Forecasting crop production using satellite-based vegetation health indices in Kansas, USA

摘要

This article shows the results of early crop yield prediction from remote-sensing data. The study was carried out in Kansas, USA. The methodology proposed allows the estimation of winter wheat (WW), sorghum and corn yields 3-4 months before harvest. The procedure uses the vegetation health (VH) indices (vegetation condition index (VCI) and temperature condition index (TCI)) computed for each pixel and week over a 21-year period (1985-2005) from the Advanced Very High Resolution Radiometer (AVHRR) data. Over this period, a strong correlation was found between crop yield and VH indices during the weather-related critical period of crop development, which controls much final crop productivity. The 3-month advanced yield forecasts were independently compared with official agricultural statistics, showing that the estimation errors for WW, sorghum and corn were 8%, 6% and 3%, respectively. Implementing the 3-4 months lead forecast in operational practice will aid farmers to mitigate weather vagaries using irrigation, diseases/insects control, application of fertilizers and so on during a growing season and will help decision-makers to regulate marketing strategies, import/export and price policies and address food security issues.