Alfalfa hay yield prediction using yield monitor data and satellite imagery - Research Paper Example

These variables can be used as yield indicators upon which crop growth models are based (Clevers et al, 1994). Accurate yield estimation relies on the availability and quality of actual crop status data. Satellite images provide a spatial resolution in spectral bands and offer coverage cycles. A number of studies have used such satellite images for monitoring of crop development (Watz et al, 1996) a characterization of natural vegetation (Saiz et al,1996). Information about the crops can be obtained from Vegetation indices (Tucker, 1979).

These are functions that measure crop reflectance in terms of spectral analysis. Studies support the fact the use of Vis from the sensors were effective in determining the spatial variation and phonological changes in vegetation, though there are several Vis defined the most commonly used VI is the NDVI(normalized difference vegetation index) (Schowengerdt,1997). Techniques to improve the use of NDVIs are being developed, for example to normalize multi-temporal NDVs derived from NOAA AVHRR data for atmospheric effects.

(Potdar et al, 1999). The reason for the variations was the low correlation coefficients. (Groten, 1993, Sharma 1993, Rosema, 1998) The main goal of agricultural crop management is to guarantee food resource for its population. Crop yield prediction ahead of harvest time and involving large regions is important for all countries. Crop yield prediction entails the application of crop growth models and crop yield models. Despite advances in crop yield prediction models, the applicability of such a model is limited to particular crops, cultivation practices, and growing conditions (Gommes, 1998).

In addition, there is recognition that an integral part of predicting yield lies in accurate identification of growing sites and measurement of crop sown area, prior to using VI based methods for predicting crop growth and yield (Gommes , 2001). Thus, the goal of our study was to validate the prediction model as per NDVI with actual yield. In fact, early research has revealed the need to obtain better spectral signatures for predicting growth and yield of crops (Sonia ,1999, Sonia et al ,2002).

This is because no single model of prediction has proven satisfactory in all conditions. For example, models like CERES (Larrabee et al, 1985), WOFOST (Diepen et al,1989) and EPIC (Williams et al, 1984), is limited due to the fact they were targeted towards specific research methods. The leading crop simulation models have been deemed to complex for wide acceptable forecasting purposes (Gommes, 1998). Remote Sensing has been used to furnish input data for models. Spectral vegetation indices correct the atmospheric and soil spectral effects on remotely sensed data.

(Broge et al,2002)(McDonald et al ,1983)(Tucker,1979)(Tucker et al, 1985)(Tucker et al, 1991)(Unganai et al, 1988)(Williams, et al, 1991). This Vegetation Index (VI) is a measure of total green biomass at any given time has been related to crop yield (Potdar et al, 1993). The normalised difference vegetation index (NDVI) can be computed from red and near infrared reflectance data available from LANDSAT-TM, SPOT, ARTEMIS, NOAA-AVHRR satellite images (Heilkema et al, 1990). The NDVI derived