Testing Soil-Routing versus Canopy-Loss Controls on Post-Fire Streamflow Response in the Humid Southern Appalachians Using DHSVM

Yiyun Song, Duke University Durham North Carolina, United States. yiyun.song@duke.edu

Brad Murray, Duke University Durham North Carolina, United States. abmurray@duke.edu

Ram Oren, Duke University Durham North Carolina, United States. ramoren@duke.edu

Brandon Hays, Duke University Durham North Carolina, United States. brandon.hays@duke.edu

Post-fire streamflow increases are commonly attributed to either: 1) soil effects on runoff generation and routing, such as hydrophobicity and reduced infiltration; or 2) canopy loss, which reduces interception and transpiration. In humid, heavily forested mountain watersheds, however, the relative importance of these controls remains uncertain. Following the 2016 Southern Appalachian wildfires, observations in paired watersheds, with each pair consisting of burned and unburned neighboring watersheds, present the opportunity to test which set of mechanisms plays the most important role in this type of landscape. In one pair, during the first post-fire year streamflow in the burned watershed was about 2.7 times that in its unburned reference, while another pair showed little post-fire divergence.

To test the mechanisms behind this contrast, we analyzed model experiments using the Distributed Hydrology Soil Vegetation Model (DHSVM). Results indicate that burn-severity effects on canopy loss, expressed through leaf area index (LAI) reduction, and the resulting decreases in interception and transpiration, produce the largest and most persistent shifts in modeled flow-duration curves. In contrast, perturbations to soil parameters alone are insufficient to reproduce the observed burned–unburned separation. Analysis of 5-minute outlet hydrographs further weakens the case for an explanation centering on soil effects: the burned watershed that exhibits the strong first-year cumulative streamflow response does not consistently exhibit greater flashiness or event-scale peak response than the corresponding unburned watershed, despite its stronger first-year streamflow response.

In parallel, we are developing a historical MODIS LAI/NDVI-based analysis to constrain the timing and persistence of post-fire canopy deficit and to evaluate whether modeled LAI-loss trajectories are consistent with observed canopy-recovery dynamics. Preliminary results suggest that nonlinear ecological responses to fire-induced canopy loss could explain why the streamflow was approximately equal in burned and unburned members of one pair during the first post-fire year, and in both pairs during the second year. Even though observations show that the LAI was reduced in both burned watersheds in both years, the reduction was more pronounced in one case: the first year in the burned watershed that exhibited anomalously high. Previous work has shown that if LAI loss is moderate, the dense forests in this region do not exhibit reduced transpiration rates, as the remaining leaves transpire at higher rates. Forest transpiration rates within this regime remain at energy-limited ceiling. However, if LAI loss exceeds a threshold value, transpiration rates fall. We hypothesize that in the burned watershed exhibiting anomalously high streamflow, canopy loss effects were severe enough during the first year to surpass the threshold and reduce transpiration, while in other cases the threshold was not exceeded.

2026 CSDMS meeting-086

Testing Soil-Routing versus Canopy-Loss Controls on Post-Fire Streamflow Response in the Humid Southern Appalachians Using DHSVM