Conditioned Latin Hypercube Sampling to Accomplish Multiple Research Objectives in the Powder River Basin, WY.

The Powder River Basin of Wyoming has been extensively impacted by coal-bed methane production. The overall project goal was to assess the utility of available hyperspectral imagery and lidar derived digital terrain model( DTM)s of the Beaver Creek Watershed for addressing multiple research objectives: 1) model hydrologic functions, 2) digitally map soil properties, specifically soil organic carbon, 3) evaluate the impact of linear disturbances (e.g. roads) on soil properties and rangeland health, and 4) identify Sage Grouse habitat suitability. To accomplish these various objectives, an unbiased, representative, and flexible sampling scheme was needed. We utilized a conditioned Latin Hypercube sampling (cLHS) technique to randomly select sample locations representative of the maximum variability of key environmental covariates. These covariates were remotely sensed topographic and spectral data related to soil, landscape, and vegetation properties, including topographic wetness index (TWI) and Landsat band ratios. While we had remotely sensed data for the entire watershed, only a portion of the watershed was accessible for sampling. The cLHS technique facilitated the comparison of biophysical characteristics of the entire watershed vs. the accessible area. Because each of the research objectives required different numbers of samples and different levels of sampling intensity (i.e., soil sampling only vs. soil sampling and vegetation characterization plots), we used cLHS to select a hierarchical set of nested samples from 50 to 10 sample points. Multivariate regression and Random Forests inference models, DTM derivatives and hyperspectral imagery, field data, and soil properties determined in the laboratory will be used to predict the spatial distribution of soil and vegetative properties in the watershed for a range of sample sizes. The nested cLHS sampling design was adaptable considering the different sampling intensities required, as well as unanticipated logistical constraints in the field that reduced the number of samples collected. This design will allow us to evaluate the efficacy of different sizes of cLHS-selected samples in meeting multiple objectives across the watershed.