Modelling and measurement of soil moisture content based on a remote sensing method for applications in Semi-arid tropics.

Thumbnail Image



Journal Title

Journal ISSN

Volume Title


University of Bonn


Soil moisture plays a very crucial role in land surface processes It should therefore be monitored with the same accuracy and frequency as other important environmental variables Two approaches are used for estimation of soil moisture content, namely, modelling and measurement, either in-situ or remote. Integration of modelling and measurements may provide the best solution towards estimation of soil moisture content The utility of ground-based thermal infrared remote sensing method for the estimation of nearsurface soil water content was tested under tropical semi-arid agricultural conditions in Morogoro, Tanzania, East Africa Field experiments were conducted between January and August, 1997 at a bare soil site Regression relationships between the daily maximum surface soil temperature minus air temperature (TDMax) and weighted-average soil water contents to different depths in the soil profile were developed based on the measured data Better correlations were obtained for the top 0-5 and 0-15 cm layers of the soil, with coefficients of determination ofO 81 and 0.78, respectively Use of "Normalized TDMax" as well as cloudncsscover- correctcd "Normalized TDMax” (TDaMax) resulted in even better coefficients of determination (c g , 0 95 for the 5 cm depth) A physically based model of coupled flow of heat and water in the soil (SUAHEAT) was developed The SUAHEAT model was tested by comparing its simulated soil water contents and soil temperatures with those measured at the bare soil site Both qualitative as well as quantitative methods were used to evaluate the model performance, for the calibration and validation phases For the calibration phase, the average values of the mean absolute difference (MAD) of soil water content were 0.06, 0 05, 0 05, and 0.03 m7nP for the 5, 15, 30, and 45 cm depths, respectively The corresponding average values of the root mean square difference (RMSD) of soil water content were 0.07, 0 06. 0 05, and 0 03 m’/m’ for the same depths, respectively As for the validation phase, the average MAD values of soil water content were 0 09, 0.05, 0 08, and 0 17 nP/nP for the same depths, respectively. The corresponding average RMSD values for the same depths were 0 12. 0 08, 0 10, and 0 19 nP/nP, respectively The unusually large errors (at the 45 cm depth) during the validation phase could be attributed to the overestimation of soil water content values during very wet conditions arising from the calibration equation used. The performance of the model in the simulation of surface, near-surface, and profile soil temperatures was also both qualitatively and quantitatively evaluated In the calibration phase, the MAD values were 2 8, 1 1,0 5, and 0 3 °C for the 5, 15. 30, and 45 cm depths, respectively. The corresponding RMSD values for the same depths were 3 5, 1.3, 0.6, and 0.4 °C, respectively. For the validation phase, the MAD values for the same depths were 2.3, 1 2, 0 7, and 0.4 °C, respectively. The corresponding RMSD values for the same depths were 2.7, 1.4, 0.8, and 0.5 °C, respectively. Generally, the errors obtained with the use of the SUAHEAT model are comparable to values obtained by other researchers elsewhere using similar models. Integration of the model with remotely sensed surface soil water content and temperature data was effected through use of the initial profiles of soil water content and temperature. The initial profile data were derived from the regression relationships between TDMax and surface soil water content on the one hand, and between the surface and profile soil temperatures on the other, respectively. Model simulation results obtained using the remotely sensed initial conditions indicated that it is feasible to use remotely sensed data (one value of TDMax per day) to initialize the model Similarly, extrapolation of the remotely sensed near-surface (0-5 cm depth) soil water content to values at greater depths in the soil profile was shown to be feasible.



Remote Sensing, Soil Moisture, Measurement-Soil moisture, Modelling, Semi-arid tropics