Hydrological Applications of Multi-source Soil Moisture Products

  • Moonhyuk Kwon

Student thesis: Doctoral ThesisDoctor of Philosophy (PhD)

Abstract

Soil moisture (SM) is a key state variable in understanding the hydrologic processes, including runoff, infiltration, drought and crop growth. Thus, acquiring accurate SM information has been a priority in hydrology; however gathering such data remains a challenge for many areas of the world with respect to their spatio-temporal aspects. This challenge has contributed to the popularity of using complementary tools such as satellite retrievals and land surface models. This thesis examines multi-source SM data and further enhances their practical use in SM simulation.
First, the exponential filter method is introduced to estimate a root zone SM based on satellite surface SM. Subsequently, a cumulative distribution function matching approach is applied, not only to address the inevitable systematic biases between in-situ and satellite SM but also to determine an ideal temporal combination. The performance of each bias-correction group is validated through a cross-validation procedure. However, a major issue in using satellite SM data for practical applications is their coarse spatial resolution. Therefore, a multivariate stochastic SM estimation approach, based on a Gaussian-mixture nonstationary hidden Markov model, is introduced to spatially disaggregate the satellite SM data for multiple locations. It is revealed that the mean correlation coefficient of the proposed model is significantly greater than that of an ordinary regression model.
The second part of this thesis focuses on expanding the applicability of SM data in hydrological applications. I introduce a hybrid modelling framework by incorporating SM state variables obtained from the Tank model and multi-satellite sensors via a machine learning based regression technique. The enhanced performance of the hybrid model over a conventional model (the Tank model) is especially apparent in the simulation of low flows; this indicates that even though the overall contribution to runoff prediction is not significant, satellite SM products appear to help capture distinct features of the rainfall-runoff process. ERA-Interim SM data are then employed to identify the spatial and temporal characteristics of agricultural drought stemming from SM deficit. Furthermore, the copula-based Multivariate Standardized Drought Index is exploited to explicitly determine the interdependence and interaction between precipitation and SM deficiency.
Date of Award25 Jun 2019
Original languageEnglish
Awarding Institution
  • The University of Bristol
SupervisorDawei Han (Supervisor) & Theo Tryfonas (Supervisor)

Keywords

  • soil moisture
  • Remote sensing
  • bias correction
  • Spatial downscaling
  • Rainfall-runoff model
  • Support vector machine
  • Multivariate drought index
  • Clustering analysis

Cite this

'