Optimization of agricultural field workability predictions for improved risk management

Improved prediction of the days available for field work, or field working days (FWDs), is an important consideration for adapting farming systems to increased weather variability. We developed modeling approaches to estimate robust soil moisture thresholds for FWDs. We used simulated soil moisture to train the model on the same type of data that would be used for FWD forecasting (prediction). These new models were tested against previously suggested thresholds for field workability. Model 1 used historical field work and weather records from three crop research centers in a logistic regression model. A soil moisture threshold of1.10 times the plastic limit (1.10PL) was identified. Model 2 identified statewide soil moisture and temperature thresholds by optimizing the root mean square error of the predicted number ofweekly statewide FWDs across a 52-yr data set. The resulting thresholds of either 0.88PLor 0.73FC (field capacity) and an average temperature requirement of at least 6°C yielded statistically smaller absolute errors for the state average FWDs and in eight of the nine crop reporting districts. The Model 2 thresholds also eliminated systematic overprediction present in previous thresholds. These results demonstrate that immutable theoretical thresholds for FWDs based on field-measured soil moisture can be suboptimal for prediction at a larger spatial scale due to consistent bias.