Title: Technique for ranking potential predictor layers for use in remote sensing analysis

Author: Lister, Andrew; Hoppus, Mike; Riemann, Rachel

Year: 2004

Publication: In: Proceedings, Society of American Foresters 2003 national convention; 2003 October 25-29; Buffalo, NY. Beshesda, MD: Society of American Foresters: 402-409.

Abstract: Spatial modeling using GIS-based predictor layers often requires that extraneous predictors be culled before conducting analysis. In some cases, using extraneous predictor layers might improve model accuracy but at the expense of increasing complexity and interpretability. In other cases, using extraneous layers can dilute the relationship between predictors and target variables that the modeling technique seeks to exploit. The current study seeks an automated method whereby a ranking of potential predictor data can be obtained so that the researcher can quantitatively justify including or excluding certain predictors. In our example, we seek to determine the relative strength of the relationship between various Landsat ETM satellite spectral layers and total basal area on 380 study plots established by the USDA Forest Service's Forest Inventory and Analysis unit.