Vegetation damage and forest cover influence streamflow response to Hurricane Maria in Puerto Rico

Jazlynn Hall and Maria Uriarte, Ecology, Evolution, and Environmental Biology, Columbia University, New York, NY, Martha Scholl, National Research Program, U.S. Geological Survey, James B. Shanley and Serena Matt, NH-VT Water Science Center, USGS, Montpelier, VT

Background/Question/Methods
Cyclonic storms are powerful disturbance events that bring large rainfall volumes and cause widespread vegetation damage. In September 2017, Puerto Rico was hit by a severe hurricane, Maria (category 4). Forest damage is expected to reduce initial interception and evapotranspiration (ET) while regrowth will gradually increase water demand; impacts on streamflow regimes are not well understood. Here we use total daily streamflow data from 13 USGS gaged watersheds in Puerto Rico between 2013 and 2018 to examine the magnitude and duration of storm impacts on streamflow regimes. We also ask whether these impacts can be attributed to watershed-level vegetation damage (canopy height loss) and forest cover, after accounting for climatic factors (PET, VPD). Our approach also accounts for spatiotemporal differences in rainfall, the main driver of streamflow patterns. We compared post- to pre-hurricane streamflow responses to NOAA GHCND rainfall. For each calendar month and watershed, we used linear mixed models with varying intercept and slope to compare pre- and post-hurricane streamflow for (a) no rainfall conditions (intercept) and (b) the relationship between streamflow and rainfall (slope). We then compared the magnitude of change in slopes and intercepts and asked whether they could be attributed to forest cover and vegetation damage.
Results/Conclusions
The mixed model accounted for 70% of streamflow variability. In the month following Hurricane Maria (i.e., post-hurricane, Oct 2017), all watersheds experienced elevated streamflow under zero-rainfall conditions (higher intercepts). These effects lasted 3-16 months, reflecting lower canopy interception and reduction in water demands of heavily damaged vegetation, followed by high demand from regrowth. Streamflow-rainfall relationships (regression slopes) decreased post-hurricane in 10 of the 13 watersheds. In Puerto Rico’s steep topography, a high proportion of rainfall rapidly enters streams, so pre-hurricane slopes were high. As post-storm intercepts increased from lower vegetation interception and uptake, regression slopes decreased. Preliminary results show that moderately forested watersheds (50-60% forest, n = 3) with high forest damage experienced greater post-hurricane (Oct 2017) increases in intercepts than less damaged watersheds. Increases in post-hurricane intercepts were lower for highly forested watersheds (greater than 80% forest, n = 5) than moderately forested watersheds, and there was no difference in intercept between damaged and undamaged watersheds. This suggests that forests, however damaged, increase resistance to storm impacts on water resources. Our results add to previous studies showing how interactions between vegetation damage and forest cover influence streamflow regimes and increase understanding of forest damage effects on watershed function.