Assimilating in situ soil moisture measurements into the DSSAT-CSM using a Kalman filter

Abstract

With the ability to monitor soil moisture in time comes the opportunity to develop ways to incorporate these measurements into predictive models, without compromising or overriding the model physics. The importance of soil moisture to the growth of crops is well understood and because of this it is recognized as one of the more important parts of crop modeling programs. This research focused on improvements to the Decision Support System for Agrotechnology Transfer Cropping System Model (DSSAT-CSM) based on the accuracy of soil moisture estimates. To accomplish this, data assimilation techniques were implemented to process the uncertainty of the model related to state variables and the uncertainty found within in situ soil moisture measurements. Consideration of soil parameter uncertainty, which influences model estimates of soil moisture and model output, was taken into account using a Monte Carlo approach. A Kalman filter was used to combine the model estimates of soil moisture with in situ soil moisture measurements, while varying several important soil parameters in the model using a Monte Carlo approach. Covariances for the Kalman filter were calculated for the model and measurements based on the models standard deviation of soil moisture estimates and the standard deviation of the in situ soil moisture measurements. Data for this study was obtained from a research study conducted on irrigated wheat during the winters of 2003-04 and 2004-05 in Maricopa, Arizona in which thorough field and crop data were collected. The uncertainty of soil parameters was only moderately captured by the Monte Carlo approach for assimilation into the top layer of the soil profile. Improvement resulted for data assimilation of soil moisture through the reduction of the error between the measured and simulated