A Comparison of Vector and Raster GIS Methods for Calculating Landscape Metrics Used in Environmental Assessments
|Authors:||Timothy G. Wade, James D. Wickham, Maliha S. Nash, Anne C. Neale, Kurt H. Riitters, K. Bruce Jones|
|Station:||Southern Research Station|
|Source:||Photogrammetric Engineering & Remote Sensing. Vol. 69, No. 12, December 2003, pp. 1399–1405.|
GIS-based measurements that combine native raster and native vector data are commonly used in environmental assessments. Most of these measurements can be calculated using either raster or vector data formats and processing methods. Raster processes are more commonly used because they can be significantly faster computationally than vector, but error is introduced in converting vector data to raster. This conversion error has been widely studied and quantified, but the impact on environmental assessment results has not been investigated. We examined four GIS-based measurements commonly used in environmental assessments for approximately 1000 watersheds in the state of Maryland and Washington, D.C. Each metric was calculated using vector and raster methods, and estimated values were compared using a paired t-test, Spearman rank correlation, and cluster analyses. Paired t-tests were used to determine the statistical significance of quantitative differences between methods, and Spearman rank correlation and cluster analyses were used to evaluate the impact of the differences on environmental assessments. Paired t-test results indicated significant quantitative differences between methods for three of the four metrics. However, Spearman ranks and cluster analyses indicated that the quantitative differences would not affect environmental assessment results. Spearman rank correlations between vector and raster values were greater than 0.98 for all comparisons. Cluster analyses resulted in identical assignment for 88 percent to over 98 percent of watersheds analyzed among vector and various raster methods.