Determining the feasibility of collecting high-resolution ground-based remotely sensed data and issues of scale for use in agriculture

Abstract

A ground-based remote sensing system was attached to a linear move irrigation system and successfully collected pixels at an approximate density of 1/meter 2. This low-resolution data was used to create 1-meter resolution images in near real time over a 1-hectare cotton field. A new method using GIS and spatial statistics (kriging) was successfully developed for evaluating the 1-meter images and simulate 2 through 7 meter resolution for determining the effects of scale on data collection for crop management as applied to precision agriculture. The images collected reliably predicted nitrogen and water stress in the field and demonstrated how scale from 1 to 7 meters affects reliability of measuring water and nitrogen stress. A 2X2 Latin square water and nitrogen experiment on cotton consisting of optimal and low nitrogen and water treatments was conducted within 4 replicates of the 4 treatments. The remotely sensed data were used to develop images of the plot to ascertain the ability to detecting nitrogen and water stress. Nitrogen stress was evaluated using the canopy chlorophyll content index while water stress was evaluated using the difference between canopy and air temperature. Four days of field images collected in 1999 at a 1-meter resolution were selected for evaluation. The days represent, one day prior to water and nitrogen treatments, two days of little to moderate nitrogen stress, and one day with severe nitrogen stress and moderate water stress. The image analysis incorporated standard statistics, kriging, and fractals. The 1-meter data was used to produce images with grids of 2 through 10 meters. Standard statistics were used to analyze the four days by grid size. The results indicated no difference in the mean in the data for any grid size within a treatment for either water or nitrogen; however, CV generally decreased with grid size. Kriging was used to evaluate the data for pretreatment day and stressed day for one plot representing each of the four treatments. Data for 1, 3, 5, and 7 meters resolution was kriged and compared to the 1-meter grid to determine reproducibility. It was determined that for temperature it is difficult to reproduce finer resolution data, especially in stressed plots. The nitrogen indice was reproducible to a high degree of accuracy for grids as large as 7 meters. Fractal analysis was used to evaluate the kriged data. The results were mixed in that numbers for some plots increased as grid size increased, and decreased as expected for others.