Files
Abstract
The U.S. Department of Agriculture (USDA) has been exploring methods of using remote sensing as a basis for improving survey methodology. This paper discusses digital techniques employing computer classification of pixels with ground enumerated livestock inventories for the State of Iowa during 1978. The methods of analyses include discriminant functions for classification of LANDSAT tapes and regression methods for making estimates of livestock numbers based on double sampling employing an area sampling frame. The results were much less promising than for acreage estimates, but similar to results for crop yield forecasts. These small gains in estimating efficiencies may have, in part, been due to the time interval between the dependent variable (livestock numbers) and independent variables (classified pixels). However, the combined gains in crop acreages, yields, and production when added to modest gains expected for livestock indicate that the combined economic benefits for agriculture are important.