Assessment Of Satellite-Based Microwave Soil Moisture And Hydrological Data Assimilation

With the global climate change and population explosion, extreme events are more likely to occur, such as droughts, floods, and heat waves. Globally, the climate system and hydrological cycle of the Earth are extremely large and complex. Soil moisture, as one of the key state variable to determine the coupling system between land and atmospheric surfaces, its spatial and temporal characteristics play an important role in improving the understandings of global climate system. Thanks to the development of satellite-based sensors and computer technology, the task of detecting and retrieving various climate variables can be effectively done using the satellite remote sensing data. Specifically, active and passive microwave remote sensing have been extensively utilized for observing global-scale surface soil moisture. Despite the fact that multiple global satellite-based microwave soil moisture observations can provide a nearly forty-years long time series data, their qualities are however miscellaneous. Thus, an accurate evaluation and spatial-temporal characterization of multiple satellite-derived microwave soil moisture is of vital importance. Meanwhile, the observations can only provide instantaneous values, which might not be effective in studying the hydrological processes and forecasting applications. To address the problems mentioned above, this thesis aims to propose a theoretical framework to evaluate the accuracy of satellite-based surface soil moisture and perform effective hydrological data assimilation. The main contributions are summarized as follows:Based on the data assimilation framework, the state-of-the-art of microwave soil moisture retrievals, hydrological numerical models, and data assimilation algorithms have been summarized in a systematic way. By further insightful analysis of the advantages and shortcomings of their current practical applications, the research goal and tasks of this thesis are proposed.Multiple satellite-based microwave sensors can provide numerous observations of soil moisture. Typically, these satellite are orbiting in a sun-synchronous orbit and can provide two observations daily through AM and PM overpasses. However, the applications have been mainly focusing on the observations from the night time and/or early morning. The evaluation of multiple satellite-based microwave soil moisture has been conducted with the acquisition time as a dividing criterion, including active and passive microwave. The differences of thermal conditions between land surface and vegetation canopy as well as the vegetation water content-existed with different acquisition times-are investigated. To be specific, the mainland of United States is selected as the study region. Soil moisture estimates generated from active and passive microwave sensors and land surface model are collocated together, and the impacts of acquisition time on the observational errors have been investigated via a large-scale evaluation technique, namely, triple collocation. Furthermore, a direct comparison against in-situ measurements have been applied for cross-validation. Through two independent evaluation methods, the relationship between satellite-based microwave soil moisture observational errors and acquisition time, sensor technology, land cover, and retrieval algorithms have been investigated. The spatial distribution of observation accuracy has also been demonstrated under different land surface conditions.The satellite-based soil moisture observational errors are usually assumed to follow Gaussian distribution during validations and applications, such as the hydrological data assimilation. However, the soil moisture retrievals may have also periodic errors as a result of the land surface spatial heterogeneity and the periodic sampling of satellite swaths. Therefore, the physical mechanism of periodic errors in passive microwave satellite-based soil moisture has been analyzed in the spectral domain. Synthetic experiment, site validation, and global assessment have been conducted to investigate the reasons of the occurrence of periodic errors. The in-situ soil moisture measurements and high spatial resolution land cover are used to study the relationship between periodic error and the land surface spatial heterogeneity. By using the longest continuously recorded satellite-based soil moisture product, a periodogram method is utilized to investigate the power spectral density of the high frequency and low frequency regimes of soil moisture. With the proposed high frequency or peak detection method, the global spatial distribution map of periodic error can be obtained. Furthermore, by using directly observed brightness temperature (TB) data from the satellite, the derived-TB parameters are calculated. Finally, a spatial heterogeneity index is proposed to characterize the land surface spatial variability, and to analyze the areas where periodic errors exist.Through combining the multi-source observations together with physical hydrologic models, the hydrologic data assimilation can alleviate the uncertainties of heterogeneous data from various sources, yielding land surface hydrologic state estimations and predictions that are consistent in physics and have continuous spatial-temporal characteristics. Though the ensemble Kalman filter and its variants have been extensively explored in data assimilation, their practical applications are still ongoing. In this thesis, the Babaohe basin located in upper Heihe watershed in China is chosen as the study area. Specifically, the geophysical and meteorological forcing inputs are gathered, then the sensitive parameters of the hydrologic model are calibrated using streamflow observation, yielding a base model for the hydrologic data assimilation. Focusing on solving the problems encountered in data assimilation of ensemble Kalman smoother and complex hydrological model, a semi-distributed hydrological model is selected. The inflation factor and localization methods have been included to improve the estimation of background error covariance, resulting in enhanced capability of the ensemble Kalman smoother in handling large-scale and high-dimensional systems. Through the assimilation of surface soil moisture observations, soil moisture estimations within deeper and root-zone soil layers as well as key hydrologic prognostic variables are improved. Finally, the impacts of spatial heterogeneous input data and parameter and the enhanced data assimilation algorithm on the outcomes of data assimilation have been investigated, including precipitation, soil type, land cover information and etc.