Model calculated soil moisture or precipitation for yield modeling?

Corn yield is determined by a variety of physical (e.g. temperature, water availability), economic (e.g. crop prices), and management (e.g. production inputs, planting decisions) factors. Precipitation is often used as a proxy for water availability in corn yield models. In some instances, precipitation overestimates actual water availability. In other instances, precipitation underestimates water availability. The actual amount of water available to plants is determined by water in the soil. Although soil moisture is a more direct measure of water available to plants it is not often used in yield modeling. There are several reasons why soil moisture measurements or model estimates of soil moisture are not commonly used in yield models: 1) soil moisture is difficult and expensive to measure, 2) existing models for soil moisture require detailed soil properties and climatological inputs, 3) proxies for soil moisture are complicated to calculate and often require detailed inputs, and 4) existing soil moisture models operate at spatial scales which are inappropriate for yield modeling.

In this paper, we describe a process model for soil moisture which uses only basic climatological information (e.g. daily Tmax, Tmin, Prcp) and coarse resolution (e.g. 1 km) soil parameters to calculate soil moisture throughout the growing season for corn. The model estimated soil moisture is compiled relative to the phasic development of corn as determined by the CERES-Maize crop weather model. The calculated soil moisture variables are then used in conjunction with economic and management variables to estimate a multiple regression yield model for corn.

The accuracy of the model calculated soil moisture is limited by the simplicity of the inputs used by the process model. Indeed, when compared to actual observations of soil moisture for locations in Illinois, the model accounts for, at maximum, about 50% of the overall variation in soil moisture. However, within the framework of yield modeling, the soil moisture estimates outperform yield models developed using precipitation to proxy water availability. Tests of predictive accuracy are used to compare the competing models. Additionally, the sensitivity of the results to the soil parameter resolution is evaluated and found to be robust for the yield modeling in this analysis.