Error estimation of objective analysis of surface observations

Abstract

As part of the Localized Aviation Model Output Statistics (MOS) Program, the Meteorological Development Laboratory is analyzing surface data reports on an hourly basis. The Bergthórsson-Cressman-Döös-Glahn analysis program that is being used for gridding the MOS forecasts has been tailored to analyze surface observations. These analyses are available in the National Digital Guidance Database (NDGD) over the conterminous United States. This database is on the same grid as, and is interoperable with, the National Digital Forecast Database. It is desired to know the errors involved in these analyses. Whereas the actual errors are unknowable for several reasons, they can be estimated. On any given analysis, one would expect the error at a specific location to be a function of some knowable parameters, such as distances between the reporting locations and the grid points, the terrain roughness, the density of reporting locations, and the variability of the data values-all in the immediate vicinity. We have made analyses of surface temperature and dewpoint over the conterminous United States every fifth hour for one year. On each analysis, 20 land stations and one water station were randomly withheld from the available data. For each withheld datum, the analysis value at that site was estimated by bilinear interpolation. The differences between these interpolated values and the actual observations were related to knowable parameters through least squares regression-one relationship for land and another for water-for each variable. These regression equations were then applied, respectively, to each land and water grid point for a specific hour. This produced an estimate of the error of the analysis for each grid point for that specific time. These grids of errors are, along with the analyses, available in the NDGD. This paper describes the process and shows the results.

Trust and trustworthy artificial intelligence: A research agenda for AI in the environmental sciences

“…AI has been applied to weather forecasting for decades (e.g., Glahn & Dallavelle, 2000; Glahn & Im, 2011; Malone, 1955). For example, AI has been used to forecast the probability of tornadoes, hail, damaging wind, and hurricane intensity (Billet et al., 1997; Demuth et al., 2006; Gagne et al., 2017; Kitzmiller et al., 1995; McGovern et al., 2014; Williams, 2014; cf.…”