Moving toward climate-informed agricultural decision support – can we use PRISM data for more than just monthly averages?

Decision support for agriculture is enormously varied in target application and complexity, ranging from simple worksheets to near real-time forecast systems requiring significant computational and manpower resources. Until recently, most such decision support systems have been constructed with little consideration of the variability of local climate, much less climate change. This situation is evolving, but there are multiple challenges in either retrofitting existing decision support systems or developing new ones to incorporate climate variability or climate forecasts.

One of the long-standing challenges has been the unavailability of continuous, high-quality precipitation data for most of the land surface, even in the contiguous U.S. Such continuous time series are necessary as the climatological basis for the development of probabilistic climate-informed decision support (e.g., probability of exceedance curves). The development of the PRISM Climate Group's monthly precipitation data appears to have filled much of that need and, as advertised on their web site, “PRISM is the USDA's official climatological data.”

However, an initial evaluation of PRISM monthly precipitation time series versus contributing NOAA/NCDC COOP data has raised questions concerning the accuracy of the variance of the PRISM monthly precipitation time series, even in relatively low-relief Oklahoma. A more thorough examination of the PRISM precipitation time series is being conducted in comparison to independent precipitation data collected from a spatially dense network of sites in central Oklahoma over the last several decades, supplemented by data from a collocated Oklahoma Mesonet site and nearby COOP stations. Our primary concern is whether the PRISM data can provide representative statistics beyond the monthly 4-km means, including probability of exceedance curves.

Preliminary results confirm data quality issues with the COOP data (as reported by others), but also suggest that QA/QC procedures used by PRISM on COOP and other input data may be limited in their success. We will outline our analysis approach, and present results to date and preliminary conclusions.