Representing extremes in a daily gridded precipitation analysis over the United States: Impacts of station density, resolution, and gridding methods

This study focuses on errors in extreme precipitation in gridded station products incurred during the upscaling of station measurements to a grid, referred to as representativeness errors. Gridded precipitation station analyses are valuable observational data sources with a wide variety of applications, including model validation. The representativeness errors associated with two gridding methods are presented, consistent with either a point or areal average interpretation of model output, and it is shown that they differ significantly (up to 30%). An experiment is conducted to determine the errors associated with station density, through repeated gridding of station data within the United States using subsequently fewer stations. Two distinct error responses to reduced station density are found, which are attributed to differences in the spatial homogeneity of precipitation distributions. The error responses characterize the eastern and westernUnited States, which are respectively more and less homogeneous. As the station density decreases, the influence of stations farther from the analysis point increases, and therefore, if the distributions are inhomogeneous in space, the analysis point is influenced by stations with very different precipitation distributions. Finally, ranges of potential percent representativeness errors of the median and extreme precipitation across theUnited States are created for high-resolution (0.25°) and low-resolution areal averaged (0.9° lat 3 1.25° lon) precipitation fields. For example, the range of the representativeness errors is estimated, for annual extreme precipitation, to be from 116% to -12% in the low-resolution data, when station density is 5 stations per 0.9° lat × 1.25° lon grid box.