Spatiotemporal super-resolution mapping (STSRM) aims to generate fine-resolution (FR) land cover maps from coarse-resolution (CR) observations by leveraging prior FR land cover maps. However, most existing STSRM methods still cannot effectively exploit the spatial details contained in prior FR information, which limits the recovery of fine structures in the predicted FR results. Here, we propose STICCD, a Spatiotemporal Feature Integration and Cascaded Change Decoding framework, to couple bitemporal CR fraction variations with prior FR spatial details for target-time FR land cover reconstruction. In STICCD, an adaptive cross-source spatiotemporal feature integration module is designed to jointly model the interactions between bitemporal CR fraction variations and prior FR spatial details, thereby constructing a unified representation for subsequent decoding. On this basis, a change localization and transition decoding strategy is introduced to progressively recover land cover transitions. The binary localization branch first identifies candidate changed regions to reduce the interference of the dominant unchanged background, and the transition decoding branch further determines class-to-class transitions within these regions. The final target-time FR land cover map is then reconstructed by updating the prior FR land cover map according to the predicted changes. Experiments on simulated datasets derived from the US National Land Cover Database and on real mangrove mapping using Sentinel-2 and Planet imagery demonstrate that STICCD consistently outperforms competing methods in both quantitative accuracy and spatial reconstruction quality, particularly in preserving fine-scale structures, complex boundaries, and land cover transitions.

Surface water monitoring is essential for ecological security and water management. Owing to the limitations in the spatial and temporal resolution of commonly used remote sensing data, it is challenging to achieve accurate water monitoring relying on individual data, especially for small water bodies. This study integrated Planet-NICFI images with long-term Landsat images to monitor surface water dynamics in Zimbabwe. A deep learning segmentation model was developed to generate high-resolution binary water maps from Planet-NICFI images, which were linked to Landsat spectral features to construct a random forest regression model for producing water fraction series. The fraction series was evaluated by Global Surface Water product and applied to analyze the dynamics of the water area. Results showed that the segmentation model achieved an IoU of 0.91, while the random forest regression model attained an R2 of 0.89 and an RMSE of 0.09. In Zimbabwe, a total of 58246 water bodies were identified, showing a superior detection capability than global surface water products. 26.23% of regions exhibited a significant increasing trend, while 15.37% showed a decreasing trend from 1994 to 2024. This study indicated that integrating high-resolution binary maps with long-term fraction series can better monitor the spatial and temporal dynamics of surface water.

Wetland ecosystems have suffered from prolonged and severe global degradation. In recent years, wetland restoration has made significant contributions to addressing this issue. However, restoration interventions have induced complex and dynamic shifts in plant community composition. Conventional remote sensing approaches often fail to achieve fine-scale monitoring of wetland plant restoration when access to high-cost remote sensing data, such as hyperspectral imagery, is limited. To address this challenge, this study proposes a fine-scale emergent plant mapping method that integrates spatiotemporal-spectral fusion for resolution enhancement with a transformer-based classifier utilizing high-dimensional features. The study employs a spatiotemporal-spectral fusion model, TemPanSharpening net, to improve the spatial resolution of long-term multispectral image sequences. Subsequently, multiple spectral features are selected and conveyed to a Transformer variant classification model. This approach is applied to map 2 m resolution annual dynamics of emergent plant communities in the Honghu Lake South, China. Compared to conventional approaches, our method significantly enhances mapping granularity with an overall accuracy of 88.21%, and reveals that 9.5% of the carbon storage might be overlooked. This research overcomes the limitations of fine-scale emergent plant monitoring under constrained imaging conditions. It provides technical support for accurately monitoring the effectiveness of wetland plant restoration.

Against the backdrop of accelerating global climate change and urbanization, urban land cover change has emerged as a critical indicator for understanding the dynamic evolution of cities and the transformation of urban ecosystems. This study proposes a data-driven framework for fine-scale urban land cover change assessment based on the UASFNet model, enabling high-precision evaluation of urban land cover dynamics. The approach first performs preprocessing and co-registration of bi-temporal remote sensing images from the study area, and applies the trained UASFNet model to identify urban land cover types and extract land cover information for each temporal phase. The Analytic Hierarchy Process (AHP) is then employed to determine the weights of various indicator factors. By integrating building disturbance, greenbelt disturbance, and road disturbance indices, the framework quantitatively evaluates the intensity of land cover change at both pixel and regional scales. Experimental results across three benchmark datasets, consisting of high-resolution sub-meter RGB urban remote sensing imagery, demonstrate that UASFNet achieves superior segmentation accuracy, with mean Intersection over Union (mIoU) values of 91.52%, 93.31%, and 88.90%, substantially outperforming several state-of-the-art baseline models. Spatial analysis of the Langfang urban area (2017-2023) reveals a marked increase in impervious surface coverage (+16.86%) and a sharp decline in greenbelt (-40%), with the urban landscape exhibiting a multi-core, belt-like expansion pattern oriented toward newly developed districts. The proposed framework not only enhances the interpretability and generalization of remote sensing models in complex urban environments but also provides a scalable analytical tool to support urban spatial planning, ecological conservation, and sustainable city governance.

ABSTRACT Understanding how limited energy constrains fish populations in fragile high‐altitude lakes is essential for sustainable fisheries management. This study developed a satellite‐based framework that integrated MODIS‐derived chlorophyll a data with a vertically generalised production model. This framework was used to map Fish Potential Production (FPP) and establish a novel Fish Carrying Capacity Index (FCCI) for the endemic naked carp (Gymnocypris przewalskii) in Qinghai Lake, China, from 2002 to 2024. Over the 23‐year period, lake‐wide fish production increased 50‐fold (from 2592 to 127,500 t) alongside an upward trend in FPP driven by primary production. Seasonally, FPP per unit area peaked in summer (July–August: 30–50 g/m2) and declined in spring (May–June: 0–30 g/m2). Spatially, the highest values occurred near riverbank and tributary estuaries, whereas central waters remained low. The FCCI revealed significant heterogeneity; the northwestern regions experienced high food demand pressure (FCCI > 0.5), while the southeastern areas were underutilised (FCCI < 0.3). As the lake‐wide FCCI never exceeded 0.6, the current level of primary productivity can support further strategic restocking, provided that releases are redirected to the southeast to relieve pressure on the northwest. This study demonstrates how remote sensing can be used to balance fish conservation and production goals in sensitive plateau ecosystems.

Africa is one of the most rapidly urbanizing regions in the 21st century. Understanding the development of Urban Green Spaces (UGS) is crucial for ecological environmental protection and the well-being of urban populations. However, high-resolution, long-term remote sensing data for monitoring and analyzing urban green spaces in Africa remains lacking. To address this gap, this study conducts a precise and comprehensive analysis of the longterm UGS dynamics in major African cities during the 21st century using Landsat images. An improved vegetation index, KNDVI_c, is proposed to estimate Fractional Vegetation Cover (FVC) for assessing UGS conditions. Additionally, six natural and anthropogenic driving factors are identified to quantify their impacts on UGS across Africa. The results indicate that from 2000 to 2024, although UGS development in Africa has improved to a certain extent, overall volatility remains relatively high, accompanied by significant differentiation and complex spatiotemporal changes. African UGS is influenced by the interaction and coupling between natural environmental and socioeconomic factors, with the role of socioeconomic factors becoming increasingly prominent over the study period. The corresponding driving mechanisms vary across regions due to differences in local climatic conditions and socioeconomic development levels. In summary, this study fills a critical gap in the long-term, high-resolution systematic investigation of African UGS and provides a scientific foundation for formulating sustainable urban development policies in African cities.

Soil moisture (SM) products such as SMAP have a coarse spatial resolution, which limits their applicability in agricultural management and drought monitoring. Downscaling techniques can overcome this limitation by enhancing the spatial resolution of SM data. However, mainstream methods rarely address both spatial heterogeneity and the high-dimensional nonlinear coupling between SM and environmental factors. To address this gap, this study proposes an Improved Geographically Weighted Random Forest (IGWRF) model that integrates local adaptation and global generalization for high-accuracy SM downscaling. This study focuses on Kenya in East Africa, where we downscale the 9 km SMAP SM product to 1 km and evaluate IGWRF against traditional RF and GWRF using in-situ measurements. The results show that: (1) all methods produced results strongly correlated with the original SMAP SM (R > 0.9) and significantly enriched spatial detail and texture; (2) in complex terrain, GWRF and IGWRF achieved higher accuracy than RF, with IGWRF showing the greatest consistency with in-situ measurements (R = 0.771); (3) relative to RF, IGWRF increased R by 4.5 % and reduced RMSE and ubRMSE by 7 % and 5.8 %, respectively, demonstrating the superior performance of IGWRF. The study confirms that the IGWRF effectively captures the spatial heterogeneity of SM and addresses the challenge of jointly modeling spatial heterogeneity and nonlinear relationships in SM downscaling, significantly improving the accuracy of downscaled results. This research provides high-resolution (1 km) SM data to support agricultural decision-making and water resource management in drought-prone regions of Africa, filling a critical gap in fine-scale SM products across the continent.

Despite containing the world's second-largest tropical forest, the nuances of forest dynamics in the Congo Basin remain relatively less studied compared to other major tropical regions. While emerging evidence highlights smallholder clearing in the Congo Basin, the dynamic change mechanism of small-scale forest loss (i.e., clearing size <= 1.8 ha) and post-disturbance recovery is still not well understood, yet critical for the biodiversity conservation and carbon sequestration. Here, we presented a comprehensive quantification of forest loss and post-loss recovery in the Congo Basin by developing a new small-scale forest change product using all available Landsat and Sentinel-2 records from 2000 to 2021, improving upon previous products of global forest change (GFC) and JRC tropical moist forest (TMF) cover change. We detected extensive, small-scale, fragmented forest loss with an area of 24.54 +/- 1.70 Mha and found that 76.5% of the total forest loss belonged to small-scale clearing and 93.6% occurred within 4 km of human settlements. Of the forest loss, only similar to 13.21% subsequently experienced high-quality recovery, with greater proximity to human settlements associated with lower recovery. Protected areas decreased the forest loss rate, but the corresponding post-loss recovery was even less than that in non-protected areas, indicating that protected areas are only partially efficient and highlighting the vulnerability and exposure of the vast coverage of non-protected areas. Nevertheless, near-natural recovery of annual forest loss, defined as post-loss forest recovery without any further disturbances following the initial loss, holds the potential to achieve approximately four times greater high-quality canopy-level recovery magnitude (i.e., 50.9%) than the current levels of recovery, which underscores the importance of preventing additional disturbances after forest loss. This research provides essential evidence with which to guide the protection of the primary rainforest and reinforce policies to enhance post-loss recovery.

Climate change and population growth present significant challenges to global food security, underscoring the critical importance of sustainable and efficient agricultural production. Crop rotation is a key agricultural practice that enhances food production, improves soil fertility, reduces pest and disease pressure, and maintains agro-ecological balance. The complexity and diversity of cropping patterns, particularly in the fragmented farmland of southern China, limit the availability of high-resolution crop rotation maps in precision agriculture. To improve the consistency between cropping intensity (CI) estimation and crop pattern (CP) mapping, this study developed a hierarchical framework for extracting cropland, CI, and CP from remotely sensed images. Using the Google Earth Engine (GEE) platform, a 10-m binary cropland/non-cropland map was first generated from the time-series Normalized Difference Vegetation Index (NDVI). Then, CI was derived within cropland regions using an intelligent algorithm that counts the number of growth cycles. Finally, taking advantage of crop phenology and CI constraints, nine cropping patterns were extracted from a diversified cropping region. Comparing with field survey data, the results revealed overall accuracies of 98.97%, 96.47%, and 87.92% for the cropland/non-cropland map, cropping intensity map, and cropping pattern map, respectively. These findings demonstrate the reliability of the generated maps and the potential of the proposed framework for revealing diverse cropping patterns in complex cropping regions.

Accurate monitoring of forest disturbance and recovery dynamics is essential for understanding ecosystem resilience. However, reliable monitoring remains challenging in heterogeneous tropical landscapes using satellite observations. In this study, we developed a deep learning–based spectral unmixing approach to estimate annual fractional vegetation cover (FVC) of trees, shrubs, herbaceous species, and bare ground across Ethiopia using dense Landsat and Sentinel-2A/B time series. The proposed attention-based bidirectional long short-term memory (Bi-LSTM) model effectively captured spectral–temporal patterns and achieved robust performance, with mean absolute errors (MAEs) below 14

Accurate representation of floodplain topography is essential for hydrological modeling, ecosystem assessment, and flood risk estimation. This study proposes a novel approach to produce high-resolution floodplain topography maps, integrating inundation frequency map derived from Sentinel-2 imagery and river cross-sectional terrain from Ice,Cloud, and Land Elevation Satellite-2 (ICESat-2). While satellite-derived inundation frequency (IF) maps provide indirect elevation information, they are susceptible to slope-induced bias in river floodplains. To eliminate the spatial heterogeneity introduced by gradient along the river channel, a quantitative relationship between floodplain elevation and inundation frequency was established, incorporating slope correction. Specifically, we calculated the slope gradient sequence of the river in the study area to construct a slope correction lookup table (LUT). Then, we calibrated the elevation of each sample point by eliminating the slope gradient based on the correction origin. This study uses the lower Han River floodplain as a case study, covering approximately 100 km of river reach with an elevation variation of about 5 m. The proposed method yielded RMSEs of 1.039 m, 1.364 m, and 1.375 m for the Han River, Mississippi River, and Han River in-situ validation, respectively, indicating high predictive accuracy and strong robustness across validation settings. The stable model performances were present by comparing the algorithmic accuracy across 10- meter, 5- meter, 2.5-meter spatial resolutions of inundation frequency. Different slope data is also suit for this slope-corrected method, such as SWORD-derived slope information. The model error.This method requires no in-situ measurements, making it broadly applicable to data-scarce or inaccessible floodplains worldwide.

Wetland is a critical and complex ecosystem but highly susceptible to human activities. It has experienced significant change along with the increase of human intervention since decades ago. Currently, remote sensing-based mapping is believed an effective method for monitoring wetland changes. However, long-term wetland mapping is difficult due to poor historical data quality and wetland complexity, resulting in the triangular contradiction of long-term, high-resolution, and comprehensive classification systems. To address these challenges, we first modified the single-image-super-resolution method Residual Channel Attention Network (RCAN) and combined it with the famous spatiotemporal fusion model STARFM to enhance the Landsat images to 10 m resolution, which enabled more wetland objects to be identified. Then we extracted water bodies using UNet and classified them by object-based shape indices. Next, we applied recursive feature elimination and random forest classifier to map vegetated wetlands. Our overall classification accuracy reached 85.65%. We observed that the wetland in Jianghan Plain expanded significantly since 2000, with ponds showing the most notable growth rate (>114.04%). Aquatic crops make up a large portion of the wetland area, remaining stable with slow growth since 2000, while lake and surrounding herbaceous wetland restoration has shown notable progress in recent years.

Study area: Tana River Basin in Kenya. Study focus: Water use efficiency (WUE) is a pivotal variable to understand the balance of ecosystem carbon gain and water loss. However, the periodic seasonal aridity chronically affects the ecosystem water and carbon cycling processes. Therefore, for improving the water-carbon coupling in regions with seasonal hydroclimatic variability, it is pivotal to conducting a comprehensive understanding of both WUE and its seasonal aridity resilience (SAR). Based on the quantifying evaluation of WUE and its SAR, this study analyzed their spatiotemporal variation and influencing factors with Theil-Sen's slope, Mann-Kendall test, and interpretable machine learning models. New hydrological insights for the region: (1) Around 73.46 % areas of the basin presented a decreasing trend of SAR, indicating the negative impact of seasonal aridity on water-carbon coupling is intensifying; (2) The sub-humid/humid zone and farmland presented the highest WUE (2.00 and 2.09 g C kg(-1) H2O) but the lowest SAR (0.92 and 1.01) among climate zones and biomes, respectively; (3) At the basin scale, WUE was mainly influenced by vegetation composition, leaf area index, and temperature, while solar radiation, aridity index and leaf area index were the dominant factors for SAR. Moreover, the dominant factors in different climate zone and biome were varying. These insights provide scientific supports for the management of ecosystem water-carbon coupling in regions with seasonal hydroclimatic variability.

Landslide hazard chains pose significant threats in mountainous areas worldwide, yet their cascading effects remain insufficiently studied. This study proposes an integrated framework to systematically assess the landslide-landslide dam-outburst flood hazard chain in mountainous river systems. First, landslide susceptibility is assessed through a random forest model incorporating 11 static environmental and geological factors. The surface deformation rate derived from SABS-InSAR technology is incorporated as a dynamic factor to improve classification accuracy. Second, motion trajectories of rock masses in high-risk zones are identified by Rockfall Analyst model to predict potential river blockages by landslide dams, and key geometric parameters of the landslide dams are predicted using a predictive model. Third, the 2D HEC-RAS model is used to simulate outburst flood evolution. Results reveal that: (1) incorporating surface deformation rate as a dynamic factor significantly improves the predictive accuracy of landslide susceptibility assessment; (2) landslide-induced outburst floods exhibit greater destructive potential and more complex inundation dynamics than conventional mountain flash floods; and (3) the outburst flood propagation process exhibits three sequential phases defined by the Outburst Flood Arrival Time (FAT): initial rapid advancement phase, intermediate lateral diffusion phase, and mature floodplain development phase. These phases represent critical temporal thresholds for initiating timely downstream evacuation. This study contributes to the advancement of early warning systems aimed at protecting downstream communities from outburst floods triggered by landslide hazard chains. It enables researchers to better analyze the complex dynamics of such cascading events and to develop effective risk reduction strategies applicable in vulnerable regions.

Surface water resources are widely distributed and undergo rapid fluctuations, necessitating large-scale, highfrequency monitoring. Remote sensing technologies provide critical data for this purpose, but challenges such as data gaps and contamination hinder the effective observation of surface water dynamics at fine temporal scales. This limitation can obscure the recording of short-term water variations, ultimately leading to misclassified inundation extents. This study aims to develop a framework for large-scale and short-interval surface water dynamic monitoring using Sentinel-1/2 data, and generate surface water dynamic product for detailed analysis of water distribution and changes in East Africa (EA). Specifically, we proposed a novel water mapping algorithm including water extraction, integration and filtering techniques for Sentinel-1/2 data to map semimonthly surface water dynamic across EA. We then used a simple similarity-based gap-filling method to fill the data gaps in these water maps. Using this framework, we generated semimonthly and seamless surface water dynamic product covering EA from 2017 to 2023. A comprehensive spatiotemporal analysis of surface water distribution and dynamics in EA was then conducted using the product. The results showed that the water mapping algorithm achieved an overall accuracy of 0.9746, with precision (0.9815) higher than recall (0.9706). The gap-filling algorithm proved highly robust, with overall accuracy exceeding 0.98 under different scenarios. The spatial distribution of surface water in EA is heterogeneous, with dominant permanent water area (66.57 %), followed by temporary water area (22.04 %), and seasonal water area (11.39 %). The overall surface water area in EA shows fluctuation, with an increase from 2017 to 2021, followed by a decrease from 2021 to 2023. By incorporating SAR data and increasing observation frequency, our product revealed finer-scale surface water dynamics than previous product. This large-scale, short-interval mapping framework provides new insights for regional and global water resource monitoring, while the EA dataset serves as a key reference for water management in the region.

Accurate measurement of river width from remote-sensing images enables many applications like river discharge inversion and the associated carbon flux. Traditional river width measurement methods rely on hard classification of water and land, which, while effective for wide rivers, may exhibit low accuracy for small rivers. This study presents a novel fraction image-based deep learning model for measuring river width, with particular emphasis on small rivers. The model consists of two main steps: first, the water fraction image, representing the water area percentage in each pixel, is estimated from original multi-spectral remote sensing images, second, a convolutional neural network is trained to directly predict river width from the water fraction image. The convolutional neural network is trained on simulated water fraction images with corresponding river width values derived by down-sampling river masks from the Global River Width from Landsat dataset. Validation results from 188 global rivers indicate that the model significantly improves the river width quantification compared to methods based on hard classification, with a mean absolute error of 22 m and a root mean squared error of 43 m using Landsat images with 30 m spatial resolution. This approach offers a new, more accurate method for measuring the widths of small rivers from space.

Flash flood, a type of the surface runoff that rises up and down sharply in small watersheds in the hilly and mountainous areas, is a difficulty in flood control and disaster prevention. As a basic unit of people's social life, community is the most widely involved unit in disaster risk management. Flood-prone communities are characterized by geographical dispersion and confinement. This study carries out a research on disaster resilience and time-varying effects from the perspective of flash flood, and proposes a multidisciplinary integrated approach: (1) A multilevel interpretive structure model of the influencing factors of community resilience based on coupling decision mathematical model was constructed to analyze the differential influencing factors of community resilience; (2) The information diffusion method was used to quantitatively analyze and rank the community resilience of flash flood disasters; (3) Rand Index was used to quantitatively estimate the magnitude of community resilience and analyze the characteristics of its time-varying effect curve. Taking the flood-prone areas in northern Guangdong Province as the research area, this study analyzes three types of communities, including urban, rural, and suburban. The results show that the community resilience indicator system from the perspective of flash flood impact is a multi-dimensional and multi-layered complex network system. The direct influencing factors of community resilience in different types of community are different, the construction of water supply and drainage facilities and the flood emergency drills, as the fundamental influencing factors of community resilience, play an essential role in enhancing community resilience to flash floods. Since most of the surveyed communities are located in areas with frequent flash floods, residents have a high awareness of disaster prevention and mitigation, and rural show the higher community resilience than the sample of urban and suburban. With the expansion of inundation, the resilience of all three types of communities shows a consistent trend of "sharp decrease - rapid increase - slow decrease". This study can provide an important reference for enhancing the community resilience in the flood-prone areas, and also improving their disaster prevention and mitigation capabilities. This comprehensive analysis method can provide scientific support to the refined prevention and mitigation of other types of disasters.

Quantification of phosphorus transport and fate is necessary for effective mitigation to water body eutrophication issues but challenging in a large river basin due to the spatial heterogeneity of landscape patterns, especially in data-limited regions. This study presents a new estimation framework, taking usage of multi-source data sets locally and globally, to generate high-resolution maps of phosphorus sources, retention and delivery for a large river basin of Asia, the Yangtze River basin. Results show that 65%-73% phosphorus delivered to the river network and the Yangtze estuary originate from 26% basin areas, which are dominated by sources including urban living, livestock cultivation and cropland farming. 57% (30%-73%) phosphorus from sources retains in the basin rather than delivers to the estuary. Most (about 70%) retention occurs in main tributaries, lakes and reservoirs, posing high freshwater eutrophication risk. Small wetlands are identified as key terrestrial retention landscapes for diffuse sources, which serve as promising buffers if their functions are enhanced with proper restoration and better management. The new framework and its application support detailed understanding of phosphorus source, retention and delivery in a large river basin and provide insights into the impact of landscape patterns that helps enhancing eutrophication mitigation strategies.

Climate-induced shifts in the composition and structure of alpine vegetation cover, both expansion and reduction, are altering alpine ecosystem functions. However, accurately quantifying variations over large-scale regions requires a detailed characterization of the fine-scale mosaic vegetation covers. In this study, we employed a regression-based unmixing model using synthetic data to develop a multi-temporal machine learning model aimed to estimate the fractions of alpine plant functional types (PFTs) from 1984 to 2024 in the Yarlung Zangbo River Basin (YZRB), China. The estimated cover fractions for tree cover, shrub cover and herbaceous cover had mean absolute errors of 10.36%, 14.06% and 13.38%, respectively. The variations in the fractions of each alpine PFT revealed a slight increase in tree cover and shrub cover, alongside a contraction in herbaceous cover. Specifically, tree cover and shrub cover expanded by +1.54% and +1.83% per decade, respectively, while herbaceous cover declined at a rate of 1.98% per decade. These variations were predominantly observed at higher elevations (4000-6000 m), on shaded aspects, and on lower slopes. The variations in these fractions are also positively correlated with air temperature and soil moisture in most regions. This study provides new insights into vegetation cover shifts in this ecologically sensitive region.