Wildfire risk to water supply catchments: A Monte Carlo simulation model

Abstract

Wildfires create burnt areas which may be vulnerable to increased erosion when affected by high intensity or long duration storm rainfall. This vulnerability is most apparent in the first year after fire and reduces as soil and vegetation recover. Post-fire erosion events may deliver large quantities of sediment and associated contaminants to streams and reservoirs, potentially resulting in water that is undeliverable to cities and towns. A suitable model to determine the magnitude and likelihood of large sediment loads initiated by the random combination of fire and storm rainfall in southeastern Australian forests does not currently exist. A new model is being developed that determines the risk of these high-magnitude sediment loads based on the combined probability of fire and storm rainfall events. The model combines deterministic models for fire spread, erosion and sediment delivery processes within an annual Monte Carlo simulation of fire ignitions and storm rainfall. By simulating fire behaviour and analysing historical fire and weather records, it is possible to estimate the probability, extent and severity of a large range of possible wildfires. This range will be combined with a range of possible storm rainfalls to give the risk of large sediment producing conditions. The Monte Carlo model is initially to be implemented in Melbourne's two largest water supply catchments, the Upper Yarra and the Thompson. These forested catchments provide Melbourne with approximately 80% of its water supply. Current work on this model is focused on the fire modelling and this paper will deal exclusively with this part of the system. While the fire modelling method is complete, the fire model is currently in the process of calibration and validation. Fire behaviour is simulated using the PHOENIX fire behaviour model developed at the University of Melbourne. PHOENIX is presently used by Victoria's Department of Sustainability and Environment to simulate fire behaviour and model fire management scenarios. Simulated fire spread is determined by a weather time series which may also be used to calculate an instantaneous Forest Fire Danger Index (FFDI). A revised FFDI was used by PHOENIX to incorporate changes in fuel moisture in response to the variation of solar radiation over the course of the day. Wildfires were simulated using historical fire weather for the 344 worst fire days on record. These 344 days compose a partial duration series of severe fire weather days containing all daily peak FFDI days greater than the lowest annual maximum FFDI on record. These fires were individually ignited on a 6km grid covering a large area of central Victoria. These simulations provided a dataset of over 80,000 possible fires to inform the Monte Carlo model. The described approach to modelling fire behaviour will deliver the dataset required by the overall Monte Carlo model while working within the limitations of PHOENIX and the available historical weather and fire records.