Shunlin Liang’s research while affiliated with The University of Hong Kong and other places

Soil moisture (SM) data records longer than 30 years are critical for climate change research and various applications. However, only a few such long-term global SM datasets exist, and they often suffer from large biases, low spatial resolution, or spatiotemporal incompleteness. Here, we generated a consistent and seamless global SM product from 1982 to 2021 using deep learning (DL) by integrating four decades of Advanced Very High Resolution Radiometer (AVHRR) albedo and land surface temperature products with multi-source datasets. Considering the temporal autocorrelation of SM, we explored two types of DL models that are adept at processing sequential data, including three long short-term memory (LSTM)-based models, i.e., the basic LSTM, Bidirectional LSTM (Bi-LSTM), and Attention-based LSTM (AtLSTM), as well as a Transformer model. We also compared the performance of the DL models with the tree-based eXtreme Gradient Boosting (XGBoost) model, known for its high efficiency and accuracy. Our results show that all four DL models outperformed the benchmark XGBoost model, particularly at high SM levels (> 0.4 m3 m-3). The AtLSTM model achieved the highest accuracy on the test set, with a coefficient of determination (R2) of 0.987 and root mean square error (RMSE) of 0.011 m3 m-3. These results suggest that utilizing temporal information as well as adding an attention module can effectively enhance the estimation accuracy of SM. Subsequent analysis of attention weights revealed that the AtLSTM model could automatically learn the necessary temporal information from adjacent positions in the sequence, which is critical for accurate SM estimation. The best-performing AtLSTM model was then adopted to produce a four-decade seamless global SM dataset at 5 km spatial resolution, denoted as the GLASS-AVHRR SM product. Validation of the GLASS-AVHRR SM product using 45 independent International Soil Moisture Network (ISMN) stations prior to 2000 yielded a median correlation coefficient (R) of 0.73 and unbiased RMSE (ubRMSE) of 0.041 m3 m-3. When validated against SM datasets from three post-2000 field-scale COsmic-ray Soil Moisture Observing System (COSMOS) networks, the median R values ranged from 0.63 to 0.79, and the median ubRMSE values ranged from 0.044 to 0.065 m3 m-3. Further validation across 22 upscaled 9 km Soil Moisture Active Passive (SMAP) core validation sites indicated that it could well capture the temporal variations in measured SM and remained unaffected by the large wet biases present in the input European reanalysis (ERA5-Land) SM product. Moreover, characterized by complete spatial coverage and low biases, this four-decade, 5 km GLASS-AVHRR SM product exhibited high spatial and temporal consistency with the 1 km GLASS-MODIS SM product, and contained much richer spatial details than both the long-term ERA5-Land SM product (0.1°) and European Space Agency Climate Change Initiative combined SM product (0.25°). The annual average GLASS-AVHRR SM dataset from 1982 to 2021 is available at https://doi.org/10.5281/zenodo.14198201 (Zhang et al., 2024), and the complete product can be freely downloaded from https://glass.hku.hk/casual/GLASS_AVHRR_SM/.

Grazing is a significant anthropogenic disturbance to grasslands, impacting their function and composition, and affecting carbon budgets and greenhouse gas emissions. However, accurate evaluations of grazing impacts are limited by the absence of long-term high-resolution grazing intensity data (i.e., the number of livestock per unit area). This study utilized census livestock data and a satellite-based vegetation index to develop the first Long-term High-resolution Grazing Intensity (LHGI) dataset of grassland in seven pastoral provinces in western China from 1980 to 2022. The LHGI dataset effectively captured spatial variations in grazing intensity, with validation at 73 sites showing a correlation coefficient (R²) of 0.78. The county-level validation showed an averaged R² values of 0.73 ± 0.03 from 1980 to 2022. This dataset serves as a vital resource for estimating grassland carbon cycling and livestock system CH4 emissions, as well as contributing to grassland management.

Leaf Area Index (LAI) is a critical parameter in terrestrial ecosystems, with high spatial resolution data being extensively utilized in various research studies. However, LAI data under future scenarios are typically only available at 1° or coarser spatial resolutions. In this study, we generated a dataset of 0.05° LAI (F0.05D-LAI) from 1983–2100 in a high spatial resolution using the LAI Downscaling Network (LAIDN) model driven by inputs including air temperature, relative humidity, precipitation, and topography data. The dataset spans the historical period (1983–2014) and future scenarios (2015–2100, including SSP-126, SSP-245, SSP-370, and SSP-585) with a monthly interval. It achieves high accuracy (R² = 0.887, RMSE = 0.340) and captures fine spatial details across various climate zones and terrain types, indicating a slightly greening trend under future scenarios. F0.05D-LAI is the first high-resolution LAI dataset and reveals the potential vegetation variation under future scenarios in China, which benefits vegetation studies and model development in earth and environmental sciences across present and future periods.

Estimating gross primary production (GPP) of terrestrial ecosystems is important for understanding the terrestrial carbon cycle. However, existed nationwide GPP datasets are primarily driven by coarse spatial resolutions (≥500 m) remotely sensed data, which fails to capture the spatial heterogeneity of GPP across different ecosystem types at land surface. This paper introduces a new GPP dataset, Hi-GLASS GPP v1, with a fine spatial resolution (30-m) and monthly temporal resolution from 2016 to 2020 in China. The Hi-GLASS GPP v1 dataset is generated from 30-m Landsat data using a process based light use efficiency model. The Hi-GLASS GPP v1 model integrates a detailed map of maize plantations, a crucial C4 crop in China known for its higher photosynthetic efficiency compared to C3 crops. This inclusion helps correct the underestimation of GPP that typically occurs when all croplands are categorized as C3. The Hi-GLASS GPP v1 dataset demonstrates a robust correlation with GPP data derived from eddy covariance towers, thereby enabling a more accurate assessment of terrestrial carbon sequestration across China.

Satellite-derived vegetation indices (VIs) have been extensively used in monitoring vegetation dynamics at local, regional, and global scales. While numerous studies have explored various factors influencing VIs, a remarkable knowledge gap persists concerning their applicability in mountain areas with complex topographic variations. Here we bridge this gap by conducting a comprehensive evaluation of the topographic effects on ten widely used VIs. We used three evaluation strategies, including: (i) an analytic radiative transfer model; (ii) a 3D ray-tracing radiative transfer model; and (iii) Moderate Resolution Imaging Spectroradiometer (MODIS) products. The two radiative transfer models provided theoretical evaluation results under specific terrain conditions, aiding in the first exploration of the interactions of both shadow and spatial scale effects on VIs. The MODIS-based evaluation quantified the discrepancies in VIs between MODIS-Terra and MODIS-Aqua over flat and rugged terrains, providing new insights into real satellite data across different temporal scales (i.e., from daily to multiple years). Our evaluation results were consistent across these three strategies, revealing three key findings. (i) The normalized difference vegetation index (NDVI) generally outperformed the other VIs, yet all VIs did not perform well in shadow areas (e.g., with a mean relative error (MRE) of 14.7% for NDVI in non-shadow areas and 26.1% in shadow areas). (ii) The topographic impacts exist at multiple spatiotemporal scales. For example, the MREs of NDVI reached 28.5% and 11.1% at 30 m and 3 km resolutions, respectively. The quarterly and annual VIs deviations between MODIS-Terra and MODIS-Aqua also increased with slope. (iii) We found the topography-induced interannual variations in multiple VIs both in simulated data and MODIS data. VIs trend deviations between MODIS-Terra and MODIS-Aqua over the Tibetan Plateau from 2003 to 2020 increased as the slope steepened (i.e., NDVI and enhanced vegetation index (EVI) trend deviations generally doubled). Overall, the sun-target-sensor geometry changes induced by topography, causing shadows in mountains along with obstructions in sensor observations, compromised the reliability of VIs in these terrains. Our study underscores the considerable impacts of topography, particularly shadow effects, on multiple VIs at various spatiotemporal scales, highlighting the imperative of cautious application of VIs-based trend calculation in mountains.

The complex topography in mountainous regions, exemplified by factors like slope aspect, leads to noticeable variations in vegetation patterns, which are fundamental for understanding mountain ecosystems. However, a consistent global-scale quantification of topography's influence on vegetation patterns is still lacking. Here, we utilize two phenological metrics as proxies for vegetation-maximum vegetation greenness and seasonal greenness amplitude-computed from Sentinel-2 images, to quantify differences across three topographic factors: slope aspect, steepness, and elevation within each 0.15°×0.15° mountain grid. Our mapping reveals clear geographic patterns indicating that topography strongly influences vegetation in arid and polar ecosystems, with an influence approximately 1.9 times higher than in temperate ecosystems. Topography is also important in humid regions, as demonstrated by diverse vegetation types growing on different slope aspects, steepness levels, and elevations. Additionally, the impacts of slope aspect, steepness, and elevation vary within the same region. In 25.9% of mountain grids, slope aspect causes the largest difference in vegetation patterns, while elevation and steepness account for 43.1% and 31%, respectively. Our study highlights the hotspot areas where topographic effects on vegetation patterns are most pronounced, enabling researchers to focus on these regions for better parameterization of Earth system models.