| Early research in agriculture used remotely sensed data mostly for the identification of spectral signatures, where crop type/area classification would depend on data acquired from hand-held or truck-mounted instruments. Through time different approaches were taken for crop type/area classification, such as a systematic sampling of inventory ground data that were used later for training and testing using image processing techniques. Later, technology such as the global positioning system (GPS) and geographic information systems (GIS) were used for application to precision agriculture. These new tools provided a better meaning to understand and analyze the complex variability of the crop-soil-atmosphere system to estimate crop yields.; The present research used data collected in the Management System Evaluation Area (MSEA) in 1998 and two Landsat thematic Mapper images (July and August) to analyze the crop variability. Ground truth parameters, such as chlorophyll, leaf area index (LAI), and electricity conductivity, were collected throughout the growing season. In addition, vegetation indexes (VI) such as the Normalized difference vegetation index (NDVI), simple vegetation index (SVI), soil adjusted ratio vegetation index (SARVI), were computed for the two images. Both ground truth data and VI's were statistically analyzed with yield measurements taken with an on-the-go yield monitor to estimate a best fit yield model for use with soybeans and corn.; The correlation analysis within a Landsat pixel reported SVI52 (r = 0.62), SVI53 (r = 0.56), and SVI54 (r = 0.53) as the most significant relationships. The results from the ground truth data vs. on-the-go yields reported total clay (%) (r = 0.90), leaf area (r = 0.76), and tissue plant analysis (r = 0.73); A stepwise regression analysis was performed using the Landsat TM images and the VI's selected. A series of linear models were evaluated taking into account the Landsat TM and yield while varying the scale (1 pixel to 16 30-m pixels). The same stepwise regression analysis were performed but adding the Digital Elevation Model (DEM) and the Electrical Conductivity (EC) which resulted in the best coefficient of determination (R2 = 0.94, R 2 = 0.98, and R2 = 0.96 for plots 100, 200, and 300, respectively).; A stepwise regression analysis was also explored with standardized yields. The resulted models allowed exploring the use of a crop independent model.; The corn and soybeans yield models developed with the 1998 data were used in two extra images (1991 and 1992) to test the models... |
