High-resolution albedo has the advantage of a higher spatial scale from tens to hundreds of meters, which can fill the gaps of albedo applications from the global scale to the regional scale and can solve problems related to land use change and ecosystems. The Sentinel-2 satellite provides high-resolution observations in the visible-to-NIR bands, giving possibilities to generate a high-resolution surface albedo at 10 m. This study attempted to evaluate the performance of the four data-driven machine learning algorithms (i.e., random forest (RF), artificial neural network (ANN), k-nearest neighbor (KNN), and XGBoost (XGBT)) for the generation of a Sentinel-2 albedo over flat and rugged terrain. First, we used the RossThick-LiSparseR model and the 3D discrete anisotropic radiative transfer (DART) model to build the narrowband surface reflectance and broadband surface albedo, which acted as the training and testing datasets over flat and rugged terrain. Second, we used the training and testing datasets to drive the four machine learning models, and evaluated the performance of these machine learning models for the generation of Sentinel-2 albedo. Finally, we used the four machine learning models to generate a Sentinel-2 albedo and compared them with in situ albedos to show the models’ application potentials. The results show that these machine learning models have great performance in estimating Sentinel-2 albedos at a 10 m spatial scale. The comparison with in situ albedos shows that the random forest model outperformed the others in estimating a high-resolution surface albedo based on Sentinel-2 datasets over the flat and rugged terrain, with an RMSE smaller than 0.0308 and R² larger than 0.9472.

Anisotropic canopy reflectance plays a crucial role in estimating vegetation biophysical parameters, whereas soil reflectance anisotropy affects canopy reflectance. However, woodland canopy bidirectional reflectance distribution function (BRDF) models considering soil anisotropy are far from universal, especially for the BRDF models of mountain forest. In this study, a mountain forest canopy model, named geometric-optical and mutual shadowing and scattering from arbitrarily inclined-leaves model coupled with topography (GOSAILT), was extended to consider the soil anisotropic reflectance characteristics by introducing the simple soil directional (SSD) reflectance model. The modified GOSAILT model (named GOSAILT-SSD) was evaluated using unmanned aerial vehicle (UAV) field observations and discrete anisotropic radiative transfer (DART) simulations. Then, the effects of Lambertian soil assumption on simulating the vi-directional reflectance factor (BRF) were evaluated across different fractions of vegetation cover (Cv), view zenith angles (VZA), solar zenith angles (SZA), and spectral bands with the GOSAILT-SSD model. The evaluation results, with the DART simulations, show that the performance of the GOSAILT-SSD model in simulating canopy BRF is significantly improved, with decreasing RMSE, from 0.027 to 0.017 for the red band and 0.051 to 0.037 for the near-infrared (NIR) band. Meanwhile, the GOSAILT-SSD simulations show high consistency with UAV multi-angular observations (R² = 0.97). Besides, it is also found that the BRF simulation errors caused by Lambertian soil assumption are too large to be neglected, with a maximum relative bias of about 45% for the red band. This inappropriate assumption results in a remarkable BRF underestimation near the hot spot direction and an obvious BRF overestimation for large VZA in the solar principal plane (PP). Meanwhile, this simulation bias decreases with the increase of fraction of vegetation cover. This study provides an effective technique to improve the capability of the mountain forest canopy BRDF model by considering the soil anisotropic characteristics for advancing the modeling of radiative transfer (RT) processes over rugged terrain.

Semi-empirical bidirectional reflectance distribution function (BRDF) models are developed based on various spatial-resolution pixels. Because of its simplicity and physical significance, it is widely used in medium- and low-spatial-resolution quantitative remote sensing. With the emergence of high-spatial-resolution remote sensing data and the lack of high-spatial-resolution BRDF models, semi-empirical BRDF models have also been directly applied to high-spatial-resolution qualitative and quantitative remote sensing research. However, whether semi-empirical BRDF models can be directly applied to pixels with high resolution remains unclear. To answer this question, this letter quantitatively evaluates the applicability of semi-empirical BRDF models for remote sensing data with 0.5-30 m spatial resolution based on WIDAS multi-angular observation dataset obtained during the HiWATER experiment in 2012. The results demonstrate that the semi-empirical BRDF models are not applicable at the 0.5 m pixel scale but are applicable at the 10 m pixel scale. There is a transitional pixel scale from not applicable to applicable between 0.5 m and 10 m. We define this scale as the optimal minimum pixel scale (OMS) of semi-empirical BRDF models. The OMS is related to the spatial structure of vegetation scene, and it is highly consistent with the canopy characteristic scale calculated based on the semivariogram method (R2=0.901). Therefore, the range of the semivariogram can be used to estimate the OMS to answer the question of which scale semi-empirical BRDF models are applicable to high-spatial-resolution images.

Hyperspectral anomaly detection is a crucial technique for recognizing abnormal pixels in hyperspectral images (HSIs), that is, those with distinct spectral characteristics from those of the surrounding background. Traditional methods always fall short in effectively leveraging the information regarding the spectral and spatial aspects of the dataset simultaneously, limiting their detection performances. This article proposes a novel framework using U-Net, termed hybrid convolution and transformer-based U-Net (HCT-Unet), which integrates convolution with a multihead attention mechanism in Transformer for enhanced hyperspectral anomaly detection. To ensure a more comprehensive understanding of spatial and spectral interactions, the HCT-Unet architecture capitalizes on the strengths of local feature extraction of convolutional layers and the capabilities of the long-range dependency modeling of Transformers. A key innovation of this framework is an error attention mechanism, which facilitates adaptive multiscale feature fusion and enhances the feature representation capacity. Furthermore, a new anomaly score calculation method is proposed, which combines reconstruction error with the pixelwise structural similarity index (SSIM) to determine pixel anomaly from both local structural preservation and global spectral consistency perspectives. Experiments carried out on seven different hyperspectral datasets reveal that the proposed method consistently outperforms the widely accepted state-of-the-art methods in hyperspectral anomaly detection.

Glacial lakes are among the most sensitive indicators of climate change. With the ongoing climate warming, the number and area of glacial lakes are dramatically increased, particularly in High Mountain Asia (HMA), thereby concomitantly arousing the overall number and frequency of glacial lake outburst flood (GLOF) events as well, which pose a threat to vulnerable communities. Regular glacial lake monitoring is needed to detect the abnormal changes that could signify instabilities. However, the existing methods rely on the scoring regressions from combinations of glacial lake properties to assess the outburst risks of glacial lakes in a large scale, neglecting the spatial structures of lake scenes and property dependencies which may result in diverging hazard estimates. In this work, we conducted an appraisal of GLOF risks in HMA by reasoning the association between the glacial lake spatial and property data. Concretely, the cascade of spatial features and property features were classified with semi-supervised mode to further identify the lake that shares a similar pattern with the outburst of glacial lakes. As a result, our model yielded 88.26% validation accuracy in historical GLOF events from 1990 to 2022 in HMA. We thus evaluated the glacial lakes with our model in HMA in 2020 and finally characterized 49 glacial lakes as high-risk lakes. From the lake investigation, 4 glacial lakes that had been outburst but were not mentioned previously in HMA from 1990 to 2020 were provided. Moreover, among the dependency analysis of lake properties, the pairs of (subregion, precipitation) and (distance, lake type) show a strong dependency on the state of glacial lakes. No significant dependency, by contrast, exists between the area change rate and the other properties.

Lakes are among the most frequently occurring features observed in remote sensing imagery on Earth, which construct the part of water resource and supplying a great deal of life downstream. To assess water coverage and its regional or global impacts, identifying lakes in remote sensing imagery is a prerequisite as well in such lake-related studies. However, the diverse environmental conditions under which lakes form, the presence of mixed sediments leads to considerable spatial and spectral heterogeneity, making it challenging for existing methods to accurately detect heterogeneous water bodies. In this work, we proposed a weakly-supervised contrastive representation method for finding the inherent universal natures of lakes with sediments. Our model comprised two modules, spatial segmentation module and spectral contrastive module for capturing the spatial and spectral features of lakes jointly. Then the patterns of lakes were compared with similarities to unlabeled samples and final lake area was extracted by decision making process. We tested our model on Landsat-8 OLI imagery, and our model yielded F1 score of 79.86±0.94 and IoU of 70.85±0.91, which showed great superiority to other competitors, especially in these lakes with sediments.