|Sudduth, Kenneth - Ken|
Submitted to: Precision Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/25/2007
Publication Date: 11/20/2008
Citation: Sadler, E.J., Sudduth, K.A., Jones, J.W. 2008. Separating spatial and temporal sources of variation for model testing in precision agriculture. Precision Agriculture. 8:297-310.
Interpretive Summary: The information-intensive field of precision agriculture provides both a need for and application of process-based computer simulation models of crop growth and yield. To be truly useful in precision agriculture, models must be capable of simulating both year-to-year and place-to-place differences in yield. However, no real consensus has emerged on objective methods to evaluate model performance independently in time and space. Most researchers have attempted to evaluate performance of entire datasets, with time and space combined. This can provide misrepresentative indicators of model performance. Two methods to isolate and independently test model performance for time-based and space-based variation were developed. Analyses of two datasets illustrate the importance of separately testing them, and also show additional information that can be obtained when this is done. Ultimately, isolating sources of variation provides increased confidence in measures of model performance. The guidelines and methods included in this paper can be used by researchers to improve decision support for a wide range of modeling objectives in precision agriculture.
Technical Abstract: Applying crop simulation models to precision agriculture appears to be a matter of developing spatial suites of input parameters and running a model for each set. Extensive modeling literature has reported independent tests in multiple combinations of variety, soils, and climate, which has been generally considered substantial validation of the performance of the models. However, most literature tests evaluated model performance in simulating mean yields across multiple plots in yield trials, whereas precision agriculture challenges models to simulate not just the mean, but also spatial variation in yield. No real consensus has emerged regarding exactly how to test model performance, nor of what performance constitutes success. In addition, common measures of goodness of fit suffer from dependence on the range of variation in the independent variable. When multiple sources of variation, for example inter-annual and spatial, are combined in a test, commonly used performance measures may fail to support the hypotheses represented in a paper’s objectives. Our objectives are to examine objectives common in precision agriculture, to discuss expectations of model performance, and to compare several traditional and some alternative measures of model performance. These issues are illustrated with examples that illustrate the limitations and strengths of the performance measures. By synthesis and example, we provide guidelines and structure for future precision agriculture modeling efforts.