Post-wildfire Streamflow Reconstruction Using Artificial Neural Network Model

Abstract:

The increasing number, size, and severity of wildfires in forested water-stressed catchments in the western Unites States may affect the hydrological response of these watersheds, thereby impacting the underlying hydrogeochemical processes, e.g., flow regimes and mineral weathering. These conditions may prevail for extended periods of time. High severity wildfires in conifer forests are thought to be responsible for the majority of long-term landscape denudation (e.g., erosion). Forest fires may result in a dramatic decrease in vegetation coverage, changes of soil properties, and alterations to the hydrologic cycle, i.e., decreased evapotranspiration and infiltration, and increased watershed yields. Furthermore, post-wildfire debris flows, a turbulent flowing mixture of sediment and liquid, may jeopardize gauging stations creating gaps in flow records. These missing intervals in discharge time series pose big challenge for data interpretation and analyses in post-wildfire watershed research. Therefore, reconstruction of this missing data becomes an obvious need in a wide range of hydrologic investigations.

In 2013, the Thompson Ridge wildfire burned three headwater catchments in the Jemez River Basin Critical Zone Observatory (JRB-CZO) within the Valles Caldera in northern New Mexico. When the first summer monsoon storms occurred, a post-wildfire debris-flow damaged the gauging structures of two of the headwater catchments in the JRB-CZO, creating a gap of approximately three years of missing discharge data. In this study, we built an artificial neural network (ANN) model to reconstruct these missing flows and discuss it within the context of previously proposed gap filling techniques. The main advantage of ANN is its ability to extract nonlinear relationships between the inputs and outputs. The input variables utilized in the model are discharge from two nearby stations as well as total precipitation, snow water equivalent, and temperature.