Analyze Results and Identify Trends from the Historical Period (1971 - 2005) and Future Period (2006 - 2099)

This tutorial was not designed to provide guidance for applying downscaled climate model projections to inform an analysis. Each Federal, State, and Local agency should follow agency specific guidance for including possible future climate change information in hydrologic analysis. This tutorial highlights tools in HEC-HMS that aid modelers in making use of gridded meteorologic datasets and utilizing the new Ensemble compute option to organize many simulations. 

The information in this section was created by processing results in the ensemble compute DSS files. The information shown represent what you could do with the results from HEC-HMS model simulations. At this point, it took approximately 7 hours of effort to gather data, configure the HEC-HMS project, and develop a calibrated model to apply against the CMIP5 Climate Projection datasets. Running the 3 ensemble analyses took approximately an hour and a half. Processing, analyzing, and displaying the simulation results took additional effort; however, time savings could be made with standard templates for analyzing and displaying results. 

There are more results included in the output files than the ones included below. The following table contains the HEC-HMS results and time steps used for comparisons. None of the results from the HEC-HMS model were bias correct when running the future period simulations. For the most part, results from the MF_TuleR_S20 subbasin were used for comparisons. The MF_TuleR_S20 subbasin is a high elevation subbasin and represents the runoff response for a majority of the watershed. The TuleR_S10 subbasin is included in the Precipitation Frequency comparison. You could easily process these results and compute total watershed average precipitation, temperature, and SWE information instead of evaluating results from individual subbasins. 

HEC-DSSVue was used to convert output results to the appropriate time step and extract monthly average time-series and annual maximum values. The following figure shows a portion of the CMIP5_historical DSS file. Notice the F-Part pathname indicates the simulation run used for each ensemble member. You can see for the ReservoirInflow element, the 3-hour results were converted to 1-day average flow and then the peak value for each year was extracted.  

Before jumping into the results, below are a few observations from evaluating results.

Historical Period (1971 – 2005)

1. There is a bias in the average annual precipitation between the CMIP5 Climate Projection datasets and the Livneh dataset.
2. There is a large bias in extreme 3-day annual maximum precipitation between the CMIP5 Climate Projection datasets and the Livneh dataset for the MF_TuleR_S20 subbasin.
3. The bias in precipitation translated to a bias in average monthly/annual flow (at the ReservoirInflow element which is the outlet of the 390 square mile watershed).

Future Periods (2006 – 2036, 2037 – 2067, and 2068 – 2099) compared to results from the historical period (1970 – 2005)

1. The trend in monthly/annual precipitation shows slightly lower values in future periods for the RCP 4.5 scenario and slightly larger values for the RCP 8.5 scenario (MF_TuleR_S20 subbasin).
2. For the MF_TuleR_S20 subbasin, the trend in SWE shows a decrease in March’s average SWE from 12.1 inches (1970 – 2005) to 4.4 inches for the RCP 4.5 scenario and 1.4 inches for the RCP 8.5 scenario (both values are averages from the 2068 – 2099 period). The average March temperature for the 1970 – 2005 period was 40.4 Degree F, and for the 2068 – 2099 period the average March temperature was 44.2 Degree F for the RCP 4.5 scenario and 47.2 Degree F for the RCP 8.5 scenario.
3. The trend in monthly/annual flow at the ReservoirInflow element shows an annual decrease in runoff for the RCP 4.5 scenario but higher winter flows. The RCP 8.5 scenario show an increase in annual average flow and winter flows due to much less snow accumulation; precipitation falls mostly as rain. Both RCP scenarios show a shift in the runoff pattern, this pattern shift is most likely due to less snow accumulation and faster melt.

Historical Period Caparisons (1971 – 2005)

Precipitation

The following table and figure contain the monthly and annual average accumulated precipitation for the MF_TuleR_S20 subbasin. Generally, the CMIP5 Climate Projection datasets contains more annual precipitation than the Livneh dataset. The average of the CMIP5 Climate Projection datasets shows more than one inch of precipitation than the Livneh dataset during the months of December, January, and April.

The following figures show the annual maximum 3-day precipitation frequency plots for the MF_TuleR_S20 and TuleR_S10 subbasins. The data was plotted using the Weibull plotting position formula (35 years of data). Notice the CMIP5 Climate Projection datasets generated much higher annual maximum 3-day precipitation than the Livneh dataset. There is much better agreement between the CMIP5 and Livneh precipitation in the TuleR_S10 subbasin. 

Snow Water Equivalent

The following table and figure contain the monthly and annual average SWE for the MF_TuleR_S20 subbasin. Some of the CMIP5 Climate Projection datasets generate more SWE than the Livneh dataset. This is expected since the monthly average and peak 3-day precipitation showed the CMIP5 Climate Projection datasets had a precipitation biased higher when compared to the Livneh dataset. 

Reservoir Inflow

The following table and figure show monthly average flow at the ReservoirInflow element (into the reservoir). As expected, simulations using the CMIP5 Climate Projection datasets have a larger average flow than the model using the Livneh dataset. 

Future Period Caparisons (2006 – 2099)

The future period comparisons were created by averaging results across all ensemble members. In addition, the future period was broken up into three time windows, 2006-2036, 2037-2067, and 2068-2099. Results from these time windows are compared against CMIP5 Climate Projection results from the historical period, 1971-2005. 

Precipitation

The following tables and figures contain monthly/annual precipitation at the MF_TuleR_S20 subbasin, for both RCP scenarios. The trend in monthly/annual precipitation shows slightly lower values in future periods for the RCP 4.5 scenario and slightly larger values for the RCP 8.5 scenario (for the MF_TuleR_S20 subbasin). Most of the increase in precipitation for the RCP 8.5 scenario occurred in January and February. 

Snow Water Equivalent

The following tables and figures contain monthly/annual SWE for the MF_TuleR_S20 subbasin, for both RCP scenarios. The trend in SWE shows a decrease in March’s average SWE from 12.1 inches (1970 – 2005) to 4.4 inches for the RCP 4.5 scenario and 1.4 inches for the RCP 8.5 scenario (both values are averages from the 2068 – 2099 period). If interested in how the SWE varies spatially within subbasins, you could turn on the Spatial Results options for the simulations and plot the SWE, and other results, within the basin model. The program will compute and display information for each grid cell in the basin model.

There is published information that temperature index type snow models overpredict melt for future periods as shown in this paper; https://westernsnowconference.org/sites/westernsnowconference.org/PDFs/2014Raleigh.pdf. You could reduce the meltrate function (relationship between degree days above freezing and the meltrate) for the future period simulations if there was information about the amount of overprediction generated when only considering temperature and not the full energy budget. It is likely most precipitation in this example watershed fell as liquid precipitation in later years for the RCP 8.5 scenario. The average March temperature for the 1970 – 2005 period was 40.4 Degree F, and for the 2068 – 2099 period the average March temperature was 44.2 Degree F for the RCP 4.5 scenario and 47.2 Degree F for the RCP 8.5 scenario.

Reservoir Inflow

The following tables and figures contain monthly/annual flow for the ReservoirInflow element, for both RCP scenarios. The trend in monthly/annual flow at the ReservoirInflow element shows an annual decrease in runoff for the RCP 4.5 scenario but higher winter flows. The RCP 8.5 scenario show an increase in annual average flow and winter flows due to much less snow accumulation; precipitation falls mostly as rain. There was an increase in future precipitation in the RCP 8.5 scenario, especially in January and February. Both RCP scenarios show a shift in the runoff pattern, this pattern shift is most likely due to less snow accumulation.