Characterization of hydrology and sediment transport following drought and wildfire in Cache Creek, California

The worst drought in California in over 1,200 years occurred between 2012-2017 (Griffin, 2014), depleting surface water and groundwater supply and drying out the soils past wilting point. In the summer of 2015, the Jerusalem and Rocky fires burned roughly 40,000 acres within the Cache Creek watershed. To fully characterize the post-fire effects in the Cache Creek watershed, an hourly model of streamflow and sediment transport was developed using the Hydrological Simulation Program – FORTRAN (HSPF). This model requires air temperature, precipitation, and potential evapotranspiration as climate inputs. Hourly station data are sparse in the area and may not capture the variability of elevation and local climatology patterns within the watershed.

A technique used previously to spatially-interpolate daily-climate station data has improved the characterization of local and regional climate patterns on a daily scale in areas with sparse data (Flint et al., 2014). This technique was extended to hourly observed data to produce spatially-varying climate inputs for the Cache Creek hydrologic model to run as a continuous multi-year simulation with hourly time steps. Monthly PRISM grids were used in a two-step scaling method with climate Gradient and Inverse Distance Squared (GIDS) maps (Nalder and Wein, 1998) to develop daily grids, then the daily grids were used to scale hourly climate GIDS maps. This method captures the temporal variability at each climate station yet preserves the regional monthly spatial structure of the PRISM data.

Hydrologic calibration used data from water year 2015, and validation used the same parameters for water year 2016. The model was run through water year 2017 to characterize the effects of wildfire on hydrology and sediment transport. For final simulations, the model was run at an hourly time step from June 2014 through September 2017 to ensure a model initiation period of 4 months prior to the target simulation period used for analysis. Sediment parameters were initially set using the existing Sacramento River Basin model for this sub-watershed area and then iteratively adjusted in the calibration process. To simulate a fire across the landscape, sediment parameters for water years 2016-17 were further modified for burned sub-basins to represent post-fire vegetation and soils in 2016, then partial recovery in 2017.

Results were inconclusive for drought and wildfire effects on runoff. Modeled peak flows generally underpredicted observed peak flows; however, the modeled storm volumes were only slightly under or over the observed storm volumes. Sediment transport was sensitive to the watershed disturbances and R^2 values for daily mean suspended concentrations (SSC) and sediment discharge were 0.70 and 0.75, respectively. Simulated hourly values correlated less strongly with observed instantaneous SSC and sediment discharge (R^2 values of 0.56 and 0.46, respectively).