the expected response to a 10-year rainfall depth over each entire 2049 catchment. Substantially increased flow peaks and runoff volumes were estimated. The 2050 simplicity of a map presentation style (figure CS1-5) facilitated use by planners, 2051 designers, and emergency response officials. Results were provided at 96 computation 2052 points over the fire extent and also provided as attributes in ArcGIS shapefiles. In 2053 addition to simulated flood peaks, time-to-peak estimates were provided to give 2054 emergency response personnel estimates of the expected flood response times. For 2055 detailed results, refer to Appendix A of the project report (Yochum 2012b) and the 2056 accompanying poster (Yochum 2012a). 2057 2058 Substantially higher peak flows and flood volumes are predicted for post-fire conditions. 2059 In many catchments, post fire conditions were predicted to cause a 50- or 100-year flood 2060 (pre-fire recurrence) to result from a 10-year rain event on burned landscapes, similar to 2061 measured fire runoff responses (Conedera et al. 2003). If it is estimated that the fire 2062 impacts on runoff in each of these catchments will be substantial for at least 5 years, the 2063 risk of a 10-year rainfall event over each point in these catchments over those 5 years of 2064 destabilization is 41 percent, with resulting 50- to 100-year floods (pre-fire recurrence). If 2065 a ten year recovery period is assumed, the risk increases to 65 percent. However, the 2066 expected severity of flooding lowers with increasing catchment size, due to the the small 2067 spatial extent of typical convective storms. 2068

2069 Due to their simplicity, peak flow enhancement ratios (Qpost/Qpre) can be a preferred 2070 method for communicating flood enhancement predictions. Predicted ratios for the 25-

DRAFT Technical Note, April 2015 C1-9 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2071 year rain event are provided (figure CS1-6). Peak flow enhancement ratios are higher for 2072 more frequent rain events (2- and 10-year storms) and lower for less frequent events (50- 2073 and 100-year storms), with this same pattern observed with field-collected data (Moody 2074 and Martin 2001). This method only provides meaningful results where the pre-fire peak 2075 flow is greater than zero for the rainfall event of interest. 2076 2077 Figure CS1- 4. Example estimated pre- and post-fire hydrographs for the 10-year rain event.

2078

DRAFT Technical Note, April 2015 C1-10 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2079 Figure CS1- 5. Example map with pre- and post-fire flood prediction estimates for the Poudre Park 2080 area.

2081 2082 Figure CS1- 6. Estimated 25-year peak flow enhancement ratios for the High Park Fire.

2083 2084 Comparison with Regression Predictions 2085 Table CS1-2 illustrates USGS regression modeling results (from Streamstats) compared 2086 to CN modeling results at a number of locations, for the 10- and 25-year events. 2087 Considering the large expected prediction errors of the USGS regression equations in this

DRAFT Technical Note, April 2015 C1-11 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2088 area (typically 140%), the results are reasonably comparable. The greatest differences in 2089 prediction are typically in the largest catchments. Differences in the results are likely due 2090 to limited data available for selecting CN values for post-fire conditions, questionable CN 2091 model appropriateness in forested watersheds, and possible inaccurate rainfall depths and 2092 aerial reduction factors. Additionally, the regression technique does not account for 2093 relevant hydrologic processes, such as variable soil infiltration capacity, vegetative type, 2094 ground cover condition, and stream flow attenuation. 2095 2096 Table CS1- 2. Comparison of CN modeling with USGS regression results from StreamStats.

Ar e a 10-yr flow (cfs) 25-yr flow (cfs) Ar e a 10-yr flow (cfs) 25-yr flow (cfs) Point (mi2) CN Reg CN Reg Point (mi2) CN Reg CN Reg MC-2 3.33 119 212 249 360 BG 1.22 97 95 177 151 MC-5 6.61 272 335 588 574 UN-3 0.28 21 40 42 61 BC-8 23.2 44 407 119 650 RdC-2 5.05 131 231 284 382 BC-13 42.9 244 599 474 974 RdC-5 13.2 180 445 479 764 BC-18 55.0 332 741 729 1230 RdC-6 16.2 168 512 478 885 HlG-3 10.9 207 248 446 387 RC-2 3.00 172 170 334 277 HlG-5 21.8 210 348 466 554 RC-4 8.16 282 316 549 531 LC-1 1.20 23 96 60 152 SG-1 3.09 103 120 216 183 LC-4 6.97 168 277 354 460 SG-2 1.19 45 73 96 111 CG 2.00 89 92 178 139 SG-4 5.99 191 182 395 283 2097 2098 Accuracy and Limitations 2099 As with all hydrologic modeling, the results provided by these simulations are 2100 approximate. Comparisons with flood peaks estimated using the USGS regression 2101 equations (table 2) indicate that the pre-fire runoff simulations are more-or-less 2102 reasonable overall, but that the modeling often provides estimates that may be inaccurate, 2103 especially as catchment size increases. Post-fire runoff prediction using the CN technique 2104 is hampered by the very little available field data to enable reliable selection of CN 2105 values from measured rainfall and runoff in burned catchments. Additionally, the use of 2106 dated rainfall depths and areal reduction factors (Miller et al. 1973) in orographic-forced 2107 mountainous watersheds adds an additional layer of uncertainty to the estimates. 2108 2109 Additionally, greater infiltration is indicated by the USFS soil survey than the adjacent 2110 NRCS survey, with infiltration commonly increasing by a step at the survey boundary

DRAFT Technical Note, April 2015 C1-12 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2111 and a large area with HSG A indicated. In the most problematic areas zero runoff is 2112 predicted for pre-fire conditions in some catchments during the 10- and 25-year rain 2113 events. This problem may be due to shallow, permeable soils over bedrock dominating 2114 the USFS soil survey classification methodology, but the true reason for this 2115 inconsistency is unknown. As a result, the modeling may be underpredicting runoff and 2116 overestimating flood response ratios in catchments draining the Arapaho-Roosevelt soil 2117 survey area, especially for more frequent (shallower) rainfall events. 2118 2119 More fundamentally, the reliability of the CN method for predicting peak flow from 2120 forested, mountainous watersheds is questionable, due to doubtful appropriateness of the 2121 CN method and shifting streamflow generation processes between and pre- and post-fire 2122 conditions. This issue is discussed in the main body of this technical note under the 2123 section Limitations of the CN Method. 2124 2125 Despite these shortcomings, due to its relative simplicity and achievable data 2126 requirements for large scales, the CN method is a preferred tool for predicting flood 2127 responses of wildfire areas. The modeling performed for the High Park Fire has 2128 substantial value for identifying areas of greatest threat to life, property and 2129 infrastructure. Peak flow ratios provide an excellent tool for communicating expected 2130 increases in runoff with agencies, first responders, and the public. And peak flow and 2131 runoff volume estimates are still required for sizing infrastructure improvements. The 2132 unknown extent of uncertainty in the estimates needs to be effectively communicated to 2133 provide assurances that involved parties use the estimates with caution. Research and 2134 technical guidance is needed to develop and communicate more robust methods for flood 2135 prediction from wildfire-impacted landscapes. 2136 2137 Conclusions 2138 Using the NRCS Curve Number method, peak flow predictions were made for streams 2139 draining the High Park Fire area, for both pre-fire and post-fire conditions. Watershed 2140 maps for each modeled catchment were developed, illustrating computation points, peak 2141 discharges, peak flow enhancement ratios, soil burn severity, and stream outlet

DRAFT Technical Note, April 2015 C1-13 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2142 hydrographs. While many relevant hydrologic mechanisms were simulated, including 2143 rainfall depth and spatial extent, variation in runoff by soil burn severity, vegetation type 2144 and soil conductivity, variable lag times, and stream attenuation, the questionable 2145 reliability of the CN method in forested watersheds, for both post- and pre- fire 2146 conditions, adds varying uncertainty to the estimates. Research is needed to address this 2147 uncertainty, especially since lives are often at risk. In the meantime, wildfires will occur 2148 and methods need to be available to predict the expected flood response from burned 2149 watersheds. This case study provides an example of one such approach. 2150 2151 Acknowledgements 2152 Appreciation is expressed to High Park Fire BAER team, for their initial analyses, 2153 Brandon Stone and Colorado State University for the utilized BARC soil burn severity 2154 interpretation, and Randy McKinley of the USGS and Eric Schroder of the USFS for the 2155 presented soil burn severity interpretation. Appreciation is also expressed to John 2156 Andrews for his contributions to the post-fire CN assignment compilation, John Norman 2157 for GIS analysis assistance and soils expertise, and Kari Sever and Sam Streeter for field 2158 data collection. 2159 2160 References 2161 BAER 2012. High Park Fire Burned Area Emergency Response Report. USDA USFS 2162 and NRCS, Larimer County CO Department of Transporation. 2163 2164 Capesius, J.P., and Stephens, V.C. 2009. Regional Regression Equations for Estimation 2165 of Natural Streamflow Statistics in Colorado. U.S. Department of Interior, U.S. 2166 Geological Survey, Scientific Investigations Report 2009-5136. 2167 2168 Conedera, M., Peter, L., Marxer, P., Forster, F., Rickenmann, D., and Re, L. 2003. 2169 Consequences of forest fires on the hydrogeological response of mountain catchments: A 2170 case study of the Riale Buffaga, Ticino, Switzerland. Earth Surface Processes and 2171 Landforms 28, 117-129. 2172

DRAFT Technical Note, April 2015 C1-14 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2173 Garen, D.C., and Moore, D.S. 2005. Curve Number Hydrology in Water Quality 2174 Modeling: Uses, Abuses and Future Directions. Journal of the American Water 2175 Resources Association 42(2), 377-388. 2176 2177 Goodrich, D.C., Canfield, H.E., Burns, I.S., Semmens, D.J., Miller, S.N., Hernandez, M., 2178 Levick, L.R., Guertin, D.P., and Kepner, W.G. 2005. Rapid Post-Fire Hydrologic 2179 Watershed Assessment Using the AGWA GIS-Based Hydrologic Modeling Tool. In: 2180 Moglen, G.E., ed. Proceedings of Managing Watersheds for Human and Natural Impacts: 2181 Engineering, Ecological, and Economic Challenges, July 19-22, 2005, Williamsburg, 2182 VA, USA. American Society of Civil Engineers, Alexandria, VA. 2183 2184 Hawkins, R.H. 1993. Asymptotic Determination of Runoff Curve Numbers from Data. 2185 Journal of Irrigation and Drainage Engineering 119, 334-345. 2186 2187 Hawkins, R.H., Greenberg, R.J. 1990 WILDCAT4 Flow Model. School of Renewable 2188 Natural Resources, University of Arizona, Tucson, AZ. BLM contact: Dan Muller, 2189 Denver, CO. 2190 2191 Hudak, A.T., Robichaud, P.R., Evans, J.S., Clark, J., Lannom, K., Morgan, P., and Stone, 2192 C. 2004. Field Validation of Burned Area Reflectance Classification (BARC) Products 2193 for Post Fire Assessment. Remote Sensing for Field Users: Proceedings of the Tenth 2194 Forest Service Remote Sensing Applications Conference, Salt Lake City, Utah, April 5-9 2195 2004. 2196 2197 Livingston, R.K., Earles, T.A., and Wright, K.R. 2005. Los Alamos Post-Fire Watershed 2198 Recovery: A Curve-Number-Based Evaluation. In: Moglen, G.E., ed. Proceedings of 2199 Managing Watersheds for Human and Natural Impacts: Engineering, Ecological, and 2200 Economic Challenges, July 19-22, 2005, Williamsburg, VA, USA. American Society of 2201 Civil Engineers, Alexandria, VA. 2202 2203 Maidment, D.R. 1992. Handbook of Hydrology McGraw-Hill, Inc.

DRAFT Technical Note, April 2015 C1-15 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2204 2205 Moody, J.A., and Martin, D.A. 2001. Post-Fire, Rainfall-Intensity-Peak Discharge 2206 Relations for Three Mountainous Watersheds in the Western USA. Hydrological 2207 Processes 15, 2981-2993. 2208 2209 Miller, J.F., Frederick, R.H., and Tracey, R.J. 1973. Precipitation-Frequency Atlas of the 2210 Western United States. U.S. Department of Commerce, National Oceanic and 2211 Atmospheric Administration, National Weather Service, NOAA Atlas 2, Vol 3, Silver 2212 Spring, Maryland. 2213 2214 Rallison, R.E. 1980. Origin and Evolution of the SCS Runoff Equation. Proceedings of 2215 the Symposium on Watershed Management; American Society of Civil Engineers, Boise, 2216 ID. 2217 2218 Springer, E.P., and Hawkins, R.H. 2005. Curve Number and Peakflow Responses 2219 Following the 2220 Cerro Grande Fire on a Small Watershed. In: Moglen, G.E., ed. Proceedings of Managing 2221 Watersheds for Human and Natural Impacts: Engineering, Ecological, and Economic 2222 Challenges, July 19-22, 2005, Williamsburg, VA, USA. American Society of Civil 2223 Engineers, Alexandria, VA. 2224 2225 Tedela, N.H., McCutcheon, S.C., Rasmussen, T.C., Hawkins, R.H., Swank, W.T., 2226 Campbell, J.L., Adams, M.B., Jackson, C.R., and Tollner, E.W. 2012. Runoff Curve 2227 Numbers for 10 Small Forested Watersheds in the Mountains of the Eastern United 2228 States. Journal of Hydrologic Engineering, 17(11), 1188-1198. 2229 2230 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2231 Service), 2004a. 2232 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 9, Hydrologic Soil 2233 Cover Complexes. 2234

DRAFT Technical Note, April 2015 C1-16 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2235 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2236 Service), 2004b. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 2237 10, Estimation of Direct Runoff from Storm Rainfall. 2238 2239 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2240 Service), 2010. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 2241 15, Time of Concentration. 2242 2243 Woodward, D. 2005. Personal Communications (email). Retired National Hydraulic 2244 Engineer, USDA Natural Resources Conservation Service, Washington D.C. 2245 2246 Wright, K.R., Earles, T.A., and Pemberton, E. 2005. Mesa Verde Bircher Fire 2247 Hydrological Impact. In: Moglen, G.E., ed. Proceedings of Managing Watersheds for 2248 Human and Natural Impacts: Engineering, Ecological, and Economic Challenges, July 2249 19-22, 2005, Williamsburg, VA, USA. American Society of Civil Engineers, Alexandria, 2250 VA. 2251 2252 Yochum, S.E. 2012a. Poster: Increased Flood Potential of Steams Draining the High Park 2253 Fire, with Ratios of Predicted Post/Pre Fire Peak Flows for the 1-Hour, 25-Year Rain 2254 Event. U.S. 2255 Department of Agriculture, Natural Resources Conservation Service, Colorado State 2256 Office. 2257 2258 Yochum, S.E. 2012b. High Park Fire: Increased Flood Potential Analysis. U.S. 2259 Department of Agriculture, Natural Resources Conservation Service, Colorado State 2260 Office. 2261 2262 Yochum, S., and Bledsoe, B. 2010. Flow resistance in high-gradient streams. Proceedings 2263 of 4th Interagency Hydraulic Modeling Conference, June 21 – July 1st, 2010, Las Vegas, 2264 NV, USA. 2265

DRAFT Technical Note, April 2015 C1-17 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 1: High Park Fire Colorado 2266 (This page intentionally blank)

DRAFT Technical Note, April 2015 C1-18 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2267 Case Study 2: Bitterroot Wildfires, Montana 2268 2269 Background 2270 Southwestern Montana in the year 2000 endured severe climatic drought conditions. Late 2271 that summer wildfires raged through the area, with the Bitterroot National Forest hit 2272 particularly hard. See figures CS2-1 through CS1-3 for location and fire severity 2273 mapping. 2274 2275 Figure CS2- 1. State of Montana, with inset of Bitterroot River valley

2276 2277 Figure CS2- 2. Bitterroot River Watershed with inset of area of the wildfires in the year 2000.

2278

DRAFT Technical Note, April 2015 C2-1 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2279 Figure CS2- 3. Bitterroot fire severity for the fires of 2000 (Mahlum et al. 2011)

2280 2281 These fires were largely started by lightning strikes, as shown in figure CS2-4. The 2282 damaged area included 244,000 acres on USDA National Forest lands and 49,000 acres 2283 of state-owned and private land. The severity of the fires ranged from low (some under- 2284 story destroyed, minimal impact on tree canopy) to high (complete destruction of all 2285 living plant material). See figure CS2-3 for fire severity and figure CS2-5 for photo of 2286 fire in progress. 2287

DRAFT Technical Note, April 2015 C2-2 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2288 Figure CS2- 4. Lightning strikes in the Bitterroot Valley of Montana, initiating the wildfires of 2000.

2289 2290 2291 Figure CS2- 5. One of the most poignant wildfire scenes ever captured, from the Bitterroot in 2000.

2292 Photo by John McColgan, USDA Forest Service 2293 2294 After the fires, specialists from the USDA Forest Service and Natural Resources 2295 Conservation Service (NRCS) evaluated potential hydrologic impacts to area residents 2296 and infrastructure. The NRCS effort included the Emergency Watershed Protection 2297 Program (EWP). (For more detail on EWP, see the technical note section “The role of 2298 NRCS in post-fire assessments and modeling.” For the Bitterroot fires of 2000 an

DRAFT Technical Note, April 2015 C2-3 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2299 immediate technical challenge arose because traditional agency hydrologic methods were 2300 not designed to address post-wildfire conditions. A literature search for existing 2301 procedures yielded none that could be expeditiously applied. As a result, the NRCS 2302 Montana office developed a simple but accurate hydrologic evaluation method which 2303 would be more broadly applicable to post-wildfire watershed protection projects. The 2304 spreadsheet tool FIRE HYDRO (Cerrelli, 2005) resulted from this effort. 2305 2306 A key factor which made apparent the need for such a tool was the lack, in existing 2307 NRCS engineering handbooks, of guidance specifically related to post-wildfire 2308 conditions. The runoff curve number (RCN) methodology is described in NRCS 2309 National Engineering Handbook (NEH), Part 630, Hydrology, Chapter 9, Hydrologic 2310 soil-cover complexes (2004a), and Chapter 10, Estimation of direct runoff from storm 2311 rainfall (2004b). The new spreadsheet tool was to retain this guidance, given its 2312 familiarity to most NRCS engineers. However, the challenge was to provide correct CNs 2313 for the various land covers burned by wildfires of differing severities. Additionally, the 2314 analysis used to determine times of concentration for the burned areas needed to be 2315 examined for post-fire effects. As the FIRE HYDRO spreadsheet has been used by both 2316 NRCS and Forest Service personnel after numerous wildfires since the year 2000, and in 2317 other states besides Montana, this case study will discuss its original development and 2318 use for evaluating the Bitterroot fires. 2319 2320 Methods 2321 The envisioned goal of the spreadsheet tool was to provide a simplistic method, readily 2322 familiar to USDA field engineers that could be run on laptop computers at remote 2323 locations. In evaluating a wildfire, NRCS field engineers must quickly determine 2324 appropriate conservation practices and areal extents, which could be applied to reduce the 2325 impact of runoff and debris coming from the burn areas. The Emergency Watershed 2326 Protection program (EWP) and the Burn Area Emergency Rehabilitation program 2327 (BAER) are administered for recovery work and typically activated quickly, during and 2328 shortly after an event. (See the technical note for more information on both EWP and

DRAFT Technical Note, April 2015 C2-4 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2329 BAER.) Within these programs, projects must be evaluated, designed, contracted, and 2330 constructed quickly or they will not be considered for agency funding. 2331 2332 The hydrologic principles found in the NRCS Engineering Field Handbook, Chapter 2, 2333 Estimating Runoff (USDA-NRCS 1989) were considered appropriate to meet the needs 2334 for this simple but accurate analysis of the wildfire areas. These procedures are 2335 abbreviated EFH-2. 2336 2337 At the time of this study (2000), EFH-2 was only available as a manual method that 2338 required the user to look to the tabular and graphical listings within the guidance to 2339 determine the runoff volume for the singular watershed under investigation . The EFH-2 2340 peak discharge analysis requires the user to specify the design rainfall distribution type 2341 (defined by NRCS into four categories nationwide) that is appropriate for the project area 2342 under investigation. The NRCS Montana office has its own criteria for determining the 2343 appropriate selection of design rainfall distribution based on the project area’s ratio of the 2344 6-hour to 24-hour rainfall amounts for the desired recurrence event. The spreadsheet 2345 methodology was created in Microsoft Excel and named FIRE HYDRO, incorporating 2346 NRCS RCN technology and EFH-2 peak discharge graphical solution in conjunction with 2347 the NRCS Montana design rainfall selection criterion. 2348 2349 EFH-2 provides a set of curves for each rainfall distribution region in the US. Each curve

2350 pertains to an Ia/P ratio (where Ia refers to the CN runoff initial abstraction and P is total 2351 rainfall, both units in inches). To obtain peak discharge, one selects the correct curve,

2352 given Ia/P, then enters the x-axis with time of concentration, and reads unit peak 2353 discharge on the y-axis. For FIRE HYDRO an algebraic function was derived for each of 2354 these curves and provided in the Excel file to solve for predicted peak discharge. 2355 2356 Among the technical issues that needed to be addressed in the creation of FIRE HYDRO 2357 were proper selection of burn area runoff curve numbers and times of concentration, and

DRAFT Technical Note, April 2015 C2-5 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2358 soil hydrophobicity. These issues will be discussed in greater detail below. The FIRE 2359 HYDRO 2360 methodology analyzes four different watershed conditions. They are as follows: 2361 • Pre-Fire (watershed significantly healed from any previous fires) 2362 • Post-Fire 1 (Immediately after the event, including consideration of hydrophobic 2363 soil) 2364 • Post-Fire 2 (After some emergence of vegetation and breakdown of 2365 hydrophobicity) 2366 • Post-Fire 3 (After estimated or expected vegetation recovery in next growing 2367 season) 2368 2369 These four conditions were evaluated to give the designer a greater feel for the range of 2370 runoff conditions that were, are, and will be present for the watershed being investigated. 2371 This information helps the designer better understand the change in relative risk over 2372 time on various alternative hydrologic control measures considered for installation, 2373 including the “do nothing” alternative. 2374 2375 For the Bitterroot fires, conditions were monitored for both immediately post-fire 2376 conditions and later post-fire conditions. Figure CS2-6 shows how runoff shortly after 2377 the burn eroded the soil surface and was retarded by the conservation practice of silt 2378 fencing. 2379

DRAFT Technical Note, April 2015 C2-6 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2380 Figure CS2- 6. Fully loaded silt fence placed in the aftermath of the 2000 Bitterroot Valley Complex 2381 wildfire.

2382 2383 Figure CS2-7 shows a watershed in the early stages of recovery. Figures CS2-8 through 2384 CS2-10 show Bitterroot watersheds after several years of recovery. 2385 2386 Figure CS2- 7. Laird Creek in 2001 within what was the Bitterroot Fire Valley Complex. 2387

2388 Figure CS2- 8. Sheafman Point in October 2008. 2389

DRAFT Technical Note, April 2015 C2-7 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2390 Figure CS2- 9. Glen Lake trail in the summer of 2012.

2391 2392 Figure CS2- 10. The landscape around Glen Lake in the summer of 2012.

2393 2394 2395 Runoff Curve Number (RCN) Estimation for Burn Areas 2396 The FIRE HYDRO method is used to evaluate individual watersheds by treating all land 2397 use/soil conditions as a composite homogeneous collection. This assumption contains 2398 inherent problems when widely varying runoff conditions exist in the watershed being 2399 modeled. A literature search for recommended RCNs for wildfire burn areas specifically 2400 related to climate and vegetation types of southwestern Montana yielded nothing.

DRAFT Technical Note, April 2015 C2-8 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2401 Consequently, select NRCS Montana engineers created the following guidance for 2402 selection of RCN based on burn severity and hydrologic soil grouping (HSG) specific to 2403 the Bitterroot wildfire vicinity (Tables CS2-1 through CS2-4.) 2404 2405 Table CS2- 1. Runoff curve numbers for high severity burned areas of southwest Montana* Hydrologic soil group High burn severity CN A 64 B 78 C 85 D 88 2406 *High Severity Burn Areas assumed to have attained a minimum of 30% ground cover consisting of 2407 vegetation, duff, thick ash, or woody debris by June of the following year. 2408 2409 Table CS2- 2. Recommended runoff curve numbers for moderate severity burned areas RCN for given hydrologic soil group* Cover type A B C D Herbaceous mixture of grass, weeds, and low- 71 81 89 growing brush, with brush the minor element Oak-aspen mountain brush mixture of oak 48 57 63 brush, aspen, mountain mahogany, bitter brush, maple Pinyon-juniper pinyon, juniper or both, grass 58 73 80 understory Sagebrush 51 63 70 grass understory Desert shrub major plants include saltbush, greasewood, creosotebush, 55 72 81 86 blackbrush, bursage, palo verde, mesquite, and cactus. 2410 *Moderate burn area RCNs are for cover types in Fair condition. This table is from EFH-2, for 2411 arid & semiarid rangelands, which stated that curve numbers for HSG A were developed only for 2412 desert shrub cover type. 2413

DRAFT Technical Note, April 2015 C2-9 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2414 Table CS2- 3. Recommended RCNs for low severity burned and unburned areas with north and east 2415 facing slopes. RCN for given hydrologic soil group* Cover type A B C D Herbaceous mixture of grass, weeds, and low- 62 74 85 growing brush, with brush the minor element Oak-aspen mountain brush mixture of oak 30 41 48 brush, aspen, mountain mahogany, bitter brush, maple Pinyon-juniper pinyon, juniper or both, grass 41 61 71 understory Sagebrush 35 47 55 grass understory Desert shrub major plants include saltbush, greasewood, creosotebush, 49 68 79 84 blackbrush, bursage, palo verde, mesquite, and cactus. 2416 *Moderate burn area RCNs are for cover types in Fair condition. This table is from EFH-2, for 2417 arid & semiarid rangelands, which stated that curve numbers for HSG A were developed only for 2418 desert shrub cover type. 2419 2420

DRAFT Technical Note, April 2015 C2-10 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2421 Table CS2- 4. Recommended RCNs for low severity burned and unburned areas with south and 2422 west facing slopes. RCN for given hydrologic soil group* Cover type A B C D Herbaceous mixture of grass, weeds, and low- 67 78 87 growing brush, with brush the minor element Oak-aspen mountain brush mixture of oak 39 49 56 brush, aspen, mountain mahogany, bitter brush, maple Pinyon-juniper pinyon, juniper or both, grass 50 67 76 understory Sagebrush 43 55 63 grass understory Desert shrub major plants include saltbush, greasewood, creosotebush, 52 70 80 85 blackbrush, bursage, palo verde, mesquite, and cactus. 2423 *Low burn and unburned area RCNs are for cover types between Fair and Good condition, with values 2424 averaged from tables CS2-2 and CS2-3. 2425 2426 These recommended RCNs (for all burn severity types listed) were arrived at by 2427 consensus amongst three NRCS Montana engineers with hydrologic evaluation 2428 experience. See figures CS2-11 and CS2-12 for examples of data available for the 2429 Bitterroot fires that used in establishing the areal extent of different burn severities. The 2430 basic premise of the three NRCS Montana engineers was to establish a logical fit for burn 2431 area RCN values within the existing accepted NRCS RCN/land use table. There was no 2432 time between the fire occurrence and the need for installation of exigency runoff 2433 protection work for gauging or other verification procedures. Storms that occurred the 2434 following spring and summer, events of approximately 2-year to 5-year recurrence, 24- 2435 hour duration, did not produce runoff which led to the failure of any protection practices 2436 installed using these RCN values. 2437

DRAFT Technical Note, April 2015 C2-11 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2438 Figure CS2- 11. Satellite imagery used to detail burn severity of the 2000 Bitterroot Valley Complex 2439 wildfire.

2440 2441 Figure CS2- 12. Burn severity map and rain gage data for a rain event that happened a year after 2442 the 2000 Bitterroot Valley Complex wildfire.

DRAFT Technical Note, April 2015 C2-12 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2443 2444 Hydrophobic Soils 2445 Hydrophobicity in soils can be induced by wildfire. It requires certain soil textures and 2446 certain plant types to be burned in order to create this condition of significant water 2447 repellency. In brief, the condition can be found in more moderately coarse soils that have 2448 a deep plant litter mat and experience a severe burn. The resulting “waxy gas” that is 2449 created then permeates and coats the upper soil layer making it water repellent. More 2450 detail on soil hydrophobicity is provided in the technical note sections “Hydrophobicity” 2451 and “Assessment of post-fire soil hydrophobic conditions.” 2452 2453 Runoff rates and volumes from hydrophobic soils can become extremely high. The 2454 engineer must make best estimates as to how widespread and persistent hydrophobicity 2455 exists after a fire. Assigning a design RCN to account for hydrophobicity also depends 2456 on the expected duration of this changed runoff condition. For example, the timing of 2457 next normal rain season compared to hydrophobic abatement recovery time should be 2458 considered. Field-testing is called for, which involves applying water to the mineral soil 2459 (below the ash layer) and checking for infiltration with time. See also the technical note 2460 section “Accounting for hydrophobicity with infiltration parameters.” In addition, Tables 2461 7 and 8 of the technical note show Forest Service derived curve numbers in relation to 2462 soil hydrophobicity. 2463 2464 For the Bitterroot fires of 2000 the NRCS Montana engineers gave all hydrophobic soils 2465 a runoff curve number of 94. It is important to consider the window of time to which 2466 hydrophobic conditions are expected to occur. Soil hydrophobicity tends to breakdown 2467 over time as the water repellent layer is fractured through either plant growth activity or 2468 water action (freeze/thaw or dew and desiccation). These two processes occur very 2469 typically in Montana but may not be so prominent in other states. 2470 2471 Experience gained from the Bitterroot EWP effort, while considering hydrophobic soil 2472 properties for effective RCN assignment, yielded some interesting findings. Significant

DRAFT Technical Note, April 2015 C2-13 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2473 field reconnaissance, largely by Forest Service personnel, was dedicated to evaluating the 2474 water-repellent properties of soils in the fire-affected watersheds. These investigations in 2475 the Bitterroot burn found areas with enough hydrophobicity to repel 50-70% of incoming 2476 water. It should be noted that even unburned forest soils can exhibit hydrophobicity. 2477 Tests in the Bitterroot fire area of adjacent unburned areas found that the soils would 2478 repel 40-60% of incoming water. It was thought that the drought that had been prevalent 2479 throughout the area in the summer of 2000 created a “tightening effect” of drought- 2480 affected soils, which created some water-repellency within the soil. Since the burned and 2481 unburned rates of water repellency seemed so similar it was determined that the fire had 2482 not induced a significant increase in hydrophobicity. Therefore, the analysis of the 2483 Bitterroot fires using FIRE HYDRO shortly after the event did not use CNs that 2484 attempted to account for hydrophobic conditions. Observations of subsequent rainfall 2485 events tended to support the earlier suspicions, as large increases in runoff that might be 2486 expected from significantly hydrophobic areas was generally not in evidence. 2487

2488 Time of Concentration (Tc) and Assumed Watershed Shape 2489 The engineers and technicians assigned to perform the watershed hydrologic evaluations 2490 of the Bitterroot wildfire area were not provided detailed topographic mapping of the site. 2491 They were provided GIS data that included total watershed area as well as percentages of 2492 which experienced low, moderate, or high severity burns along with soils data that 2493 included land slopes within the dominant soil classes found. It should be noted that, 2494 although today’s engineers are often skilled 2495 GIS users, the rapid turn-around time for the Bitterroot fires of 2000 and the skill set of 2496 the engineers who happened to be available at the time, precluded use of GIS to 2497 determine the longest flowpath lengths of subareas of the fire directly. Thus, an 2498 assumption was needed that would provide an estimate of watershed longest flow path 2499 based on drainage area. An assumed rectangular shape watershed with dimensions of 2500 one unit wide by two units long was made to produce the area stated for the watershed in 2501 the GIS data report. The longest flow path was taken as the longitudinal line down the

2502 center of this rectangle with a bend towards a corner at the upper third. The Tc was

DRAFT Technical Note, April 2015 C2-14 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2503 calculated using the Lag Method found in EFM-2. (See also the technical note section 2504 “Watershed lag method.” This method used RCN, flow length, and average watershed

2505 slope (based on values from the soils data in the GIS data) as its basis. The post-fire Tc

2506 was assumed to stay equal to the pre-fire Tc even though there was a reduction in plant 2507 cover (therefore reduced flow friction leading to higher flow velocities) after the fire that 2508 should logically allow a faster Tc to develop. The assumption here though was that the 2509 excessive amount of available debris after the fire would cause blockages in the flowpath, 2510 a kind of debris dam, which would serve to attenuate the flows. The degree of flow 2511 attenuation after a fire may vary tremendously from one area to another or one watershed 2512 to another, but it was felt that a reasonable estimate would be to use the same time of 2513 concentration for both pre-fire and post-fire conditions. 2514 2515 Results and Discussion 2516 The FIRE HYDRO spreadsheet was much utilized by engineers and technicians involved 2517 with the 2000 Bitterroot Wildfire recovery operations. NRCS personnel designed many 2518 runoff diversion practices to protect homes from the anticipated increased rainfall runoff 2519 brought about by changes in watershed conditions due to the wildfires. Forest Service 2520 personnel used FIRE HYDRO to design for the enlargement of culverts to provide 2521 capacity for the anticipated increase in runoff and also to allow for fish passage. All of 2522 the practices designed using FIRE HYDRO were able to satisfactorily withstand rainfall- 2523 runoff events in the following years without any notable failures being reported. 2524 2525 It should be noted that many homes which benefitted from conservation practices after 2526 this event happened to be situated on the only reasonably flat land around, that being 2527 surrounded by steep mountainous land. Such land was near level because of being 2528 located on an alluvial fan created by deposition of sediments from previous historic 2529 wildfires. So in effect, the homes were prime targets for heavy post fire sediment and 2530 debris deposition, if not for the successful functioning of diversion practices (most 2531 typically in the form of concrete highway dividers placed slightly down the slope and 2532 above the area of interest).

DRAFT Technical Note, April 2015 C2-15 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2533 2534 One key feature that should be considered in establishing the proper design RCN for post 2535 wildfire hydrologic evaluations is the anticipated recovery time of the watershed towards 2536 its pre-fire condition. RCN increases as burn severity worsens. This leads to higher 2537 estimated runoff from the various design rainfalls used for sizing runoff control practices. 2538 These, in turn, lead to higher project costs. Nevertheless, in the seasons that follow the 2539 wildfire landscape recovery tends to return hydrologic processes toward pre-fire 2540 conditions. The longer the healing time, the less the design RCN should diverge from 2541 that used immediately after the fire. It is incumbent upon the engineer in charge of 2542 assigning the recommended design RCNs for various land covers and burn severities to 2543 consider both the risk of failure versus storm frequency. Design RCNs based on recovery 2544 of the land cover can vary greatly in the seasons following a wildfire. 2545 2546 The Bitterroot fires of 2000 and the use of FIRE HYDRO to design conservation 2547 practices happened to coincide with disagreements among engineers about proper post- 2548 fire design storm recurrence intervals. The concern stemmed from the idea that the 2549 notable increase in RCN value as the result of a fire would be short-lived and ground 2550 cover recovery would decrease the RCN in the seasons following the wildfire. The EFH- 2551 2 method that FIRE HYDRO is based on is very sensitive to the selected design RCN. It 2552 was thought that if all designs, including those for low hazard structures, were based on 2553 the accepted storm recurrence interval (25-yr for most practices) overly conservative and 2554 costly designs would result, since the watersheds would heal significantly in the early 2555 post-fire years. The NRCS Montana State Conservation Engineer issued a guidance 2556 letter to all EWP engineers and technicians relating to the selection of recurrence interval 2557 design storms to be used for the statewide EWP effort. Low hazard scenarios were to be 2558 designed for non-damaging passage of flows of the 25-year pre-fire peak discharge or 2559 post-fire 10-year peak, whichever was greater. High hazard scenarios were to be dealt 2560 with on a case-by-case basis, in consultation with the state engineer. 2561

DRAFT Technical Note, April 2015 C2-16 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana

2562 The assumption for using the same Tc both pre-fire and post-fire should be more 2563 thoroughly investigated to determine its validity. As previously mentioned, this 2564 assumption was based on the idea that the added debris and sediment load from the burn 2565 would create increased opportunity for flow blockages to occur thereby slowing the 2566 runoff process from what might initially be expected from a higher RCN value. 2567 Improvements to post-fire hydrologic methods such as FIRE HYDRO would benefit 2568 from research in this area. Onsite inspection of the burn area during a prominent runoff 2569 event should provide evidence of the validity of the assumption that post-fire debris 2570 offsets the loss of vegetation in runoff speed. Evidence of flow blockages may exist and 2571 their potential to eventually breach after a storm. The ideal investigation would gage 2572 rainfall and runoff both before and after a wildfire. This would provide the most accurate

2573 assessment of wildfire influence to Tc and storm runoff volume and rate. 2574 2575 Complex issues should be acknowledged when attempting to model post wildfire 2576 hydrology. One of the most impactful is the time variance of RCN value. Clearly, the 2577 effective RCN for burned watersheds will have increased drastically. As the watershed 2578 heals, however, the RCN reduces toward the pre-fire condition. But at what rate? It 2579 might not be economically proper to assume the worst case RCN (immediately after the 2580 fire) for all hydrologic modeling related to design of runoff control practices in these 2581 areas. An accurate assessment of RCN change with time would help decision makers 2582 develop a sensible strategy for setting proper design parameters. Peter Robichaud, 2583 USDA Forest Service, has spent considerable time investigating and monitoring post 2584 wildfire watersheds. His findings indicate a tremendous increase in sheet erosion taking 2585 place on the landscape the year following the fire but then cutting back drastically each 2586 year thereafter. Evidently, growth of grass on the bare ground provides abatement to 2587 raindrop impact on the soil thereby stemming erosion about a year or two after the fire. 2588 The technical note references Robichaud’s work in three articles, (Foltz et al. 2009, 2589 Hudak et al. 2004, and Robichaud et al. 2008). The question remains as to whether the 2590 runoff amount has also significantly decreased in this timeframe. (The dominant driving 2591 mechanism of curve number change over time may not be raindrop impact).

DRAFT Technical Note, April 2015 C2-17 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2592 2593 Determining the reduction of post-wildfire RCNs over time to their pre-fire values may 2594 best be investigated by gaging the watershed runoff or possibly by documenting field 2595 indicators of stream flow levels along stream banks from various rainfall events. 2596 Optimally, these post-fire gage records would be compared to those kept from the same 2597 area prior to the fire. 2598 2599 One convenient option might be to gage watersheds for which good quality USGS 2600 predictive regression equations (for, say, peak discharge) exist, and that have also 2601 experienced wildfire. These areas could then be similarly analyzed for newly developed 2602 equations to reveal what trends have become apparent. RCNs could perhaps then be 2603 assigned within the NRCS hydrology models that produce results in line with the USGS 2604 gaged watersheds. 2605 2606 Limitations 2607 Sound hydrologic judgment is called for in utilizing FIRE HYDRO. It is based on the 2608 NRCS curve number runoff equation and subject to all of the assumptions pertaining to 2609 that method. The selection of appropriate (and likely weighted) RCN should be based on 2610 the best information available (pre-fire land cover/condition and fire severity distribution 2611 in the watershed being evaluated). The predicted peak discharges from this model are 2612 quite sensitive to the RCN used. The RCNs listed in this paper were refined from 2613 existing tables in EFM-2 and altered by the judgment of NRCS Montana engineers to 2614 address the conditions found in western Montana. Other areas may require similar 2615 scrutiny to best define appropriate RCN usage. The decision of whether or not it is 2616 appropriate to weight the various RCN attributed sub-areas and treat them as one 2617 homogeneous area is left to the designer. A more refined modeling approach might be to 2618 use WinTR-20 (USDA-NRCS 2004c) or one of the hydrologic models discussed in the 2619 other case studies of this technical note. Consultation with experienced users of these 2620 programs is warranted to assure accurate assessments. 2621

DRAFT Technical Note, April 2015 C2-18 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2622 Another way to handle non-homogeneous watersheds with FIRE HYDRO would be to 2623 determine peak flowrates per unit area for various pre and post burn curve numbers for 2624 each of the various recurrence events. Watersheds that do not vary too widely could 2625 benefit by this procedure. For example, the watershed would be modeled in pre-fire 2626 condition and a pre-fire flowrate per unit area determined. Then only the area of the 2627 portion burned severely could be analyzed and a severe burn flowrate per unit area 2628 determined. If the burned area of this hypothetical watershed burn was severe only, then 2629 the total predicted peak would be the severe burn area times the severe flowrate per unit 2630 area, plus the unburned area times the unburned flowrate per unit area. Watersheds with a 2631 wide mosaic of burns (and resultant varying RCNs) would not be handled in this way, 2632 since they would not fit the homogenous assumption that the spreadsheet utilizes. Again, 2633 sound judgment is called for, by the designer, to ensure that accurate modeling 2634 procedures are followed. 2635 2636 Though the drainage area is limited to 2000 acres by the parent method (EFM-2) of FIRE 2637 HYDRO, some of the burn areas are much larger than this. The results of FIRE HYDRO 2638 were compared to those from a more detailed method (TR-20) and found to be reasonably 2639 close for watersheds in the 5-10 square mile range. Consequently, no limit on drainage 2640 area was imposed on FIRE HYDRO. The user is cautioned to exercise judgment in the 2641 appropriate use of FIRE HYDRO to make sure that sound principles of hydrology are 2642 being modeled properly for the conditions present in the watershed. The assumption of

2643 rectangular watershed shape, used in the Tc computation, should no longer be necessary, 2644 since today’s widely available GIS programs generally produce longest flowpaths easily. 2645 2646 References 2647 BAER 2012. High Park Fire Burned Area Emergency Response Report. USDA USFS 2648 and NRCS, Larimer County CO Department of Transportation. 2649 2650 Cerrelli, G.A. 2005. FIRE HYDRO, a simplified method for predicting peak discharges 2651 to assist in the design of flood protection measures western wildfires. In Proceedings of

DRAFT Technical Note, April 2015 C2-19 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 2: Bitterroot Wildfires, Montana 2652 the ASCE 2005 Watershed Management Conference, July 2005, Williamsburg VA. pp. 2653 935-941. 2654 2655 Mahlum, S.K., Eby, L.A., Young, M.K., Clancy, C.G., and Jakober, M. 2011. Effects of 2656 wildfire on stream temperatures in the Bitterroot River Basin, Montana. International 2657 Journal of Wildland Fire, 20, 240-247. 2658 2659 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2660 Service), 2004a. 2661 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 9, Hydrologic Soil 2662 Cover Complexes. 2663 2664 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2665 Service), 2004b. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 2666 10, Estimation of Direct Runoff from Storm Rainfall. 2667 2668 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2669 Service), 2004c. WinTR-20 User Guide. Available at http://go.usa.gov/KoZ/. Accessed 2670 in April 2013. 2671 2672 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 2673 Service), 1984. NRCS National Engineering Handbook, Part 650 Engineering Field 2674 Handbook, Chapter 2, Estimating Runoff. 2675

DRAFT Technical Note, April 2015 C2-20 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2676 Case Study 3: West Fork Complex Fire, Colorado 2677 2678 Background 2679 The West Fork Complex Wildfire, in the San Juan Mountains west of the community of 2680 South Fork, burned more than 109,000 acres between June 9th and July 12th, 2013. See 2681 figure CS3-1 for the general location and figure CS3-2 for the area of the West Fork 2682 Complex fire. 2683 2684 Figure CS3- 1. State of Colorado with zoom into the upper Rio Grande watershed 2685 2686 Figure CS3- 2. Area of the West Fork Complex wildfire, west of South Fork CO. 2687 2688 The fire area was composed of three separate fires: the Papoose Fire in the Rio Grande 2689 watershed, the West Fork Fire in the Rio Grande and San Juan watersheds, and the 2690 Windy Pass Fire in the San Juan watershed. Together, these three fires are known as the 2691 West Fork Complex fire. See figure CS3-3 for the individual fire perimeters of the West 2692 Fork Complex. 2693

DRAFT Technical Note, April 2015 C3-1 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2694 Figure CS3- 3. Aerial extent of three fires known together as the West Fork Complex fire.

2695 2696 Wildfires cause hydrologic shifts for a number of years. Substantially increased runoff 2697 and sediment production result from the loss of vegetation and soil cover, as well as from 2698 soil hydrophobicity. Lack of interception by vegetation, reduced soil infiltration, loss of 2699 surface roughness and ground litter, and increased hydrophobicity, all combine to shift 2700 the rainfall response from infiltration-dominated to surface runoff-dominated processes. 2701 For example, watershed impacts due to recent wildfire caused a Swiss catchment to 2702 produce runoff in the range of 100-year to 200-year recurrence interval from a 10-year 2703 rainfall event. This extreme runoff increase was due to changes in infiltration capacity 2704 (Conedera et al. 2003). It should be noted that this phenomenon may have some scale 2705 dependency, with greater runoff enhancement in smaller catchments and tendencies 2706 towards overestimation in larger catchments (Stoof et al. 2011).

DRAFT Technical Note, April 2015 C3-2 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2707 2708 Hydrophobicity, which tends to be more prevalent with increased sand content and lower 2709 soil water content, has been found to weaken within a few months of a fire but persist for 2710 at least 22 months in ponderosa and lodgepole pine forests of the Colorado Front Range 2711 (Huffman et al. 2001). Post-fire sediment yield is most dependent on ground cover, with 2712 percent ground cover explaining more than 80 percent of the variability in sediment yield 2713 (Benavides-Solorio and MacDonald 2001). Accounting for soil burn severity is hence 2714 fundamental for predicting sediment yield increases. 2715 2716 The West Fork Complex Wildfire Burned Area Emergency Response (BAER) report 2717 presented an initial hydrologic analysis of flood increases to be expected from the fire. 2718 However, this assessment was performed with a number of simplifications to meet the 2719 aggressive timeline dictated by the BAER process. A more detailed peak flow analysis 2720 was performed to assess the variation and magnitude of increased flooding to be expected 2721 from the burned catchments. This analysis implemented the NRCS curve number (CN) 2722 runoff methodology. 2723 2724 Additionally, estimates of increased post-fire sediment yield help prioritize locations 2725 where hillslope and riparian mitigation may be needed. Such conservation practices 2726 minimize impacts of wildfire on such sensitive infrastructure as municipal water supply, 2727 irrigation diversions, and culvert conveyance. Predictions of post-fire sediment yield rely 2728 on mathematical models such as Revised Universal Soil Loss Equation RUSLE (Renard 2729 et al., 1997), Water Erosion Prediction Project WEPP (Elliot, 2004) and GeoWEPP 2730 (Renschler, 2003), as well as professional judgment (Robichaud et al., 2000). These 2731 methods have varying advantages and disadvantages for estimating the spatial 2732 distribution of post-fire soil loss, but all methods can require large amounts of time and 2733 energy to estimate soil loss and its associated risks over large spatial extents. With 2734 wildfires becoming more pronounced in the wildland/urban interface, rapid watershed 2735 management actions to protect sociological concerns, water quality, and ecosystem health 2736 are needed. This need for a rapid response to evaluate and manage post fire soil loss has 2737 increased the interest in using Geographic Information System (GIS) technology to

DRAFT Technical Note, April 2015 C3-3 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2738 spatially model post fire sediment yields. Many toolsets have been produced that use the 2739 above models as engines to spatially estimate soil loss rates. 2740 2741 This case study details hydrologic analyses performed within the Rio Grande basin to 2742 assess the expected magnitude of flood increases in populated areas at risk of loss of life 2743 and property. Additionally, erosion modeling was performed to evaluate the expected 2744 increase in sediment expected from the burn areas. These results were provided to 2745 stakeholders to assist with planning and prioritizing recovery efforts. 2746 2747 Methods 2748 Rainfall-runoff modeling was performed to simulate the expected flood response of the 2749 streams draining the wildfire areas. Additionally, predictions of post-fire sediment yield 2750 were also developed. The methods implemented to provide these predictions are 2751 presented below. 2752 2753 Hydrologic modeling for the West Fork Complex Fire was performed using the program 2754 HEC-HMS from the U.S. Army Corps of Engineers Hydrologic Engineering Center. The 2755 NRCS curve number (CN) technique for estimating direct runoff from rain events, 2756 combined with the NRCS dimensionless unit hydrograph method, was used in this 2757 analysis. Catchments, modeled stream channels, and pour points are presented in figure 2758 CS3-4. The average modeled catchment size is 2.4 square miles. 2759 2760 The CN is a simple catchment-scale method that gives simplified results at a stream 2761 outlet, with more accurate results expected for larger, higher-intensity rain events. The 2762 method is documented is in the NRCS National Engineering Handbook, Section 4, 2763 Hydrology, Chapters 9 and 10 (USDA-NRCS 2004a, USDA-NRCS 2004b), in Rallison 2764 (1980), as well as in numerous other publications. However, little quantitative 2765 information has been published of the database on which it was developed (Maidment 2766 1992). In general, the method was developed for rural watersheds in various parts of the 2767 United States, within 24 states, was developed for single storms (not continuous or partial

DRAFT Technical Note, April 2015 C3-4 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2768 storm simulation), and was not intended to recreate a specific response from an actual 2769 storm (Rallison, 1980). 2770 2771 Figure CS3- 4. Soil burn severity, model pour points, sub-watersheds, and stream channels

2772

2773 The initial abstraction (Ia) of the CN method has been traditionally assumed to be 0.2 of 2774 the storage coefficient, S. (For more on the CN method see the technical note section 2775 Runoff curve numbers.) To reflect the decreased storage of a fire impacted soil surface 2776 (due to a reduction of depression storage from the elimination of soil litter), the initial 2777 abstraction was assumed to be 0.1S for post wildfire conditions in catchments that were 2778 substantially burned (>50% moderate + severe soil burn severity). Catchments that were

2779 not substantially burned were modeled with the standard Ia of 0.2S. For the high CN post-

2780 wildfire conditions the impact on CNs of the Ia adjustment was ignored, since smaller

2781 shifts in CN due to changes in Ia can be expected (Woodward et al. 2003). 2782

DRAFT Technical Note, April 2015 C3-5 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2783 Runoff Curve Number (CN) Estimation 2784 CNs were assigned throughout the modeled catchments according to hydrologic soil 2785 group, vegetative type, and soil burn severity. Soil burn severity is a dominant factor in 2786 CN assignments in burned areas. (See the technical note.) Hydrophobicity was assumed 2787 to be minimal, since the next substantial rain events (after this analysis was performed) 2788 would likely occur a year after the fire. The average catchment CN was computed using 2789 an aerial averaging methodology. Hence, catchment size was limited to areas with similar 2790 runoff characteristics, to provide the most reliable results. As catchment size increased, 2791 CNs were computed for adjacent and serial catchments, and flows were routed 2792 downstream and combined with lower catchments to predict flow at downstream points 2793 of interest. This was necessary due to the larger basins draining the fire areas, as well as 2794 to account for catchment shape and stream channel attenuation. 2795 2796 Soil burn severity is the principle driver for increasing flow in runoff predictions. For this 2797 modeling, soil burn severity was measured using the BARC process from satellite data 2798 collected on 7/18/2013 (figure CS3-4). BARC uses reflectance recorded in satellite 2799 images to quantify soil burn severity. For defining soil burn severity, BARC images have 2800 the advantage of being comprehensive and relatively-rapidly developable. (See the 2801 technical note for more information on both BARC and soil burn severity.) However, 2802 comparison with field-collected data has indicated that this remotely-sensed product can 2803 be more indicative of post-fire vegetative condition than soil condition, especially in low 2804 to moderately burned areas (Hudak et al. 2004). 2805 2806 Hydrologic soil group (HSG) classification was selected using soils data published in the 2807 NRCS SSURGO (Soil Survey Geographic) database. Using this method, soil are 2808 classified as being either A, B, C, or D type, where A allows the most infiltration and 2809 least runoff and D allows the least infiltration and greatest runoff. Type A soils are 2810 prevalent in many areas of the fire. Dual hydrologic groups (i.e. A/D, B/D) are indicated 2811 in portions of the catchments. These dual classifications are for certain wet soils that 2812 could be adequately drained, with the first letter indicating the wet condition and the

DRAFT Technical Note, April 2015 C3-6 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2813 second indicating the undrained condition. In these cases, an undrained condition was 2814 assumed. 2815 2816 Vegetation type, from SWReGAP (Southwest Regional Gap Analysis Project) land cover 2817 mapping, was included in the assignment of CNs used for the modeling. The dominant 2818 vegetation types within the stream drainages of the fire areas were spruce-fir, aspen, and 2819 alpine meadow and tundra. 2820 2821 Curve numbers were assigned by polygons that had unique values of hydrologic soil 2822 group, vegetation type, and soil burn severity. Using primarily a compilation modified 2823 from the values used in the High Park (Case Study 1 of this technical note) and Black 2824 Forest Wildfire analyses, the implemented CN values are provided in table CS3-1. A fair 2825 ground cover condition was generally assumed for the unburned values abstracted from 2826 USDA-NRCS (2004a), though a good ground condition was assumed for 2827 herbaceous/grassland. The CN values for burned conditions were primarily compiled 2828 from various grey literature and unpublished sources; they should be considered 2829 approximate. 2830 2831 Table CS3- 1. CN assignments implemented in West Fork Complex Fire hydrologic modeling.

A HSG B HSG C HSG D HSG Unburned Unburned Unburned Unburned Moderate Moderate Moderate Moderate High High High High Cover Description Low Low Low Low

Herbaceous, Pasture, Alpine Meadow, Park 49 55 67 77 61 68 80 86 74 81 88 89 82 86 92 95 Oak-aspen—mountain brush mixture of oak brush, 45 52 65 77 48 55 65 86 57 70 80 89 63 70 80 92 aspen, mountain mahogany, bitter brush, maple, and other brush Ponderosa pine-juniper (grass understory) 49 57 65 77 58 65 75 86 73 78 83 89 80 85 90 92 Sagebrush (grass understory) 46 54 65 77 51 60 75 86 63 70 80 89 70 75 85 92 Lodgepole Pine Forest 49 57 65 77 60 65 70 86 73 78 83 89 79 83 87 92 Bare soil 77 77 77 77 86 86 86 86 91 91 91 91 94 94 94 94 Wetland 98 98 98 98 98 98 98 98 98 98 98 98 98 98 98 98

2832 2833 Rainfall 2834 Rainfall depths used in the modeling were extracted from NOAA Atlas 14, Vol 8 (Perica 2835 et al. 2013). 6-hour rainfall durations and NRCS Type II rainfall distributions were

DRAFT Technical Note, April 2015 C3-7 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2836 assumed. Three rainfall depths were used in the analysis, reflecting the variable spatial 2837 and elevation extent of the catchments draining the fire area. However, for each specific 2838 watershed analyzed, identical rainfall depths were used for all catchments. The specific 2839 rainfall depths for the 10-, 25-, 50- and 100-year floods are provided in table CS3-2. The 2840 implemented rainfall distributions for the 25-year event are shown in figure CS3-5. Note 2841 that the majority of the rainfall is assumed to fall within a half hour of the 6-hour duration 2842 event. 2843 2844 Table CS3- 2. Rainfall depths (inches) implemented in the modeling. Return Interval

10-year 25-year 50-year 100-year Zone A 2.17 2.59 2.93 3.28 Zone B 1.79 2.18 2.48 2.77 Zone C 1.99 2.48 2.91 3.40 2845 2846 Figure CS3- 5. 25-year cumulative rainfall distributions used in modeling.

3.0

Zone A 2.5 Zone B

2.0 Zone C

1.5

1.0

0.5 Precipitation (inches) Depth Precipitation 0.0 0 1 2 3 4 5 6

Time (hours)

2847 2848 For catchments with drainages areas greater than 11 square miles (≤ 0.98 reduction), an 2849 areal reduction factor was applied as detailed in Miller et al. 1973. Reduction varied from 2850 0.98 (Elk Creek) to 0.94 (Trout Creek). When applied, this area reduction was

DRAFT Technical Note, April 2015 C3-8 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2851 implemented in all catchments; flow may be underpredicted in the smaller, upper 2852 catchments of such drainages. 2853 2854 Lag Time 2855 Lag time (L), which is required to generate a hydrograph using the NRCS unit 2856 hydrograph methodology, was computed using the watershed lag method (USDA-NRCS 2857 2010). This equation is:

0.7 2858 lS0.8 ( +1) 2859 L = (eq. CS3-1) 1900Y 0.5 2860 2861 where l is flow length (ft), Y is average watershed land slope (%), and S is maximum 2862 potential retention (in). S may be calculated from equation CS1-2: 2863 1000 2864 S = −10 (eq. CS3-2) 2865 cn' 2866 where cn’ is the retardance factor and is approximately equal to the CN. This method 2867 allows the computation of differing lag times for pre- and post-fire conditions, reflecting 2868 the actual physical mechanism of more rapid flow response during post-fire conditions. 2869 Note that equation CS3-1 is equivalent to the time of concenration equation in the 2870 Technical Note main body (equation 1) since lag is considered to be 0.6 of the time of 2871 concentration. 2872 2873 Flow Routing 2874 A Muskingum-Cunge procedure was used to route flow from upper catchments to the 2875 stream outlets. This 1-dimensional method, embedded in HEC-HMS, allows for flow 2876 attenuation in the computations but does not provide a numerical solution of the full 2877 unsteady flow routing equations, as provided in such computational models as HEC- 2878 RAS. In each reach, flow routing was estimated using a single simplified cross section, 2879 channel slope, and Manning’s n roughness estimates. Manning’s n was selected using a 2880 visual estimation procedure, with a quality control step to assure that subcritical or

DRAFT Technical Note, April 2015 C3-9 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2881 approximately critical velocity was maintained, reflecting an assumption that existing or 2882 new channel bedform development prevents reach-average supercritical flow. 2883 2884 Sediment Bulking 2885 A simple multiplication factor was applied to the post-fire flood predictions to account 2886 for sediment bulking in the wildfire-induced floods. For burned catchments, this 2887 multiplication factor was assumed to be 1.25 if the severe + moderate soil burn severity 2888 aerial extent was greater than 50%, and 1.1 for catchments with between 15 and 50 % soil 2889 burn severity. 2890 2891 Erosion Modeling 2892 The RUSLE (Revised Universal Soil Loss Equation) model was chosen to estimate pre- 2893 and post-fire sediment related issues for the burn area because it is widely used, can be 2894 quick to execute with straight-forward factors and has a large amount of supporting 2895 documentation. The RUSLE models (pre- and post-fire) are based on a spatial version of 2896 RUSLE outlined in the Assessment of Threats to Riparian Ecosystems in the Western 2897 U.S. (ATREW) [Theobald et al., 2010] report. More recently, Litschert et al. 2014 used 2898 similar methods to investigate climate change effects on wildfire frequency and soil loss 2899 in the Southern Rockies Ecoregion. The ATREW methods entail calculating RUSLE 2900 (Equation CS3-3) the standard way, using widely available fine resolution spatial datasets 2901 to approximate the six RUSLE factors. The ATREW report also provides guidance on 2902 parameterizing the RUSLE C and P factors based on commonly used landcover datasets 2903 (e.g., USGS National Landcover Dataset and USFS Existing Vegetation Dataset), as well 2904 as equations that scale GIS based terrain analysis for the L and S factors. 2905 2906 A= RKLSCP ⋅ ⋅⋅⋅⋅ (eq. CS3-3) 2907 2908 where: 2909 A = average annual unit-area (tons/ha/yr) 2910 R = rain erosivity factor (MJ mm/(ha h yr)) 2911 K = soil erosivity factor (tons ha h/(ha MJ mm))

DRAFT Technical Note, April 2015 C3-10 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2912 L = slope length factor (m) 2913 S = slope steepness factor 2914 C = cover management factor (>= 0) 2915 P = management factor (>= 0) 2916 2917 The advantages of the ATREW approach are: 2918 • the simple model parameterization uses nationwide spatial datasets 2919 • the sedimentation rate raster allows better spatial resolution of erosion estimation 2920 • spatial evaluation of sediment yield helps prioritize soil treatment zones and 2921 emergency resource allocation 2922 2923 The sediment modeling entailed four general steps: 2924 1) collection of geospatial dataset for the greater burn area (table CS3-3) 2925 2) development of spatial RUSLE factors for pre- and post-fire conditions 2926 3) calculation of RUSLE for pre- and post-fire scenarios (using ArcGIS Raster 2927 Calculator) 2928 4) attribution of computation points and values at risk with pre- and post-fire 2929 sedimentation rates. 2930 2931 Pre-fire and post-fire sedimentation rate estimates were executed in GIS using terrain 2932 analysis tools to calculate slope length (L) and steepness (S) factors with simple map 2933 algebra statements. Computation of rainfall erosivity (R), soil erodibility (K) and cover 2934 management (C) factors used ancillary spatial datasets. The soil/cover management (P) 2935 factor was not incorporated into the analysis due to a lack of spatial information on 2936 management activity in the burn area. For each of the five RUSLE factors used, a 10- 2937 meter resolution raster dataset was generated. The five RUSLE factor rasters where 2938 multiplied together to calculate the local (cell level) sedimentation rate. These local rate 2939 values were accumulated downslope via a flow direction raster (figure CS3-6) and 2940 averaged by the contributing area above each raster cell. This results in the final 2941 sedimentation rate raster with values representing the average cumulative sedimentation 2942 rate in tons per year over 30 years for each scenario.

DRAFT Technical Note, April 2015 C3-11 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2943 2944 Table CS3- 3. Spatial datasets used for the six RUSLE factors. RUSLE Website Description factor

EPA EMAP RUSLE factors http://www.epa.gov/esd/land- R summarized to HUC 8 units for the sci/emap_west_browser/pages/wemap_mm_sl_table.htm#mapnav Western United States. NRCS SSURGO data down-loaded K http://datagateway.nrcs.usda.gov for El Paso County CO 10x10 meter elevation model L and S http://viewer.nationalmap.gov/viewer downloaded for HUC 8 areas associated with the burn area 30x30 meter landcover dataset for C http://earth.gis.usu.edu/swgap/landcover.html NV, AZ, UT and CO Parameter not used in analysis due P to lack of good spatial data 2945 2946 Figure CS3- 6. Flow direction raster cell example*

2947 *A. shows the eight directions of flow from center cell. 2948 B. shows the possible flow direction raster values. 2949 C. shows an example cell with a north flow direction. 2950 D. shows the flow direction raster value of the cell in C. 2951 E. shows a flow direction grid with the color scheme of A. 2952 F. shows a flow accumulation raster, with each cell encoding the number of cells that drain to it. 2953

DRAFT Technical Note, April 2015 C3-12 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2954 The rain erosivity (R) factor raster was generated by converting an EPA EMAP HUC 8 2955 polygon shapefile to a raster containing R factor values. Due to a lack of information 2956 about how wildfire changes rain erosivity values and the R factor raster was held constant 2957 between the pre- and post-fire scenarios. In addition, the EPA EMAP values are based 2958 on 30-year averages. 2959 2960 The development of soil erodibility (K) factor raster entailed summarizing KFFACT 2961 (SSURGO table attribute) to NRCS SSURGO map units and then rasterizing the map 2962 units in the same manner as the R factor. KFFACT (property of a soil horizon) was 2963 summarized to map unit delineations by calculating a depth/area weighted average based 2964 on horizon depths up to 15 centimeters and the component percent within a map unit. 2965 This was accomplished through queries developed in the SSURGO database downloaded 2966 from the USDA Geospatial Data Gateway website (see table CS3-3). The K factor raster 2967 was held constant for both pre- and post-fire scenarios even though burn severity alters 2968 soil erodibility. Altering soil erodibility based on burn severity between scenarios could 2969 be incorporated in future models but would require additional research and 2970 parameterization. 2971 2972 The L and S factors were calculated jointly (LS) using basic terrain analysis methods 2973 outlined in Theobald et al. (2010) using a 10 meter elevation model. These methods 2974 include calculating a percent slope, aspect (radians) and accumulated upslope length. The 2975 accumulated upslope length process entailed accumulating number of contributing raster 2976 cells to a given cell based on the overland flow paths from the flow direction raster 2977 (figure CS3-6). The resulting slope, aspect, and upslope length rasters were transformed 2978 using equations developed by Winchell et al. (2008), given herein as equation CS3-4, 2979 and Nearing (1997), given herein as equation CS3-6. These equations scale the values 2980 derived from the above terrain analysis to better fit within the framework of the RUSLE 2981 equation and ensure that the units are correct. 2982 m+1 2983 ( AD+−21) Am+ 2984 LS= S  (eq. CS3-4) xDmm∗∗+2 22.13 m  DRAFT Technical Note, April 2015 C3-13 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 2985 2986 2987 where: L = transformed slope length at the given raster cell, S = RUSLE slope factor of 2988 the cell 2989 A = contributing drainage area to the cell (m2), D = raster cell size (10m) 2990 x = aspect transformation (equation CS3-5) 2991 m = percent slope transformation (equation CS3-6) 2992 2993 x = sin αα + cos (eq. CS3-5) 2994 2995 where: α = land aspect of the raster cell in radians, clockwise from north 2996 β 11.16sinθ 2997 m = where: β = (eq. CS3-6) 1+ β θ + 2998 3sin 1.68 2999 and where θ = land slope (in percent) 3000 3001 The raster of the slope length (LS) factor was developed using the Nearing (1997) 3002 equation (given herein as equation CS3-7) in conjunction with equations CS3-5 and CS3- 3003 6, which use aspect and slope dynamics to adjust for inaccuracies caused by scale. This is 3004 necessary because the raster values of the area draining to each cell (A, in square meters) 3005 can get very large and inflate sedimentation estimates. The raster for S factor was 3006 developed by transforming percent slope using equation CS3-7. That equation scales 3007 slope values to reduce inflated soil loss calculations, especially for slopes greater than 3008 50%. The L and S factor rasters where multiplied together to using ArcGIS Raster 3009 Calculator to create the raster for LS factor. As with the R and K factors the L and S 3010 factors were held constant between the pre- and post-fire models. 3011 17 3012 = + (eq. CS3-7) S 1.5 (2.3− 6.1sinθ ) 3013 1+ e 3014 where θ = land slope (in percent) 3015

DRAFT Technical Note, April 2015 C3-14 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3016 The C factor parameterization for the pre- and post-fire scenarios was developed using 3017 various source tables from different documents related to RUSLE. The pre-fire scenario 3018 parameterization involved developing a lookup table that assigns the existing landcover 3019 types (Southwest ReGAP) within the greater burn area to their associated C factors. (See 3020 table CS3-4). The table was compiled by Theobald et al. (2010) for the ATERW report 3021 and provides a broad spectrum of landcovers found in most landcover datasets and can be 3022 modified based on local knowledge. The post-fire C factor parameterization entailed 3023 modifying the pre-burn C factor raster based on burn severity classes derived from the 3024 Burned Area Reflectance Classification (BARC) image. This process consisted of 3025 assigning the BARC burn severity classes C factor values (low burn = 1.03, moderate 3026 burn = 2.25 and high burn 3.75) (Larsen et al., 2007) that were then used to modify the 3027 pre-fire C factors by summing the two rasters together. Larsen et al. (2007) estimated that 3028 high burn severity area C factors changed by four hundred percent but didn’t estimate 3029 moderate and low burn severity changes. For these, C factors changes where estimated 3030 using professional judgment. 3031

DRAFT Technical Note, April 2015 C3-15 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3032 Table CS3- 4. C factors for common land covers (Theobald et al. 2010) Land cover C Factor Citation Barren 1.0000 Toy and Foster 1998 Coniferous Forest 0.0020 Breiby 2006 Deciduous Forest 0.0010 Breiby 2006 Deciduous Shrubland 0.0250 Breiby 2006 Dense Grassland 0.0800 Dawen et al. 2003 Floodplain Forest 0.0100 Breiby 2006 Lowland Coniferous Forest 0.0025 Breiby 2006 Lowland Deciduous Forest 0.0015 Breiby 2006 Marsh/Riparian/Wetland 0.0010 Breiby 2006 Medium-tall grassland 0.0120 Breiby 2006 Mixed Forest 0.0010 Breiby 2006 Mixed Forest woodland 0.0020 Breiby 2006 Open Water/Exposed Rock 0.0000 Breiby 2006; McCuen 1998 Shrubland Other 0.0290 McQuen 1998 Snow field 0.0010 Dawen et al. 2003 Sparse Grassland 0.2000 Dawen et al. 2003

Aggregate mining 1.0000 Guobin et al. 2006 Asphalt 0.0001 Toy and Foster, 1998 Cultivated Crops Irrigated 0.2400 McCuen, 1998 Developed General 0.0030 Guobin et al. 2006 Developed Suburban 0.0020 Guobin et al. 2006 Developed Urban 0.0010 Guobin et al. 2006 Fallow 1.0000 McCuen, 1998 General Cropland 0.5000 Dawen et al., 2003 Gravel 0.2000 Toy and Foster, 1998 Industrial 0.0050 Guobin et al. 2006 Mixed Urban 0.0040 Guobin et al. 2006 Paddy field 0.1000 Dawen et al., 2003 Pasture Hay 0.1400 McCuen, 1998 Recreational Grasses 0.0080 McCuen, 1998 Small Grains 0.2300 McCuen, 1998

3033 3034 The final estimates of sedimentation rate for the pre- and post-fire scenarios were 3035 generated by multiplying the 5 factors, accumulating the those values downslope via the 3036 flow direction raster, and calculating an area-weighted sedimentation rate. The latter is 3037 determined by dividing the accumulated sedimentation rate by the total accumulated 3038 drainage area. Figures CS3-7 shows how sediment yield varies in different zones on a 3039 hillslope. The transport of sediment from areas where it originates (steep slopes or burned

DRAFT Technical Note, April 2015 C3-16 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3040 areas) dampens with decrease in slope, increase in flow distance, or interruption by 3041 unburned areas. Figure CS3-8 is an example of an area-weighted sediment yield raster, 3042 showing that the high loss areas (red) dampened with flattening slope and distance. 3043 3044 Figure CS3- 7. Conceptual diagram of sediment yield estimation on hillslope zone breaks

3045

DRAFT Technical Note, April 2015 C3-17 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3046 Figure CS3- 8. Example of an area-weighted RUSLE sediement yield raster

3047 3048 Results and Discussion 3049 Hydrologic modeling was performed to develop estimates of increased flood hazard and 3050 erosion potential for streams draining the West Fork Complex Wildfire. For several key 3051 catchments, at locations where there are threats to residences, hydrographs such as in 3052 figure CS3-9 show the expected response to a 25-year rainfall depth over each entire 3053 catchment. Substantially increased flow peaks, runoff volumes, and sediment were 3054 estimated. Using a map presentation style (figures CS3-10) that is simple for planners, 3055 designers, and emergency response officials to utilize, results were provided at pour 3056 points throughout and downstream of the fire extents. Results were also provided as 3057 attributes in ArcGIS shapefiles. For detailed results, refer to the project report (Yochum 3058 and Norman 2014). 3059

DRAFT Technical Note, April 2015 C3-18 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3060 Figure CS3- 9. Example of estimated pre- and post-fire hydrographs for the 25-year event.

3061 3062 Figure CS3- 10. Example map providing pre- and post-fire flood predictions for the S. F. of the Rio 3063 Grande.

3064

DRAFT Technical Note, April 2015 C3-19 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3065 Runoff Modeling Results 3066 In many catchments, post-fire conditions are predicted to cause a 50- or 100-year (pre- 3067 fire) flood to result from a 10-year rain event on burned landscapes, similar to actual 3068 measured fire runoff responses (Conedera et al. 2003). Peak flow magnification ratios for 3069 the 25-year rainfall event are provided (figure CS3-11). These ratios are computed as the 3070 post-fire peak flow divided by the pre-fire peak flow; a value of 3 indicates that 3 times 3071 the peak flow is predicted from the burned landscape for the identical rain event. 3072 3073 Figure CS3- 11. Peak flow magnification ratios for the West Fork Complex wildfire.

3074 3075 The fire results in a 41 percent risk that a 10-year rain event will cause runoff that, pre- 3076 fire, would have been 50- to 100-year in recurrence. These substantial fire impacts on 3077 runoff may persist for at least 5 years. However, as catchment size increases the small 3078 spatial extent of typical convective storms will reduce the severity of the flood effects 3079 from these storms. 3080 3081

DRAFT Technical Note, April 2015 C3-20 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3082 Sediment Modeling Results 3083 Sediment modeling was intended to provide estimated sediment magnification ratios at 3084 hydrologic pour points, as well as approximations of sediment yield. Sediment 3085 magnification ratios were calculated by taking the ratio between the predicted post- and 3086 pre-fire sediment flux values (tons/yr) to compute the magnitude of change. The true 3087 magnification ratios and quantities for the pre- and post-fire scenarios are unknown, and 3088 these estimates are based on a lumped modeling approach (RUSLE) that provides values 3089 with substantial expected error. However, the results do provide a reasonable framework 3090 for making management decisions. An overview of the sediment magnification ratios is 3091 provided in figure CS3-12. 3092 3093 Figure CS3- 12. Sediment magnification ratios for the West Fork Complex wildfire.

3094 3095 The sedimentation analysis for the entire West Fork Complex Wildfire predicts that the 3096 Upper Rio Grande River will experience on average 130 times more sediment annually 3097 above pre-fire conditions (approximately 1,030,000 tons/yr). This is based on sediment 3098 rates estimated for the 27 outlet pour points that are hydrologically connected to the burn

DRAFT Technical Note, April 2015 C3-21 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3099 area. These pour point estimates account for all sediment production and transport that 3100 occurs upstream of them. The Papoose Wildfire accounts for 65% of the total sediment 3101 yields in the West Fork Complex with 70% of its outlets having sediment magnification 3102 ratios greater than 100. The West Fork Wildfire area has lower sediment magnification 3103 ratios with 50% of its outlets having ratios higher than 100, but it has the highest yielding 3104 catchment (Hope Creek HC) which accounts for about 18% of the total sediment yield of 3105 the entire West Fork Fire Complex area. 3106 3107 The majority of sediment produced by the Papoose Wildfire area is from two large 3108 catchments (Little Squaw Creek and Trout Creek) which together account for about 50% 3109 of its total sediment yield. The other 17 outlet pour points do not account for as much 3110 sediment on an individual basis, but several (CIC-1, PC-1, UN-1, UN-2 and WC-1 in 3111 figure CS3-4) have sedimentation magnification rates that exceed 500 times pre-fire 3112 conditions. These catchments are small, but have steep slopes with high proportions of 3113 moderate and high burn severities which could pose a mass movement risk. The pour 3114 points that make up Trout Creek all have sediment magnification ratios greater than 100 3115 with TC-6 (figure CS3-4) having the highest sedimentation rate. The Little Squaw Creek 3116 catchment of the Upper Rio Grande has magnification ratios that increase dramatically 3117 downstream with a large bump in sediment yield at the outlet point (LSC-4). The Texas 3118 Creek catchment has magnification ratios that increase up to the midpoint of the 3119 catchment (TxC-3) and then decreases to the outlet point with sediment yields still 3120 increasing steadily downstream. 3121 3122 The majority of sediment produced by the West Fork Wildfire is dominated by two 3123 catchments (Goose Creek, and Hope Creek) that together produce 75% of the its total 3124 sediment yield. The analysis for Hope Creek predicts sediment magnification ratios 3125 greater than 400 with a peak ratio of 605 at HC-2, dropping back to 483 at the outlet. This 3126 catchment along with LF-4, CbC-1, and DC-3 (figure CS3-4) connect to the South Fork 3127 of the Rio Grande River which will increase its sediment yield by approximately 250,000 3128 tons annually. The Goose Creek analysis predicts that sediment magnification ratios and 3129 yields increase dramatically to pour point GC-4 and then decrease by half at the outlet

DRAFT Technical Note, April 2015 C3-22 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3130 pour point at Lake Humphreys, where a post-fire sediment yield of approximately 85,400 3131 tons/year was simulated. The other four outlets DC-3, TCE-3, EC-3 and Lec-5 have 3132 moderately low sediment magnification ratios, with a total yield of approximately 3133 390,000 tons/year. 3134 3135 Modeling Limitations 3136 As with all hydrologic modeling, the results provided by these rainfall-runoff and 3137 sediment models are approximate. The best use of these results is on a comparative basis, 3138 to help identify points of concern and develop mitigation priorities. Absolute values of 3139 runoff and sediment yield are approximate and use of these values for infrastructure 3140 sizing and estimating reservoir sedimentation rates should be done with caution. Specific 3141 limitations in the modeling methods are discussed below. 3142 3143 Rainfall-Runoff Modeling Limitations 3144 Post-fire runoff prediction using the CN technique is hampered by the very little available 3145 field data available to reliably select CN values from measured rainfall and runoff in 3146 burned catchments. With CN values for burned conditions primarily compiled from 3147 various grey literature and unpublished sources, they should be considered approximate. 3148 As a result, consistent application of these values result in model results that are more 3149 dependable in a comparative manner. Catchment-scale research of wildfire-impacted 3150 areas is needed to develop more robust post-wildfire CN estimates. 3151 3152 An additional complication arises from the peak rate factor imbedded in the NRCS 3153 dimensionless unit hydrograph method. Within HEC-HMS, this peak rate factor cannot 3154 be altered from the default value, despite an understanding that this factor tends to 3155 increase with steeper watersheds. The ability to adjust the peak rate factor (an option in 3156 WinTR-20) would be a valuable addition to HEC-HMS. For more information, see the 3157 technical note section “NRCS dimensionless unit hydrograph.” 3158 3159 Most fundamentally, the reliability of the CN method for predicting peak flow from 3160 forested, mountainous watersheds is questionable. Forested watersheds in unburned

DRAFT Technical Note, April 2015 C3-23 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3161 conditions may be dominated by saturation-excess overland flow (figure CS3-13), where 3162 runoff is produced from relatively small and variable portions of a catchment when 3163 rainfall depths exceed the soil capacity to retain water. Newly burned catchments, on the 3164 other hand, may be dominated by infiltration-excess (Hortonian) overland flow, where 3165 surface runoff is generated when rainfall intensity is greater than soil infiltration capacity, 3166 and flow runs down the hillslope surface. Evidence of this surface runoff is provided by 3167 such features as surface rilling on freshly-burned hillslopes. Rainfall-runoff modeling 3168 performed in the San Dimas Experimental Forest (Chen et al. 2013) found that pre-fire 3169 runoff predictions were more accurate using the CN method, while KINEROS2 3170 performed better for post-fire conditions. These results suggest fundamental shifts in 3171 runoff mechanisms between pre- and post-fire conditions, complicating modeling 3172 strategies. 3173 3174 Figure CS3- 13. Infiltration excell and saturation excess overland flow (Laboratory of Ecohydrology,

3175 EPFL) 3176 3177 Despite this shortcoming, due to its relative simplicity and achievable data requirements 3178 on large scales, as well as reasonable results when qualitatively compared to actual post- 3179 fire runoff events, the CN method is often considered a preferred tool for predicting flow 3180 response of wildfire areas. Alternatives approaches can be problematic, due to the need to 3181 quantify infiltration rates, the effects of soil litter and hydrophobicity on retention and 3182 infiltration for pre- and post-fire conditions, as well as vegetation canopy rainfall 3183 interception. This accounting would need to be performed at a spatial resolution

DRAFT Technical Note, April 2015 C3-24 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3184 appropriate for obtaining the needed flood predictions throughout the entire wildfire 3185 extent, which can be a very challenging task with the limited time and resources allotted 3186 to practitioners for prediction. 3187 3188 Sediment Modeling Limitations 3189 The ATREW method used to estimate pre- and post-fire sediment yields was originally 3190 developed to estimate sedimentation rates to riparian areas based on different landuse and 3191 climate change scenarios for the Western U.S. (See Theobald et al. 2010.) This scaling 3192 issue along with limited information on parameterizing C factors for burned vegetation, 3193 inherent errors associated with the spatial data and the lumped approach of the RUSLE 3194 equation makes predicted sedimentation rates unreliable. This was evident when Larsen 3195 et al. (2007) evaluated spatial versions of RUSLE and WEPP performance using 3196 established plots for fires on the Front Range of Colorado and found that the models 3197 performed poorly due to unaccountable variability not incorporated into the models. To 3198 address this weakness, it is preferred that sediment yield modeling focus on the 3199 magnitude of change in soil loss and not just to amount produced post-fire. The 3200 magnitude values are based on an annual average increase in sediment yields for 30 3201 years. This means that actual sediment yields for a given year will fluctuate around the 3202 predicted value, but over 30 years may be close to the predicted annual increase in 3203 sediment. Nevertheless, the ATREW method, used to estimate pre- and post-fire 3204 sedimentation rates and the magnitude of sedimentation post-fire, provides a robust 3205 framework for managers to help prioritize and guide timely post-fire mitigation activities. 3206 3207 Conclusions 3208 Using the Curve Number and RUSLE methods, peak flow and erosion potential 3209 predictions were made for streams draining the West Fork Complex wildfire area, for 3210 both pre- and post-wildfire conditions. Watershed maps for each modeled watershed were 3211 developed, illustrating pour points, peak discharges, peak flow and sediment 3212 magnification ratios, soil burn severity, and stream outlet hydrographs. While many 3213 relevant hydrologic mechanisms were simulated in the rainfall-runoff modeling, 3214 including rainfall depth and spatial extent; variation in runoff by soil burn severity,

DRAFT Technical Note, April 2015 C3-25 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3215 vegetation type and soil conductivity; variable lag times; and stream attenuation, the 3216 questionable reliability of the CN method in forested watersheds, for both post- and pre- 3217 fire conditions, adds a level of undefined uncertainty to the estimates. 3218 3219 The pre- and post-fire sedimentation analysis used landcover solely as a proxy for 3220 changes in soil erodibility, rainfall erosivity and cover post-fire. This allows for a rapid 3221 response in estimating sediment related issues post-fire, but could be enhanced if soil 3222 erodibility and rainfall erosivity were incorporated. Additional research is needed to 3223 address these uncertainties, especially since lives are often at risk. In the meantime, 3224 wildfires will occur and methods need to be available to predict the expected flood 3225 response and sediment release from burned watersheds. This case study provides 3226 examples for simulation approaches that can be relatively-rapidly deployed after a 3227 wildfire. 3228 3229 Acknowledgements 3230 Appreciation is expressed to Dan Moore, Hydraulic Engineer at the West National 3231 Technology Support Center, for peer review. Appreciation is also expressed to Rodney 3232 Clark, NRCS Area Engineer, and Beau Temple, NRCS Soil Conservation Technician, for 3233 field data collection activities documenting channel condition for the modeled streams. 3234 Additionally, the U.S. Forest Service is appreciated for providing heat perimeter and soil 3235 burn severity mapping data. 3236 3237 References 3238 Benavides-Solario, J.D., and MacDonald, L.H. 2001. Post-fire runoff and erosion from 3239 simulated rainfall on small plots, Colorado Front Range. Hydrological Processes, 15, 3240 2931-2952. 3241 3242 Breiby, T. 2006. Assessment of soil erosion risk within a subwatershed using GIS and 3243 RUSLE with a comparative analysis of the use of STATSGO and SSURGO soil 3244 databases. Volume 8, Papers in Resource Analysis. Saint Mary’s University of Minnesota 3245 Central Services Press. Winona, MN. 22 pages.

DRAFT Technical Note, April 2015 C3-26 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3246 3247 Capesius, J.P., and Stephens, V.C. 2009. Regional Regression Equations for Estimation 3248 of Natural Streamflow Statistics in Colorado. U.S. Department of Interior, U.S. 3249 Geological Survey, Scientific Investigations Report 2009-5136. 3250 3251 Chen, L., Berli, M., Karletta, C. 2013. Examining modeling approaches for the rainfall- 3252 runoff process in wildfire-affected watersheds: using San Dimas Experimental Forest. 3253 Journal of the American Water Resources Association 1-16. 3254 3255 Conedera, M., Peter, L., Marxer, P., Forster, F., Rickenmann, D., and Re, L. 2003. 3256 Consequences of forest fires on the hydrogeological response of mountain catchments: A 3257 case study of the Riale Buffaga, Ticino, Switzerland. Earth Surface Processes and 3258 Landforms 28, 117-129. 3259 3260 Dawen, T., Shinjiro, K., Tailkan, O., Toshio, K., and Katumi, M. 2003. Global potential 3261 soil erosion with reference to land use and climate changes. Hydrologic Processes 17, 3262 2913-2928. 3263 3264 Elliot, W.J., Elliot, A.V., Qiong, W., and Laflen, J.M. 1991. Validation of the WEPP 3265 model with rill erosion plot data, paper presented at the 1991 ASAE International Winter 3266 Meeting, Am. Soc. Of Agric. Eng., Chicago, Ill, Dec. 1991. 3267 3268 Guobin, T., Shulin, C., and Donald, M.K. 2006. Modeling the impacts of no-till practice 3269 on soil erosion and sediment yield with RUSLE, SEDD, and ArcView GIS. Soil & 3270 Tillage Research 85, 38-49. 3271 3272 Hudak, A.T., Robichaud, P.R., Evans, J.S., Clark, J., Lannom, K., Morgan, P., and Stone, 3273 C. 2004. Field Validation of Burned Area Reflectance Classification (BARC) Products 3274 for Post Fire Assessment. Remote Sensing for Field Users: Proceedings of the Tenth 3275 Forest Service Remote Sensing Applications Conference, Salt Lake City, Utah, April 5-9 3276 2004.

DRAFT Technical Note, April 2015 C3-27 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3277 3278 Huffman, E.L., MacDonald, L.H., and Stednick, J.D. 2001. Strength and persistence of 3279 fire-induced soil hydrophobicity under ponderosa and lodgepole pine, Colorado Front 3280 Range. Hydrologic Processes, 15, 2877-2892. 3281 3282 Larsen, I. J., and MacDonald, L. H. 2007. Predicting postfire sediment yields at the 3283 hillslope scale: Testing RUSLE and Disturbed WEPP, Water Resources Res., 43 3284 (W11412): 1-18. 3285 3286 Litschert, S.E., Theobald, D.M., and Brown, T.C. 2014. Effects of climate change and 3287 wildfire on soil loss in Southern Rockies Ecoregion. Catena 118, 206-219. 3288 3289 Maidment, D.R. 1992. Handbook of Hydrology McGraw-Hill, Inc. 3290 3291 McCuen, R.H. 1998. Hydrologic analysis and design. Second edition. Prentice-Hall, Inc., 3292 Upper Saddle River, New Jersey. 3293 3294 Miller, J.F., Frederick, R.H., and Tracey, R.J. 1973. Precipitation-Frequency Atlas of the 3295 Western United States. U.S. Department of Commerce, National Oceanic and 3296 Atmospheric Administration, National Weather Service, NOAA Atlas 2, Vol 3, Silver 3297 Spring, Maryland. 3298 3299 Nearing, M.A. 1997. A single continuous function for slope steepness influence on soil 3300 loss. Soil Science Society of America 61(3): 917-919. 3301 3302 Perica, S., Martin, D., Pavlovic, S., Roy, I., St. Laurent, M., Trypaluk, C., Unruh, D., 3303 Yekta, M., Bonnin, G. 2013. NOAA Atlas 14 Volume 8 Version 2, Precipitation- 3304 Frequency Atlas of the United States, Midwestern States. NOAA, National Weather 3305 Service, Silver Spring, MD. 3306

DRAFT Technical Note, April 2015 C3-28 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3307 Rallison, R.E. 1980. Origin and Evolution of the SCS Runoff Equation. Proceedings of 3308 the Symposium on Watershed Management; American Society of Civil Engineers, Boise, 3309 ID. 3310 3311 Renard, K.G., Foster, G.R., Weesies, G.A., McCool, D.K., and Yoder, D.C. 1997. 3312 Predicting soil erosion by water: A guide to conservation planning with the revised 3313 universal soil loss equation (RUSLE), Agric. Handb. 703, 404 pp., USDA, Washington, 3314 D. C. 3315 3316 Renschler, C. S. 2003. Designing geo-spatial interfaces to scale process models: The 3317 GeoWepp approach, Hydrol. Processes, 17, 1005-1017. 3318 3319 Robichaud, P.R. 2000. Fire effects on infiltration rates after prescribed fire in northern 3320 Rocky Mountain forests, USA, J. Hydrol., 231-232, 220-229. 3321 3322 Stoof, C.R., Vervoort, R.W., Iwema, J., van den Elsen, E., Ferreira, A.J.D., Ritsema, C.J. 3323 2011. Hydrological response of a small catchment burned by an experimental fire. 3324 Hydrology and Earth System Sciences Discussions 8, 4053-4098. 3325 3326 Theobald, D.M., Merritt, D.M., and Norman, J.B., III. 2010. Assessment of threats to 3327 riparian ecosystems in the Western U.S. A report presented to The Western 3328 Environmental Threats Assessment Center, Prineville, OR, by the USDA Stream Systems 3329 Technology Center and Colorado State University, Fort Collins, 61p. 3330 3331 Toy, T.J, and Foster, G.R., co-editors. 1998. Guidelines for the use of the Revised 3332 Universal Soil Loss Equation (RUSLE) version 1.06 on mined lands, construction sites, 3333 and reclaimed lands. Office of Surface Mining and Reclamation (OSM). Western 3334 Regional Coordinating Center, Denver, Colorado. 3335 3336 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 3337 Service), 2004a.

DRAFT Technical Note, April 2015 C3-29 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 3: West Fork Complex Fire, Colorado 3338 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 9, Hydrologic Soil 3339 Cover Complexes. 3340 3341 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 3342 Service), 2004b. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 3343 10, Estimation of Direct Runoff from Storm Rainfall. 3344 3345 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 3346 Service), 2010. NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 3347 15, Time of Concentration. 3348 3349 USGS National Gap Analysis Program. 2004. Provisional digital land cover map for the 3350 Southwestern United States. Version 1.0. RS/GIS Laboratory, College of Natural 3351 Resources, Utah State University. 3352 3353 Winchell, M.F., Jackson, S.H., Wadley, A.M., and Srinivasan, R. 2008. Extension and 3354 validation of a geographic information system-based method for calculating the Revised 3355 Universal Soil Loss Equation length-slope factor for erosion risk assessments in large 3356 watersheds. Journal of Soil and Water Conservation 63(3): 105-111. 3357 3358 Woodward, D. 2005. Personal Communications (email). Retired National Hydraulic 3359 Engineer, USDA Natural Resources Conservation Service, Washington D.C. 3360 3361 Yochum, S., and Norman, J. 2014. West Fork Complex Fire: Potential Increase in 3362 Flooding and Erosion. USDA Natural Resources Conservation Service, Colorado State 3363 Office, Denver, CO. 3364

DRAFT Technical Note, April 2015 C3-30 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3365 Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3366 3367 Background 3368 Lightning sparked several outbreaks of wildfire in May 2012 that joined to became the 3369 largest in New Mexico recorded history. Lack of road access and the steep terrain of the 3370 Gila Wilderness hampered containment efforts, allowing the individual fires to grow 3371 rapidly. By the end of May, they had become one large wildfire known as the 3372 Whitewater-Baldy Complex. 3373 3374 The rugged landscape was not the only reason these fires escaped control. The watershed 3375 had already been under extreme drought conditions, with the two previous winters 3376 recording very low snowfall. Air temperatures at the time of the outbreak were well 3377 above average, and high winds contributed to the merging of the fires. In early June the 3378 Whitewater-Baldy Complex was only about 18 percent contained. By mid-June 3379 containment increased to 56%. For about three months the wildfire burned Ponderosa, 3380 Pinon/Juniper, and mixed conifer forests with relatively low burn temperatures. Rainfall 3381 in mid-July finally helped fire fighters gain momentum, with 95% containment attained 3382 in late July. 3383 3384 The wildfire had burned about 465 square miles. The USDA Forest Service estimated 3385 the final burn severity for the area within the fire perimeter to be 14% high, 12% 3386 moderate, 55% low or unburned, and 20% unknown (due to inadequate satellite 3387 imagery). 3388 3389 This case study presents a hydrologic analysis of the flood potential of Whitewater Creek 3390 and possible hazard to the community of Glenwood NM. As shown in figure CS4-4, 3391 Glenwood is located near the confluence of Whitewater Creek and the San Francisco 3392 River. Considering the drainage area upstream of Glenwood, the percent of the 3393 watershed burned was 34% high severity, 21% medium severity, and 13% low severity. 3394

DRAFT Technical Note, April 2015 C4-1 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3395 Figure CS4- 1. Location of concern: State of New Mexico with county boundaries, USFS Gila 3396 National Forest in green, Gila Wilderness area in yellow.

3397 Inset shown in figure CS4-2 3398 3399 Figure CS4- 2. Location of concern, with Whitewater-Baldy Complex Fire location in red.

3400 Iinset shown in figure CS4-3 3401

DRAFT Technical Note, April 2015 C4-2 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3402 Figure CS4- 3. Location of concern, showing fire location in red, nearby communities and highways.

3403 Map provided by the USGS Gila National Forest. 3404 3405 Figure CS4- 4. Whitewater Creek watershed and the community of Glenwood NM.

3406

DRAFT Technical Note, April 2015 C4-3 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3407 3408 NRCS Involvement 3409 The NRCS Emergency Watershed Protection (EWP) program was implemented for the 3410 installation of monitoring gages in the Whitewater-Baldy Complex burn area, to provide 3411 early warning to downstream communities of flood potential. The intensity of the 3412 wildfire and its large area prompted the USDA Forest Service, in cooperation with the 3413 United States Geological Survey (USGS), to seek EWP funding from NRCS. The NRCS 3414 New Mexico State office worked with the New Mexico Department of Homeland 3415 Security & Emergency Management, which acted as the EWP local sponsor. USGS 3416 installed a streamflow gage on Whitewater Creek, near the Catwalk Recreational Area 3417 (figure CS4-5). That station also received a precipitation gage. The existing NRCS 3418 Snotel site is shown in figure CS4-5, along with additional newly installed gages. 3419 3420 Figure CS4- 5. Gages in the vicinity of Whitewater Creek.

3421 1 = Catwalk streamflow and precipitation (activated 4 July 2012 by USGS) 3422 2 = Hummingbird Saddle precipitation (activated 29 June 2012 by USGS) 3423 3 = Silver Creek Divide SNOTEL site (activated 1 Oct 1978 by NRCS) 3424 4 = Mogollon Baldy Lookout precipitation (activated 28 June 2012 by USGS) 3425 5 = Bear Wallow Lookout precipitation (activated 29 June 2012 by USGS) 3426

DRAFT Technical Note, April 2015 C4-4 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3427 NRCS performed a hydrologic analysis to assess flood potential for the community of 3428 Glenwood. In recent years, New Mexico residents had witnessed flash flooding after 3429 wildfire. The watersheds of the 2011 Las Conchas fire near Los Alamos experienced 3430 monsoon rains within weeks of the end of the fire, which sent floodwater, sediment, and 3431 debris rapidly downstream. That wildfire burned 245 square miles and held the record 3432 for the largest in recorded New Mexico history, until eclipsed by the Whitewater-Baldy 3433 Complex wildfire, almost twice as large. 3434 3435 The USDA Forest Service, through the BAER team process, also provided a hydrologic 3436 assessment. In addition, the USGS held a workshop in early July 2012 for Glenwood 3437 residents to learn how to access the early warning data from the new gages. The data are 3438 on the web at: http://nm.water.usgs.gov/wildfire/. 3439 3440 The NRCS hydrologic modeling effort, documented in this case study, examined pre-fire 3441 floods and post-fire floods, both immediately after the fire and one year later. Data from 3442 the newly installed precipitation and streamflow gages was used to verify that the 3443 hydrologic model produced reasonable results. 3444 3445 Figure CS4- 6. Burned trees near Hummingbird Saddle, Gila National Forest,

3446 Photo by Mike Sanders, USGS.

DRAFT Technical Note, April 2015 C4-5 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3447 3448 Methods and Application 3449 As an alternative to the often used runoff curve number (CN) method, this case study 3450 modeled infiltration of rainfall by the process-based Green & Ampt (GA) method. Both 3451 the CN and GA methods are discussed in the main body of this Technical Note. It should 3452 be noted that CN runoff is not directly comparable to GA infiltration, in that the CN 3453 runoff equation accounts for more phenomena than infiltration. These include losses due 3454 to transpiration and surface ponding, which, if using GA, must be modeled separately. 3455 3456 In this case study, runoff transformation was also performed differently than the NRCS 3457 dimensionless unit hydrograph (UH). The standard NRCS UH peak rate factor of 484, 3458 determined with Technical Note equation 9, is usually too low for mountain streams. The 3459 NRCS National Engineering Manual hydrology chapter 15 (USDA-NRCS 2010) 3460 provides dimensionless unit hydrographs for peak factors up to 600, which may be used 3461 in the hydrology program WinTR-20. However, even a peak factor of 600 is often too 3462 low for mountain streams in the unburned condition, and the effect of wildfire is to raise 3463 it further. Technical Note figure 17 shows the pre-fire and post-fire unit hydrographs for 3464 one sub-basin of Whitewater Creek (this case study), demonstrating that the peak factor is 3465 raised by the fire event from 720 to 769. 3466 3467 Athough runoff transformation for Whitewater Creek also used the unit hydrograph 3468 technique, the shape and peak of the UH were determined with time-area histograms. 3469 Neither peak factor nor time of concentration were needed. The hydrology model HEC- 3470 HMS (USCOE-HEC 2013) accepts input of user-derived unit hydrographs, as well as the 3471 GA infiltration method. The following list summarizes the modeling options used in this 3472 case study: 3473 • choice of hydrologic computer model: HEC-HMS (USCOE-HEC 2013) 3474 • runoff transformation: synthetic UH derived by GIS analysis of time-area 3475 histograms 3476 • infiltration loss method: Green & Ampt equation

DRAFT Technical Note, April 2015 C4-6 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3477 • initial abstraction, ponding: estimates of loss due to vegetation canopy, surface 3478 detention 3479 • baseflow: pre-flood estimates based on subarea size, post-flood estimates based 3480 on exponential recession curve. 3481 • rainfall events examined: 2-year, 25-year, and 100-year, under three scenarios 3482 (pre-fire, post-fire immediately, and post-fire one year later). 3483 • sediment bulking estimated 3484 3485 Choice of hydrologic computer model: HEC-HMS 3486 The model HEC-HMS (USCOE 2013) features many options for estimating flood 3487 hydrographs. The runoff transformation options include the NRCS dimensionless UH, 3488 but without optional peak factors other than the standard 484. Additional synthetic UH 3489 options are the Snyder and Clark UHs, which require some form of peaking factor 3490 estimation. This case study demonstrates a user-derived synthetic unit hydrograph, in the 3491 S curve format. (See figure CS4-7 for the HEC-HMS input window which summarizes 3492 the modeling options.) 3493

DRAFT Technical Note, April 2015 C4-7 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3494 Figure CS4- 7. HEC-HMS input window for modeling options in a single sub-basin.

3495 3496 The Whitewater Creek watershed upstream of Glenwood NM is shown in figure CS4-8. 3497 For each of the forty-one subareas shown a GIS time-area analysis was performed. More 3498 detail is provided for a single sub-basin (figure CS4-9, the red inset of figure CS4-8). 3499 3500 Figure CS4- 8. Whitewater Creek hydrologic subareas.

DRAFT Technical Note, April 2015 C4-8 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3501 3502 Figure CS4- 9. Upper Whitewater subarea (called “Baldy Fork”)

3503 3504 Time-area histogram and synthetic UH Estimation 3505 Determining runoff depth over time at the outlet of a given subarea requires an estimate 3506 of flow travel time from every part of the drainage area. Even small subareas vary 3507 considerably in flow slopes, flow cross-sectional areas, and surface roughnesses. Travel 3508 time computation requires spatial accounting of these variables as they affect flow 3509 velocity, and then the accumulation of time as flows with these varying velocities 3510 traverse the distances to the outlet. 3511 3512 Given the three standard flow segment types (overland, shallow-concentrated, and stream 3513 channel), equations and tables from the Technical Note on the Velocity Method may be 3514 used. A time-area histogram, however, requires not merely time of concentration, but 3515 flow time estimation from many subdivisions of the drainage area. Such an analysis is 3516 greatly facilitated by the use of GIS. 3517

DRAFT Technical Note, April 2015 C4-9 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3518 To develop a time-area histogram for Baldy Fork (figure CS4-9) using GIS, the flow time 3519 from each GIS raster cell in the subarea to the outlet was determined. The cells were 3520 grouped into five minute time bands. Flow velocity in each cell depended on whether the 3521 flow segment type was primarily overland, shallow-concentrated, or stream channel. 3522 Within these segment types, surface roughness was specified. To create a unit 3523 hydrograph the time band GIS attribute data was copied into a spreadsheet, as will be 3524 detailed below. 3525 3526 Roughness in GIS 3527 The land roughness was determined, given pre-fire land cover and post-fire burn severity, 3528 as shown in Tables CS4-1 through CS4-3. For typical overland and shallow-concentrated 3529 flow roughness values, see Technical Note Tables 10 and 11, respectively. Three 3530 roughness rasters were created, one for each fire condition. The raster grid sizes, based 3531 on the available digital elevation model (DEM) were ten meters square. 3532 3533 Table CS4- 1. Pre-fire roughness assumptions. Flow segment Land use assumptions n value Overland Grass, short-grass prairie 0.15

Shallow-concentrated Steep, rocky, less depth flow 0.08

Channel Steep, rocky, less debris 0.045 3534 3535 Table CS4- 2. Post-fire (immediately) roughness assumptions. Flow segment Land use assumptions n value Bare ground (severe, moderate burn) 0.011 Overland Bare ground (low burn) 0.08 Bare ground (no burn) 0.15

Shallow-concentrated Steep, rocky, more depth flow 0.06

Channel Steep, rocky, more debris 0.05 3536 3537 3538

DRAFT Technical Note, April 2015 C4-10 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3539 Table CS4- 3. Post-fire (1 year) roughness assumptions. Flow segment Land assumptions n value Bare ground (severe, moderate burn) 0.03 Overland Bare ground (low burn) 0.10 Bare ground (no burn) 0.15 3540 3541 To set the flow segment type of each raster cell, the typical GIS hydrologic function flow 3542 accumulation was used. In GIS, the flow accumulation value of each cell is the number 3543 of cells upstream of it (which therefore drain to it). The flow segment type for each ten 3544 meter square cell was determined using the following assumptions: 3545 • Overland flow extends for about the first 100 feet of drainage (three cells). 3546 • Shallow-concentrated flow extends for the next 1000 feet (thirty cells). 3547 • Channel flow is assumed for any flow accumulation of more than thirty-three 3548 cells. 3549 3550 These GIS operations are performed with the raster calculator in ArcMap (ESRI 2010). 3551 An example of a raster calculator command is shown in figure CS4-10, circled in red. 3552 Note that the function Con is short for Conditional (an if statement). Given the flow 3553 accumulation layer flowAcc, the command may be translated into plain English, as 3554 follows. “If flowAcc is less than or equal to three (an overland flow segment) then set the 3555 roughness to 0.15. Otherwise, if flowAcc is less than or equal to 33 (a shallow- 3556 concentrated flow segment), set roughness to 0.08. If neither of those two conditions 3557 exists (flowAcc is greater than 33, a channel flow segment), then set the roughness to 3558 0.045. Note in figure CS4-10 that the output raster is named “roughPreFire”. 3559

DRAFT Technical Note, April 2015 C4-11 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3560 Figure CS4- 10. ArcMap Raster Calculator expression creating a pre-fire roughness layer.

3561 3562 To assess the burn condition of each raster cell, the burn severity shape file provided by 3563 USDA Forest Service was transformed into a raster layer within GIS. In that layer, 3564 (figure CS4-11) each cell with a high severity is assigned the number 4, medium 3, low 2, 3565 and unburned 1. For the pre-fire condition, using the data in Table CS4-1, the ArcMap 3566 Raster command is shown in Figure CS4-10. Note that the roughness layers may be 3567 created for the entire watershed, not each subarea separately. 3568 3569 For the condition immediately post-fire, using Table CS4-2, the raster command slightly 3570 more complicated: 3571 3572 Con("rufPreFire" == 0.08,0.06,Con("rufPreFire" == 0.045,0.05,Con("burnRaster" >= 3573 3,0.011,Con("burnRaster" == 2,0.08,"rufPreFire")))) 3574 3575 In plain English, this command says, “If the roughPreFire cell is 0.08 (i.e., a shallow- 3576 concentrated flow cell) then set the roughness to 0.06, otherwise, if the cell in 3577 roughPreFire is 0.045 (i.e., a channel cell) then set the roughness to 0.05. If neither of 3578 those conditions is true, 3579 then if the burn raster cell is greater than or equal to 3 (moderate or severe) then set the 3580 roughness to 0.011, otherwise if the burn raster cell is 2 then set the roughness to 0.08.

DRAFT Technical Note, April 2015 C4-12 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3581 Finally, if none of those conditions is true, then set the new raster value to the same as the 3582 old raster value (unburned overland flow cells). 3583 Figure CS4- 11. Burn raster (red = high, orange = medium, yellow = low, green = no burn).

3584 3585 3586 Flow slope in GIS 3587 ArcMap (ESRI 2010) has a surface slope function that does not produce flow slope 3588 because it uses average elevations in a raster cell neighborhood. To obtain flow slope for 3589 each cell, the required information is elevation drop from that cell to the lowest neighbor, 3590 and the flow length between the two cells. Since flow in a three by three neighborhood 3591 of cells (figure CS4-12) may be either straight or diagonal, the flow length is one of two 3592 values. To create a flow distance raster in feet, given a 10 meter DEM, the two length 3593 values are either 46.40 feet (diagonal) or 32.81 feet. Figure CS4-12 shows the cell values 3594 created by the hydrology command flow direction. The following raster command to 3595 obtain flow length examines the flow direction raster for whether the flow is straight or 3596 diagonal: 3597

DRAFT Technical Note, April 2015 C4-13 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3598 Con("flowDir" == 1,32.81,Con("flowDir" == 4,32.81,Con("flowDir" == 3599 16,32.81,Con("flowDir" == 64,32.81,46.4)))) 3600 3601 Figure CS4- 12. Flow direction codes

3602 3603 To obtain elevation difference, the ArcToolbox function “Focal Statistics” may be used. 3604 Among the available statistics is the elevation drop to the lowest neighbor. Then a flow 3605 slope raster is obtained with a raster calculation dividing that elevation difference by the 3606 flow length. One final procedure is to prevent zero slopes by evaluating the flow slope 3607 raster and setting any zeroes to some minimum value. 3608 3609 Velocity relationship derivations 3610 In addition to the velocity methods discussed in the Technical Note, other references such 3611 as USCOE (2013) and MacArthur and DeVries (1993) discuss flow velocity calculations 3612 for different channel segment types. Falling under the general category of kinematic 3613 wave routing, these techniques transform Manning’s equation into the general form 3614 shown in equation 1, where 3615 y is flow depth, a is flow cross-sectional area, S is flow slope, and n is Manning’s 3616 roughness 3617 coefficient: 3618 5/3 1.486 S 3619 Q = ay where: a = (eq. CS4-1) 3620 n 3621 Given possible flow areas, often estimated by assuming a flow width, equation CS4-1 3622 can, for overland flow, be formulated as equation CS4-2: 3623 φ S 1.486y2/3 Voverland = φ = n DRAFT Technical Note, April 2015w C4-14 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3624 where: (eq. CS4-2) 3625 3626 Note that the velocity equations in Technical Note Table 11 are also given in this form. 3627 Table CS4-4 shows the range of phi values for overland flow. 3628 3629 Table CS4- 4. Values of φ, given assumptions of overland flow width and depth. 3630 y W φ (in) (ft) 3631 0.25 2.0 0.0563632 0.25 2.5 0.0453633 0.25 3.0 0.0383634 0.25 3.5 0.0323635 0.25 4.0 0.0283636 0.50 2.0 0.0893637 0.50 2.5 0.0713638 0.50 3.0 0.0603639 0.50 3.5 0.0513640 0.50 4.0 0.0453641 3642 3643 For shallow-concentrated flow (collector channels), assuming a triangular channel shape, 3644 MacArthur and DeVries (1993) develop an equation in the same form (equation CS4-3). 3645 Table CS4-5 shows the range of phi values for shallow-concentrated flow, given side- 3646 slope z in length horizontal divided by length vertical.

3647 1/3 z 3648 φ S where: φ = 0.94 A1/3 (eq. CS4-3) Vshallow = 2 3649 n 1+ z 3650 3651

DRAFT Technical Note, April 2015 C4-15 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3652 Table CS4- 5. Shallow-concentrated flow parameters. 3653 y z φ (ft) (V:H) 3654 0.5 3 0.5723655 0.5 4 0.5803656 0.5 5 0.5843657 1.0 3 0.9083658 1.0 4 0.9213659 1.0 5 0.9283660 3661 3662 For channel flow segments, assuming a trapezoidal shape, MacArthur and DeVries 3663 (1993) developed an equation of similar form, but more complicated because trapezoidal 3664 flow area is dependent on both channel bottom width and side slopes, in addition to 3665 depth. In a watershed, the equation also varies considerably with the drainage area. In 3666 other words, the flow area varies directly with drainage area. In kinematic wave routing 3667 models, the flow depth and area are computed along the channel, with a time-step change 3668 in flow depth. For a time-area histogram, an overall velocity for each cell is needed, 3669 regardless of what time during a storm. (Recall the UH assumption that the time is 3670 independent of the depth of direct runoff.) 3671 3672 One way to estimate flow area for any given GIS raster cell is to use regional regression 3673 equations, if available. For many areas of the country equations which relate bankfull 3674 flow area to watershed area have been developed. For the Gila Wilderness region 3675 Moody, Wirtanen, and Yard (2003) derived the following: 3676 0.512 3677 AB= 4.78 r2 = 0.9163 (eq. CS4-4) 3678 3679 where: A = flow cross-sectional area (ft2) 3680 B = watershed area (mi2) 3681

DRAFT Technical Note, April 2015 C4-16 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3682 Although the equation has a relatively high correlation, the authors included their data, 3683 which shows that many much larger watersheds were used than the subarea sizes within 3684 Whitewater Creek. To obtain a more applicable equation for such small subareas the 3685 authors’ data was culled for only the small watershed sizes, 23 total, and the following 3686 equation derived, for use in Whitewater Creek (where A and B are as defined above): 3687 3688 AB= exp( 0.5036ln + 1.595 ) r2 = 0.88 (eq. CS4-5) 3689 3690 To obtain channel flow velocity, based on a trapezoidal channel shape (figure CS4-13), 3691 the following equation was used: 3692 φ S 1/3 3693 V = where: φµ = 1.486 A (eq. CS4-6) channel n 3694 3695 Figure CS4- 13. Trapezoidal channel dimensions.

3696 3697 The µ in equation CS4-5 is derived from the choice of trapezoidal side slopes and the 3698 relationship between bottom width and depth, with a range of values shown in table CS4- 3699 6: 3700 3701

DRAFT Technical Note, April 2015 C4-17 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3702 Table CS4- 6. Trapezoidal channel parameters. 3703 z w / y µ 3704 2 2 0.457 3705 2 3 0.447 2 4 0.4373706 3 2 0.4163707 3 3 0.4103708 3 4 0.4033709 4 2 0.3853710 4 3 0.3813711 4 4 0.3763712 3713 3714 Velocity relationships used in Whitewater Creek 3715 Tables CS4-4, CS4-5, and CS4-6 show that variations in estimated flow depth have a 3716 relatively minor effect on velocity, especially in comparison to the effect of roughness. 3717 For Whitewater Creek, overland flow depth was assumed 0.25 inches, width 3 feet. With 3718 these assumptions, equation CS4-2 becomes equation CS4-7 below. 3719 0.5 3720 0.038 S (eq. CS4-7) Voverland = 3721 n 3722 For the shallow-concentrated flow segments, depth was assumed to be 0.5 feet and z was 3723 assumed to be 4, which yields (from equation CS4-3 and Table CS4-5) equation CS4-8: 3724 3725 0.58 S (eq. CS4-8) Vcollector = 3726 n 3727 For the channel flow segements, a separate raster layer for flow cross-sectional area was 3728 obtained using equation CS4-5. Then, assuming a trapezoidal channel, two different 3729 relationships were derived depending on slope. For a flow slope greater than or equal to 3730 0.05, width-depth ratio was assumed 3 and z = 2. For flow slope less than 0.05, width- 3731 depth ratio was assumed 5 and z = 3. Then, using equation CS4-6 and Table CS4-6, the 3732 following equations were derived:

DRAFT Technical Note, April 2015 C4-18 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3733 3734 steeper channels: 0.664 S (eq. CS4-9) VA= 1/3 3735 channel n 3736 3737 less steep channels: 0.589 S 1/3 (eq. CS4-10) VAchannel = 3738 n 3739 3740 When creating the velocity rasters, the possibility exists that velocity will come out 3741 unreasonably high due to very steep slopes, such as vertical drops, or zero due to flat 3742 slopes. A better travel time estimate was obtained for Whitewater Creek by restricting 3743 the maximum and minimum velocities to 15 and 0.10 feet per second, respectively. 3744 3745 Travel time in GIS 3746 In ArcGIS, the hydrology tool Flow Length is used to obtain the flow travel time of any 3747 cell to its outlet. This is possible because the tool can apply an optional weight raster, 3748 which may contain such parameters as unit conversions and channel sinuosity, as shown 3749 in equation CS4-11. The time in minutes is obtained using a weight raster that includes a 3750 sinuosity of 1.1, a conversion of meters to feet (since the tool is using the 10m DEM), 3751 and from seconds to minutes. 3752 1.1∗ 3.28083 3753 = ∗ where: weight = (eq. CS4-11) time flowLength weight flowVelocity ∗60 3754 3755 The resulting time raster can then be reallocated to group the cells into time bands of, say, 3756 five minutes. The 5-minute time band raster for Baldy Fork subarea is shown in Figure 3757 CS4-14. 3758

DRAFT Technical Note, April 2015 C4-19 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3759 Figure CS4- 14. Twenty-one five-minute travel time bands for Baldy Fork, with Attribute Table.

3760 3761 The Attribute Table shows the number of cells in each of 21 time bands in Baldy Fork. 3762 Since there are 21 five-minute bands the longest drainage time (or time of concentration) 3763 may be estimated at 1 hour and 45 minutes. However, the hydrologist can note the 3764 number of cells in those later time bands. Sometimes they become so few that a better 3765 estimate of time of concentration is taken to be earlier. For Baldy Fork, one might select 3766 timeband 18, or 90 minutes, because later time bands have only a total of 61 cells among 3767 them. Note that a separate travel time raster is produced for the three scenarios, pre-fire, 3768 post-fire immediately, and post-fire after one year. 3769 3770 Spreadsheet derivation of time-area histogram and unit hydrograph 3771 To obtain a time-area histogram the five-minute time band attribute table is copied into a 3772 spreadsheet. This procedure is described in many hydrology textbooks, including 3773 Bedient, Huber, and Vieux (2012). Since the cells are 100 square meters each, the area 3774 drained by each time band can be converted to any convenient unit, such as acres, in the

DRAFT Technical Note, April 2015 C4-20 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3775 spreadsheet. The resulting time-area histogram for Baldy Fork pre-fire is shown in figure 3776 CS4-15. To obtain a unit hydrograph, a duration must be chosen and one inch of 3777 effective rainfall applied within that timeframe. The runoff from each time band is 3778 computed, lagged, and summed. 3779 3780 Figure CS4- 15. Pre-fire time-area histogram for Baldy Fork.

3781 3782 A screenshot of the spreadsheet is shown in figure CS4-16. The flow values in column 3783 G, in acre-inches, are obtained by summing the lagged flow values in columns K through 3784 P. Those flow values are obtained by multiplying the 5-minute increment of effective 3785 rainfall (0.167 inches) by the area in acres of each time band (from the GIS attribute 3786 table, column B). Then, in column H, the flow is converted to cfs. The 30 minute unit 3787 hydrograph peak (cell H14) is 1817 cfs, the peak that would be expected from Baldy Fork 3788 if one inch of effective rainfall was evenly distributed over the subarea in thirty minutes. 3789 Figure CS4-17 shows the thirty minute unit hydrgraphs for Baldy Fork under the three 3790 scenarios.

DRAFT Technical Note, April 2015 C4-21 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3791 Figure CS4- 16. Pre-fire spreadsheet. (30-minute unit hydrograph in column H)

3792 3793 Figure CS4- 17. Unit Hydrographs, 30-minute duration for Baldy Fork.

3794 3795 The 30-minute unit hydrograph is applicable when the duration of excess rainfall from a 3796 given event is thirty minutes long. To enable a unit hydrograph compilation to apply for 3797 any duration of excess rainfall, a so-called S-curve is derived by lagging the 30-minute 3798 unit hydrographs in successive 30-minute periods, as shown in figure CS4-18.

DRAFT Technical Note, April 2015 C4-22 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3799 3800 Figure CS4- 18. S-curve from lagged and summed 30-minute unit hydrographs

3801 3802 The maximum of the S-curve is the unit flow for the subarea—the amount of runoff if a 3803 full inch of effective rainfall runs off the entire watershed area in thirty minutes. In the 3804 case of Baldy Fork, since the total area is 57,124,070 square feet the unit flow is:. 3805 57,124,070 × 1 1 1 3806 × × = 2,644 30 2 12 60 𝑓𝑓𝑓𝑓 𝑖𝑖𝑖𝑖 𝑓𝑓𝑓𝑓 𝑚𝑚𝑚𝑚𝑚𝑚 3807 𝑐𝑐𝑐𝑐𝑐𝑐 𝑚𝑚𝑚𝑚𝑚𝑚 𝑖𝑖𝑖𝑖 𝑠𝑠𝑠𝑠𝑠𝑠 3808 The hydrology model HEC-HMS requires the S-curve to be given as a percent of lag 3809 versus percent of unit flow. Watershed lag is the time from the center of the effective 3810 rainfall duration to the runoff peak. The lag for Baldy Fork can be seen from the 30- 3811 minute unit hydrograph, figure CS4-17, to go from 15 minutes (center of rainfall) to 45 3812 minutes (peak) or 30 minutes. The S-Curve, reformulated for input into HEC-HMS is 3813 shown in figure CS4-19. The values along the x-axis come from the times of figure CS4- 3814 18 divided by the lag time, and the values of the y-axis come from the flow values of 3815 figure CS4-18 divided by the unit flow. 3816

DRAFT Technical Note, April 2015 C4-23 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3817 This section has shown the development of the UH S-curve input to HecHMS for one 3818 subarea in the Whitewater Creek watershed. The same procedures were followed to 3819 produce S-curves for every subarea in the model for the three scenarios. 3820 Figure CS4- 19. Baldy Fork S-curves for HecHMS, three scenarios

3821 3822 Infiltration Loss Method: Green & Ampt Equation 3823 The process-based Green & Ampt method of infiltration uses soil properties to model 3824 progressive loss of surface water into the soil and the change in soil moisture capacity. 3825 Although GA is a point-model, the more spatially homogeneous the soil in a given 3826 subarea the more applicable this method. HEC-HMS does have the ability to do a grid- 3827 based Green & Ampt infiltration, modeling from cell to cell within a subarea, but this 3828 requires more data than is available for the Whitewater Creek basin. Being within the 3829 Gila Wilderness, spatially distributed soil data is not available. For the entire watershed, 3830 the Green & Ampt parameters were estimated based on soil type and using the Technical 3831 Note Table 12. The assumed soil texture was “loamy sand” with porosity of 0.42 cubic 3832 inches of pore space per cubic inch of soil. The initial moisture content was assumed to 3833 be near “field capacity” or 15 percent of the porosity. Converting from the metric units 3834 of Table12, the wetting front suction was estimated at 2.4 inches and the hydraulic 3835 conductivity 1.6 inches per hour.

DRAFT Technical Note, April 2015 C4-24 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3836 3837 A final specification for this method is the percent impervious surface of the subarea. 3838 This was set at zero for the pre-fire condition. For post-wildfire conditions, the percent 3839 imperviousness was used as a way to account for soil hydrophobicity. However, water- 3840 repellent soil is not the same as a parking lot pavement. It is consistent in neither spatial 3841 extent, nor depth or layer thickness. In addition, higher rainfall rates can sometimes 3842 penetrate hydrophobic soils and infiltrate more than lower rainfall rates. The 3843 hydrophobicity will then generally return upon drying. Finally, high soil burn severity is 3844 not a fool-proof predictor of the location of hydrophobic soil. 3845 3846 For modeling soil hydrophobicity in the Whitewater Creek watershed, the percent 3847 imperviousness for the immediately post-fire condition for each subarea was set to 65% 3848 of the high soil burn severity area for the 25-year and more frequent rainfall events. For 3849 the 100-year rainfall event, the imperviousness was set to 30% of the high burn severity 3850 area. For the post-fire 1-year later condition, the imperviousness was set to 45% of the 3851 high burn severity for the 25-year rainfall and more frequent events. For the 100-year 3852 rainfall event, the imperviousness was set to 15% of the high burn severity. 3853 3854 Initial Abstraction, Surface Ponding, Canopy Loss 3855 The CN method includes an initial abstraction which is considered to represent 3856 interception by vegetation, surface storage, and infiltration prior to runoff. This initial 3857 abstraction is not changeable by the user and is a function of the potential maximum loss 3858 to infiltration, ultimately a fraction of the user choice of curve number. While the CN 3859 method is provided as an option by HEC-HMS, the program also provides explicit 3860 canopy loss and surface loss methods. Since the Green & Ampt loss method was chosen 3861 for the Whitewater Creek analysis, further losses due to vegetation and surface 3862 depressions needed to be modeled. 3863 3864 The options for these losses are not highly sophisticated in HEC-HMS. A maximum 3865 retention is specified by the user and an initial condition. The HEC-HMS user manual 3866 (USCOE-HEC 2013) states that use of a canopy method is generally not necessary for

DRAFT Technical Note, April 2015 C4-25 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3867 storm events, whereas continuous simulation modeling would take into account the 3868 evaporation of intercepted moisture between storms. However, forest canopy is not an 3869 insignificant moisture interceptor. The flashiness of Southwestern US storm events 3870 contributes to raising the significance of canopy. In addition, for wildfire events in 3871 forests the near total loss of canopy represents a widespread landscape change. For 3872 Whitewater Creek pre-fire conditions the canopy initial storage was set to zero and the 3873 maximum storage set to 0.05 inches. For post-wildfire conditions the canopy 3874 interception was set to none. 3875 3876 Surface storage is a separate option in HEC-HMS, represents loss of incoming rainfall to 3877 surface depressions, and may be used to represent the retention of moisture in the forest 3878 litter or duff layer. For Whitewater Creek pre-fire condition the initial storage was set to 3879 zero and maximum loss set to 0.03 inches. For post-fire conditions the maximum loss 3880 was set to 0.01 inches. 3881 3882 Baseflow 3883 HEC-HMS provides several options for modeling baseflow. The recession method is 3884 common for flood simulations, by which the initial baseflow is specified, along with a 3885 constant to modify the rate of exponential recession, and a threshold at which the 3886 baseflow is reset after the event. In the arid Southwest US baseflow is generally less 3887 significant when the concern is storm-event flood peaks. However, the Whitewater 3888 Creek analysis included the recession baseflow option, with the initial baseflow set to 3889 0.25 cfs per square mile of subbasin drainage area. 3890 3891 Rainfall events examined 3892 The NOAA National Weather Service Precipitation Frequency Data Server (NOAA 3893 2013) was recently updated for New Mexico. The website (see reference) allows the user 3894 to select a given precise location on a map and provides the total storm values for various 3895 recurrence intervals and durations. Often the 24-hour duration values are used. In the 3896 flashy thunderstorm prone Southwestern US, a shorter duration may be more applicable.

DRAFT Technical Note, April 2015 C4-26 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3897 Since the BAER hydrologic analysis for the Gila Wilderness used a 6-hour duration, this 3898 was selected for the current analysis. 3899 3900 The total rainfall values shown in Table CS4-7 show that there is some orographic effect 3901 within the watershed, lowering rainfall values with descending elevation toward the 3902 outlet. The first flood hydrograph analysis of this case study assumed that these rainfall 3903 amounts would fall on all subareas at the same time. This is unrealistic but provides 3904 conservatively high peaks and larger volume hydrographs. 3905 glad 3906 Table CS4- 7. Whitewater Creek watershed 6-hour duration storm totals. 3907 2-year 25-year 100-year Watershed area (in) (in) (in)3908 Upper Whitewater 1.54 2.73 3.493909 SF Whitewater 1.42 2.49 3.193910 Lower Whitewater 1.20 2.12 2.733911 3912 3913 3914 A second and third set of flood hydrograph analyses considered the three recurrence 3915 interval 6-hour rainfalls from Table CS4-7 with an areal reduction factor obtained from 3916 figure CS4-21. The second set of analyses centered the storms over the upper 3917 Whitewater Creek portion of the watershed and the third set centered the storms over the 3918 SF Whitewater sub-watershed. For these areally reduced analyses, each basin received a 3919 rainfall rate lowered in proportion to their distance from the assumed center of the storm. 3920 3921 Figure CS4-20, developed for Walnet Gulch AZ, is assumed applicable to Whitewater 3922 Creek, in the same geographic region, about 180 miles away. As Whitewater Creek is 3923 about 23 km wide, east to west, the areal reduction from figure CS4-20 is significant. 3924 The largest area on the graph is 300 square kilometers, which corresponds to a radius of 3925 about 10 kilometers. Therefore, many of the sub-basins of Whitewater Creek will be 3926 farther away from the storm center than can be analyzed with the figure. Rather than 3927 reduce the rainfall of more remote subareas to zero, the analyses applied minimum values

DRAFT Technical Note, April 2015 C4-27 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3928 from figure. The selected storm centerings for the areally reduced analyses are shown in 3929 figure CS4-21. The red star marks the upper Whitewater centering and the red circular 3930 line shows the 10km radius for that storm. The blue star marks the storm centering in the 3931 SF Whitewater sub-basin, with blue circular line showing the 10km radius from that 3932 point. 3933 3934 Figure CS4- 20. Areal reduction factors developed for 6-hour duration storms at Walnut Gulch 3935 AZ Experimental Watershed (modified, from Osborn, Lane, and Myers 1980).

3936 3937 To distribute the rainfall totals over time, the standard SCS Type II rainfall distribution, 3938 considered applicable in New Mexico, was used. Although the standard distribution 3939 pertains to a 24-hour duration, it is easily modified by proportioning for shorter durations, 3940 including the 6-hour. 3941

DRAFT Technical Note, April 2015 C4-28 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3942 Figure CS4- 21. Two storm centerings for Whitewater Creek and 10km radii.

3943 3944 Sediment Bulking 3945 This analysis did not attempt to model sedimentation. As discussed in the main body of 3946 the Technical Note, the maximum sediment concentration short of “debris flow” is 3947 considered about twenty percent. A USGS report specifically concerning the Whitewater 3948 Baldy Complex wildfire (Tillery, Matherne, and Verdin, 2012) estimated high 3949 probabilities in Whitewater Creek of debris flows from 30-minute duration rainfall events 3950 for recurrence intervals of 2-year, 10-year, and 25-year. A shorter duration storm of 3951 higher intensity is generally required to generate debris flow. 3952 3953 Figure CS4-22 is a screenshot from Plate 3 of Tillery, Matherne, and Verdin (2012) 3954 which shows their overall debris flow hazard findings for Whitewater Creek. Notice the 3955 main channel of Whitewater Creek in blue and the community of Glenwood at the 3956 downstream end circled in red. Sub-basins labeled 92 and 93 are the upper Whitewater, 3957 sub-basin 91 is the SF Whitewater, and sub-basin 90 is the upper Little Whitewater. 3958 These of brown color are the highest hazard for debris flow in response to a 25-year 30- 3959 minute rainfall event (a total storm rainfall of about 1.54 inches).

DRAFT Technical Note, April 2015 C4-29 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3960 3961 Figure CS4- 22. Whitewater debris flow hazard (Tillery, Matherne, and Verdin (2012).

3962 3963 As a conservative estimate of possible sediment bulking for the events examined in this 3964 analysis, with six-hour duration, the sub-basins in brown from figure CS4-23 were 3965 considered 20 percent bulked and the sub-basins in yellow, 10 percent bulked. 3966 3967 Results and Discussion 3968 Figure CS4-23 shows the four locations for which modeling data is reported herein. The 3969 model user can receive results for the outlet of any sub-basin (shown in the figure by 3970 different pastel colors). The output is reported for three recurrence intervals, 2-year, 25- 3971 year, and 100-year, for the three scenarios, pre-fire, post-fire immediately, and post-fire 3972 one year later. In addition, three types of storm centerings were considered. One for 3973 which no areal reduction is applied and the rainfall values of Table CS4-7 are applied at 3974 the same time over the entire watershed. The second and third centerings are as shown in 3975 figure CS4-21, over the upper Whitewater subarea or over the SF Whitewater, with areal 3976 reductions determined from the curves of figure CS4-20.

DRAFT Technical Note, April 2015 C4-30 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3977 3978 Figure CS4- 23. Locations of reported HEC-HMS modeling output.

3979 1 = Whitewater Creek outlet at Glenwood NM 3980 2 = USGS gage “Whitewater Creek at Catwalk” (installed 2012) 3981 3 = Upper Whitewater Creek just above the confluence of SF Whitewater 3982 4 = SF Whitewater just above the confluence of the Whitewater. 3983 3984 Table CS4-8 shows HEC-HMS modeling results for the rainfall events with no areal 3985 reduction. (These runs include the slight reduction, shown in Table CS4-7, due to the 3986 fact that the data provided by NOAA, for three general areas of the watershed, lessens for 3987 each recurrence interval as location changes toward the west.) 3988

DRAFT Technical Note, April 2015 C4-31 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 3989 Table CS4- 8. HecHMS output: hydrograph peaks at four locations (no areal reduction)

Post-fire peak discharge Post-fire peak Pre-fire peak (immediately discharge Storm Location discharge (cfs) following the (One year later) fire)) (cfs) (cfs) 100-yr abv SF 9,839 14,595 12,492 100-yr SF at mouth 3,618 5,492 4,818 100-yr gage 13,119 18,861 16,160 100-yr Glenwood 14,304 20,970 17,768 25-yr abv SF 3,386 12,324 9,387 25-yr SF at mouth 1,087 3,843 3,064 25-yr gage 4,371 15,285 11,824 25-yr Glenwood 4,326 16,229 12,312 2-yr abv SF 969 8,611 6,051 2-yr SF at mouth 661 3,125 2,323 2-yr gage 1,712 10,958 7,932 2-yr Glenwood 1,781 12,035 8,613 3990 3991 These peaks are considered conservatively high, since no areal reduction was applied and 3992 the nature of storms in the Gila Wilderness area is more localized convective 3993 thunderstorms. In addition, although the unit hydrographs did change between the two 3994 post-fire scenarios, due to the consideration of some vegetation regrowth, the majority of 3995 the drop in the peak for the one year later condition came from assumptions about the 3996 reduction in hydrophobic soils. No field verification was completed for this analysis and 3997 the BAER report (USDA Forest Service, 2012) did not discuss the extent of hydrophobic 3998 soil. Estimates for this analysis were made using the soil burn severity GIS shapefiles 3999 and from the findings of Tillery, Matherne, and Verdin (2012). Figure CS4-24 shows the 4000 full hydrographs for the 100-year no areal reduction scneario at the Whitewater Creek 4001 outlet, near Glenwood. 4002

DRAFT Technical Note, April 2015 C4-32 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4003 Figure CS4- 24. HecHMS results, 100-year storm peaks at Glenwood (no areal reduction).

4004 4005 The areal reductions, as discussed above came from Osborn, Lane, and Myers, (1980), in 4006 particular their examination of Walnut Gulch AZ. That paper also examined a watershed 4007 near Alamogordo NM, which showed less areal reduction than Walnut Gulch. However, 4008 the Gila Wilderness area is considered closer geographically to Walnut Gulch than to 4009 Alamogordo. Nevertheless, the areal reductions may be less at Whitewater Creek than 4010 what were applied for the results reported in Table CS4-9. Note that the 100-year pre-fire 4011 peak at Glenwood between Tables CS4-8 and CS4-9 drops from 14,304 cfs to 2,904 cfs. 4012 The post-fire reductions in peak are less dramatic, with the post-fire immediately peak at 4013 Glenwood dropping from 20,970 cfs to 11,826 cfs. 4014 4015

DRAFT Technical Note, April 2015 C4-33 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4016 Table CS4- 9. HecHMS output: hydrograph peaks at four locations (areal reductions applied, with 4017 centerings as shown in figure CS4-22).

Upper Whitewater centering SF Whitewater centering

One

(cfs)

(cfs) (cfs) (cfs)

Storm Location

year later) year later) fire peak (immediately (immediately peak fire fire peak (immediately (immediately peak fire - - fire ( peak discharge fire (One peak discharge fire peak discharge (cfs) discharge peak fire fire peak discharge (cfs) discharge peak fire following fire) fire) following - - - - following fire) fire) following Post Post Pre Pre Post Post 100-yr abv SF 3,126 10,708 8,140 943 7,346 5,162 100-yr SF at mouth 3 1,534 1,052 1,967 4,741 3,875 100-yr gage 3,118 11,785 8,886 2,230 11,010 8,119 100-yr Glenwood 2,904 11,826 8,808 2,259 11,952 8,672 25-yr abv SF 763 7,274 5,010 12 5,022 3,218 25-yr SF at mouth 3 1,338 914 290 2,582 1,889 25-yr gage 753 8,247 5,634 290 7,103 4,793 25-yr Glenwood 682 8,355 5,645 282 7,770 5,177 4018

DRAFT Technical Note, April 2015 C4-34 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4019 Figure CS4- 25. HecHMS results, 100-year storm peaks at Glenwood (areal reduction with centering 4020 over the upper Whitewood sub-basin)

4021 4022 Comparing the results in Table CS-9 between the two storm centerings varies depending 4023 on recurrence interval. For the 100-year flood, the pre-fire peak at Glenwood is 22% less 4024 if the 4025 same storm is centered over the SF Whitewater rather than the upper Whitewater, but the 4026 post-fire immediately results show a 1% greater peak at Glenwood if the storm is 4027 centered over the SF Whitewater. This is a direct result of the burn severity between the 4028 two sub-basins. As the sub-basins heal, the results show that the Glenwood 100-year 4029 peak moves toward being greater if the storm is centered over the upper Whitewater. 4030 Figure CS4-26 shows the post-fire results for the 100-year peaks at Glenwood under both 4031 centering conditions. Figure CS4-27 shows the same results for the 25-year recurrence 4032 interval. Note that storms centered over the SF Whitewater sub-basin peak at Glenwood 4033 about 15 minutes earlier than if they were centered over the upper Whitewater. 4034

DRAFT Technical Note, April 2015 C4-35 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4035 4Figure CS4- 26. 100-year post-fire hydrographs at Glenwood between storm centerings in the 4036 upper Whitewater sub-basin and the SF Whitewater sub-basin.

4037 4038 Figure CS4- 27. 25-year post-fire hydrographs at Glenwood between storm centerings in the upper 4039 Whitewater sub-basin and the SF Whitewater sub-basin.

DRAFT Technical Note, April 2015 C4-36 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4040 4041 Sediment bulked peak flows have been computed as mentioned above, with the upstream 4042 end (about one third) of the Upper Whitewater sub-basin contributing a 20 percent 4043 sediment discharge, the lower end 10 percent, and the entire SF Whitewater sub-basin 4044 contributing 20 percent. These results are shown in Table CS4-10. 4045 4046 Table CS4- 10. Sediment bulked flows for Whitewater Creek. Upper centering SF centering

(cfs) (cfs)

Storm Location (cfs) (cfs) later) later) fire peak (immediately (immediately peak fire fire peak (immediately (immediately peak fire - - following fire) fire) following following fire) fire) following k discharge (One k discharge pea year fire fire (One peak discharge year - - Post Post Post Post

100-yr abv SF 11,848 8,272 7,478 6,190 100-yr SF at mouth 1,534 1,052 4,741 3,875 100-yr gage 13,591 10,236 11,536 10,095 100-yr Glenwood 13,632 10,158 12,478 10,648 25-yr abv SF 7,975 5,011 5,024 3,921 25-yr SF at mouth 1,534 1,052 4,741 3,875 25-yr gage 9,532 6,518 7,163 6,012 25-yr Glenwood 9,641 6,529 7,829 6,396 4047 4048 Modeling observed data 4049 September 2013 brought heavy rainfall to the Rockies from Wyoming to New Mexico. 4050 The extensive damage in and around Boulder, Lyons, and Estes Park Colorado grabbed 4051 national headlines. But storms of similar magnitude also visited the Gila Wilderness 4052 watersheds that had been previously damaged by the Whitewater-Baldy Complex fire. 4053

DRAFT Technical Note, April 2015 C4-37 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4054 The largest rainfall totals in the Whitewater Creek watershed were recorded between 14 4055 and 15 September 2013. The NOAA Precipitation-Frequency data server (NOAA 2013) 4056 data for a storm centering over the confluence of the SF Whitewater is shown in Table 4057 CS4-11, along with the rainfall totals (in blue) for the event recorded at the Catwalk gage. 4058 4059 Table CS4- 11. NOAA Precip-Frequency values and Sep 2013 rainfall event from Catwalk gage Storm centering at SF Whitewater confluence (rainfall in inches) Recurrence--> Duration 50- 100- 200- 500-year 1000-year gage 14-15 Sep 2013 year year year 1-hour 1.83 2.04 2.24 2.5 2.71 3.01 9-10 pm 2-hour 2.02 2.27 2.52 2.86 3.14 4.09 9-11 pm 3-hour 2.12 2.38 2.66 3.04 3.35 6.10 9-midnight 6-hour 2.48 2.8 3.14 3.61 3.99 6.85 9pm-3am 4060 4061 For all four durations shown in the table the event was more rare than a 1,000 year 4062 recurrence. Evidence of storm areal extent is provided by the fact that the Hummingbird 4063 Saddle gage during this event received minimal precipitation. Between 9 pm and 4064 midnight, while the Catwalk gage was recording 6.10 inches of rainfall, the 4065 Hummingbird Saddle gage received zero. For the three hours after midnight, rainfall at 4066 Calwalk dropped off considerably (0.75 inches, total) and Hummingbird Saddle recorded 4067 a similar total (0.93 inches). Figure CS4-28 shows the storm event for which modeling 4068 results are presented below. 4069

DRAFT Technical Note, April 2015 C4-38 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4070 Figure CS4- 28. Whitewater Creek NM flood event of 14-15 Sep 2013, gaged rainfall and streamflow

4071 4072 To have two gages in a post-wildfire watershed may be considered a great luxury, but the 4073 data will be far from adequate to calibrate a hydrologic model. However, such data can 4074 be used to show whether the hydrologic model is reasonable. For example, three runs 4075 with different assumed storm centerings were made for this event. The same areal 4076 reduction factors as shown in figure CS4-21 were used (although the rarest event on that 4077 graph is 100-year, compared to the observed event, more rare than a 1000-year 4078 recurrence). The areal reduction for the observed event could be expected to show a 4079 steeper drop off than the 100-year event in the figure. This is not contradicted by the 4080 concurrent rainfall at Hummingbird Saddle. To estimate sediment bulking, the extreme 4081 rarity and magnitude of the event was considered to justify a 20% estimate for both 4082 brown and yellow sub-watersheds from figure CS4-23. The event occurred about a year 4083 after the wildfire, so some healing could have occurred. However, this event could have 4084 produced debris flow, and uncertainty of sediment transport may be considered 4085 significant for this observed event. To create sediment hydrographs, the Data Storage 4086 System (DSS) functionality of HEC-HMS was employed. A separate program HEC- 4087 DSSVue provides hydrograph manipulation tools, with which sediment hydrographs 4088 were made by multiplying individual sub-basin flow hydrographs by 20%. The database 4089 preserves the timing, so that the sub-basin bulked hydrographs can be summed for an

DRAFT Technical Note, April 2015 C4-39 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4090 outlet such as the Catwalk gage. Only two subwatershed sediment hydrographs were 4091 required (as can be seen by examining figure CS4-23). One for the upper Whitewater 4092 and one for the entire SF Whitewater. These were combined to create a total sediment 4093 hydrograph at the gage. Figure CS4-29 shows modeled versus observed flows. 4094 4095 Figure CS4- 29. Whitewater Creek flood 14-15 Sep 2013, observed and modeled

4096 4097 Conclusions 4098 The post-fire hydrograph results are highly dependent on the amount of hydrophobic soil 4099 and the timing of its breakdown and return to pre-fire conditions. This is impossible to 4100 predict accurately without extensive field investigation. The burn severity mapping can 4101 be considered to at least rule out hydrophobicity in unburned and low burned areas. The 4102 extent to which hydrophobicity truly excludes infiltration in highly burned areas should 4103 be investigated. Although the findings of the research literature indicate that post- 4104 wildfire hydrophobic soils can persist for over five years, critical questions remain for 4105 hydrologic modeling. For example, does hydrophobicity immediately post-fire 4106 completely cover a landscape or only partially? And does the breakdown of hydrophobic 4107 effects on runoff occur more rapidly than the complete disappearance of the phenomenon 4108 because of the non-homogeneity on a landscape? 4109

DRAFT Technical Note, April 2015 C4-40 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4110 The second major issue for hydrologic modeling, particularly in the arid Southwestern 4111 US, is the extent of areal reduction in any given area. A literature search for this analysis 4112 turned up very little, and studies by NOAA that were scheduled to be underway by 2006 4113 apparently have not been completed. 4114 4115 In considering the methodology differences between this case study and others, which 4116 use NRCS Curve Number and dimensionless unit hydrographs, a number of salient points 4117 arise. Practitioners accounting for high burn severity and/or hydrophobicity have found 4118 that CN values should be set very high, usually around 95. See tables 5, 7, 8, and 9 in 4119 this document. This need to approach the very upper limit of CNs may have been 4120 necessary not only because hydrophobicity does, indeed, somewhat mimic a pavement, 4121 but also because the standard NRCS peak factor was not adjusted. The standard peak 4122 factor is usually too low for steep mountain streams. However, the practice of running up 4123 the CN to account for hydrophobicity has no less scientific merit than setting sub-basin 4124 percents of imperviousness, as was done in this case study. The need for some way to 4125 rapidly assess hydrophobic effects becomes apparent. 4126 4127 Although the time-area histogram methodology demonstrated in this case study precludes 4128 the need to determine peak factors, the GIS manipulations require analysis time, which 4129 can be a function of the GIS experience of the hydrologist. Perhaps scripts could be 4130 written to speed up the process, but the greater the number of sub-areas the more analysis 4131 time needed to create time-area based UHs. However, the numerous GIS layers available 4132 for use in post-wildfire analyses, especially regarding soil burn severity, should 4133 encourage the hydrologist to increase GIS proficiency. 4134 4135 References 4136 Bedient, P.B., Huber, W.C., and Vieux, B.E. 2012. Hydrology and Floodplain Analysis, 4137 Prentice Hall, NJ, 5th Edition, 801 pages. 4138 4139 Beven, K.J. 2001. How far can we go in distributed hydrologic modelling? Hydrology 4140 and Earth Systems Sciences 5, 1-12.

DRAFT Technical Note, April 2015 C4-41 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4141 4142 ESRI (Environmental Systems Resource Institute). 2010. ArcMap 10.0 ESRI, Redlands, 4143 California. 4144 4145 MacArthur, R., and DeVries, J.J. (1993) Introduction and application of kinematic wave 4146 routing techniques using HEC-1. U.S. Army Corps of Engineers Hydrologic Engineering 4147 Center, Davis CA, 68 pages. 4148 4149 Moody. T., Wirtanen, M., and Yard, S.N. (2003) Regional relationships for bankfull 4150 stage in natural channels of the arid Southwest. Report from Natural Channel Design, 4151 Inc., Flagstaff AZ, 38 pages. 4152 4153 NOAA National Weather Service Precipitation Frequency Data Server 2013. Available 4154 at 4155 http://dipper.nws.noaa.gov/hdsc/pfds/. Accessed Sep 2013. 4156 4157 Oreskes, N., Shrader-Frechette, K., and Belitz, K. (1994) Verification, validation, and 4158 confirmation of numerical models in the earth sciences. Science 263, 641-646. 4159 4160 Osborn, H.B., Lane, L.J., and Myers, V.A. (1980) Rainfall/watershed relationships for 4161 Southwestern Thunderstorms. Transactions ASAE 23(1), 82-91. 4162 4163 Tillery, A.C., Matherne, A.M., and Verdin, K.L. (2012) Estimated probability of 4164 postwildfire debris flows in the 2012 Whitewater-Baldy fire burn area, Southwestern 4165 New Mexico. USGS Open File Report 2012-1188, 11 pages. 4166 4167 USCOE-HEC (U.S. Army Corps of Engineers, Hydrologic Engineering Center), 2013. 4168 HEC-HMS Technical Reference Manual. Available at 4169 http://www.hec.usace.army.mil/software/hec-hms/. Accessed in April, 2013. 4170

DRAFT Technical Note, April 2015 C4-42 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4171 USDA Forest Service (2012) Glenwood, Reserve and Wilderness Ranger Districts, 4172 BAER Report Executive Summary for the Whitewater Baldy Complex fire, including 4173 hydrology appendices. 4174 4175 USDA-NRCS (U.S. Department of Agriculture-Natural Resources Conservation 4176 Service), 2010. 4177 NRCS National Engineering Handbook, Part 630 Hydrology, Chapter 15, Time of 4178 Concentration. 4179 4180

DRAFT Technical Note, April 2015 C4-43 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 4: Whitewater Creek, Gila Wilderness, New Mexico 4181 (This page intentionally blank)

DRAFT Technical Note, April 2015 C4-44 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4182 Case Study 5: Saratoga Springs, UT 4183 4184 Predicting and comparing measured bulking and peak discharge using 4185 multiple methods for post fire hydrologic and sedimentation analysis on 4186 the Dump Fire in Saratoga Springs 4187 By: Nathaniel Todea, USDA Natural Resources Conservation Service Hydraulic Engineer, 4188 125 South State Street Room 4010, Salt Lake City, Utah 84106, 801.524.4573, 4189 [email protected] 4190 In Proceedings of the SEDHYD 2015 Joint Sedimentation and Hydrologic Modeling 4191 Conference, Reno, NV, April, 2015. 4192 4193 Abstract 4194 As part of the 2012 Utah fire season, analysis of the Dump Fire was conducted to design 4195 sediment basins. The event was estimated to be a 1.25-inch storm that lasted 25 minutes and 4196 dropped an estimated bedload of 70,000 tons of material, which damaged houses, inundated 4197 basements, and overtopped a small basin. The event is comparable to two times the 100-year 4198 (1% chance event) flow. The Dump Fire was analyzed as a part of the United States 4199 Department of Agriculture Natural Resources Conservation Service Emergency Watershed 4200 Protection Program. 4201 4202 The Dump Fire watershed is located near Saratoga Springs, Utah, which is on the eastern 4203 edge of the Basin and Range physiographic region. It was analyzed using the United State 4204 Department of Agriculture Agricultural Research Service Automated Geospatial Watershed 4205 Assessment Tool (USDA-ARS AGWA). The runoff curve number (CN) and derived 4206 hydrologic characteristics were calibrated using local stream gage networks, regression 4207 equations (USGS StreamStats), National Oceanic and Atmospheric Administration (NOAA), 4208 National Weather Service NOAA Atlas 14 rainfall distribution, and modified cumulative 4209 Kirpich time of concentration methods for the pre-fire condition. 4210 4211 Furthermore, specific papers and their methods were analyzed and compared for 4212 modifications to CN based on burn severity and reduction of cover, changes to lag time, and

DRAFT Technical Note, April 2015 CS5-1 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4213 other basin characteristics. Peak discharge and area of basin burned based on lithology and 4214 peak discharge was considered. Lag times were changed due to relative increases in CN. 4215 Fire related debris flow volumes from the Western U.S. regression model were used to 4216 compare final results. 4217 4218 Of the 70,000 tons of bedload material, about 15% made it to a housing development 4219 downstream. Typically, the ratio of sands and colloidal to bedload was estimated at 10:1 or 4220 3:1 ratios. Since the AGWA value was within reason for the total sediment and sands as a 4221 percentage (i.e. 10%), it was assumed to be comparable to total bedload in this case. Overall, 4222 AGWA was found to be a reliable to tool for sedimentation / bulking values in this situation. 4223 4224 Background 4225 In late June 2012 wildfire burned the watershed above the Utah communities of Saratoga 4226 Springs and Eagle Mountain, about 40 miles south of Salt Lake City. Reported to have been 4227 sparked by target shooters, the fire burned approximately 6,000 acres and required the 4228 evacuation of an estimated 9,000 residents. No serious injuries or damages were reported as 4229 a result of the event, which became known at the Dump Fire. Local residents protested the 4230 name, which came about because the fire was started near an old dump. This case study will 4231 refer to the event as the Saratoga Springs Fire. See Figure CS5-1 for location. 4232 4233 The wildfire prompted the City of Saratoga Springs to request assistance from the NRCS 4234 through the Emergency Watershed Protection Program (EWP). Storm damages following an 4235 early September rainfall event (only about a month after the fire) occurred before counter- 4236 measures could be installed. NRCS performed a post-fire hydrologic analysis in support of 4237 the design of a sediment basin to protect residents from the accelerated erosion and 4238 sedimentation caused by the fire. Figures 2 and 3 illustrate the burning watershed from the 4239 point of view of the community of Eagle Mountain. 4240 4241 The storm of 1 September 2012 was centered over an unnamed tributary and Israel Canyon, 4242 which drain into Saratoga Springs. Local officials reported that the rainfall was 1.25 inches 4243 over a 25-minute duration. NRCS engineers estimated that the subsequent runoff deposited a 4244 bedload estimated at 70,000 tons. The mud slurry damaged houses, inundated basements,

DRAFT Technical Note, April 2015 CS5-2 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4245 and filled and overtopped a small debris basin. The event was comparable to two times the 4246 100-year (1% chance event) flow. The flow direction into the residential areas of Saratoga 4247 Springs is shown in Figure CS5-4. The drainages discharge onto an alluvial fan that slopes 4248 into residential development where mud slurry, small boulders, cobbles, and gravel were 4249 deposited during the storm (Figures CS5-5 and CS5-6). This debris damage was caused by 4250 the vulnerability of the watershed immediately following the fire, which increased erosion 4251 and mudflows in the steep gradient alluvial fan. 4252 4253 Figure CS5- 1. State of Utah and fire location, about 40 miles south of Salt Lake City.

4254 4255 Figure CS5- 2. Burning watershed from Eagle Mountain, UT, 21 June 2012.

4256

DRAFT Technical Note, April 2015 CS5-3 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4257 Photo by Cindi Braby, Eagle Mountain, UT 4258 4259

DRAFT Technical Note, April 2015 CS5-4 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4260 Figure CS5- 3. Burning watershed from Eagle Mountain, UT, 22 June 2012.

4261 Photo by Cindi Braby, Eagle Mountain, UT. 4262 4263 Figure CS5- 4. Burned watershed flow direction into Saratoga Springs, UT from 1 September 2012 4264 event.

4265 4266

DRAFT Technical Note, April 2015 CS5-5 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4267 Figure CS5- 5. Sediment laden flow through residential neighborhood.

4268 Photo courtesy of Utah County. 4269 4270 Figure CS5- 6. Typical sediment composition in residential area.

4271 4272 Photo courtesy of City of Saratoga Springs. 4273 4274 Methods 4275 As part of the post-fire hydrologic analysis for the Saratoga Springs fire, a number of 4276 computer programs were used. Runoff hydrographs were determined using the NRCS 4277 hydrology program WinTR-20 (USDA-NRCS(a) 2004). The model Automated Geospatial 4278 Watershed Assessment Tool, or AGWA, (Goodrich et al. 2006 and USDA-ARS 2014) was

DRAFT Technical Note, April 2015 CS5-6 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4279 used to determine sedimentation rates. AGWA was created by the USDA Agricultural 4280 Research Service, Southwest Watershed Research Center in Tuscon AZ. It combines 4281 previously existing models KINEROS2 (Smith et al. 1995, Woolhiser et al. 1990) and SWAT 4282 (Arnold et al. 2012). 4283 4284 The runoff curve number (CN) and derived hydrologic characteristics from local stream gage 4285 networks and regression equations (USGS StreamStats) were used to determine the logical 4286 pre-fire inputs. NOAA Atlas 14 rainfall distribution and modified cumulative Kirpich time 4287 of concentration methods were also used. 4288

4289 Changes to the time of concentration, Tc, and CN inputs into WinTR-20 and AGWA were 4290 based on past studies for post-fire analysis. Goodrich et al., (2005) provides support for 4291 modification of CN values, given reduction of cover and burn severity. McLin et al. (2001) 4292 provides a method to estimate change in lag time associated to relative increases in CN. 4293 4294 Two methods were used to analyze the viability of derived post-fire peaks and debris flow 4295 volumes. These include the Cannon and Gartner (2005) regression equations for estimating 4296 peak debris flow, given burn area and lithology of burn area, basin gradient, and storm 4297 rainfall; and the Gartner et al. (2008) regression equations for estimation of debris flow 4298 volumes. 4299 4300 Typically, the ratio of sands and colloidal grain sizes to bedload is estimated at 10:1 or 3:1 4301 ratios. Since the AGWA value was within reason for the total sediment and sands as a 4302 percentage (i.e. 10%), it was assumed to be comparable to total bedload in this case. 4303 4304 For the hydrologic analysis of the burned watershed above Saratoga Springs, the entire 4305 watershed was assumed to have experienced moderate burn severity. 4306 4307 An initial estimate of pre-fire sedimentation conditions was made using the map of Bridges 4308 (1973) which shows estimated yearly sediment yield and a breakdown between sheet and rill 4309 erosion versus channel and gully erosion. For the Saratoga watershed, pre-fire sediment

DRAFT Technical Note, April 2015 CS5-7 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4310 yield is estimated to range between 0.1-0.2 acre-feet per square mile per year. with sources 4311 being 60% sheet and rill and 40% channel and gully. 4312 4313 Pre-fire flow ranges 4314 The first step in estimating the pre-fire watershed condition was to review existing gages in 4315 the area to determine reasonable pre-fire flows for the 25-, 50-, and 100-year (4%-, 2%- and 4316 1%-chance) events. After this was completed, WinTR-20 was used with NOAA Atlas 4317 distributions and a modified cumulative Kirpich equation. Runoff curve numbers were 4318 modified between ground cover conditions and used USDA – NRCS National Engineering 4319 Handbook, Part 630, Hydrology, Chapter 9, Hydrologic Soil-Cover Complexes (USDA 4320 NRCS(b), 2004), the United States Geologic Survey (USGS) National Land Cover Database 4321 (NLCD) (Fry et al, 2011), and the NRCS SSURGO database (USDA-NRCS(c), 2012). 4322 4323 Stream gages 4324 Six nearby stream gages were analyzed to help estimate peak flows for the pre-fire watershed 4325 above Saratoga Springs. The stream gages are listed in Table CS5-1 and are shown on the 4326 Figure CS5-7 map. Their statistics were taken from the StreamStats report appendix 4327 (Kenney et al. 2007). Discharge per unit area (cubic feet per second / square mile, CSM) was 4328 computed for each probability and graphed (Figure CS5-8). 4329 4330 Table CS5- 1. Six stream gages near Saratoga Springs, UT Drainage # on fig. CS5-8 USGS # gage name (sq. mi.) 1 10172790 5.77 Settlement Canyon nr Tooele UT 2 10172805 5.38 N. Willow Creek nr Grantsville UT 3 10172800 4.19 S. Willow Creek nr Grantsville UT 4 10166430 26.5 W. Canyon Creek nr Cedar Fort UT Clover Creek abv Big Hollow nr Clover 5 10172765 6.7 UT 6 10172910 16.8 Settlement Creek abv Resvr nr Tooele UT 4331

DRAFT Technical Note, April 2015 CS5-8 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4332 Figure CS5- 7. Location of stream gages and study area

4333 4334 Figure CS5- 8. Stream gage regression output converted to discharge per unit area

4335 4336 Of the six nearby gaged watersheds, numbers 1 through 3 (Table CS5-1) are most similar in 4337 drainage area, although the modeled watershed area above Saratoga Springs is 2.5 square

DRAFT Technical Note, April 2015 CS5-9 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4338 miles, which is about half the size of three nearby gages. In Figure CS5-8, the graphs of 4339 those three gages are the higher ones: blue, green, and violet. 4340

4341 Tc Pre-fire 4342 The upper range CSM exceedance probability was used to determine the pre-fire inputs into

4343 WinTR-20. The Tc was estimated to be 1 hour, or an average velocity of 6.0 feet per second. 4344 The upper elevation of the watershed is 7,500 feet above mean sea level, the watershed outlet 4345 is at 4,875 feet above mean sea level, and the longest flow path is 3.8 miles in length. 4346 4347 CN Pre-fire 4348 The CN look-up values were adjusted pending NEH, part 630, Chapter 9 ground cover

4349 conditions (USDA-NRCS(b), 2004) based on NLCD and SSURGO data. The generated Tc, 4350 adjusted CN, and NOAA Atlas 14 rainfall distribution were entered into WinTR-20. The CN 4351 was adjusted until the WinTR-20 output and calculated CSM matched the range of CSM of 4352 nearby stream gages. 4353 4354 Post-fire peaks and volumes of sediment 4355 The burned watershed above Saratoga Springs has a total drainage area of 4.91 square miles. 4356 For WinTR-20 analysis, the burned area was divided into three subareas, as shown in Figure 4357 CS5-9. Subareas 1 and 2 converge and provide outlet to the Saratoga Springs residential 4358 areas shown in Figures CS5-4 through CS5-6. Subarea 1 is known as Israel Canyon. See 4359 Tables CS5-2 and CS5-3 for WinTR-20 basic input related to these subareas, including 4360 selected pre-fire and post-fire CN. 4361

DRAFT Technical Note, April 2015 CS5-10 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4362 Table CS5- 2. Burned watershed subarea input to WinTr-20.

Tc Tc Drainage CN CN Subarea (Pre-fire) (Post-fire) (sq. mil) (Pre-fire) (Post-fire) (hrs) (hrs) 1 1.81 62 74 0.92 0.85 2 0.60 74 80 0.40 0.37 1 + 2 2.50 65 75 1.00 0.92 4363 4364 Figure CS5- 9. Burned zone (red) and sub-areas upstream of Saratoga Springs.

4365 4366 4367 CN Post-fire modification 4368 Post-fire CN were selected based on Goodrich (et al. 2005), which stated that, “there [is] a 4369 15% reduction in cover for low-severity burns, a 32% reduction for moderate-severity burns, 4370 and a 50% reduction for high-severity burns” (Figure 10). 4371 4372

DRAFT Technical Note, April 2015 CS5-11 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4373 Figure CS5- 10. Curve number from cover, for hydrologic soil groups (Goodrich et al. 2005)

4374 4375 4376 Table CS5-3 shows the CN increase from pre-fire to post-fire CN used in this study. The 4377 table associates these with the standard National Land Cover Dataset (NLCD) and hydrologic 4378 soil groupings (hsg). 4379 4380 Table CS5- 3. Increase in runoff curve number from pre-fire to post-fire conditions. Burn severity NLCD Cover name HSG A HSG B HSG C HSG D 41 Deciduous Forest 4 5 3 2 42 Evergreen Forest 4 16 10 8 Low 43 Mixed Forest 4 5 3 2 51 Shrubland 2 2 1 1 41 Deciduous Forest 10 10 5 5 42 Evergreen Forest 10 21 12 11 Moderate 43 Mixed Forest 10 10 5 5 51 Shrubland 5 5 3 2 41 Deciduous Forest 15 16 8 7 42 Evergreen Forest 15 27 15 13 High 43 Mixed Forest 15 16 8 7 51 Shrubland 10 11 6 3 4381 4382 Time of concentration post-fire modification

4383 Tc was adjusted using McLin et al. (2001), which suggest that lag time decreases from the 4384 pre-burn to post-burn condition as a result of increase in CN (Figure CS5-11). Note that this

DRAFT Technical Note, April 2015 CS5-12 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4385 depends on channel blockages caused by the fire, and frequency that the watershed 4386 experiences wildfire. Channel blockages can possibly increase lag times. In this case, 4387 however, the watershed cover is generally mixed with deciduous forest and low-lying shrubs 4388 and no channel blockages were assumed. Furthermore, the roughness of the watershed was 4389 assumed to decrease as a result of fire. For this case study, the following rule of thumb was 4390 adopted for changes in runoff velocity and associated change in time of concentration: 4391 velocity increases 0.5 feet per second for low severity burns, 1.0 feet per second for moderate 4392 severity burns, and 1.5 feet per second for high severity burns. 4393 4394 Figure CS5- 11. Interdependency of CN and lag time, Cerro Grande Wildfire (McLin et al. 2001)

4395 4396 *Relative change is defined as the sum of pre-fire and post-fire values divided by pre-fire values. 4397 4398 AGWA 4399 The AGWA model was used to model sediment rates. AGWA has a GIS interface and uses 4400 one of two routines to route runoff using the kinematic wave equation in the sub-model 4401 KINEROS2. In AGWA the landscape, including land uses and management practices, are 4402 handled by the sub-model SWAT. Table 4 shows output parameters for KINEROS and 4403 SWAT. Although AGWA produces both flow hydrographs and sedimentation rates, only the 4404 latter was used for the current study. 4405

DRAFT Technical Note, April 2015 CS5-13 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4406 Table CS5- 4. Output variables available in AGWA. KINEROS -  Infiltration (mm; m3/km),  SWAT- Channel Discharge (m3/day),  Infiltration (in; ac-ft/mi),  Runoff (mm), ET (mm),  Percolation (mm), Surface Runoff (m3),  Sediment yield (kg/ha),  runoff (mm),  Transmission loss (mm),  Peak flow (m3/s), Peak flow (mm/hr),  Water yield (mm),  Sediment yield (t/ha), Sediment discharge (kg/s)  Precipitation (mm) 4407 4408 Bulking 4409 Another way to estimate sedimentation is to consider typical runoff bulking factors for 4410 recently burned watersheds. This was done to further support concentration volumes that 4411 were deposited on the alluvial fan. Considering a 20% bulking factor, an event 4412 sedimentation volume can be estimated and applied to WinTR-20 post fire results. 4413 4414 Western regional equation 4415 The empirical Western U.S. regression model to estimate fire-related debris-flow volumes 4416 Gartner et al. (2008) was used to estimate post fire sediment. The equation used is presented 4417 below (Equation 1). The Western U.S. regression model to estimate fire-related debris-flow 4418 volumes Gartner et al. (2008) was taken from the Giraud and Castleton (2009) investigation. 4419 = 0.59( ) + 0.65 / + 0.18 / + 7.21 (eq. CS5- 1) 1 2 1 2 4420 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙 𝐵𝐵 𝑅𝑅 4421 where: V = volume (cubic meters) 4422 S = basin area with slopes greater than or equal to 30% (square kilometers) 4423 B = basin area burned at moderate and high severity (square kilometers), and 4424 R = total storm rainfall (millimeters) 4425 4426 Comparative analysis to existing studies 4427 Figure CS5-12 from McLin et al. (2001) illustrates that the change in peak discharge per unit 4428 area caused by wildfire can be quite large, therefore the pre and post fire ranges from Figure 4429 CS5-12 were considered. 4430

DRAFT Technical Note, April 2015 CS5-14 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4431 Figure CS5- 12. Comparison of observed and simulated pre-and post-fire peak discharges per unit 4432 drainage basin area in New Mexico. (McLin et al, 2001)

4433 4434 Regional equations 4435 The regression equations of Cannon and Gartner (2005) were also used to estimate post-fire 4436 flow. Since the Saratoga Springs watershed is sedimentary, the regression equation (eq. 2) 4437 from Cannon and Gartner (2005) was used to compute an estimated peak discharge. Since 4438 the equation units are SI, conversions are required. For the Saratoga Springs watershed, 2.5 4439 square miles converts to 6.47 square kilometers. 4440

0.4 2 4441 QApb= 17 R = 0.42 (eq. 2) 4442 4443 Bridges map 4444 The Bridges (1973) map entitled “Estimated Sediment Yield Rates for the State of Utah” 4445 references many data sources as part of the map, including: 1) Great Basin Upper Colorado 4446 and Lower Colorado Regions, Comprehensive Framework Study, Appendices VIII, Water 4447 Management, June 1971, Pacific Southwest Inter-Agency Committee/Water Resources 4448 Council, 2) Utah State soils map and soil descriptions, 3) Reservoir Surveys by SCS and 4449 USBR, 4) suspended load measurements by USGS, USGR and SCS, 5) Watershed studies by 4450 SCS, and 6) General knowledge of the state from regular SCS program work. The author 4451 notes, “Do not use these rates to determine sediment yields at specific sites. Large variations 4452 in sediment rates may occur within the delineated areas”. 4453

DRAFT Technical Note, April 2015 CS5-15 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4454 According to the United States Department of Agriculture, Soil Conservation Service map 4455 (Bridges, 1973) the sediment yield for the Saratoga watershed ranges between 0.1-0.2 acre- 4456 feet/square mile/year with a 60% sheet and rill erosion and 40% channel and gully erosion. 4457 4458 The range of erosion rates for the Saratoga Springs watershed from the 1973 map are plotted 4459 (Figure 13). On the graph, the red squares represent the acre-feet/square mile/year rate that 4460 correlates to the tons for 2.5 square miles, pre-burn condition. The blue diamonds on the 4461 graph show values assuming a 0.45 acre-feet/square mile/year rate plotted in total tons, a 4462 conservative pre-fire condition. Finally, the green triangles represent the breakdown of 4463 erosion types (60% sheet / rill erosion, 60% other colloidal material, 40% channel and gully 4464 erosion), a post-fire condition. 4465 4466 Figure CS5- 13. Range of annual erosion rates from Bridges (1973).

35000.0 30000.0 25000.0 (60-60-40) Conservative Tons Total (tons) 20000.0 Tons ~0.45

Tons 15000.0 10000.0 5000.0 Conversion (@ 2.5 sq.mi) sq.mi) 2.5 (@ Conversion 0.0 0.1 1 10 4467 Acre feet /square mile/year range from 1973 map 4468 4469 Results 4470 WinTR-20 Pre and Post Fire: The WinTR-20 provides pre-fire flow calculations that 4471 correlated well to the USGS stream gages CSM provided in Figure 8. Table CS5-5 below 4472 illustrates these results. The lower return intervals from WinTR-20 pre-fire results are higher 4473 and the higher return intervals are lower than USGS calculated CSM. For estimation 4474 purposes, the upper results for the 25 year (4% chance), 50 year (2% chance) and 100 year 4475 (1% chance) will be focused on during the rest of the paper. Table CS5-6 reflects the pre- 4476 and post- fire peak discharges. 4477

DRAFT Technical Note, April 2015 CS5-16 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4478 Table CS5- 5. Cubic feet per second per square mile (CSM) from WinTR-20 pre-fire results and selected 4479 stream gages. 2-year 5-year 10-year 25-year 50-year 100-year (50%- (20%- (10%- (4%- (2%- (1%- chance) chance) chance) chance) chance) chance) (CSM) (CSM) (CSM) (CSM) (CSM) (CSM) WinTR-20 Pre-fire 0 2.5 6.8 28.4 58 100.8 Observed CSM ~5 12 <20 30 45 60 from Figure 8 4480 4481 Table CS5- 6. Burned watershed pre-fire and post-fire peak flow output from WinTR-20. 2-year 5-year 10-year 25-year 50-year 100-year Subarea (cfs) (cfs) (cfs) (cfs) (cfs) (cfs) 1 (pre-fire) 0 0 0 25 65 127 1 (post-fire) 14 57 118 234 359 514 2 (pre-fire) 6 29 62 126 195 280 2 (post-fire) 39 92 149 248 347 466 1+2 (pre-fire) 0 6 17 71 145 252 1+2 (post-fire) 26 92 179 342 516 728 4482 4483 The WinTR-20 output in Figure 14 shows the considerable increase in runoff peaks and 4484 volumes due to the fire. The peaks are predicted to more than double, with the 25-year (4% 4485 chance) event (red dashed for post-fire versus black dashed for pre-fire) rising from 71 cfs to 4486 342 cfs. The runoff volume (represented by the area under each curve and in Table CS5-7) is 4487 predicted to increase runoff from the pre-fire to post-fire event, 192% to 122% for the 25- 4488 year to 100-year events. 4489

DRAFT Technical Note, April 2015 CS5-17 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4490 Figure CS5- 14. Burned watershed pre-fire and post-fire hydrographs from WinTR-20.

4491 4492 4493 Table CS5- 7. Storm totals runoff for various recurrence intervals, input to WinTR-20 and post-fire 4494 runoff values for Subarea 1+2. 2-year 5-yr 10-yr 25-yr 50-yr 100-yr 200-yr 500-yr Percent (50%- (25%- (10%- (4%- (2%- (1%- (0.5%- (0.2%- Chance chance) chance) chance) chance) chance) chance) chance) chance) Pre-Fire runoff (inches) 0.01 0.03 0.07 0.13 0.19 0.27 0.35 0.48 Subarea 1+2 Post Fire runoff (inches) 0.09 0.17 0.25 0.38 0.48 0.60 0.73 0.92 Subarea 1+2 Pre Fire - Runoff in acre 1 4 9 17 25 36 47 64 feet over 2.5 sq.mi Post Fire - Runoff in acre 12 23 33 51 64 80 97 123 feet over 2.5 sq.mi 4495

DRAFT Technical Note, April 2015 CS5-18 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4496 AGWA 4497 The sediment for the 25-year (4% chance) storm was estimated to be 10,303 tons and the 4498 100-year (1% chance event) storm was estimated to be 27,897 tons. 4499 4500 Bulking 4501 An event sedimentation volume was estimated considering a 20% bulking factor on the 4502 WinTR-20 post fire results. The information in Table CS5-8 was derived by taking the post- 4503 fire clear flow hydrographs of Figure CS5-14, from WinTR-20, and considering the event 4504 sedimentation rate to be 20% at each time-step. The area under the sedimentation 4505 hydrograph provided a volume, which was converted to tons by assuming a sediment unit 4506 weight of 108 pounds per cubic foot in Table CS5-8. 4507 4508 Table CS5- 8. Burned watershed post-fire event total sediment runoff in tons. Event totals 25-yr 50-yr 100-yr (4% chance) (2% chance) (1% chance) (tons) (tons) (tons) Sediment 23,705 29,943 37,429 4509 4510 Western U.S. regression model 4511 The empirical Western U.S. regression model to estimate fire-related debris-flow volumes 4512 Gartner et al. (2008) was estimated to be between 34,550 – 51,182 tons for the 2- to 100-year 4513 (50% to 1% chance) rainfall events. These numbers are high since the model assumes that 4514 watersheds typically have moderate to high severity burns. 4515 4516 Table CS5-9 shows the WinTR-20 pre-fire and post-fire peaks (in cfs per square mile) for 4517 comparison with Figure CS5-12. The current case study results would plot generally lower 4518 on the figure than the New Mexico watersheds, the WinTR-20 results are considered 4519 reasonable. 4520 4521

DRAFT Technical Note, April 2015 CS5-19 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4522 Table CS5- 9. Saratoga Springs burned watershed pre-fire and post-fire peaks (CSM) from WinTR-20. 25-year 50-year 100-year Flood (4% chance) (2% chance) (1% chance) Pre-fire (CSM) 28.4 58 100.8 Post-fire (CSM 136.8 206.4 291.2 4523 4524 Regional equations

4525 The regression equations of Cannon and Gartner (2005) after inserting 6.47 as Ab into the

4526 equation results in a Qp of 35.88 cubic meters per second, or a Qp of 1267 cfs after converting 4527 to English units . This results in 506 CSM; and the WinTR-20 post-fire result is 728 cfs or 4528 291 CSM. 4529 4530 Conclusions 4531 The review of USGS stream gages and making modification to the CN to have both values 4532 correlate was an attempt to familiarize one on potential results, pre-fire. Once this was done

4533 a post-fire CN was applied and Tc was lowered. This resulted in both higher peaks and 4534 bigger volumes of runoff. AGWA, a 20% bulking to WinTR-20 post fire results, and the 4535 Western Regional Equation overall correlated well. These results were compared to McLin 4536 et al (2001) pre- and post- fire peak discharges per unit drainage basin area and to a range or 4537 potential erosion values, Bridges sedimentation map (1973). 4538 4539 Typically, one would think of using the peak discharge of post-fire analysis using WinTR-20,

4540 while increasing the CN and decreasing the Tc. The peak discharge during the flood event 4541 and high water marks were observed as being in range for the post-fire WinTR-20 results. 4542 The Cannon and Gartner (2005) estimate of over 1,200 cfs may be reasonable and could be 4543 used for preliminary or lower and upper bound limits; the lower limit at 75% of the total, and 4544 the upper at 100% of the total. 4545 4546 The range of sediment, bulking and mud slurries were in a range of 10,000 tons to over 4547 50,000 tons for the post-fire 25- to 100-year (4% to 1% chance) events. The AGWA range is 4548 between 10,000 and 50,000 tons for the 25- to 100-year (50% to 1% chance) post-fire event. 4549 The bulking at 20% is 23,000 to 37,000 tons for the post-fire event. The western regional

DRAFT Technical Note, April 2015 CS5-20 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4550 equation produced a number of 34,600 – 51,200 tons. Between the sand, colloidal, and 4551 bedload the percentages can vary. For the sediment, sand and colloidal the numbers above 4552 could still be a low estimate. Bedload at either 1:3 or 1:10 ratio could be still be small. 4553 However, due to the system being flushed by previous storm events, these ratios may be 4554 accurate. A range of 5,000 to 17,000 tons could be accounted for bedload. 4555 4556 Acknowledgements 4557 The author appreciates the assistance, review, and support of Claudia Hoeft, Dan Moore, 4558 Steve Yochum, Kent Sutcliffe, Norm Evenstad, Bronson Smart, the AGWA ARS staff, and 4559 Utah USDA NRCS. 4560 4561 4562 References 4563 Arnold, J. G., D. N. Moriasi, P. W. Gassman, K. C. Abbaspour, M. J. White, R. Srinivasan, 4564 C. Santhi, R. D. Harmel, A. van Griensven, M. W. Van Liew, N. Kannan, M. K. Jha, SWAT: 4565 Model Use, Calibration and Validation. Transactions of the ASABE. 55(4): 1491-1508. 4566 4567 Bridges, B.L. 1973. Map: Estimated sediment yield for the State of Utah. Erosion and 4568 Sedimentation, Western US Water Plan, prepared by SCS August 1973. 4569 4570 Cannon, S.H. and Gartner, J.E. 2005. Wildfire-related debris flow from a hazards 4571 perspective, in Jakob, M. and Hungr, O., Debris-flows hazards and related phenomena: 4572 Chichester, United Kingdom, Springer-Praxis Books, p. 362–385. 4573 4574 Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C., Herold, N., and 4575 Wickham, J., 2011. Completion of the 2006 National Land Cover Database for the 4576 Conterminous United States, PE&RS, Vol. 77(9):858-864. 4577 4578 Gartner, J.E., Cannon, S.H., Santi, P.M., and deWolfe, V.G. 2008. Empirical models to 4579 predict the volumes of debris flows generated by recently burned basins in the western U.S. 4580 Geomorphology, 96, 339-354. 4581

DRAFT Technical Note, April 2015 CS5-21 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4582 Giraud, R.E., and Castleton, J.J. 2009. Estimation of Potential Debris-Flow volumes for 4583 Centerville Canyon, Davis County, Utah. Report of Investigation 267 Utah Geological 4584 Survey, Utah Department of Natural Resources, Salt Lake City, Utah. 4585 4586 Goodrich, D.C., H. E. Canfield, I.S. Burns, D.J. Semmens, S.N. Miller, M. Hernandez, L.R. 4587 Levick, D.P. Guertin, and W.G. Kepner. 2005. Rapid Post-Fire Hydrologic Watershed 4588 Assessment using the AGWA GIS based Hydrologic Modeling Tool. Proc. ASCE Watershed 4589 Manage. Conf., July 19-22, Williamsburg, VA. 4590 4591 Kenney, T.A., Wilkowske, C.D., and Wright, S.J. 2007. Methods for estimating magnitude 4592 and frequency of peak flows for natural streams in Utah: U.S. Geological Survey Scientific 4593 Investigations Report 2007-5158, 28 p. 4594 4595 McLin, S. G., Springer, E. P., and Lane, L. J. 2001. Predicting floodplain boundary changes 4596 following the Cerro Grande wildfire. Hydrological Proc., 15(15): 2967-2980. 4597 4598 Smith, R.E., Goodrich, D.C., and Quinton, J.N. 1995. Dynamic, distributed simulation of 4599 watershed erosion: The KINEROS2 and EUROSEM models, Journal of Soil and Water 4600 Conservation, 50(5):517-520. 4601 4602 USDA-ARS (U.S. Department of Agriculture-Agricultural Research Service). 2014. AGWA 4603 (Automated Geospatial Watershed Assessment Tool) website, through the ARS Southwest 4604 Watershed Research Center, Tuscon AZ, http://www.tucson.ars.ag.gov/agwa/. 4605 4606 USDA-NRCS(a) (U.S. Department of Agriculture-Natural Resources Conservation Service), 4607 2004. WinTR-20 User Guide. Available at http://go.usa.gov/KoZ/. Accessed in April 2013. 4608 4609 USDA-NRCS(b), 2004. National Engineering Handbook, Part 630 Hydrology, Chapter 9, 4610 Hydrologic Soil-Cover Complexes. 4611 4612 USDA-NRCS(c), 2012. Soil Survey Geographic (SSURGO) Database. Available online at 4613 http://sdmdataaccess.nrcs.usda.gov/. Accessed [2012].

DRAFT Technical Note, April 2015 CS5-22 DRAFT Technical Note Hydrologic Analyses of Post-Wildfire Conditions Case Study 5: Saratoga Springs, UT 4614 4615 Woolhiser, D.A., Smith, R.E., and Goodrich, D.C. 1990. KINEROS, A kinematic runoff and 4616 erosion model: Documentation and User Manual. U.S. Department of Agriculture, 4617 Agricultural Research Service, ARS-77, 130 pp. 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635

DRAFT Technical Note, April 2015 CS5-23