Saturday, 15 July 2006

Space-time Kalman Filtering of Soil Redistribution.

Gerard B.M. Heuvelink1, Jeroen M. Schoorl1, Tom Veldkamp1, and Dan J. Pennock2. (1) Wageningen University and Research Centre, P.O. Box 37, Wageningen, 6700 AA, Netherlands, (2) University of Saskatchewan, 51 Campus Drive, Saskatoon, SK S7N 5A8, Canada

Soil redistribution is the net result of erosion and sedimentation. Assessment of soil redistribution in a given landscape over a given period of time may be done using process-based and empirical approaches. Process-based approaches rely on knowledge of how environmental processes acting in the landscape cause soil to move from one place to another. Empirical approaches rely on observations of soil redistribution, which may be interpolated in space and time using (geo)statistical methods. In this presentation we use space-time Kalman filtering to combine these two basic approaches. The Kalman filter operates recursively to predict forward one step at a time the soil redistribution from the predicted soil redistribution at the previous time and the observations at the current time. The case study that is used to illustrate the methodology concerns a seven hectare part of the Hepburn research site, located on the hummocky till plains of Saskatchewan, Canada. Tillage erosion causes soil to move downward along the steepest gradient, whereby the amount of soil loss per year is assumed linearly related to slope angle. Observations of cumulative soil redistribution from 1963 to 2000 were derived using cesium 137 as a tracer. In total 99 observations were taken, using a regular sampling design with a grid mesh of 25 m. The soil redistribution observations differed meaningfully from the deterministic model predictions (R2=0.389), causing the Kalman filter to make a marked adjustment to the soil redistribution map. The adjustment was particularly strong along the transportation route near the measurement locations. Use of the space-time Kalman filter to predict soil redistribution is attractive because it makes optimum use of process knowledge and observations, but routine use of the technique is hampered by the computational load and by parameterisation problems. Sensitivity analyses showed that the model results are most sensitive to the system noise. Future research must therefore be directed to realistic assessment of the errors inflicted by the assumptions and simplifications of the soil redistribution model.

Back to 3.2A Environmental Impacts of Soil Erosion - Measuring and Modeling On- and Off-Site Damages of Soil Erosion - Poster
Back to WCSS

Back to The 18th World Congress of Soil Science (July 9-15, 2006)