PURC: The Digital Methodology for Mapping Soils
Location
Eccles Conference Center
Event Website
http://water.usu.edu/
Start Date
3-27-2006 10:45 AM
End Date
3-27-2006 11:50 AM
Description
Vast areas of the earth need new or updated soil survey data, particularly on public lands facing pressure for rapid energy development. However, traditional methods of soil survey are inefficient, expensive, and often inaccurate. Soil mapping has generally been based on the conceptual models that unique soils are the products of unique sets of environmental covariates, including climate, vegetation, relief, and parent material. We developed and tested a methodology that uses spatially explicit digital data to represent soil-forming factors to to predict and map soil distribution: Pedogenic Understanding Raster Classification (PURC) methodology. Our first study area was in the Powder River Basin of Wyoming, which is under intense pressure for coal-bed methane development. Topographic data derived from digital elevation models (DEMs) and Landsat ETM spectral data were selected to represent soil-forming factors of relief (relative elevation to the Powder River, slope, compound topographic index), vegetation (Fractional vegetation derived for the Normalized Difference Vegetation Index), and parent material (Landsat spectral band ratios). These digital data were analyzed using commercially available GIS and image processing software. Unsupervised, supervised, and simple knowledge-based classifications were used in the preliminary stage to develop visual representations of soil-landscape patterns and to plan for field data collection. As more was learned about the survey area from data collection, a knowledge-based decision-tree classification model was built and refined. The resulting maps were evaluated qualitatively by local experts and quantitatively using accuracy assessment, and showed good agreement between predicted and observed map units. We then transferred PURC to a second soil survey project area in the Green River Basin of Wyoming and incorporated classification tree analysis as an alternative for predicting the distribution of soil classes on the landscape. The Green River Basin had greater variability in parent material, lower erosion rates, and a colder climate than the Powder River Basin, and is under pressure for natural gas development. Unsupervised classification techniques were used in the preliminary stage of the methodology to recognize existing soil-landscape patterns and to develop an initial sampling plan. Knowledge-based classification and classification tree analysis were used to develop models predicting soil delineations and quantifying soil map unit concepts. The output images generated from the knowledge-based and classification tree models produced similar predictions of soil patterns across the landscape, with slightly different predictions of soil map units. However, the knowledge-based model was much more time-intensive than the classification trees, and failed to classify all pixels within the study area even after multiple iterations of the model. Although finals maps were slightly different, both models were transparent and could be further refined as additional data becomes available or as land-use needs change. In general, the PURC approach appears to provide a logical framework for the collection and analysis of spatially explicit, digital data that can be used to predict soil distribution on the landscape more accurately, consistently, and efficiently than traditional soil survey methods.
PURC: The Digital Methodology for Mapping Soils
Eccles Conference Center
Vast areas of the earth need new or updated soil survey data, particularly on public lands facing pressure for rapid energy development. However, traditional methods of soil survey are inefficient, expensive, and often inaccurate. Soil mapping has generally been based on the conceptual models that unique soils are the products of unique sets of environmental covariates, including climate, vegetation, relief, and parent material. We developed and tested a methodology that uses spatially explicit digital data to represent soil-forming factors to to predict and map soil distribution: Pedogenic Understanding Raster Classification (PURC) methodology. Our first study area was in the Powder River Basin of Wyoming, which is under intense pressure for coal-bed methane development. Topographic data derived from digital elevation models (DEMs) and Landsat ETM spectral data were selected to represent soil-forming factors of relief (relative elevation to the Powder River, slope, compound topographic index), vegetation (Fractional vegetation derived for the Normalized Difference Vegetation Index), and parent material (Landsat spectral band ratios). These digital data were analyzed using commercially available GIS and image processing software. Unsupervised, supervised, and simple knowledge-based classifications were used in the preliminary stage to develop visual representations of soil-landscape patterns and to plan for field data collection. As more was learned about the survey area from data collection, a knowledge-based decision-tree classification model was built and refined. The resulting maps were evaluated qualitatively by local experts and quantitatively using accuracy assessment, and showed good agreement between predicted and observed map units. We then transferred PURC to a second soil survey project area in the Green River Basin of Wyoming and incorporated classification tree analysis as an alternative for predicting the distribution of soil classes on the landscape. The Green River Basin had greater variability in parent material, lower erosion rates, and a colder climate than the Powder River Basin, and is under pressure for natural gas development. Unsupervised classification techniques were used in the preliminary stage of the methodology to recognize existing soil-landscape patterns and to develop an initial sampling plan. Knowledge-based classification and classification tree analysis were used to develop models predicting soil delineations and quantifying soil map unit concepts. The output images generated from the knowledge-based and classification tree models produced similar predictions of soil patterns across the landscape, with slightly different predictions of soil map units. However, the knowledge-based model was much more time-intensive than the classification trees, and failed to classify all pixels within the study area even after multiple iterations of the model. Although finals maps were slightly different, both models were transparent and could be further refined as additional data becomes available or as land-use needs change. In general, the PURC approach appears to provide a logical framework for the collection and analysis of spatially explicit, digital data that can be used to predict soil distribution on the landscape more accurately, consistently, and efficiently than traditional soil survey methods.
https://digitalcommons.usu.edu/runoff/2006/AllPosters/4