Regionally optimized fire parameterizations using feed-forward neural networks

Abstract The fire weather index (FWI) is a widely used metric for fire danger based on meteorological observations. However, even though it simulates the nonlinear relationship between the meteorological observations and fire intensity, the ability of the FWI to accurately represent global satellite-derived fire intensity observations is limited due to its empirical formulation based on a specific region. In this study, we propose a regionally-fitted fire parameterization method using feed-forward neural networks (FFNNs) to understand the nonlinear relationship between the meteorological variables and the fire intensity, which eventually improves the parameterization accuracy. These FFNNs for each grid point utilize daily-averaged meteorological variables (2 m relative humidity (RH2m), precipitation (PRCP), 2 m temperature, and wind speed) as inputs to estimate the satellite-derived fire radiative power (FRP) values. Applying the proposed FFNNs for fire parameterization during the 2001–2020 period revealed a marked enhancement in cross-validated skill compared to fire intensity estimation based on the FWI. This improvement was particularly notable across East Asia, Russia, the eastern US, southern South America, and central Africa. The sensitivity experiments demonstrated that the RH2m is the most critical variable in estimating the FRP. Conversely, the FWI-based estimations were primarily influenced by PRCP. The FFNNs accurately captured the observed nonlinear RH2m-FRP and PRCP-FRP relationship compared to that of the FWI-based estimations.