Field Sampling for Agricultural Model Input and Parameterization.

Agricultural system models can adequately reflect the average field situation if model parameters, boundary conditions and measurements of state variables and functional properties are included in an appropriate quality and quantity. Only a limited number of studies have been focused on the ability of a model to represent site-specific processes beyond capturing the relevant field-average agronomic processes properly. Experimentalists and model developers can benefit from each other if both communicate opportunities and needs for model input to support site-specific estimations. Experimentally, it is necessary to measure relevant spatially continuous changes of soil properties and their range of representativity. Once this range is known, spatial soil processes and their relationships can be identified through auto- and cross-covariance analysis. The next step is to quantify the sensitivity of a model, e.g., a crop growth model that integrates soil dynamics and biomass development processes, and the specific input requirements to capture final crop yield patterns. The aim is to derive some consistent soil and crop state variables that reflect spatial variability and substantially support crop growth prediction and its behavior in space. Among the various soil properties, spatial soil textural distribution is an important state variable. Spatial and temporal development of soil water content indicates how well the model captures highly important soil processes affecting plant growth and nutrient transport. The inclusion of remote sensing observations of the normalized difference vegetation index (NDVI) and other indices (leaf area index LAI) for estimating relative crop yield distribution in space has been evaluated in statistical models. Results show that through the integration of past developments, biomass indices improve the prediction of spatial crop yield distribution.