GEONEX Real-time hourly 1-km weather data for the conterminous US for land surface modeling using GOES/ABI data

We developed a system to produce near real-time hourly 1-km weather data (temperature, precipitation, humidity, and wind speed)for the conterminous US. The data are aimed at real-time analysis of land surface processes (i.e. water and energy cycle modeling), which require low latency and high spatial resolution. We extended the methodology developed for NASA Earth Exchange - Gridded Daily Meteorology (NEX-GDM) dataset (1 km historical climate data from 1979-2017 for the conterminous US) to the real-time hourly application. The methodology uses multiple random forest estimators incorporating several datasets to capture spatial variability across continental scales. Each dataset has its strength in a specific spatio-temporal region so that incorporation of multiple datasets is an effective strategy for representing spatial variability at the continental scale. In this study, the spatial data, which is explanatory variable, include satellite data (GOES-16/ABI, GPM/GMI, GPM constellation satellites, and MODIS), weather forecasting model data (GFS, and NAM), NEXRAD radar data, and topography data (elevation, slope, aspect, and distance from coast lines). By adding multiple channels of each sensor, the total number of explanatory variables can be more than 20. We used METAR and SNOTEL for training datasets. The latency of data production and availability is less than an hour. Preliminary analyses of the output data show reasonable spatial-temporal patterns, providing an opportunity for modeling evapotranspiration, soil moisture, fire danger, etc.