Evaluating the Accuracy of Hourly Precipitation Data in the U.S.

Hourly precipitation data in the contiguous United States is has many reporting inconsistencies. Some of these invalid reports may be due to rain gauge malfunctions; however other misreports, such as stations not reporting precipitation for an extended period of time, or stations that report unrealistic hourly rainfall amounts remain unexplained.An algorithm was developed to determine which World Meteorological Organization (WMO) stations accurately reported hourly precipitation data by comparing these stations to the nearest WMO site. The stations were divided according to their climate zone (dictated by the National Climatic Data Center) and the difference in elevation between the station and its nearest neighbor. Hourly precipitation values were summed annually from 2006-2016 and then compared to the annual sum of the nearest neighbor. Since mixed precipitation events or severe thunderstorms could cause a station to appear erroneous, the algorithm allows for three out of the ten evaluated years to fall outside of the ratio range given the location of the station.

For stations in every climate zone except the Southwest, West, and Northwest and stations with elevation differences of less than 5,000 feet, a ratio of 50-150% between the first-order station and its nearest neighbor was used to determine if a station was valid. Stations were eliminated if they had three or more out of ten years that fell outside of the range. For stations in the Southwest, West, or Northwest, or stations with elevation differences of greater than 5,000 feet, a ratio of 35-170% was used. In addition to eliminating stations with inaccurate data, stations that reported wet or dry when compared to the first-order station for nine or ten of the years were also flagged. Additional analysis will be completed to determine the geographic region that has the most accurate hourly precipitation reports, and whether the nearest neighbor approach is most appropriate for this analysis.

Preliminary results indicated that approximately 25% of WMO stations would be eliminated using the thresholds previously discussed. If the ratio ranges were compressed to 75-125% for all climate zones, excluding the Southwest, Northwest, and West and a range of 60-145% was used for these stations, more than 65% of WMO stations would be eliminated. Additionally, over 35% of stations reported a dry bias during eight of the ten years, while only 4% of stations reported a wet bias for eight of the ten years.