Fire has an important role in watershed dynamics, and it is unclear how the interaction between fire and hydrological processes will be modified in a changing climate. Detailed landscape models of fire spread and fire effects require comprehensive data, are computationally intensive, and are subject to cumulative error from uncertainties in many parameters. In contrast, statistical models draw attributes such as extent, frequency, and severity a priori from selected distributions that are estimated from current data, implicitly assuming a stationary driving process that may not hold under climate change. We are designing a relatively simple stochastic model of fire spread (WMFire) that will be coupled with the Regional Hydro-Ecological Simulation System (RHESSys), for projecting the effects of climatic change on mountain watersheds. The model is an extension of exogenously constrained dynamic percolation (ECDP), wherein spread is controlled primarily by a spread probability from burning pixels, and which has been shown to have the capacity to identify dominant controls on cross-scale properties of low-severity fire regimes. Each year RHESSys will pass projected pixel-level values of fuel, fuel moistures, wind speed and wind direction to the fire spread model. Spread probabilities will then be calculated from the fuel load, fuel moisture, and orientation of the pixel relative to the slope gradient and wind direction. The stochastic structure of the spread model will subsume the uncertainties in future patterns of fire spread, fuels and climate. WMFire is being calibrated by and evaluated against current known fire regime properties for watersheds in the Pacific Northwest (USA) using Monte Carlo inference.
Available at: http://works.bepress.com/maureen-kennedy/14/