Spatial regression models for the economic analysis of on-farm site-specific management trials

The goal of this research was to develop spatial regression methods to economically evaluate on-farm site-specific management (SSM) trials. Whether yields are collected with a combine yield monitor, a weigh wagon, or other technology, they usually have a spatial structure. Covariates including soil test information, topographical and hydrological information, and soil series are also spatially correlated. It is well-known by researchers that inference drawn from spatially correlated data is compromised thereby increasing the risk that incorrect decisions about the effectiveness of SSM technologies and its influence on farm profitability. Four essays address this issue. The first essay evaluated net returns to a variable rate nitrogen (VRN) management strategy using yield monitor data and topographic information. Partial budgets using regression coefficients estimated with four spatial regression methods are compared to partial budget results using coefficients not conditioned on spatial information. Partial budget results based on the spatial models favored the VRN strategy. The same results were achieved by all spatial models: an incorrect decision about technology choice can be made if spatial dependence was not included in the regression model. The second essay examined how value is added to manure when it is spatially managed together with other inputs such as lime, phosphorous, and potassium. The value of manure using application increased $10 acre-1 compared to following traditional uniform-rate extension recommendations for manure and fertilizer management. The third essay estimated spatial and temporal response to nitrogen (N) and P using five years of production data from a corn-soybean rotation. Corn and soybean response to P varied spatially, but was temporally stable in substantial parts of the field. Economic analysis suggests that variable rate P application can be profitable, but N application decisions are mainly a risk management issue. The final essay used data from the VRN-P study. Coefficients for a P carryover equation were estimated simultaneously with corn and soybean response to N and P using regression. The optimal time path of a spatially managed N and P application program (N-P) is compared with an N-P management strategy where these inputs are applied uniformly.