Home ITCPhD Defence Pei Zhang | Dynamics of Regional Soil Moisture and Soil Temperature Profile on the Tibetan Plateau: Observations Analysis and Modelling

PhD Defence Pei Zhang | Dynamics of Regional Soil Moisture and Soil Temperature Profile on the Tibetan Plateau: Observations Analysis and Modelling

Dynamics of Regional Soil Moisture and Soil Temperature Profile on the Tibetan Plateau: Observations Analysis and Modelling

The PhD defence of Pei Zhang will take place in the Waaier building of the University of Twente and can be followed by a live stream.
Live Stream

Pei Zhang is a PhD student in the Department of Water Resources. Promotor is prof.dr. Z. Su from the faculty ITC.

Soil moisture and soil temperature (SMST) are essential state variables in water, energy, and carbon exchanges within the soil–vegetation–atmosphere system. Quantifying the seasonal dynamics and trend changes of the SMST is important for understanding the response of hydrological cycle to meteorological variation and climate change. The Tibetan Plateau (TP), often referred to as the “Third Pole” and the “Water Tower of Asia”, has been undergoing climate changes since the early 1980s, characterized by surface warming, moisture increasing, solar dimming, and wind stilling. Previous studies have provided insights into how these changes could affect hydrology and thermal circulations over the TP. However, no research has yet examined the response of SMST to recent climate changes, primarily due to the absence of long-term SMST time series data spanning at least a decade and the deficiencies in satellite retrieval and model simulation under the complex geographical conditions of the TP.

The Tibetan Plateau observatory of plateau scale soil moisture and soil temperature (Tibet-Obs) was established in 2008 and has been operational for over ten years. This offers an opportunity for this research to analyse the dynamics of the regional SMST profile on the TP using the in situ measurements provided by Tibet-Obs and improved modelling based on ground observations. The Tibet-Obs consists of three regional-scale networks covering three typical climatic zones over the TP, namely the Maqu, Naqu, and Nagri (including Shiquanhe and Ali) networks, with a total of 26, 11, and 24 monitoring sites, respectively. However, the number of valid sites is decreasing over time due to the sensor damage or record interruption, resulting in limited continuous in situ measurements for extended periods. Hence, only SMST data from the Maqu and Shiquanhe networks are adopted in this thesis for dynamic analysis, considering both data quantity and time coverage. The two study areas are distinct and representative within the TP region, where the Maqu network is located in the eastern TP characterized by a cold-humid climate, grassland cover, and silt loam soil, while the Shiquanhe network is located in the western TP characterized by a cold-arid climate, desert terrain, and sandy soil.

At the first step, we aim to determine the uncertainties and accuracy of site-scale measurements when used as a reference for regional-scale data on the TP. To achieve this aim, an effective upscaling method should be selected to obtain the regional-scale SMST data from the site-scale measurements when using a limited number of valid sites. In Chapter 2, four spatial upscaling methods, namely arithmetic averaging, Voronoi diagram, time stability, and apparent thermal inertia, were compared using the surface SM data from the Maqu and Shiquanhe networks, respectively. The comparisons indicated that the arithmetic averaging method generally outperformed the others and was selected to generate the upscaled long-term SMST profile data in Chapter 3. The validation results demonstrated that the upscaled SM data achieved root mean square difference (RMSD) values of 0.024, 0.019, and 0.030 m3 m-3 at depths of 5, 20, and 40 cm for the Maqu network and 0.011, 0.009, and 0.010 m3 m-3 for the Shiquanhe network. Similarly, the upscaled ST data achieved RMSD values of 0.7, 0.2, and 0.3 ℃ at depths of 5, 20, and 40 cm for the Maqu network and 2.5, 1.4, and 1.8 ℃ for the Shiquanhe network.

At the second step, we aim to investigate the primary characteristics and dynamic trends of regional SMST across the TP regions in context of the climate change. Based on the regional-scale SMST dataset generated in first step, Chapter 3 presented an analysis of the seasonal variations and interannual trend changes in profile SMST dynamics in a 10-year period for the two hydro-meteorologically contrasting networks, in addition to the characteristics of freeze-thaw (F/T) cycles. The results revealed significant seasonality in the time series of both SM and ST at various depths within both network areas, with the amplitudes of their variations generally decreasing as soil depth increases. Furthermore, it can be observed that the seasonal SM variations in the cold-humid Maqu area are larger than those of the cold-arid Shiquanhe area, whereas the ST seasonality is generally stronger within the Shiquanhe area. In general, the deeper soil layers in both networks presented a later onset of freezing and earlier thawing and, thus, a shorter F/T duration in comparison to the surface layer. Additionally, Chapter 3 also showed the Mann-Kendall trend analysis on a SMST time series spanning approximately 10 years, regarding to separately the warm (from May to October) and cold seasons (from November to April), alongside the full year. The findings demonstrated that the influence of climate change on SMST dynamics was more evident in SM time series within the humid region, whereas it predominantly manifested in ST time series within the arid region.

At the third step, we aim to evaluate the performance of existing satellite- and model-based SMST products on the TP using the ground observations obtained in Chapter 3. To achieve this aim, Chapter 4 showed a comparative analysis of 12 satellite-based and 10 model-based SM products against regional SM data at depths of 5, 20, and 40 cm for the Maqu and Shiquanhe networks. Regarding to the satellite-based surface SM products, the results demonstrated that surface SM retrievals from C- and X-band brightness temperature (TB) measurements (i.e., from LPRM and JAXA AMSR2) generally showed poorer performance compared to those from L-band TB measurements (i.e., from SMAP and SMOS). Among all the products, SMAP_L3 provided the best surface SM retrievals for both networks, with the SCA-V and DCA products performing better in the arid and humid areas, respectively. As a result, the harmonized products produced based on the SMAP data (i.e., SMOSMAP and NNsm) provided enhanced accuracy compared to the SMOS-IC and LPRM/JAXA AMSR2 products. Regarding to the model-based profile SM products, the results indicated that SM estimates generally exhibited higher accuracy in the Maqu humid area compared to the Shiquanhe arid area, and performed better during the warm seasons in comparison to the cold seasons. Furthermore, Chapter 3 presented an assessment of five model-based products, namely ERA5, MERRA-2, GLDAS-2.1 CLSM, Noah, and VIC, against regional ST data at depths of 5, 20, and 40 cm for the Maqu and Shiquanhe networks. The results indicated that all five products underestimated the ST at each depth in both networks, and tended to produce earlier onset of freezing and a later start of thawing at all three depths, leading to a longer F/T duration than ground observations.

At the last step, we aim to improve the accuracy of model-based products on the TP while retaining the advantage of of satellite-based data. To achieve this aim, we pursued two different avenues building upon the assessment results in the third step. In Chapter 4, we explored the potential for enhancing the accuracy of the model-based products through the utilization of both the exponential filter (ExpF) and cumulative distribution function (CDF) matching methods in conjunction with high-performing satellite-based surface SM data. Specifically, the surface SM estimates from SoMo were rescaled using CDF matching method with SMAP_L3 surface SM data as the reference, and then the rescaled SoMo surface SM data was employed to estimate subsurface SM at depths of 20 and 40 cm using the ExpF method, resulting in a reduction of RMSD by 38% and 71% for the Maqu network, as well as 92% and 93% for the Shiquanhe network. In Chapter 5, we conducted three sets of numerical experiments using the Noah model with different soil property data, meteorological forcing data, and model parameterizations. The assessment results suggested an optimal scheme that involves utilizing the augmented Noah model with soil data sourced from in situ samples, driven by the CMFD meteorological forcing dataset. In the Maqu humid region, the scheme achieved a reduction of RMSD by 50% and 60% for SM simulations at depths of 5 and 20 cm compared to the initial Ctrl experiment, as well as 21% and 46% for ST simulations. Similarly, in the Shiquanhe arid region, the scheme achieved a reduction of RMSD by 84% and 88% for SM simulations at depths of 5 and 20 cm compared to the initial Ctrl experiment, as well as 47% and 54% for ST simulations.

Overall, this research generated a reliable SMST dataset by utilizing in situ measurements, to describe the dynamics of the SMST profile on the TP for an long-term period. Furthermore, two effective avenues for improving SMST estimates were developed and evaluated over the TP regions. Future study could focus on extending the improved SMST estimates to cover the entire TP, providing a comprehensive insight into the climatic and hydrological changes of the TP.