The impact of controlled drainage on agricultural yields: A spatial panel model using yield monitor data

This dissertation provides a modeling framework for precision agriculture data analysis over time by using Geographical Information Systems (GIS) and econometric spatial panel models, with an application estimating the financial incentives for farmers regarding controlled drainage adoption. Using field-level yield monitor data and spatial panel techniques, the yield response equation is estimated using fixed and random effects models that incorporate spatial dependence in the error term, while controlling for the topography, weather and the controlled drainage treatment. The richness of the spatial panel techniques for precision agriculture data analysis is demonstrated and the need for space-time varying independent variables is emphasized. Controlling for fixed or random effects allows disentangling the effects of spatial dependence from spatial heterogeneity and omitted variables, and thus, is necessary to properly investigate the yield response. The proposed framework can be used for future research to model crop sensor data over time, and better understand the complex phenomena studied in precision agriculture. The Environmental Quality Incentives Program for Indiana with respect to controlled drainage subsidies amounts to a $40 per acre incentive payment capped at $2000 per year for up to three years, as well as a 50% subsidy of the average cost of purchase and installation of control structures. For three of the experimental farms (Davis, Site 1 and Site 3) the decision to invest in controlled drainage technology is supported both with and without subsidy, and the additional returns range from $7,856 to $33,591 over the lifetime of the drainage management control structures, with the yield increase due to controlled drainage ranging from 40 to 97 bushels per acre. At Site 2, drainage water management will not be adopted as it yields a loss in the range of $1,764 to $2,880 with and without subsidy, respectively. Regarding the implications for drainage research, the results of this study should be interpreted with caution. This dissertation is a method-oriented investigation of the applicability of spatial panel models to the problem of crop sensor yield response estimation, with a particular application focused on estimating the impact of controlled drainage on corn yields. The inferences cannot be generalized to all the fields in the Midwest or beyond, because the analysis focuses on within-field variations. The estimated models represented a tradeoff because of data availability and the computational requirements of the spatial panel models. Ideally, spatial panel models are based on space-time varying variables. The common lack of such data in a precision agriculture setting impeded an elaboration of a more proper model of yield response, in accordance with the agronomic literature. The Davis farm was an exception in this regard, and we can therefore place most confidence in the positive yield impact associated with drainage management from this case