2007 Irrigated Lands for the Bruneau-Grandview Area: Machine Learning Generated

This raster file represents land within the Grandview-Bruneau Water Budget Area classified as either “irrigated” with a cell value of 1 or “non-irrigated” with a cell value of 0 at a 30-meter spatial resolution. These classifications were determined at the pixel level by a Random Forest supervised machine learning methodology. Random Forest models are often used to classify large datasets accurately and efficiently by assigning each pixel to one of a pre-determined set of labels or groups. The model works by using decision trees that split the data based on characteristics that make the resulting groups as different from each other as possible. The model “learns” the characteristics that correlate to each label based on manually classified data points, also known as training data. A variety of data can be supplied as input to the Random Forest model for it to use in making its classification determinations. Irrigation produces distinct signals in observational data that can be identified by machine learning algorithms. Additionally, datasets that provide the model with information on landscape characteristics that often influence whether irrigation is present are also useful. This dataset was classified by the Random Forest model using top-of-atmosphere reflectance data from Landsat 5 and Landsat 7, NASA STRM Digital Elevation data, and Height Above Nearest Drainage (HAND) data. Landsat 5, Landsat 7, NASA STRM Digital Elevation data, and HAND data are at a 30-meter spatial resolution. The Cropland Data Layer (CDL) from the United States Department of Agriculture National Agricultural Statistics Service (USDA NASS), National Agriculture Imagery Program (NAIP) data from the USDA Farm Service Agency (FSA), and U.S Fish & Wildlife Service National Wetlands Inventory data were also used in determining irrigation status for the manually classified training data points but were not used for the machine learning model predictions. The final model results were manually reviewed prior to release, however, no extensive ground truthing process was implemented. Wetlands areas identified by US Fish & Wildlife Service data were masked as non-irrigated if they had no overlapping irrigation POUs or were manually determined to have no potential irrigation to correct some of the riparian areas that were falsely classified by the model as irrigated. Gaps in Landsat 7 due to the scan-line corrector error were corrected by adding manually generated polygons for irrigated areas. “Speckling”, or small areas of incorrectly classified pixels, was reduced by a Boundary Clean smoothing technique which uses a descending sort order by size. Zones with larger total areas have a higher priority to expand into zones with smaller total areas.