1a. Objectives (from AD-416)
This research seeks to provide improved management tools and information technologies to aid commercial cotton production in the Mid South through the application of precision agriculture practices. It seeks to reduce costs of cotton production through the acquisition, storage, processing, and application of geo-referenced information. Combines remotely sensed data with insect sampling methods to improve decisions regarding control for thrips and stink bugs in cotton. Develops appropriate statistical anslysis concepts and models for precision agriculture. Develops local area networking software and decision aids for digital data collection and use in precision agriculture.
1b. Approach (from AD-416)
Statistical models will be developed for geo-referenced measurement of precision agriculture cotton production practices. Improvements to mixed model analysis methodology will be developed using archived data as well as new data from remotely sensed images and field measurements. Statistical models to analyze spatial application decisions from data that involves a disproportionate number of zero counts and has small sample sizes will be developed and improved. Light Detection and Ranging (LiDAR) and Normalized Difference Vegetation Index (NDVI) data will be combined to facilitate site-specific management of thrips and stink bugs in cotton. Concepts and software will be built upon a wireless network to manage large quantities of geo-referenced data.
3. Progress Report
Yield monitor data from 2001-2005 combined with Light Detection and Ranging (LIDAR) imaging and Veris apparent soil conductivity data have resulted in a significant improvement of 30% over quadratic regression analysis of clustered data. We compared 5 years of yield monitor data using LIDAR generated crop yield stability maps and medium resolution global positioning system (GPS) elevation maps. The medium resolution GPS maps resulted in significant misidentification of yield areas in the field compared to LIDAR maps. A 66 ha field under a corn/soybean rotation was chosen to begin research in the use of Real Time Kinetics (RTK) based yield monitor data. In 2006 corn was grown in the field and in 2007-2009 soybeans were grown. The procedure previously developed with LIDAR data to generate a crop yield stability map was used; however, RTK elevation data were used in lieu of LIDAR elevation data. Results show that RTK elevation data can be used to create statistically reliable crop yield stability maps. This is significant because LIDAR maps are not widely available for producer fields; however, many growers are now using RTK for equipment auto-steer capability. Sampling station arrays of yellow sticky cards for insect detection were geographically placed in a cotton field at five sites selected according to a categorical, pseudo-likelihood classification map derived from remote sensing information of mid-June crop vigor and LIDAR information of elevation. Cards were replaced at 4-7 day intervals during flowering. Count model regression techniques will be used to test the hypothesis that: thrips, Lygus (plant bug), and beneficial arthropod (insects) abundance differs among the zones. Correlations between sweep net counts of Lygus and counts from these sampling stations can also be made. Research was begun to examine applicability of site-specific, integrated pest management (IPM) for soybean. Sampling station arrays of yellow sticky cards were geographically placed in a large soybean field. Sites were selected based upon a categorical, pseudo-likelihood classification map derived from Veris shallow apparent electro-conductivity readings (ECa) and LIDAR information of elevation. Cards were replaced at 4-7 day intervals from first bloom to maturity. Count model regression techniques will be used to test the hypothesis that: soybean arthropod pests and beneficial numbers will differ among the classified zones. The first year of research to jointly use categorical classification of pest abundance at GPS referenced sample locations selected from a categorical, pseudo-likelihood classification map derived from remote sensing information of crop vigor and LIDAR information of elevation was completed. Results showed promise for simplifying the linkage of scouting data with the remote sensing layers for timing the application of broadcast and spatial applications of pesticides. Data collected will also be used to examine the effectiveness of a commercial field application of the topological, general linear mixed statistical model to complete an analysis of the multi-temporal pesticide applications.