RHESSys-WMFire: coupling wildfire to hydrology and vegetation to project the effects of climate change on mountain watersheds.

In mountain watersheds of the western US, fire and other disturbances interact with vegetation dynamics and hydrology to affect ecosystem structure and function. It is necessary to couple projections of hydrological processes in a changing climate with projections of fire disturbances, but to our knowledge none of the watershed-scale ecosystem models that have disturbance modules have explicit hydrological routing. This is because it is difficult to find common currency between variables associated with spatially explicit hydrology and those associated with fire spread and fire severity. We have developed WMFire, a stochastic raster-based model of fire spread that integrates fire with the Regional Hydro-Ecological Simulation System (RHESSys), a spatially distributed model of climate-water-carbon interactions that projects the effects of climatic change on mountain watersheds. We resolve mismatches between the two model systems by aggregating the dynamics of fire spread into probabilities, whereby the probability a fire spreads into a neighboring cell increases with the fuel load and moisture input from the hydrological model, and the probability further increases if the direction of spread is uphill and in line with the prevailing wind direction. In this manner WMFire is expected to replicate patterns of spread and fire regimes, not the propagation of individual fire events. Patterns of fire spread and effects simulated by WMFire are returned to RHESSys, modifying carbon source-sink dynamics and hydrological routing. Model evaluation entails a quantitative procedure to assess the agreement between simulated fire-regime properties and observations. Simulations to date have focused on historical projections to evaluate how well the coupled model replicates fire regimes of the past half-century on two watersheds: one in the HJ Andrews experimental forest in Oregon, USA, and another the watershed supplying water to Santa Fe, NM, USA. In extensive model evaluation we match the level of abstraction of fire spread to that of the hydrological model, thereby identifying and reducing sources of uncertainty both in the fire spread model and in the coupling of the fire spread model with the hydrological model.