Soil Moisture Modelling, Retrieval From Microwave Remote Sensing And Assimilation In A Tropical Watershed
The knowledge of soil moisture is of pronounced importance in various applications e.g. flood control, agricultural production and effective water resources management. These applications require the knowledge of spatial and temporal variation of the soil moisture in the watershed. There are three approaches of estimating/measuring soil moisture namely,(i) in-situ measurements,(ii) remote sensing, and(iii) hydrological modelling. The in situ techniques of measurement provide relatively accurate information at point scale but are not feasible to gather in large numbers relevant for a watershed. The soil moisture can be simulated by hydrological models at the desired spatial and temporal resolution, but these simulations would often be affected by the uncertainties in the model physics, parameters, forcing, initial and boundary conditions. The remote sensing provides an alternative to retrieve the soil moisture of the surface (top few centimeters ) layer, but even this data is limited by the spatial or temporal resolution, which is satellite dependant. Hydrological models could be improved by assimilating remotely sensed soil moisture, which requires a retrieval algorithm. In order to develop a retrieval algorithm the satellite data need to be calibrated/validated with the in-situ ground measurements. The retrieval of surface soil moisture from microwave remote sensing is sensitive to surface conditions, and hence requires calibration/validation specific to a site/region. The improvement in the hydrological variables/fluxes is sensitive to the framework adopted during the assimilation of remotely sensed data. The main focus of the study was to assess the retrieval algorithm for the surface soil moisture from both active (ENVISAT,RADARSAT-2)and passive(AMSR-E) microwave satellites in a semi-arid tropical watershed of South India. Further, the usefulness of these retrieved remotely sensed products for the estimation of recharge was investigated by developing a coupled hydrological model and an assimilation framework. A brief introduction was made in Chapter 1 on the importance of surface soil moisture and evapotranspiration in hydrology, and the feasible options available for the retrieval from microwave remote sensing. A detailed review of the literature is presented in Chapter 2 to establish the state-of-the-art on the following:(i) retrieval algorithms for the surface soil moisture from active and passive microwave remote sensing,(ii) estimation of actual evapotranspiration from optical remote sensing(MODIS),(iii) coupled surface-ground water hydrological models,(iv) estimation of soil hydraulic properties with their uncertainties, and(v) assimilation framework specific to hydrological modelling. To calibrate/validate the retrieval algorithms and to test the coupled model and the assimilation framework developed, field measurements were carried out in the BerambadI experimental watershed located in the Kabini river basin. The surface soil moisture in 50 field plots, profile soil moisture up to 1m depth in 20 field plots, and ground water level in 200 bore wells were measured. Twelve images of ENVISAT, seven teen images of RADARSAT-2, along with AMSR-E and MODIS data were used. These data pertained to different durations during the period 2008 to 2011,the details of which are given in Chapter 3. The approach for the retrieval of surface soil moisture and the associated uncertainty from active and passive microwave remote sensing is given in Chapter 4. Surface soil moisture was retrieved for six vegetation classes using the linear regression model and copulas. Three types of copulas(Clayton, Frank and Gumbel) were investigated. It was found that the ensemble mean simulated using the linear regression model and three copulas was nearly same. The copulas were found to be superior than the linear regression model when comparing the distributions of the mean of the generated ensemble. Among the copulas it was observed that the Clayton copula performed better in the lower and middle ranges of backscatter coefficient, while the Gumbel and Frank copulas were found to be superior in the upper ranges of backscatter coefficients. The range of RMSE was approximatively 4cm3cm−3 indicating that the retrieval from ENVISAT/RADARSAT-2 was good. ACDF based approach was proposed to retrieve the surface soil moisture map for the watershed with a spatial resolution of 100m x 100m ( i.e one hectare). The map of the uncertainty in the retrieved surface soil moisture was also prepared using the Clayton copula. The AMSR-E surface soil moisture product was calibrated for the watershed during the period 2008 to 2011, using the map generated from the ENVISAT/RADARSAT data. They Clayton copula was used to generate the ensemble of the corrected AMSR-E surface soil moisture. The standard deviation of the generated ensemble varied from 0.01 to 0.03cm3cm−3 ,hence the derived surface soil moisture product for Berambadi was found to be good. In the Chapter 5, a one dimensional soil moisture model was developed based on the numerical solution of the Richards’ equation using finite difference method and inverse modeling was carried out using the Generalized Likelihood Uncertainty Estimation(GLUE) approach for estimating the soil hydraulic parameters of the van Genuchten(VG) model and their uncertainty. The parameters were estimated from the two field sites(Berambadi and Wailapally watershed in South India) and from laboratory evaporation experiment for the Wailapally site. It was found that the GLUE approach was able to provide good uncertainty bounds for the soil hydraulic parameters. The uncertainty in the estimates from the field experiment was found to be higher than from the laboratory evaporation experiment for both water retention and hydraulic conductivity curves. The saturated soil moisture(θs )and shape parameter (n) of VG model estimated from the laboratory evaporation and field experiment were found to be the same, and further more they showed a lower uncertainty from both the experiments. Moreover, the residual soil moisture (θr), inverse of capillary fringe thickness (α) and saturated hydraulic conductivity( KS) showed a relatively higher uncertainty. In the Berambadi watershed ,the inverse modeling was performed in three bare field plots, and it was found that ﬁeld plots which had higher θs showed a relatively higher actual evapotranspiration (AET) and lower potential recharge. In Chapter 6, the retrieval of profile soil moisture up to 2m by assimilation of surface soil moisture was investigated by performing synthetic experiments on six soil types. The measured surface soil moisture over top 5cm depth was assimilated into the one dimensional soil moisture model to retrieve the profile soil moisture. Even though the assimilation of surface soil moisture helped in improving the profile soil moisture for the six soil types, the bias was observed. To reduce the bias, pseudo observations of profile soil moisture were generated and used in addition to the surface soil moisture in the assimilation altogether. These pseudo observations were generated using the linear relationship existing between the surface and profile soil moisture. A significant bias reduction was found to be feasible by using this method when pseudo observations beyond 75cm depth were used then there was no significant improvement. A coupled surface-ground water model was developed, which had 5 layers for the vadose zone and one layer for the ground water zone, in order to consider the major hydrological processes from ground surface to ground water table in a semi-arid watershed. The details of the coupled model were described in Chapter 7. The major aim of this model was to be able to use remotely sensed data of surface soil moisture and evapotranspiration to simulate recharge. The model was tested by applying in a lumped framework to the field data set in the Berambadi watershed for the year 2010 to 2011. The performance of the model was evaluated with the measured watershed average root zone soil moisture and ground water levels. The watershed average root zone soil moisture was obtained by averaging the field measurements from 20 plots and average ground water level was obtained by averaging the field measurement from 200 bore wells. In order to assimilate the AET into the coupled model, the daily AET at a spatial resolution of 1km was estimated from MODIS data. The AET was validated in one forested and four agricultural sites in the watershed. The validation was based on the comparison with AET simulated from water balance models. For agricultural plots the STICS (crop model) and for the forested site the COMFORT (hydrological) model were used. The AET from the MODIS showed a reasonably good match with both the forested and agricultural plots at the annual scale (for the crop model approximately 4-5 months). Model simulations were carried out with and without assimilating the remotely sensed data and the performance was evaluated. It was found that the assimilation helped in capturing the trends in deeper layer soil moisture and groundwater level. At the end, in Chapter 8 the major conclusions drawn from the various chapters are summarized.
- Civil Engineering (CiE) 
Showing items related by title, author, creator and subject.
Estimation of Root Zone Soil Hydraulic Properties by Inversion of a Crop Model using Ground or Microwave Remote Sensing Observations Sreelash, K (2018-02-09)Good estimates of soil hydraulic parameters and their distribution in a catchment is essential for crop and hydrological models. Measurements of soil properties by experimental methods are expensive and often time consuming, ...
Parate, Harshad Rameshwar (2017-10-18)The vadose zone is the unsaturated zone between the ground surface and water table. This zone is of much importance as it acts as a link between surface water and ground water. Knowledge of soil moisture in this zone is ...
Ghosh, Rohit (2018-04-06)Interannual variation of Indian summer (June-September: JJAS) monsoon rainfall (ISMR) depends on its relative intensity during early (June-July: JJ; contribution 52%) and late (August-September: AS; contribution 49%) phases. ...