A Comparison of Thiessen Polygon,Kriging, and Spline Models of Potential UV Exposure |
| |
| Abstract: | The performance accuracy of Thiessen-polygon and kriging interpolation methods available in the standard GIS packages was evaluated based on magnitude of errors in predicting potential UV exposure across the continental U.S., and the results were compared with those of the ANUSPLIN routine that runs outside typical GIS through a series of C++ and FORTRAN commands. Input data consisted of global radiation measures recorded at 215 stations, latitude, longitude, and elevation from a 30 arc-second Digital Elevation Model. The objective was to identify the most accurate prediction method for facilitating measurement of potential UV exposure at local (e.g.1km2 grid cell) and county levels. The ANUSPLIN method produced the smallest prediction errors in estimating values of potential UV exposure at 1 km2 resolution; these measurements were aggregated to the county level. We examined how much variation was lost through aggregation, as well as the potential bias associated with the possibility that some counties have predominantly north or south facing slopes. The impact of using inferior procedures on the estimates and geographic patterns of potential UV exposure was also examined. ANUSPLIN generated results that are reproducible and for which uncertainty is known. These measurements will be used in subsequent analysis of the role of UV exposure in melanoma etiology. |
| |
| Keywords: | |
|
|
