wxgenR: A Collaborative Stochastic Weather Generator Incorporating Seasonality for Climate Impacts Assessment

A stochastic weather generator was developed with the novel feature of considering seasonality of precipitation and temperature, motivated largely within the context of climate change adaptation and planning. The weather generator, wxgenR, is released as an R language package on the Comprehensive R Archive Network. wxgenR development was initiated by a hydroclimate analysis using areal average precipitation and temperature from the Lower Santa Cruz River Basin in Arizona under both historical and projected future conditions. The Lower Santa Cruz River Basin application highlights wxgenR's ability to incorporate seasonality shifts induced by climate change, such as monsoon onset and demise timing, in resulting weather simulations.

wxgenR was further tested using weather station data from nine locations in the continental United States with varied hydroclimatic regimes. In a blind cross validation, the simulations’ ensemble median root mean square error across the stations ranged between 0.03– 0.13-inches and 1.39– 2.24-°F for daily precipitation and mean temperature, respectively. Continuous ranked probability skill score for precipitation spanned 0.25– 0.51 and 0.20– 0.41 for temperature, indicating that all stations showed skillful performance when compared to the 2000– 2022 daily station specific reference climatology.

Upcoming work will use wxgenR to condition a stochastic streamflow generator. Specifically, climate change perturbed simulations of precipitation and temperature from wxgenR will be used to simulate streamflow under varied climate conditions.