Skip to content | Change text size
Jeff Walker home

PhD Thesis

Estimating Soil Moisture Profile Dynamics From Near-Surface Soil Moisture Measurements and Standard Meteorological Data

Jeffrey Phillip Walker (B. Surv. (Hons. I), B. E. (Civil) (Hons. I))


An estimate of the spatial distribution and temporal variation of soil moisture content in the top few metres of the earth's surface is important for numerous environmental studies. Soil moisture content can be determined from: (i) point measurements; (ii) soil moisture models; and (iii) remote sensing. Only a limited area can be monitored with an adequate spatial and temporal resolution using the point measurement technique, while estimates from distributed soil moisture models are generally poor. This is due to an incomplete knowledge of model physics and the large spatial and temporal variation of soil moisture that results from heterogeneity in soil properties, vegetation and precipitation. Remote sensing can be used to collect spatial data over large areas on a routine basis, providing a capability to make frequent and spatially comprehensive measurements of the near-surface soil moisture content. However this technique is limited by an infrequent satellite repeat time and the shallow depth of the soil moisture measurements, consisting of the top few centimetres at most. These upper few centimetres of the soil are the most exposed to the atmosphere, and their soil moisture content varies rapidly in response to rainfall and evaporation.

This thesis overcomes the limitations of the above approaches for determining soil moisture, by linking a physical model of soil moisture movement in both the vertical and horizontal directions, with a data assimilation technique that uses near-surface soil moisture measurements. In this way, the near-surface soil moisture measurements are interpolated in space and time between satellite overpasses, and extrapolated over the soil profile depth. The point measurements of soil moisture profiles are used for calibration of the soil moisture forecasting model, and ongoing evaluation the soil moisture profile estimation from data assimilation.

To address the poor resolution in time of remote sensing data, a water balance approach is used to model soil moisture during the inter-observation period. Using this approach, the soil moisture hydrologic model is forced using estimates of evapotranspiration and precipitation from standard meteorological data. As observations of the near-surface soil moisture content become available, they are incorporated into the soil moisture model using an assimilation technique. This has required the development of a hydrologic model specifically designed to accept remote sensing data as input. In this thesis, a theoretical model is developed for estimating the satellite observation depth for active microwave observations. Moreover, a procedure is proposed for inferring the soil moisture profile over the observation depth, from active microwave remote sensing observations.

This thesis has compared the Dirichlet boundary condition, hard-updating and Kalman-filtering assimilation schemes for estimation of the soil moisture profile. Conclusions are reached for the efficiency of these assimilation schemes, the depth over which near-surface soil moisture measurements are required, and the effect of updating interval on soil moisture profile estimation. These questions are addressed initially by a one-dimensional Richards equation soil moisture forecasting model using synthetic data. The study has shown that the Kalman-filter is superior to the hard-updating and Dirichlet boundary condition assimilation schemes. It is has also shown that the observation depth did not have a significant effect on improving the soil moisture profile estimation when using the Kalman-filter assimilation scheme. Moreover, the Kalman-filter was less susceptible to unstable updates if volumetric soil moisture was modelled as the dependent state, rather than matric head.

While suitable for the one-dimensional problem, the Richards equation model was too computationally demanding for the distributed catchment application. Hence, a computationally efficient distributed soil moisture forecasting model for both vertical and lateral redistribution of soil moisture content, based on a conceptualisation of the Buckingham-Darcy equation, was developed. Moreover, the Kalman-filter assimilation scheme was too computationally demanding for forecasting of the model covariance matrix in a spatial application. To overcome this computational burden, a Modified Kalman-filter was developed, which forecast the model covariance matrix using a dynamics simplification approach.

Both the distributed soil moisture forecasting model and the Modified Kalman-filter have been applied to a field application at the "Nerrigundah" experimental catchment. While an application of the one-dimensional version of this simplified soil moisture model has evaluated the vertical redistribution component, the catchment application has evaluated the lateral redistribution component. Moreover, the usefulness of near-surface soil moisture measurements for updating of soil moisture models to improve the prediction of soil moisture content over the soil profile has been illustrated, showing that an improved estimate of the soil moisture profile was achieved.

Power Point Presentation of Dissertation

You may freely download the following pdf icon files of my thesis:

Preface (pdf 332kB) Chapter 5 (pdf 293kB) Chapter 10* (pdf 922kB) Appendix B* (pdf 2347kB)
Chapter 1 (pdf 88kB) Chapter 6* (pdf 2983kB) Chapter 11* (pdf 453kB) Appendix C (pdf 1451kB)
Chapter 2 (pdf 6190kB) Chapter 7* (pdf 763kB) Chapter 12 (pdf 96kB) Appendix D (pdf 166kB)
Chapter 3* (pdf 396kB) Chapter 8* (pdf 1444kB) References (pdf 178kB) Appendix E (pdf 237kB)
Chapter 4 (pdf 1553kB) Chapter 9* (pdf 31577kB) Appendix A (pdf 311kB) Appendix F (pdf 3061kB)

* These pdf files contain non-perfect reproductions of some figures.