Premium
Comparison of elevation and remote sensing derived products as auxiliary data for climate surface interpolation
Author(s) -
Alvarez Otto,
Guo Qinghua,
Klinger Robert C,
Li Wenkai,
Doherty Paul
Publication year - 2014
Publication title -
international journal of climatology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.58
H-Index - 166
eISSN - 1097-0088
pISSN - 0899-8418
DOI - 10.1002/joc.3835
Subject(s) - covariate , elevation (ballistics) , digital elevation model , multivariate interpolation , environmental science , interpolation (computer graphics) , meteorology , latitude , climatology , precipitation , advanced very high resolution radiometer , climate model , remote sensing , normalized difference vegetation index , climate change , statistics , mathematics , computer science , geography , geology , bilinear interpolation , geodesy , satellite , aerospace engineering , engineering , animation , oceanography , geometry , computer graphics (images)
Climate models may be limited in their inferential use if they cannot be locally validated or do not account for spatial uncertainty. Much of the focus has gone into determining which interpolation method is best suited for creating gridded climate surfaces, which often a covariate such as elevation (Digital Elevation Model, DEM ) is used to improve the interpolation accuracy. One key area where little research has addressed is in determining which covariate best improves the accuracy in the interpolation. In this study, a comprehensive evaluation was carried out in determining which covariates were most suitable for interpolating climatic variables (e.g. precipitation, mean temperature, minimum temperature, and maximum temperature). We compiled data for each climate variable from 1950 to 1999 from approximately 500 weather stations across the Western United States (32° to 49° latitude and −124.7° to −112.9° longitude). In addition, we examined the uncertainty of the interpolated climate surface. Specifically, Thin Plate Spline ( TPS ) was used as the interpolation method since it is one of the most popular interpolation techniques to generate climate surfaces. We considered several covariates, including DEM , slope, distance to coast (Euclidean distance), aspect, solar potential, radar, and two Normalized Difference Vegetation Index ( NDVI ) products derived from Advanced Very High Resolution Radiometer ( AVHRR ) and Moderate Resolution Imaging Spectroradiometer ( MODIS ). A tenfold cross‐validation was applied to determine the uncertainty of the interpolation based on each covariate. In general, the leading covariate for precipitation was radar, while DEM was the leading covariate for maximum, mean, and minimum temperatures. A comparison to other products such as PRISM and WorldClim showed strong agreement across large geographic areas but climate surfaces generated in this study ( ClimSurf ) had greater variability at high elevation regions, such as in the Sierra Nevada Mountains.