You are here:
Application of Partial Least Square (PLS) Regression to Determine Landscape-Scale Aquatic Resources Vulnerability in the Ozark Mountains
Citation:
NASH, M. S. AND R. D. LOPEZ. Application of Partial Least Square (PLS) Regression to Determine Landscape-Scale Aquatic Resources Vulnerability in the Ozark Mountains. Joint Statistical Meeting, Vancouver, BC, CANADA, July 31 - August 05, 2010. JOURNAL OF THE AMERICAN STATISTICAL ASSOCIATION. American Statistical Association, Alexandria, VA, 3588-3597, (2010).
Impact/Purpose:
The study area is a 21,848 square kilometer area of land that encompasses the headwaters of the White River, and generally the Ozark Mountains (Figure 1). The study area contains a mix of pasture and other agriculture (e.g., poultry production facilities, cattle operations, and hay operations), forest, and urban land cover, as well as several large reservoirs (Figure 2). The White River originates in northwestern Arkansas and flows through southwestern Missouri and north-central Arkansas. The White River descends from the Ozark Mountains into Arkansas’ agricultural plain where it meanders to its confluence with the Mississippi River.
Description:
Partial least squares (PLS) analysis offers a number of advantages over the more traditionally used regression analyses applied in landscape ecology, particularly for determining the associations among multiple constituents of surface water and landscape configuration. Common data problems encountered during landscape ecological analyses may include small sample sizes, missing data values among sampled areas, a large number of predictor variables, correlated variables, and high noise-to-signal relationships. PLS attempts to account for the above data problems, by building a robust association model. We utilized PLS to predict in situ surface water Escherichia coli (E. coli) bacterial counts in the Upper White River from the associated landscape-ecological metrics in the Ozark Mountains (southwestern Missouri and northwestern Arkansas, USA). The amount of variability in E. coli counts was explained by each PLS model and reflects the composition of the contributing landscape among the watersheds analyzed. The predicted values and their confidence intervals explain how land cover type and configuration, and land use may affect the abundance of E. coli in surface waters of the Upper White River region of the Ozark Mountains.