Economics of management zone delineation in cotton precision agriculture: A multi-year analysis

The concept of precision agriculture is based on the ability to improve the management of production factors using site specific information. The optimal configuration of management zones for more precise management of farm inputs is one of the most important components of precision farming. The main objective of this chapter is to develop and describe a management zone delineation procedure based on a spatial statistic approach using precision agriculture data from cotton production in the Texas high plains from 2002 to 2004. An optimization model is utilized to evaluate the economic impact of more precisely applying N fertilizer based on these management zones versus the more traditional method of using a uniform rate for the whole field. The main input of this optimization model is a cotton yield response function estimated using a spatial econometric technique that accounts for spatial dependence in the residuals and also for spatial heterogeneity when using multi-year data (spatial panel data). The results obtained suggested that applying variable N rates based on the management zones delineated would result in higher net returns over fertilizer cost relative to the traditional uniform rate application under the single year analysis (2002). When using the multi-year spatial data (pool data from 2002, 2003, and 2004), the Variable Rate application technique (VRN) seems to be consistently more profitable when comparing it to the uniform N rate application technique (URA). These results are obtained under the assumption that the response function is stable over time (structural stability). When testing for structural stability we obtained that the response function for the different application techniques (uniform rate applications, and variable rate application) changes over time. Although this conclusion suggest that an ex post economic evaluation of the different N application techniques should be done on a yearby- year basis, we shown that the "average" (stable over time) response function estimated from multi-year data can still yield similar inferences to the year-by-year approach as long the growing conditions do not deviate substantially from the average over a period of time. In light of this result, we conjecture that an "average" response function estimated using past multi-year data can still potentially be used to undertake an ex ante evaluation of alternative N application practices for a future year (2005 in our case). The results from this ex ante analysis can potentially be used by producers to help them decide which application technique may be more beneficial for them (assuming that normal growing conditions will occur in the future year of interest).