Estimating the Probable Maximum Flood (PMF) Using an Existing HEC-HMS Model

HEC-HMS version 4.10 was used to develop this example.

Download the initial model files here - PA PMF Workshop -Existing Model_Start.zip

Introduction

This workshop provides an example for developing the Probable Maximum Flood (PMF) using an existing HEC-HMS model of the Sayers Dam watershed.  The existing model was calibrated to three historic flood events.  The original Spillway Design Flood (SDF) did not use HMR 52 to develop the PMP and it is not clear how the flood hydrograph was computed.  Consequently, use of the SDF does not allow the modeler to complete a sensitivity analysis of the hydrologic processes and assumptions associated with the reservoir Inflow Design Flood (IDF).  Therefore, reuse of the SDF contained in the reservoir regulation manual is not appropriate.

The existing Bald Eagle Creek HEC-HMS model was calibrated to the largest historic flood events on record, as shown below. 

In many cases, an existing model will not have been calibrated to the largest events on record, and adequate time for additional model calibration and validation is not available during the Periodic Assessment.  However, guidance is provided for applying the existing precipitation-runoff model to simulate the PMF and includes information about setting infiltration/loss parameters, unit hydrograph parameters, and channel routing parameters.  Conservative assumptions are appropriate during the Periodic Assessment given the high level of uncertainty regarding many of the components in the hydrologic analysis.  If a dam was found to be hydrologically deficient during the Periodic Assessment, a more detailed study would follow where conservative assumptions are no longer necessary or required. 

Three large runoff events were used to calibrate the model on occurring in June 1972 (Tropical Storm Agnes), another in January 1996 (rain-on-snow), and September 2004 (Tropical Storm Ivan).  These events are among the largest to occur in the Bald Eagle Creek watershed.  The most extreme flood event that has transpired since the construction of Sayers Dam occurred during the passage of Tropical Storm Agnes in June 1972.  In mid-June, a multi-day period of extremely heavy, long duration rainfall led to record peak discharge rates and reservoir stages (since the implementation of wide-scale streamflow gaging stations).  The Tropical Storm Agnes event still remains the largest event of record in terms of peak discharge rates and reservoir stages throughout the majority of the Susquehanna River watershed.  Within the West Branch Susquehanna watershed, three-day precipitation accumulations exceeded 10 inches in various locations with the majority of the watershed exceeding seven inches in that same time period. 

When reviewing the historic storms used to calibrate and validate the model, note the quality of the model calibration and the magnitude of the storm events.  Documenting that the storm events used to calibrate and validate the model were among the largest on record, or were relatively average flood events, will help with decisions about whether the model should be calibrated and validated to additional storms in any future analysis.

Overview

This workshop illustrates one method for developing the Inflow Design Flood (IDF) given an existing HEC-HMS model.  The workshop focuses on using the existing HEC-HMS model and making modifications appropriate for the Probable Maximum Flood (PMF) simulation. The condition of the existing HEC-HMS model might not always be clear.  It is important to review the model and understand any assumptions that might have been made when the model was developed (review over the hydrologic report).  It is also important to look over the calibration results and see whether the computed simulation does a good job at replicating historic floods and assess whether additional adjustments or calibrations need to occur.   In this workshop, you will summarize the model parameters in the existing HEC-HMS model and choose appropriate model parameters for the PMF simulation.  Once the parameters are selected, you will be "peaking" the unit hydrograph response at the dam and configure the model for a PMF simulation.  Additional information on IDF for Dams and Reservoirs can be read in Engineering Regulation 1110-8-2(FR).

Review the Model

1. Launch HEC-HMS and open the project by selecting File | Open | Browse.  Navigate over to the project and select the HEC-HMS project named PMF_Workshop_B.hms. Select the Basin Models folder to expand the Watershed Explorer and see all basin models in this project.  You should notice a separate basin model for each calibration event (1972, 1996, and 2004).  The fourth Basin Model, named PMF_UH_peak25, was created as a place holder for the final set of model parameters to be used in the PMF simulation.
2. Run all three of the historic simulations, 1972, 1996, and 2004.  One quick way to run multiple simulation is to click on Compute | Multiple Compute Click on Select All then Compute.
   
3. Compare model output at the Computation Points.  You can access the model output by expanding the simulation runs under the Results tab. All computation points are highlighted with a red box around the element icon. 
   
   Figure below shows the computed and observed flood hydrographs at the Bald Eagle at Milesburg junction for the 1972 event. 
   
   You should notice the model does a good job reproducing the observed flow for these three events.  The Nash-Sutcliffe metric shows good model performance as well by accessing the element summary table. 
   

4. In a spreadsheet, enter the loss and transform parameters from the three calibration events like the table below.
   

   	Constant Loss Rate (in/hr)
   Basin	1972	1996	2004	PMF
   Spring Creek
   Bald Eagle HW
   Reservoir Local

   According to ER 1110-8-2, loss parameters used when simulating the PMF event should “be derived to correspond to patterns favorable for rapid concentrations of runoff from the drainage basin”.  Typically, the lowest loss rates used during model calibration would be used for the PMF simulation, unless loss rates were unreasonable due to poor quality precipitation or flow observations.  ER 1110-8-2 also states that loss parameters need to be adjusted to account for wet conditions in the watershed at the beginning of the PMF simulation (assumption of a rainfall-runoff event occurring before PMF event). 
   
   Question 1: Based on the description above, what do you think the PMF constant loss rate parameters should be?
   

   Click here to expand...

   	Constant Loss Rate (in/hr)
   Basin	1972	1996	2004	PMF
   Spring Creek	0.12	0.2	0.09	0.09
   Bald Eagle HW	0.08	0.2	0.09	0.08
   Reservoir Local	0.15	0.2	0.1	0.1

   The lowest loss rates were selected for the PMF model.  There were no major concerns with the calibration results and model boundary conditions for all calibration years.

5. The initial loss should be set to a reasonable value that reflects an antecedent storm, occurring 3-5 days prior to the PMF (evapotranspiration will remove some surface and soil water, and water in the soil will percolate so that field capacity is reached).  The time of the year should be considered as well since the moisture deficit grows more quickly in warmer months.  For this workshop, the Initial Loss (IN) = 0.5 inches will be used because the PMF will likely occur between spring and fall months

6. Later steps will focus on adjustments to the unit hydrograph and reach routing parameters.  For now, enter in the table the Time of Concentration (Tc) and Storage Coefficient (R) parameters and compute the average for all three calibration events. The Tc and R for each subbasin can be seen by accessing the Global Editor Parameters | Transform | Clark Unit Hydrograph.
   

   	TC (hr)				R (hr)
   Basin	1972	1996	2004	Average	1972	1996	2004	Average
   Spring Creek
   Bald Eagle HW
   Reservoir Local

7. The routing reach has the same Muskingum K and X parameter values for all three calibration events.  For baseflow, it is important to verify that baseflow parameters do not result in unrealistic runoff volumes for the PMF simulation.  For most studies, the Ratio-to-Peak baseflow parameter used when calibrating the model to historic floods will not be appropriate for the PMF simulation.  The Ratio-to-Peak parameter controls when the baseflow recession curve begins on the falling limb of the runoff hydrograph.  The ratio-to-peak parameter should be set to a value less than 0.1 for the PMF simulation.  It is important to compare precipitation and runoff volumes for the PMF simulation.  The runoff volume should be less than total precipitation. For this workshop, the Ratio-to-Peak parameter was set to 0.05.

Hydrograph Peaking

According to ER 1110-8-2, the “reservoir inflow unit hydrograph should be peaked by 25-50 percent” to account for the fact that the model was calibrated to smaller floods. The Clark unit hydrograph method was used to transform excess precipitation to runoff hydrographs while the Muskingum channel routing routine was used to route flows from the Bald Eagle at Milesburg junction to the Sayers Dam reservoir element.  HEC-HMS can be used to “peak” the unit response at the dam by configuring a Meteorologic Model containing 1-inch of precipitation and modifying a Basin Model so that there are no losses or baseflow, only direct runoff. 

1. Create a copy of the PMF_UH_peak25 Basin Model and name the copy Peak_25P.
2. Open the Peak_25P Basin Model and set each subbasin Loss Method and Baseflow Method to None (only direct runoff will be computed).
3. Enter the Average TC and R Clark unit hydrograph parameters from table above for the Spring Creek, Bald Eagle HW, and Reservoir Local subbasins. 
4. Enter a Muskingum K of 3 hours and a Muskingum X of 0.25 for the Bald Eagle 40 routing reach.  Set the number of subreaches to 12 (an approximation for subreaches =$//<![CDATA[ \begin{array}{l}\frac{K}{\Delta t}\end{array} //]]>$).
5. Add a new Precipitation gage to the project named 1-inch 1 hour, the Units should be set to Incremental Inches and the Time Interval set to 1 Hour. Set the Start Date and Start Time to 01Jan1999 00:00 and the end date and time to 02Jan1999 01:00.  Go to the Table tab and enter 1 inch of precipitation on 01Jan 1999, 01:00.
6. Create a Meteorologic Model named 1-inch.  Make sure the Replace Missing Option is set to Set to Default and select Specified Hyetograph as the Precipitation method. Click on the Basins tab and link the Meteorologic Model to the Peak_25P Basin Model.  Finally, expand the 1-inch Meteorologic Model and select Specified Hyetograph.  Associate the 1-inch 1 hour precipitation gage to each subbasin on the Specified Hyetograph tab.
   
   
7. Create a simulation run named Peak Unit Response 25P using the Peak_25P Basin Model, 1-inch Meteorologic Model, and PMF Event Control Specifications.
8. Run the Peak Unit Response 25P simulation and view results for the Sayers Dam reservoir element. To view results, go to the Results tab and expand the PMF Peak 25P simulation. Select the “Sayers Dam” reservoir element and then select the Time-Series Table. Tabulate the flow time-series at Sayers Dam (Inflow time-series) and copy/paste the time-series into the Peaking UH spreadsheet located in the workshop folder.  The inflow time-series should be copied in Column C.  The plot will updated as shown in figure below, the peak flow for the unit hydrograph at Sayers Dam is approximately 12,000 cfs.
   
   The hydrograph computed in this step is the unit hydrograph for the watershed upstream of Sayers Dam.  For an actual analysis, it would be better to have a separate Basin Model and simulation for the base (un-peaked) scenario.
9. Adjust the TC and R parameters for the Spring Creek, Bald Eagle HW, and Reservoir Local subbasins using the same percentage.  This can be quickly performed by opening the Clark Unit Hydrograph Global Editor (Parameters | Transform | Clark Unit Hydrograph) and use the Fill option to Multiply by a Constant (select all of the rows in one column, right click, select Fill).  The Tc parameter controls the hydrograph timing and the R parameter controls the hydrograph shape (but also timing to some degree).  Both parameters are influenced by similar watershed characteristics, slope, flow path, roughness, and storage to name a few.  Use the same adjustment percentage for the Muskingum K parameter in the Bald Eagle 40 routing reach.  Rerun the Peak Unit Response 25P simulation and copy the updated Sayers Dam inflow time-series results into Column D in the spreadsheet.  The plot will update showing the original and “peaked” unit hydrograph.  Cell D2 shows the percentage that the unit hydrograph was peaked.  Continue adjusting Tc, R and Muskingum K parameters until the peak inflow increases by the desired percentage.
   

   Question 2:  What ratio did you use to reduce the TC, R, and Muskingum K parameters to increase the unit hydrograph at Sayers Dam by 25 percent?  Fill in your final TC and R values into the table.

   Click here to expand...

   The ratio used to reduce Tc, R and K was 0.8

   Basin	TC (hr)	R (hr)
   Spring Creek	10.4	10.6
   Bald Eagle HW	8.8	6.4
   Reservoir Local	6.4	6.4

   Basin	TC (hr)	R (hr)
   Spring Creek
   Bald Eagle HW
   Reservoir Local

10. Transfer the unit hydrograph and routing reach parameters to the PMF_UH_Peak25 Basin Model. Also, populate the Loss parameters for the three subbasins upstream of Sayers Dam. Use the PMF Constant Loss parameters in the Table from Question 1 above, and set the initial loss to 0.5 inches (assumption of an antecedent storm).  Enter an Impervious Area parameter for the Reservoir Local subbasin of 7.7%.  For this subbasin, the impervious area parameter is accounting for rain falling on the pool.  An increase in the impervious area is needed for the PMF simulation because the surface area of the pool is almost double when it is at the spillway crest (elevation 657ft - full flood pool) as compared to normal conservation pool levels.  The Impervious Area should be 2.3 % for the Spring Creek subbasin and 0% for the Bald Eagle HW subbasin.

11. Create a simulation named PMF Peak 25P that combines the PMF_UH_peak25 Basin Model, PMP Meteorologic Model, and PMF Event control specifications. To compute the PMF Peak 25P simulation run, select Run: PMF Peak 25P in the simulation toolbar, as shown in figure below.  Then press the Compute All Elements button (raindrop to the right of the toolbar).
    

12. Look at results for Sayers Dam by selecting the “Sayers Dam” reservoir element in the Watershed Explorer or in the Desktop and pressing the summary results button in the simulation toolbar. These buttons are only active if the simulation run has been computed and a Basin Model element is selected. You can also see results by navigating to the Results tab and expanding the PMF Peak 25P simulation. Select the “Sayers Dam” reservoir element and then select the Summary Table.

    
    Question 3:  What was the peak inflow and outflow during the PMF simulation at the “Sayers Dam” reservoir element?

    Click here to expand...

    Inflow: 241,204 cfs, Outflow: 205,174 cfs

    
    
    Question 4:  The top elevation of Sayers Dam is 683 ft.  Was Sayers Dam overtopped during the PMF simulation?  If the dam was not overtopped, how much freeboard is available?
    

    Click here to expand...

    Peak pool elevation is 676.8 ft. The dam was not overtopped. There is 6 feet of freeboard.

    
    Question 5:  The figure below shows the original Spillway Design Flood for Sayers Dam.  The Spillway Design Flood contained 21.8 inches of precipitation in 48 hours (12.6 inches in the maximum 3-hours).  The peak flow is 251,000 cfs and the runoff volume is 19.05 inches.  How is the current PMF simulation different than the original Spillway Design Flood?
    

Download the final model files here - PA PMF Workshop -Existing Model_Finish.zip