A framework for modelling suspended sediment flux following wildfire in forested water supply catchments, south-eastern Australia

Wildfire represents a potentially important threat to the security of water storage reservoirs situated in fire-prone forest catchments. From a water treatment perspective, managers are interested in exceedence probabilities for event suspended sediment loads rather than estimates of average annual loads. To address this, we aim to develop a stochastic model of wildfire and storm event probabilities that are linked to erosion response and suspended sediment loads entering reservoirs. In this paper we present a conceptual catchment-scale model framework that will form the basis for the development of this model. The study focuses on the two main forested water supply catchments that provide potable water to Melbourne, Victoria. Determining probabilities of wildfire occurrence for the study catchments presents a significant challenge, given the limited historic data available. An approach to modelling wildfire involves the use of expert opinion to quantify probabilities of either wildfire entering the catchments from different directions, internal ignition at one or more points, or no fire occurring on days with high fire risk weather. With each annual simulation a wildfire may occur based on random sampling of the wildfire occurrence probabilities. The fire front direction (or internal ignition point), combined with random sampling of fire weather data, form inputs to a deterministic fire behaviour model, which is used to generate burn area and severity spatial data for a range of scenarios. The post-fire storm event component is based on distributions for the maximum annual storm and flow event (for two rainfall event types) that may occur during the recovery period (three years). The two event types are localised, high intensity, short duration summer storms and widespread, high magnitude rainfall events that may occur over 1-3 day periods during winter and/or spring. Erosion response to the contrasting rainfall and associated flow characteristics of the event types is represented using various sub-models for key post-fire erosion processes, namely, hillslope surface erosion and delivery to streams, debris flow (runoff and mass failure-generated), and flood-driven channel erosion. Hillslopes and small tributary sub-catchments within the study catchments are divided into spatially-defined erosion response units based on modelled fire severities and forest-soil categories. Erosion process sub-models are run for response units depending on unit properties, fire severity, and rainfall event type. Sediment outputs from these response units are delivered to adjacent channels, with the extent of delivery to reservoirs dependent on in-channel/floodplain storage losses. Erosion associated with post-fire channel change along the main trunk streams is also modelled and included in the estimated total suspended load entering the reservoirs. The effect of post-fire recovery is captured through functions that represent temporal changes in response units (e.g. vegetation regrowth). Data on post-fire hillslope erosion, debris flow scour, and catchment sediment yields (1-100 km(2)) has been collected from areas similar to the water supply catchments for sub-model development and validation. Linking all model components within the catchment-scale framework will enable production of outputs in the form of distributions of the maximum annual event suspended load (for the two event types) in tributary streams entering the reservoirs. This will provide managers with the capacity to evaluate the risk to water supplies associated with post-fire suspended sediment loads and to develop appropriate strategies to mitigate the threat posed by wildfire and storm events.