The long-term estimation of grassland aboveground biomass (AGB) is important for grassland resource management in the Three-River Headwaters Region (TRHR) of China. Due to the lack of reliable grassland AGB datasets since the 1980s, the long-term spatiotemporal variation in grassland AGB in the TRHR remains unclear. In this study, we estimated AGB in the grassland of 209,897 km² using advanced very high resolution radiometer (AVHRR), MODerate-resolution Imaging Spectroradiometer (MODIS), meteorological, ancillary data during 1982–2018, and 75 AGB ground observations in the growth period of 2009 in the TRHR. To enhance the spatial representativeness of ground observations, we firstly upscaled the grassland AGB using a gradient boosting regression tree (GBRT) model from ground observations to a 1 km spatial resolution via MODIS normalized difference vegetation index (NDVI), meteorological and ancillary data, and the model produced validation results with a coefficient of determination (R²) equal to 0.76, a relative mean square error (RMSE) equal to 88.8 g C m⁻², and a bias equal to −1.6 g C m⁻² between the ground-observed and MODIS-derived upscaled AGB. Then, we upscaled grassland AGB using the same model from a 1 km to 5 km spatial resolution via AVHRR NDVI and the same data as previously mentioned with the validation accuracy (R² = 0.74, RMSE = 57.8 g C m⁻², and bias = −0.1 g C m⁻²) between the MODIS-derived reference and AVHRR-derived upscaled AGB. The annual trend of grassland AGB in the TRHR increased by 0.37 g C m⁻² (p < 0.05) on average per year during 1982–2018, which was mainly caused by vegetation greening and increased precipitation. This study provided reliable long-term (1982–2018) grassland AGB datasets to monitor the spatiotemporal variation in grassland AGB in the TRHR.

Forest anomalies (e.g., pests, deforestation, and fires) are increasingly frequent phenomena on Earth’s surface. Rapid detection of these anomalies is crucial for sustainable forest management and development. On-orbit remote sensing detection of multi-type forest anomalies using single-temporal images is one of the most promising methods for achieving it. Nevertheless, existing forest anomaly detection methods rely on time series image analysis or are designed to detect a single type of forest anomaly. In this study, a Forest Anomaly Comprehensive Index (FACI) is proposed to detect multi-type forest anomalies using single-temporal Sentinel-2 images. First, the spectral characteristics of different forest anomaly events were analyzed to obtain potential band combinations. Then, the formulation of FACI was determined using imagery simulated by the LargE-Scale remote sensing data and image Simulation framework over heterogeneous 3D scenes (LESS) model. The thresholds for FACI for different anomalies were determined using the interquartile method and 90 in situ survey samples. The accuracy of FACI was quantitatively assessed using an additional 90 in situ survey samples. Evaluation results indicated that the overall accuracy of FACI in detecting the three forest anomalies was 88.3%, with a Kappa coefficient of 0.84. The overall accuracy of existing indices (NDVI, NDWI, SAVI, BSI, and TAI) is below 80%, with Kappa coefficients less than 0.7. In the end, a case study in Ji’an, Jiangxi Province, confirmed the ability of FACI to detect different stages of pest infection, as well as deforestation and forest fires, using single-temporal satellite images. The FACI provides a promising method for the on-orbit satellite detection of multi-type forest anomalies in the future.

Recently, developments in hyperspectral image (HSI) classification have brought increasing attention to the challenges of the open-set problem. However, current open-set methods generally overlook the intra-class multimodal structure, making it difficult to comprehensively capture the global data distribution, which in turn reduces their ability to distinguish known from unknown classes. To address this, we propose a novel global distribution-aware network (GDAN) that jointly performs pixel-wise HSI classification and trustworthy uncertainty-aware identification of unknown class. First, a generative adversarial network (GAN) is employed as the backbone, enhanced with a self-attention (SA) module to capture long-range dependencies across the extensive spectral bands of hyperspectral data. Second, an interpretable open-set HSI classification framework is designed, combining GAN with Markov Chain Monte Carlo (MCMC) to model global distribution by exploring intra-class multimodal structures and estimate predictive uncertainty. In this framework, the traditionally fixed discriminator weights are reformulated as probability distributions, and posterior inference is conducted using MCMC within a Bayesian framework. Finally, accurate categories and predictive uncertainty of ground objects can be obtained through posterior sampling, while samples with high uncertainty are assigned to the unknown class, thus enabling accurate open-set HSI classification. Extensive experiments on three benchmark HSI datasets demonstrate the superiority of the proposed GDAN for open-set HSI classification, yielding overall classification accuracies of 94.6%, 92.6%, and 94.8% in the 200-sample scenario.

The timely and accurate mapping of crops over large areas is essential for alleviating food crises and formulating agricultural policies. However, most existing classical crop mapping methods usually require the whole-year historical time-series data that cannot respond quickly to the current planting information, let alone for future prediction. To address this issue, we propose a novel spatial–temporal feature and deep integration strategy for crop growth pattern prediction and early mapping (STPM). Specifically, the STPM first learns crop spatial–temporal evolving patterns from historical data to generate future remote sensing images based on the current observations. Then, a robust crop type recognition model is applied by combining the current early data with the predicted images for early crop mapping. Compared to existing spatial–temporal prediction models, our proposed model integrates local, global, and temporal multi-modal features comprehensively. Not only does it achieve the capability to predict longer sequence lengths (exceeding 100 days), but it also demonstrates a significant improvement in prediction accuracy for each time step. In addition, this paper analyses the impact of feature dimensionality and initial data length on prediction and early crop mapping accuracy, demonstrating the necessity of multi-modal feature fusion for spatial–temporal prediction of high-resolution remote sensing data and the benefits of longer initial time-series (i.e., longer crop planting time) for crop identification. In general, our method has the potential to carry out early crop mapping on a large scale and provide information to formulate changes in agricultural conditions promptly.

Landslides are common natural disasters in mountainous regions, exerting considerable influence on socioeconomic development and city construction. Landslides occur and develop rapidly, often posing a significant threat to the safety of individuals and their property. Consequently, the mapping of areas susceptible to landslides and the simulation of the development of such events are crucial for the early warning and forecasting of regional landslide occurrences, as well as for the management of associated risks. In this study, a landslide susceptibility (LS) model was developed using an ensemble machine learning (ML) approach which integrates geological and geomorphological data, hydrological data, and remote sensing data. A total of nine factors (e.g., surface deformation rates (SDF), slope, and aspect) were used to assess the susceptibility of the study area to landslides and a grading of the LS in the study area was obtained. The proposed model demonstrates high accuracy and good applicability for LS. Additionally, a simulation of the landslide process and velocity was constructed based on the principles of landslide movement and the rule-based discrete grid model. Compared with actual unmanned aerial vehicle (UAV) imagery, this simulation model has a Sørensen coefficient (SC) of 0.878, a kappa coefficient of 0.891, and a total accuracy of 94.12%. The evaluation results indicate that the model aligns well with the spatial and temporal development characteristics of landslides, thereby providing a valuable reference basis for monitoring and early warning of landslide events.

Land surface anomalies refer to various activities on the Earth’s surface that consist of short-term and sudden changes due to external disturbances. These anomalies are closely related to the safety of human life and property. Remote sensing offers irreplaceable advantages such as broad coverage, high temporal dynamics, and comprehensive observations, so it is the most effective tool for monitoring land surface anomalies and measuring their intensities. However, existing studies have limitations such as unclear sensitivity features, uncertain applicability, and a lack of quantitative expression at different scales. Therefore, this study develops a quantitative assessment framework for land surface anomaly intensity across four scales: the pixel scale, structure scale, object scale, and scene scale. This framework enables an adaptive and flexible weight determination of the intensity of land surface anomalies from a satellite perspective. Using the Chongqing fire as an example of a land surface anomaly, this study evaluates its land surface anomaly intensity. Moreover, we demonstrate the method’s applicability to other land surface anomaly events, such as floods and earthquakes. The experiments reveal that the land surface anomaly intensity evaluation framework, which is constructed based on pixel-scale, structure-scale, object-scale, and scene-scale features, can quantitatively express the land surface anomaly intensity with an accuracy of 75.25% and more effectively represent severely affected areas. The weights of the features at the four scales sequentially decrease: structure scale (0.2974), pixel scale (0.3225), object scale (0.1867), and scene scale (0.1932). The extensive application of this method to other land surface anomaly events provides accurate quantitative expressions of the land surface anomaly intensity. This remote sensing-based multiscale feature assessment method is adaptable and applicable to various land surface anomalies and offers critical decision support for land surface anomaly intensity warning systems.