The Minnesota State Climatology office maintains records from 1958 through present over a dense statewide network of precipitation observers. The network is used primarily to identify precipitation maxima on a single-event basis. In this study we used the densest portion of the network, the portion in and near Minneapolis-St. Paul, for a high-resolution analysis of annual maximum daily precipitation values. Such an analysis has implications for the accuracy the precipitation design values currently in use, which in turn has implications for the durability of, for example, our transportation infrastructure.

We analyzed data from observers within a grid of 100 km² cells that spans 120 km east-west by 80 km north-south and is centered on Minneapolis and St. Paul. We used software that facilitated human comparison of the apparent maximum observed value within a cell with all stations within that cell. After recording and checking the quality of the maximum values, we computed the mean annual maximum daily precipitation value for each cell. The resulting mean values were gridded (using Kriging) and then plotted to a map.

We originally found an axis of the highest mean maximum daily values over the heart of the study area, which suggests that the central portions of the local metropolitan region are more prone to intense rainfalls than surrounding areas. A regression analysis, however, indicated that the number of observers per cell (up to the ninth observer) strongly influenced the magnitude of the average annual maximum, with an r² of 0.94, suggesting that the “pattern” might owe more to the spatial distribution of observers than to mesoscale meteorological phenomena. Using the regression trend of 3.685 mm for every missing observer, we corrected the average annual maximum value for cells with fewer than nine observers. The new corrected data set completely reconfigures the spatial distribution of the highest maximum values; more importantly it suggests that maximum rainfall indices (such as return-period statistics) derived from coarse data sets are underestimating the true potential for heavy rainfall.