EXPLORING THE POSSIBILITIES OF USING NOAA, VHRR DATA TO FORECAST COTTON YIELD IN SWAZILAND

  • Manasah S. Mkhabela National Early Warning Unit (NEWU), Ministry of Agriculture and Cooperatives, Mbabane, Swaziland
  • Milton S. Mkhabela Department of Crop Production, Faculty of Agriculture, University of Swaziland, P.O. Luyengo, Swaziland

Abstract

A study was undertaken to investigate the possibility of using the Normalised Difference Vegetation Index (NDVI) data developed by the National Oceanic Atmospheric Administration (NOAA's), Very High Resolution Radiometer (VHRR) to predict cotton yield in Swaziland. Fourteen years of historical data of average cotton yield and NDVI were used for the study. A regression analysis was performed for 504 data set for NDVI and 14 data set for cotton yield. A linear model using the cumulative NDVI was developed to predict cotton yield. The regression analysis was performed on minimum, average and maximum NDVI. The cumulative maximum NDVI was found to be the most effective in forecasting cotton yield than either the minimum or the average NDVI despite the significance of the minimum NDVI. The regression coefficient for the cumulative maximum NDVI was 70%, while that for the minimum NDVI was 56%. The best regression period was found to be from the second dekad of February to the second dekad of April, which coincides with the flowering and boll formation stages of cotton. This means that a production forecast can be issued about 1-2 months before harvest. Such information would be very useful for early warning purposes. Although very good results were achieved from this study it has to be noted that all models have limitations. Satellites are affected by various atmospheric effects, which may compromise the quality of the data produced. However, if limitations are considered when using satellite data very useful information can be derived from satellite developed data. Due to the limited number of years (14 years) used in this study the model has to be further tested, calibrated and updated as more yield and NDVI data become available.

(UNISWA J Agric:2000 9: 13-21)
Section
Articles

Journal Identifiers


eISSN: 1021-0873