Evaluation of Seasonal Forecast Skill of Agriculturally Relevant Variables in an Agricultural Basin
Abstract
Seasonal forecasts (long-term forecasts with a lead time of one to seven months), have the potential to inform various aspects of agricultural decision making. Examples include farmer decisions (what to plant, when to plant, how to respond to projected water shortages), water market decisions (whether to use or lease water), and agency decisions (approving drought contingency funds). These decisions are invariably made under uncertainty. Forecasts of information relevant for decision making can reduce the risk/uncertainty associated with these decisions as long as they have sufficient skill and lead times. The context for decisions as well as forecast skill can have significant spatial and temporal variability; therefore, the usefulness of forecasts for decision making can vary across geographic and temporal contexts. Using the Yakima River basin in the U.S. Pacific Northwest as a test case, the objective of this work is to quantify the forecast skill of downscaled North American Multi-Model Ensemble (NMME) seasonal forecasts of agricultural-decision making relevant variables in a spatially and temporally explicit manner. We perform a hindcasting exercise by comparing downscaled NMME hindcasts over the period 1982-2019 with the corresponding gridded observations. The variables considered include forecasts of temperature, precipitation, irrigation demands, and water shortages. Irrigation demand and water shortages are estimated by driving a coupled-crop hydrology model and water management model with forecasts as input. Forecast skill is evaluated for multiple lead times with a range of metrics including bias, accuracy, reliability, resolution, sharpness, uncertainty that address diverse aspects of forecast quality and skill. Preliminary analysis suggests spatial-, temporal-, and variable-specific differences in forecast skill. There is also indication that a relative higher skill for precipitation and temperature in spatial and temporal extents that contribute most towards watershed level hydrologic and crop growth processes. The results of this work can inform assessment of the value of forecasts for diverse decision-making contexts and strategizing the design and implementation of operations forecast and decision support tools.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2020
- Bibcode:
- 2020AGUFMH170.0007K
- Keywords:
-
- 1833 Hydroclimatology;
- HYDROLOGY;
- 1880 Water management;
- HYDROLOGY;
- 1922 Forecasting;
- INFORMATICS;
- 6309 Decision making under uncertainty;
- POLICY SCIENCES & PUBLIC ISSUES