Ph.D. Thesis: Hydrometeorological Approach for Probabilistic Soil Moisture Simulation using Climate Inputs Sarit Kumar Das (10CE90P01)
ABSTRACT
Soil moisture is an important factor in terrestrial water balance and partitioning of the heat flux between sensible and latent fluxes within the near-surface boundary layer of the atmosphere. The redistribution of precipitation pattern and increase in near-surface air temperature as a result of global warming are causing significant change in soil moisture distribution across the globe. In this thesis, a probabilistic framework for soil moisture simulation, along with associated uncertainty, utilizing recently available hydrometeorological information relevant to its spatio-temporal variation is attempted for monthly and weekly scales. Simulation models are developed using either only precipitation or temperature data or a set of multiple climate variables. In case of multivariate simulation, curse of dimensionality is the major challenge towards the development of a multivariate probabilistic model. A supervised Principal Component Analysis (SPCA) technique is employed to extract the collective hydrometeorological forcing in such a way that will ensure the maximum association with the specific target variable while carrying out the dimensionality reduction. The derived collective hydrometeorological forcing is termed as Combined Hydrometeorological (CHM) index. The association between in-situ soil moisture and the CHM index is investigated and modeled at different temporal scales, e.g. monthly, weekly, etc. using the theory of copulas and its advanced variations, like, dynamic copula.
The station-wise models are transformed into the soil classification (e.g. soil textural class, Hydrologic Soil Group, HSG) based models using the Leave-One station-Out Cross Validation (LOO-CV) Scheme. The output of the simulation model is validated using output of existing simulation models and remotely sensed soil moisture data set. However, availability of the uncertainty estimates along with the simulated soil moisture through the proposed approach is valuable information for many fields of study. Finally, a simulated soil moisture map is prepared based on precipitation data adjusted for missing points using spring metaphor spatial interpolation.
Keywords:
Soil Moisture, Probabilistic simulation, Uncertainty Estimation, Hydrometeorology, Copula, Dynamic Copula, LOO-CV, HSG, Spring Metaphor