Publications

Soil moisture (SM) is a critical climate variable, and assimilating satellite SM into land surface models via land data assimilation (LDA) enhances continuous SM modeling and climate extreme monitoring. However, LDA often assumes static SM error dynamics, limiting accuracy. This study introduces a novel framework integrating triple collocation analysis (TCA) with machine learning (ML), including Light Gradient Boosting Machine (LGBM) and Deep Neural Networks (DNN), to quantify spatially and temporally continuous satellite-based SM errors globally.

Using only Soil Moisture Active Passive (SMAP) retrieval data, our TCA-based time-variant error prediction models successfully recover error information in regions where SMAP SM error data were previously unreliable. The SMAP-based model outperforms models relying on external datasets like the Global Land Data Assimilation System (GLDAS), offering a robust approach for improving LDA in data-scarce areas. This method also extends to other satellite-based geophysical datasets, broadening its applicability beyond SMAP.

‍

This study explores the potential of the Cyclone Global Navigation Satellite System (CYGNSS) mission for soil moisture (SM) monitoring, highlighting its ability to capture fine-scale SM variability through high revisit frequencies at sub-daily intervals. While CYGNSS was originally designed for tropical cyclone monitoring, its seven-year data record demonstrates significant promise in reliably monitoring diurnal SM dynamics using spaceborne L-band bistatic radar and GNSS-Reflectometry (GNSS-R) technology.

Despite this potential, SM retrieval from CYGNSS remains limited by knowledge gaps and unique challenges tied to its technical design. This study addresses these gaps by analyzing CYGNSS real-world data, synthesizing recent advancements in mitigating external uncertainties, and improving SM inversion techniques. Notably, algorithm-related challenges include accurate partitioning of coherent and incoherent signal components and correcting attenuation effects caused by vegetation and surface roughness. Data-related challenges involve variations in CYGNSS spatial footprint, temporal frequency, signal penetration depth, and incidence angle changes, as well as excessive reliance on reference SM datasets for model calibration, validation, and training. The computational demands of processing CYGNSS’s rapid multi-sampling data further complicate its operational use.

Future research directions identified in this study focus on leveraging machine learning and deep learning approaches to improve CYGNSS SM data quality and quantity. Additionally, assimilating CYGNSS-derived SM data into physical models offers promising opportunities to enhance predictions of hydroclimatic variables and extreme climate events, addressing key challenges in water resource monitoring and climate resilience.

This paper provides a novel quantitative assessment of the transformation of daily precipitation into terrestrial water storage across 121 global river basins over a two-decade period. The study introduces the average daily fraction of precipitation transformed into terrestrial water storage, leveraging enhanced terrestrial water storage statistical reconstruction and water storage data from the Gravity Recovery and Climate Experiment (GRACE) and its follow-on mission. We reveal that approximately 64% of land precipitation contributes to terrestrial water storage, with notable variations across different climatic and geographical regions. These findings offer critical insights into the interactions between precipitation, land surface processes, and climate change, providing valuable implications for hydrological modeling and future water resource management.

This study examines the impact of assimilating Soil Moisture Active Passive (SMAP) data into the Korean Integrated Model (KIM) to improve global soil moisture estimates and weather forecasts. SMAP soil moisture retrievals are integrated into the Noah land surface model using the ensemble Kalman filter through NASA’s Land Information System. Experiments from March to July 2022 show that assimilating SMAP data, particularly with anomaly-based bias correction, significantly enhances soil moisture estimates and improves weather forecasts, especially in northern Africa and West

‍

This research discusses the retrieval and validation of soil moisture data from the Advanced Scatterometer (ASCAT), focusing on its applications and challenges. ASCAT data, crucial for various uses like weather prediction and drought monitoring, are compared with similar data from other satellite missions. Validations using multiple datasets highlight dependencies on land cover and vegetation, revealing unexpected quality variations across different environments. A notable issue identified is subsurface scattering, often misattributed to other factors, impacting ASCAT data quality significantly. The study recommends masking affected data using developed indicators and masks to enhance accuracy in practical applications.

This study evaluates the performance of two L-band passive microwave satellite sensors, the Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP), in monitoring surface soil moisture (SM). The newly developed fused SM product (Fused-IB) derived from SMOS and SMAP observations is compared with the enhanced SMAP-L3 (SMAP-E) SM product against in situ SM data from the International Soil Moisture Network (ISMN) over the period 2016-2020. The comparison considers overall and seasonal performance, focusing on different land use and land cover (LULC) types. Results show that Fused-IB generally outperforms SMAP-E, especially in forested areas, due to the robust SMAP-IB algorithm and higher data coverage. Both products demonstrate higher accuracy in summer and autumn, but face increased uncertainties in forests, grasslands, and croplands during spring and winter due to vegetation growth and rainfall. This study highlights the importance of accounting for seasonal and eco-hydrological factors to improve the accuracy of SM retrieval algorithms.