Step 5: Change Linear Reservoir Baseflow Method from 2-layer to 3-layer for the delayed groundwater release

This tutorial walks you through configuring and calibrating an HEC-HMS model for long-term post-wildfire hydrology simulations.

Software Version

HEC-HMS version 4.13 was used to develop this tutorial.

Change the Linear Reservoir Baseflow Method from 2-Layer to 3-Layer

In Step 4, we determined that representing slow, deep groundwater discharge was essential to improving low‐range flows in the nine-year post-wildfire hydrology simulation. Consequently, in Step 5 we upgraded the Linear Reservoir baseflow method from two layers to three.

We then implemented and recalibrated the three-layer approach in HEC-HMS for the Arroyo Seco watershed—first fitting the January 18–19, 2010 event and then extending the continuous simulation through February 14, 2019.

In Step 5, we will

• review the calibrated input parameters and the corresponding results of the 2-layer and 3-layer Linear Reservoir baseflow method simulations,

• compare the two event model calibration runs between 2-layer and 3-layer Linear Reservoir baseflow method,

• quantify the improvement gained by the 3-layer baseflow approach, and

• identify opportunities to further refine the model for long-term post-wildfire hydrology.

Compare two calibrated results between 2-Layer and 3-Layer Linear Reservoir Baseflow method for the Long-Term Post-Wildfire hydrology Model

Select the Basin Models folder to expand the Watershed Explorer.
Select LGA_CUH_LR_Dys_R_C and LGA_CUH_LR3_Dys_R_C Basin Models and review all its methods and input parameters.
Run the LGA_CUH_LR_R_C_Event and LGA_CUH_LR3_R_C_Event simulations, calibrated to the January 18–19, 2010 event for 2-layer and 3-layer Linear Reservoir baseflow method.
Examine the basin model parameters and simulation results, then answer the questions below.

Question 1:
What are the differences between the two-layer and three-layer linear reservoir baseflow methods?
As shown in the two tables below, we added a third baseflow layer in the three-layer configuration—assigning a 10,000-hour GW3 coefficient to represent slow, deep groundwater discharge. We then calibrated the system by adjusting the initial flows and fractional contributions for Layers 1, 2, and 3.
New Linear Reservoir Baseflow input parameters with 3-layer.
Previous Linear Reservoir Baseflow input parameters with 2-layer.
Question 2:
What are the results of the calibration?
- Computed peak discharge (value, date, and time)
- Total runoff volume
- Nash–Sutcliffe Efficiency
Which model—LGA_CUH_LR3_DyS_R_C_Event or LGA_CUH_LR_DyS_R_C_Event—shows better calibration according to these metrics, and is it robust enough to run continuously for nine years?
compared to USGS Arroyo Seco gage observations, both the 3-layer and 2-layer models exhibit a strong match.
- Computed peak discharge for LGA_CUH_LR3_DyS_R_C_Event: 3,757 cfs on 18 Jan 2010 at 22:45
- Computed peak discharge for LGA_CUH_LR_DyS_R_C_Event: 3759 cfs on 18 Jan 2010 at 22:45
  Observed peak discharge: 4,230 cfs on 18 Jan 2010 at 22:30
- Total runoff volume: 0.97 in (LGA_CUH_LR3_DyS_R_C_Event) vs. 0.99 in (LGA_CUH_LR_DyS_R_C_Event) vs. 1.04 in (observed)
- Nash–Sutcliffe efficiency: 0.819 (LGA_CUH_LR3_DyS_R_C_Event) vs. 0.819 (LGA_CUH_LR_DyS_R_C_Event)
Both models achieve comparably high-quality event-based calibration. The next step is to compare their performance over a nine-year continuous simulation.
Calibration results with the 3-layer baseflow approach (LGA_CUH_LR3_DyS_R_C_Event)

Calibration results with the 2-layer baseflow approach (LGA_CUH_LR_Dyd_R_C_Event)
Run the LGA_CUH_LR3_DyS_R_C simulation (17Jan2020 - 14Feb1019) using the 3-layer baseflow method.
Examine the simulation results, then answer the questions below.

Question 3:
What are the results for 9-year simulation?
- Computed peak discharge (value, date, and time)
- Total runoff volume
- Nash–Sutcliffe Efficiency
Do these metrics show that the 3-layer baseflow model (LGA_CUH_LR3_DyS_R_C) outperforms the version with 2-layer baseflow model (LGA_CUH_LR_DyS_R_C), and is it robust enough for a nine-year continuous simulation?
Both 3-layer and 2-layer baseflow models were evaluated against USGS Arroyo Seco gage data.
Computed peak discharge for with 3-layer baseflow: 3757 cfs on 18 Jan 2010 at 22:45
Computed peak discharge for with 2-layer baseflow: 3759 cfs on 18 Jan 2010 at 22:45
Observed peak discharge: 4,620 cfs on 06 Feb 2010 at 13:45
Total runoff volume: 49.13 in (with 3-layer baseflow) vs. 45.97 in (with 2-layer) vs. 42.45 in (observed)
Nash–Sutcliffe efficiency: 0.566 (with 3-layer baseflow) vs. 0.565 (with 2-layer baseflow)
These metrics show that the both 3-layer and 2-layer baseflow models achieve a almost same Nash–Sutcliffe efficiency (0.566 vs. 0.565) and the 3-layer baseflow model produces more accurate runoff volumes (49.13 in vs. 45.97 in), while both models reproduce nearly identical peak discharges. However, this comparison alone does not definitively identify which method is best for a nine-year continuous post-fire simulation.
New Results with 3-layer baseflow model (LGA_CUH_LR3_DyS_R_C)
Previous Results with 2-layer baseflow model (LGA_CUH_LR_DyS_R_C)
Question 4:
Are you satisfied with these results for your long-term post-wildfire hydrology estimates?
As shown on the below flow-duration curves, the three‐layer baseflow / Dy canopy / Dy surface model (green) maintains the overall NSF score while tightly matching both high‐ and mid‐range flows (0–10% exceedance) and delivering the best fit for low flows in the 80–95% exceedance band—albeit with a slight overestimation of the very smallest flows. According to the summary results, its performance ranks as: "Satisfactory" for both NSE and RSR and "Very Good" for PBIAS-Volume.
While these results aren’t flawless, they rank among the strongest we’ve seen in our nine-year, 15-minute-timestep continuous simulations. They are sufficient to inform long-term post-wildfire evacuation and maintenance planning. However, to guarantee reliable long-term hydrologic estimates, we still need to integrate additional observed data and develop new HEC-HMS features for post-wildfire hydrology and debris-flow modeling. If you have any suggestions or feedback for improving the long-term post-wildfire hydrology simulation, please feel free to contact HEC or me at jay.h.pak@usace.army.mil.