Real-Time Model-Data Synthesis for Soil Moisture.

Soil water balance models that depend on an initial set of inputs are vulnerable to deviations from actual values, while models that also incorporate real-time available data have the potential to correct and improve the accuracy of soil moisture predictions. A framework with minimal user input and that encompasses real-time daily weather and soil moisture data under grassland, wheat fraction green canopy cover, and the use of a soil water balance model is proposed to forecast soil moisture in wheat cropland. Weather and actual soil moisture values are obtained from an environmental network of weather stations with more than 100 stations across the state of Oklahoma. To make use of these data, and since the soil moisture sensors of the weather stations are under grassland vegetation, a function that bridges soil moisture under grassland to cropland is used. The framework also uses the FAO-56 dual crop coefficient model to independently provide a second simulation of the soil moisture dynamics in wheat cropland. The function that links soil moisture from grassland to cropland has been developed assuming that the deviations from the mean plant available water (PAW) in grassland and cropland are proportional. This is supported by the fact that in 219 out of 366 days of the year soil moisture deviations from the mean PAW in grassland and cropland showed a positive linear relationship with r squared >0.75. Even though the model is constantly subjected to improvements, simulations using four winter wheat growing seasons resulted in RMSE values that ranged from 21.4 to 38.3 mm of PAW, considering an active root zone of 1.4 meters depth.