As part of the “Health of the Networks” Project at the National Climatic Data Center, we have designed a prototype system to ensure homogeneous daily precipitation observations in the Cooperative Observer Network. The fundamental mechanism of this system is a comparison between target sites and a set of nearest neighbors based on several established verification algorithms (frequency bias, magnitude bias, and equitable threat score). Although these algorithms are ordinarily used to compare simulated precipitation fields with observations, their designs have significant advantages over other measures of correlation. In particular, they mitigate the negative effects of the binary nature of precipitation observation and their relative rarity, especially in arid regions.

In a previous presentation, we described the performance of these indicators when applied to rainy-season (spring) observations in Iowa. Left unspecified were differences in results in other seasons and localities that could be expected to arise due to wide variations in rainfall frequency. We show results here that address this issue by extending the analyses in several ways. Analyses are shown for Oregon and Washington, and for cold-season precipitation in Iowa.Border effects are removed by allowing neighbor station sets to include stations from neighboring states.To estimate the impact of rainfall frequency and spatial consistency on system performance, simulations are performed using a range of possible precipitation regimes. The effects of rainfall frequency and spatial agreement between neighbors on the verification algorithms is discussed. We also demonstrate differences that occur when longer accumulation periods (3 and 7 days) are analyzed.