Parameter Estimation for Multiple Post-Wildfire Hydrologic Models in Southern CA, USA

Abstract:

This study focused on parameterization of post-wildfire hydrologic models needed to drive short lead-time predictions of flooding and debris flows. The need for improved prediction capability in southern California (CA) USA was highlighted by the catastrophic debris flows in January 2018 produced from the Thomas Fire burn scar. Measurements of post-wildfire soil-hydraulic properties for model parameterization are seldom available, so we investigated these parameters for southern CA. Soil-hydraulic properties, soil-water retention, and selected soil physical properties were quantified within the perimeter of the Thomas Fire.

Significant (p<0.05) decreases in field-saturated hydraulic conductivity (K_fs) and sorptivity (S) in burned areas relative to unburned areas were shown for tension infiltrometer measurements. Significant decreases in saturated soil-water content (θ_S) dominated wildfire effects on soil water-retention. By contrast, the van Genuchten parameters inverse air-entry pressure (α), pore-size distribution index (N), and residual water content did not show significant wildfire effects. Wildfire impacts on hydraulic and physical soil properties were greatest in the top 1 cm of soil, which suggests that post-fire soil property measurements should focus on the near surface. Post-fire increases in dry bulk density point to decreases in soil structure as a potential cause of reductions in K_fs and θ_s and associated changes in infiltration and soil-water retention. Soil-water repellency was most likely responsible for decreases in S. Comparison of K_fs, S, and the Green-Ampt wetting front potential (ψ_f) with measurements at other burned sites in southern CA suggests substantial similarities between sites. These similarities indicate the potential for regionalizing model parameters, thus enabling rapid post-fire assessments of flash flood and debris flow hazards using physically-based hydrologic models. Ratios of burned to unburned K_fs (0.37), S (0.36), and ψ_f (0.66) are suggested as one procedure for scaling unburned values for model parameterization. An alternative procedure could be to use typical burned values of K_fs = 20 mm hr^-1, S = 6 mm hr^-0.5, and ψ_f = 1.6 mm in southern CA. These parameterization approaches are most appropriate at the point to small plot scale (<1 m²).