Submitted to: Conference on Probability and Statistics in the Atmospheric Sciences
Publication Type: Proceedings
Publication Acceptance Date: 5/8/2000
Publication Date: 5/8/2000
Citation: N/A Interpretive Summary: Seasonal climate forecasts prepared by NOAA and other research groups are available covering the coming year, predicting shifts in total precipitation and average temperature over three-month periods. Forecast skill is still modest, but sufficient to present an opportunity to change decision-making in agricultural and natural resource management from a reactive mode to a proactive mode. There have been problems with application of these climate forecasts, some related to the unusual form of the forecasts (statements of probability). The more serious impediment is the mismatch between the spatial and temporal scales of the forecasts, and potential applications in agricultural and natural resource management. The seasonal climate forecasts are produced for large areas (about 1/3 the size of Kansas), while decisions need to be made for smaller areas such as watersheds, ranches, farms, and fields. We plan to develop methods to disaggregate the forecasts in a statistically appropriate manner, as well as develop information on the likelihood that the disaggregated forecasts will be realized. This will allow us to associate risks with possible forecast outcomes. Previous studies have shown that precipitation statistics change with size of area. The implication here is that disaggregated climate forecasts might be significantly less skillful, and therefore of limited utility. To address this question, we will develop a predictive relationship between the large area climate forecasts and all possible smaller area (disaggregated) realizations of precipitation.
Technical Abstract: The NOAA Climate Prediction Center is producing a monthly suite of climate forecasts, including three-month outlooks for temperature, precipitation, and degree days for an entire year in advance. These outlooks have potential value in support of risk-based decision making in agriculture and natural resource management. Unfortunately, the forecasts are generated for equal-area regions that represent one or more climate divisions, each about (380 km)2. Given our knowledge of significant climatological variations at smaller scales (sub-divisional), it seems reasonable to disaggregate the outlooks accordingly. As part of this effort, we are developing a methodology to track concurrent changes in the error (uncertainty) of the forecast. This paper will describe our initial methodology.