Authors: Sarah J. Becker*, United States Army Engineer Research & Development Center, Susan L. Lyon, United States Army Engineer Research & Development Center, Nicole M. Wayant, United States Army Engineer Research & Development Center
Topics: Spatial Analysis & Modeling, Geographic Information Science and Systems, Remote Sensing
Keywords: pixel size, raster, resampling, spatially disparate data, semivariogram
Session Type: Poster
Start / End Time: 1:20 PM / 3:00 PM
Room: Napoleon Foyer/Common St. Corridor, Sheraton, 3rd Floor
Presentation File: No File Uploaded
The objective of this research is to develop a technique to determine the best choice pixel size when resampling datasets containing spatially disparate pixel sizes. This is necessary because disparate pixel sizes must be unified in order for comparisons to be made across a stack of input data rasters. Most often, researchers select spatial resolutions based on their project needs and goals, generally choosing the coarsest resolution of their datasets and resampling finer resolutions to match. The challenge is in weighing two notions: 1) resampling data always introduces some degree of statistical error; and 2) information is lost whenever fine-scale data are resampled to coarser scales. To determine the best choice pixel size, semivariograms were created for each individual dataset and optimal pixel sizes were determined for the dataset. The datasets were then resampled to all of the individual optimal pixel sizes that were calculated, and Root Mean Square Errors between the original and resampled datasets at each pixel size were calculated. The pixel size that produced the lowest Root Mean Square Error across all datasets was identified as the best pixel size. This selected pixel size is intended to guide future raster-based multivariate analysis and modeling.