A spatial econometric approach to the economics of site-specific nitrogen management in corn production

Spatial technologies such as global positioning systems (GPS) and geographic information systems (GIS) increasingly form the basis for site-specific management (SSM) in crop production. This dissertation assesses the contribution of an explicit spatial econometric methodology in the estimation of crop yield functions that are used to optimize fertilizer application. The specific case study is for nitrogen (N) application to corn production in Argentina, where the implementation of variable rate technology (VRT) requires methods that use inexpensive information and that focus on the inputs and variability common to Argentine growing areas. The main objective of this research is to assess the economic value of the application of spatial regression analysis to yield monitor data as a means to optimize variable rate fertilizer strategies. Several studies have shown how input response curves can be estimated from yield monitor data with spatial regression analysis. Critics have questioned the value of such estimates because response may vary from season to season. Another objective of this research is to test the stability of the response and outline economic analysis when response estimates differ substantially from year to year. The data were drawn from three on-farm N trials done near Rio Cuarto, Argentina in 1998–99 and 2000–01. Landscape areas, elevation and remotely sensed images were used to delineate N management zones. Spatial autocorrelation and spatial heterogeneity were taken into account using a spatial error model and a groupwise heteroskedasticity model. The economic analysis maximized expected profit, given response estimates and price distributions. A partial budget was used to calculate uniform rate and VRT returns. Results show that N response and profit maximization N rates differ in different areas of the landscape; that N response and consequently profit maximization N rates in those landscape areas differ from year to year; that elevation is a potential objectively measurable alternative to landscape areas, and that profit maximization can be effectively carried out with the response functions that differ by landscape area and season. When the outcome of the two crop seasons can be predicted to some degree at N application time, VRT N can be profitable in most farms