Celestial satellite and earthly crop yield : informational content of satellite-based crop yield forecasts
Loading...
Date
2001
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Montana State University - Bozeman, College of Agriculture
Abstract
Since the late 70s, burgeoning efforts have been allocated to study the potential of monitoring crop conditions and forecasting crop yields via remote sensing from the satellite. An overwhelming majority of these studies shows that remote sensing from the satellite express high predictive power in crop forecasting. In this thesis, using satellite images to forecast wheat yield from 1989 to 2000 in six Montana Crop Reporting Districts (CRD), several statistical improvements were achieved over extant crop forecasting models. First, different weights were allowed for satellite images obtained at different points of time, accounting for the likely heterogeneous contributions of various crop phenological stages to the final crop yield. Second, crop acreage information was directly modeled. This, to some extent, alleviates the low-resolution problem of existing satellite imagery. Third, jackknife out-of-sample forecasts were generated to formally measure the well-known instability problem of using satellite imagery in crop forecasting across seasons. In addition, the satellite-based crop yield forecasts were compared with those of the U.S. Department of Agriculture (USDA), whose forecasts were based on traditional methods. It is shown that although meaningful crop forecasts can be generated from the satellite imagery late season, the additional yield information that can be extracted from the satellite tends to be limited. Because in the major wheat producing CRDs, the USDA forecasts are already very accurate and little independent information is observed in the satellite-based forecasts. Results suggest the needs to pinpoint crop phenological stages and to calibrate region-specific crop forecasting model.