Dead Fuel Moisture Climatology Dataset of California

One of the most critical variables controlling the wildfire rate of spread and ignition potential is fuel moisture, expressed as a ratio of the mass of water and the dry fuel mass. Fuel moisture controls flammability and the fire rate of spread, which makes it a critical input into fire spread models and fire danger systems. For that reason, long-term variations of the dead fuel moisture are critical for climatic studies focused on the impact of global warming on fire activity. For the purpose of short-term fuel moisture estimates, the current system running coupled atmosphere-fire forecasts (WRFx), leverages a fuel moisture data assimilation module. The fuel moisture data assimilation system (FMDA) combines surface fuel moisture observations from Remote Automatic Weather Stations (RAWS) with the output of a fuel moisture model to generate an hourly near-real-time dead fuel moisture product for the entire CONUS. The operational fuel moisture data assimilation system consists of a simplified time-lag fuel moisture model that is initialized from the equilibrium moisture content. After that, the observations are spatially interpolated using a Trend Surface Model (TSM). Then, TSM and fuel moisture model results are blended using Extended Kalman Filter (EKF), providing final fuel moisture maps used to initialize fire forecasts and compute fire danger products. Although the FMDA system works well in operational setting, it has significant limitations that make it unsuitable for generating a historical reanalysis of fuel moisture critical for climatic analyses. To overcome these problems the system has been modified to generate a reanalysis fuel moisture product, intended to provide the most accurate representation of the dead fuel moisture in the past. The first system adjustment included replacing Real-Time Mesoscale Analysis (RTMA) input data with the regional climate simulation generated using the WRF model. This has proven to especially improve rain representation which is one of the critical factors in estimating fuel moisture. Then, the number of covariates used to blend the observations was increased from 6 to 25. Finally, the Extended Kalman Filter (EKF) has been tuned to weigh more the importance of the observations compared to the time-lag fuel moisture model. The system was then executed hourly from 2000 to 2021, to generate 22 years of assimilated dead fuel moisture at 3 km resolution for the California region. The final product has been validated using observations as well as the well-established Nelson fuel moisture model run based on the same weather data. The results proves that the presented reanalysis product, leveraging the time lag fuel moisture model and the data assimilation system, provides more accurate representation of spatial and temporal variability of the 10-hour dead fuel moisture for the past 22 years than either the time lag, or the Nelson fuel moisture model.