Memorandum. Background

Transcription

1 Memorandum To: Kevin Stewart, P.E., Information Systems & Flood Warning Program Manager From: Mark Mitisek, H.I.T. Reviewed by: Kelly Close, P.E. Date: 02/11/2013 Project: Boulder Creek Hydromodel Subject: Hydromodel Evaluation for the July 30 th, 2012 storm event The purpose of this memorandum is to document and evaluate the current status of the Boulder Creek hydromodel, prior to the 2013 flood season. Ongoing model simulation, evaluation, and calibration of the model are important to accurate real-time flood warning. To evaluate the current status of the hydromodel the following tasks will be completed: Review the background of the Boulder Creek hydromodel Document observed precipitation and streamflow of July 30 th 2012 storm event Document simulated precipitation and streamflow of July 30 th 2012 storm event Evaluate the model performance using the July 30 th 2012 storm event Review and evaluate model components and model inputs Conclusions and recommendations for future model enhancements Background The original Boulder Creek model was developed by George Sabol using HEC-1. The original model included 15 sub-basins representing the headwaters of Boulder Creek to the Canyon mouth, including the entire Fourmile Creek watershed. The output from the model included S-graphs and Muskingum routing parameters for each sub-basin. In March 1990, the District developed Simplified Mountain Canyon Flash Flood Guidance for Boulder Creek based on the original USACE HEC-1 model (Sabol, 1990). This report provided the unit graph parameters and curve number information necessary to model rainfall runoff for each sub-basin. This information was used to accurately reproduce the model in an Excel spreadsheet model with similar results. The Excel based Boulder Creek hydromodel has been operating since 2002 as a part of the District s real-time flood ALERT system. In February of 2011, the Boulder Creek hydromodel was revised to include additional sub-basins used to evaluate the increase in flood potential in the area due to the Fourmile Cranyon fire. The revision of the model included the development of sub-basin specific unit hydrographs, effective rainfall estimates to reflect the change due to the fire, and revised precipitation station weights. These revisions were documented in Pre and Post-Fire Hydrology of Fourmile Creek (LRE, 2011). In May of 2012, the Boulder Creek hydromodel was evaluated using two 2011 storm events of July 7 th and July 13 th. It was observed that the hydromodel was simulating the correct rainfall runoff volume, but the timing was off by an average of 30 minutes for the estimated time-to-peak. The

2 Boulder Creek Hydromodel Evaluation Kevin Stewart, P.E. Page 2 model travel times were revised to better reflect the July 2011 storm events. These revisions were documented in Fourmile Creek Model Calibration (LRE, 2012). During the 2012 flood season the most significant storm event was July 30 th The simulation of this storm event will be used as the basis for this evaluation of the status of the models performance and help to identify additional future model enhancements, calibration, and recommendations. Model Performance of the July 30 th 2012 Storm Event The following section documents the simulated and observed precipitation and streamflow of the model from the July 30 th 2012 storm event. It is important to note that the evaluation of this event is in context of how the model has behaved compared to past events. Precipitation Table 1 and Table 2 below document the time series of the total precipitation measured at each station and the basin total for the July 30 th 2012 event occurring between 2:30 PM and 4:00 PM. Figure 1 shows the gauge adjusted radar rainfall (GARR) and cumulative rainfall total of the July 30 th 2012 storm event. Observed Precipitation The tabulated summary of July 30 th precipitation station totals below shows the event starting in the northwest portion of the watershed and continuing southeast toward Gold Run where the highest station totals of 1.06 inches were observed. The highest station totals were seen in the upper parts of the Fourmile Creek watershed and in or near the burned area. SHEF ID Table 1 July 30, 2012 observed Precipitation (inches) Time Station Total 2:30 PM 2:45 PM 3:00 PM 3:15 PM 3:30 PM 3:45 PM 4:00 PM TWIC MAGC FILC BETC SWIC LOGC GLHC SUNC PBHC SUSC JUCC BOJC STAC WRDC SFSC LEWC

3 Boulder Creek Hydromodel Evaluation Kevin Stewart, P.E. Page 3 Simulated Precipitation Precipitation station totals are ingested into the model and station assignments and basin precipitation total are calculated based on established Theissen polygon weights in each basin (Table 2). The precipitation basin totals are then used in the effective precipitation calculations and are distributed using basin specific unit hydrographs to determine runoff volumes. Similar to the observed precipitation station totals, the basin totals reflect the event starting in the northwest portion of the watershed with significant totals seen in BC-3 around 3:15 and continuing southeast toward Gold Run where the highest basin totals of 1.06 inches were observed around 3:45 in BC- 13C and BC-13D. The highest basin totals were seen in the upper parts of the Fourmile Creek watershed and in or near the burned area. Basin Table 2 July 30, 2012 Simulated Precipitation (inches) Time 2:30 PM 2:45 PM 3:00 PM 3:15 PM 3:30 PM 3:45 PM 4:00 PM Basin Total BC BC BC BC BC BC BC BC BC BC-12A BC-12B BC-12C BC-12D BC-12E BC-13A BC-13B BC-13C BC-13D BC-14A BC-14B BC-14C BC-14D BC Max GARR Precipitation Although gauge adjusted radar rainfall (GARR) is not currently an input into the model it can be used to see the temporal and spatial distribution of the July 30 th 2012 event. The storm track was again generally from the northwest to southeast, but GARR showed the storm moving along the ridgeline with the maximum 1 km grid total of the storm of 0.79 inches (Figure 1).

4 Boulder Creek Hydromodel Evaluation Kevin Stewart, P.E. Page 4 Figure 1 July 30, 2012 GARR (inches) (UDFCD, 2012) Streamflow The July 30 th, 2012 flood event was well documented by UDFCD and USGS, including peak discharge and the timing of peak flows at different design points throughout the basin. The summaries below show the observed and simulated peak discharge and timing of peaks. Peak Discharge Table 4 below is a summary of the observed and simulated peak discharge corresponding modeled design points for the storm event: Table 4 July 30, 2012 Observed Peak and Simulated Peak Design Point Observed Peak (CFS) Simulated Peak (CFS) Fourmile Creek at Salinas 278* 575 Fourmile Creek at Logan Mill Road Fourmile Creek at Orodell Boulder Creek near Orodell Boulder Creek At Bridge Boulder Creek At Canyon Mouth *Note the observed peak flow from Fourmile Creek at Salinas is questionable.

5 Boulder Creek Hydromodel Evaluation Kevin Stewart, P.E. Page 5 Time of Peak Table 5 below is a summary of the observed and simulated time of peak corresponding modeled design points for the storm event: Table 5 July 30, 2012 Observed Time of Peak and Simulated Time of Peak Design Point Observed Time of Peak Simulated Time of Peak Fourmile Creek at Salinas 4:02 PM 3:55 PM Fourmile Creek at Logan Mill Road 4:05 PM 4:00 PM Fourmile Creek at Orodell 4:45 PM 4:35 PM Boulder Creek near Orodell 11:00 PM 7:06 PM Boulder Creek At Bridge 4:52 PM 4:45 PM Boulder Creek At Canyon Mouth 5:45 PM 5:20 PM Model Evaluation of the July 30 th 2012 Storm Event The simulation of the July 30 th 2012 event showed the model simulating higher than observed rainfall runoff volumes, but the estimated time to peak has improved. The Boulder Creek hydromodel consist of four modeled components: 1. Precipitation Inputs 2. Effective Rainfall Estimates 3. Unit Hydrograph Distributions 4. Routing (timing) Each of these components is dependent on the other and the evaluation of model behavior needs to focus on both each component individually and in aggregate. Reviewing the July 30 th 2012 event confirmed the unit hydrographs and routings all provide a reasonable estimate of the July 30 th 2012 event. However, the precipitation inputs and effective rainfall estimates were likely overestimated compared with GARR estimates, thus resulting in the model simulating higher than observed rainfall runoff volumes and higher peak discharges. Precipitation Inputs Although the precipitation inputs and Theissen weighting was representative of the July 30 th 2012 event, when compared to the GARR, the total rainfall and distribution of rainfall were significantly different. Future modeling efforts may want to include the use of GARR rainfall inputs. Effective Rainfall Estimates Previous model enhancements and evaluations have resulted in a better representation of routing model components. The original anticipated post-fire curve numbers that were developed match the slower timing of the original model. As the model was recalibrated to reflect actual events, the timing was changed to reflect a quicker response from each of the burned basins as observed by the July 13, 2011 event. Although the routing parameters were revised, the effective rainfall estimates (curve numbers) were not. The result is the overestimation of peak discharges from the July 30 th 2012 event.

6 Boulder Creek Hydromodel Evaluation Kevin Stewart, P.E. Page 6 Conclusion The July 30 th 2012 event was a relatively small event, which are sometimes more difficult to characterize. The simulation of the July 30 th 2012 event showed the model simulating higher than observed rainfall runoff volumes, but the estimated time to peak has improved. There are two primary causes of the overestimated peak for this event. The first is the significant difference in precipitation inputs. When comparing Theissen weighted ALERT station precipitation inputs to the gauge adjusted radar rainfall (GARR) it is apparent that the rainfall totals and simulated distribution are significantly different. GARR showed the storm track moving along the ridgeline with the maximum 1 km grid storm of 0.79 inches. The Theissen weighted ALERT station precipitation inputs show a more general trend moving from the northwest to the southeast with maximum basin total of 1.06 inches. The second is unrepresentative effective rainfall estimates based on current watershed conditions. Recovery of the watershed has resulted in a change in the burn severity of the each basin and therefore the amount of effective precipitation. The original anticipated post-fire curve numbers were developed based on the low, moderate, and high burn severities in each basin, with some basins that were severely burned having curve numbers as high as 93. This is no longer representative of the post fire conditions within the Fourmile Creek watershed. To correctly simulate the July 30 th 2012 storm event the curve numbers should be reduced to represent low to moderate burn severity. Figures 2-6 show the simulated July 30 th 2012 storm event with small corrections to timing and reduced curve numbers. Recommendations Precipitation Inputs Currently the Boulder Creek hydromodel uses a very simple approach to representing the spatial and temporal distribution of precipitation in each basin. Although the use of Theissen polygons weights to determine basin rainfall totals is effective, it may result in an over or under estimation of actual precipitation for localized events. Alternative methods such grid based QPE and QPF should be explored to allow for an increase and accuracy of spatially and temporally distributed precipitation and will also help to increase warning times. Recommendation: Explore the feasibility of incorporating published QPF and QPE products into the existing model to better represent the temporal and spatial distribution of precipitation. Effective Rainfall Estimates The modeled simulation of the July 30 th 2012 event shows significant overestimation of peak discharge compared to observed flows. The effective rainfall volumes simulated are no longer representative of the post-fire condition of the Fourmile Creek watershed. The post-fire anticipated curve numbers that are currently used in the model represent high to moderate burn severity with curve numbers as high as 93 in some basins. Also, the revision to routing parameters

7 Boulder Creek Hydromodel Evaluation Kevin Stewart, P.E. Page 7 and timing has compounded this issue further. The result is significant overestimation of peak discharge. Recommendation: Reduce the curve numbers to represent low to moderate fire severity in the Fourmile Creek basin and continue to monitor the storm events and confirm corrected peak discharge estimates. Timing The modeled simulation of the July 30 th storm shows the routing of the storm hydrograph has improved and is more representative of the travel times in each basin. Although the timing has improved, some travel times can be revised to better represent the July 30 th 2012 storm event. In general it was observed that a small increase in travel times (1-2 minutes) for most basins better represents the timing of the July 30 th 2012 storm event. Note that the slight increase in travel time could be the result of pre-existing low flows in 2012 on Fourmile Creek. Recommendation: Continue to monitor timing and consider increasing the travel time for Fourmile Creek basins and Boulder Creek below Orodell if timing issues continue. Continued Model Evaluation The Boulder Creek hydromodel is not a static model that can be used year after year without continued evaluation and comparison to observed events. The changes to the watershed as a result of the fire are ongoing. Therefore ongoing evaluation of model performance, parameters, and inputs are important to the accurate representation of real-time events. Recommendation: Continue ongoing model simulation, evaluation, and calibration of the model. References Sabol, Simplified Mountain Canyon Flash Flood Guidance For Boulder Creek, Denver, CO: George Sabol Consulting Engineers, Inc. LRE, Mitisek, M. (2011). "Pre and Post Fire Hydrology of Fourmile Creek". Denver: Leonard Rice Engineers, Inc. LRE, Mitisek, M. (2012). "Fourmile Creek Model Calibration". Denver: Leonard Rice Engineers, Inc. UDFCD, Stewart P.E., K. (2012). "Information Services Flood Warning Program Notes". Retrieved from

8 Boulder Creek Hydromodel - July 30, 2012 Fourmile Creek Design Points Model Run 7/30/ :35 Gold Run Estimated Peak Flow = 104 FMC at Salina Estimated Peak Flow = 123 FMC at Crisman Estimated Peak Flow = 122 Estimated Time of Peak = 4:05 PM CFS 60 FMC at CO 119 Estimated Peak Flow = 101 Estimated Time of Peak = 4:40 PM :12 PM 2:24 PM 3:36 PM 4:48 PM 6:00 PM 7:12 PM 8:24 PM 9:36 PM 10:48 PM

9 Boulder Creek Hydromodel - July 30, 2012 Fourmile Creek Design Points Model Run 7/30/ :45 Gold Run Estimated Peak Flow = FMC at Salina FMC at Crisman Estimated Time of Peak = 4:05 PM CFS 80 FMC at CO 119 Estimated Peak Flow = 131 Estimated Time of Peak = 4:45 PM :12 PM 2:24 PM 3:36 PM 4:48 PM 6:00 PM 7:12 PM 8:24 PM 9:36 PM 10:48 PM

10 Boulder Creek Hydromodel - July 30, 2012 Fourmile Creek Design Points Model Run 7/30/ :55 Gold Run Estimated Peak Flow = FMC at Salina FMC at Crisman Estimated Time of Peak = 4:05 PM CFS 80 FMC at CO 119 Estimated Peak Flow = 131 Estimated Time of Peak = 4:45 PM :12 PM 2:24 PM 3:36 PM 4:48 PM 6:00 PM 7:12 PM 8:24 PM 9:36 PM 10:48 PM

11 Boulder Creek Hydromodel - July 30, 2012 Fourmile Creek Design Points Model Run 7/30/ :05 Gold Run Estimated Peak Flow = FMC at Salina FMC at Crisman Estimated Time of Peak = 4:05 PM CFS 80 FMC at CO 119 Estimated Peak Flow = 131 Estimated Time of Peak = 4:45 PM :12 PM 2:24 PM 3:36 PM 4:48 PM 6:00 PM 7:12 PM 8:24 PM 9:36 PM 10:48 PM

12 Boulder Creek Hydromodel - July 30, 2012 Fourmile Creek Design Points Model Run 7/30/ :15 Gold Run Estimated Peak Flow = FMC at Salina FMC at Crisman Estimated Time of Peak = 4:05 PM CFS 80 FMC at CO 119 Estimated Peak Flow = 131 Estimated Time of Peak = 4:45 PM :12 PM 2:24 PM 3:36 PM 4:48 PM 6:00 PM 7:12 PM 8:24 PM 9:36 PM 10:48 PM