Optimization of Evaporative Demand Models for Seasonal Drought Forecasting

Abstract:

Providing reliable seasonal drought forecasts continues to pose a major challenge for scientists, end-users, and the water resources and agricultural communities. Precipitation (Prcp) forecasts beyond weather time scales are largely unreliable, so exploring new avenues to improve seasonal drought prediction is necessary to move towards applications and decision-making based on seasonal forecasts. A recent study has shown that evaporative demand (E₀) anomaly forecasts from the Climate Forecast System Version 2 (CFSv2) are consistently more skillful than Prcp anomaly forecasts during drought events over CONUS, and E₀ drought forecasts may be particularly useful during the growing season in the farming belts of the central and Midwestern CONUS. For this recent study, we used CFSv2 reforecasts to assess the skill of E₀ and of its individual drivers (temperature, humidity, wind speed, and solar radiation), using the American Society for Civil Engineers Standardized Reference Evapotranspiration (ET₀) Equation. Moderate skill was found in ET₀, temperature, and humidity, with lesser skill in solar radiation, and no skill in wind. Therefore, forecasts of E₀ based on models with no wind or solar radiation inputs may prove to be more skillful than the ASCE ET₀. For this presentation we evaluate CFSv2 E₀ reforecasts (1982-2009) from three different E₀ models: (1) ASCE ET₀; (2) Hargreaves and Samani (ET-HS), which is estimated from maximum and minimum temperature alone; and (3) Valiantzas (ET-V), which is a modified version of the Penman method for use when wind speed data are not available (or of poor quality) and is driven only by temperature, humidity, and solar radiation. The University of Idaho’s gridded meteorological data (METDATA) were used as observations to evaluate CFSv2 and also to determine if ET₀, ET-HS, and ET-V identify similar historical drought periods. We focus specifically on CFSv2 lead times of one, two, and three months, and season one forecasts; which are time scales with moderate skill and are more likely to be used in hydro-climatic applications and decision-making.