Irrigated Vegetation Assessment for Urban Environments

Abstract

Assuring the availability of water in the southwestern United States is a major resource management problem. Irrigation of landscape vegetation within urban environments represents a large portion of the total urban water consumption for this region. Current estimates suggest that up to 50 percent of residential water is used for landscape irrigation. This paper examines the utility of Ikonos multispectral satellite imagery and expert classifier approaches for quantifying the amount and distribution of urban irrigated landscape vegetation. A decision tree, expert classifier model applied to Ikonos image and land-use GIS layer inputs was tested against conven tional image classification approaches. With all branches of the decision tree activated, percentage estimates of urban irrigated vegetation versus impervious cover differed from airborne image-derived reference data by less than 8 percent. Highest agreement was achieved using all model branches except a spatial structure rule, which utilized a texture metric derived from Ikonos 1-m panchromatic data. For this same product, proportion estimates of two growth form types (Tree/ Shrub and Grass) and impervious cover differed from reference data by less than 3 percent, and root-mean-square error (RMSE) values for all neighborhood-size sampling units were within 5 percent for all cover types. This “optimal” expert classifier product yielded areal proportion estimates and RMSE values that were approximately 2 percent closer to those of the reference data, compared to standard unsupervised classification applied to Ikonos multispectral data.