When a massive wall of ice and rock collapsed into the Rishiganga valley in Chamoli, India, in 2021, it destroyed hydropower plants, swept away bridges, and killed over 200 people. Four years later, a large ice–rock avalanche struck Blatten, Switzerland, burying most of the village yet causing only one casualty. The difference between catastrophe and survival was not luck, but preparedness, monitoring, and rapid response. In Blatten, authorities and residents acted on precursory signs of slope instability, enabling timely evacuation. This Perspective asks why such hazard preparedness remains rare in the Himalaya. We contrast Chamoli’s devastation with Blatten’s near-escape, assess progress made in the Himalaya since 2021, and examine whether the region is better prepared today. We then explore how elements of the Swiss early-warning approach, especially integrated monitoring, communication, and community-linked response, can be adapted to Himalayan realities to transform potential disasters into survivable events. Climate change is increasing the risk of ice-rock avalanches in the Himalaya while preparedness and recognition remain limited, according to a reassessment of the 2021 Chamoli disaster, lessons from the Blatten near-miss, and analysis of institutional gaps that motivate an anticipatory hazard governance framework.

Glacial lakes in the steep, high-altitude Himalaya are both hazardous and ecologically vital. They can trigger sudden glacial lake outburst floods (GLOFs) that endanger downstream communities, while also storing crucial freshwater, supporting cold-adapted biodiversity, and serving as indicators of climate change. The in-situ lake bathymetry data are extremely scarce in the Himalaya. The Chandra-Bhaga valley in the Western Himalaya hosts several high-risk lakes that remain largely data-deficient. This limits the ability to model GLOF hazards, assess risks to downstream populations, and evaluate ecosystem vulnerability in this semi-arid, high-elevation landscape, which includes Ramsar-designated wetlands. Using an unmanned surface vehicle (USV), we acquired high-resolution bathymetry for three priority lakes: two very high-risk proglacial lakes (Gepang Gath and Kadu Nala) and the high-altitude Ramsar wetland Chandra Tal, enabling detailed mapping of lake-bed topography and robust estimation of water storage. Our findings show that commonly used empirical formulas for estimating glacial lake volume are substantially biased in this region, prompting more in-situ surveys and data sharing. This unique dataset can be valuable to glacio-hydrologists, geomorphologists, climate scientists, GLOF hazard modelers, conservation biologists, and disaster risk planners.

This work presents a spatiotemporal analysis of 5729 individual end-of-summer snowline altitude (SLA) observations. These data points were derived from manual mapping of 2257 glaciers across four periods (1977, 1994, 2009, and 2019) in the trans-Himalayan region of Ladakh. The study shows that snowline altitudes (SLAs) have risen on over 85% of glaciers investigated. The average SLA changed from a slight decrease of -0.93 m a(-)& sup1; between 1977 and 1994 to an increase of +8.47 m a(-)& sup1; between 2009 and 2019. Rising SLAs were accompanied by a 0.03 degrees C a-1 increase in regional mean annual air temperature between 1979 and 2019. Spatial variability in the rate of SLA change from basin to basin suggests that the sensitivity of glacier SLAs to the general rise in temperature is modulated by other climate factors, such as proportions of solid vs. liquid precipitation. Our analysis reveals that in the Suru and Zanskar basins, the regional mean SLA has risen to a level that now exceeds the maximum elevation of approximately 15% of the investigated glaciers. In these cases, the snowline has overshot glacier highest elevations by up to 350 m, effectively eliminating their accumulation zones. Thus, these ice masses have a negative mass balance and, no longer sustainable under current climatic conditions, will disappear.

The future of High Mountain Himalayan Communities (HMHC) is increasingly precarious due to significant threats from retreating glaciers, diminishing snow cover, and thawing permafrost, which historically provided a stable and reliable source of freshwater. In response to declining water availability, indigenous communities are adopting innovative agro-pastoral strategies, including upslope migration. In the Western Himalayas, local populations are increasingly relocating to higher valleys in search of water and cultivable land, often settling atop rock glaciers. This study presents the first systematic field-based investigation of two such inhabited relict rock glaciers (IRRGs; Komic and Chicham) in the western Indian Himalaya, using geomorphological analysis and electrical resistivity tomography (ERT) to examine their internal structure. Isotopic analysis of meltwater was also conducted to determine its primary source. Field investigations revealed the presence of significant perennial supra-wetlands/ponds on the two landforms, even during the peak ablation season, serving as the source of livelihood for the inhabited community on these IRRGs. Since these two IRRGs are located in a periglacial environment, we conducted eight ERT profiles along the two IRRGs to assess the presence/absence of ice (permafrost) and determine the origin of water. The internal structure of the landforms exhibited well-defined, confined super-saturated zones, possibly developed by the thawing of internal palaeo-permafrost ice and sustained by infiltrated permafrost-thawed water from upstream. However, accelerating, elevation-dependent warming and lesser seasonal snowfall place these HMHC under severe threat to water security, with potentially catastrophic consequences for their livelihoods and survival. By providing the first comprehensive field-based assessment of IRRGs in this region, the study establishes these landforms as valuable modern analogues for other under-researched areas of the Himalayas, offering a crucial foundation for future research on water reserves, permafrost dynamics, and community adaptation strategies.

A recent study by Dandabathula et al. attributes the 5 August 2025 Dharali disaster to an ice-patch collapse based largely on satellite imagery. Here, we examine remote sensing and process-attribution uncertainties in that interpretation. The proposed mechanism by Dandabathula et al. lacks spectral validation, geomorphic consistency, volumetric support, and geophysical corroboration. Available independent observations instead indicate rainfall-triggered mobilisation of unconsolidated paraglacial sediments, underscoring the need for rigorous process attribution in Himalayan hazard assessment.

Small alpine lakes are rapidly emerging as an overlooked yet critical component of cryospheric risk in high-mountain regions worldwide. Hazard and risk assessments have long prioritized large alpine lakes, but the accelerating formation of small lakes in unstable, deglaciating terrain reveals a major blind spot in monitoring, modelling and policy frameworks. Here we argue that despite their modest size, these lakes have already triggered destructive outburst floods, disproportionately affecting remote and marginalized communities with limited adaptive capacity. Advancing mountain sustainability therefore requires integrating these consequential hazards into climate and disaster governance to ensure equitable, risk-informed resilience in a warming world. Outburst floods from small alpine lakes pose a growing risk to mountain communities, which are often already marginalized and lack adaptation options. Cryosphere degradation accelerates these threats, adding urgency to integrate small high-mountain lakes into climate and disaster policy.

High-mountain environments are some of the most fragile and difficult environments to manage. The local communities are always under stress due to lack of sufficient land resources for their sustenance and frequent occurrences of landslides further complicate the situation. Landslide detection in high-mountains is increasingly critical as climate-induced deglaciation, permafrost thaw, and shifting precipitation patterns drive slope instability and cascading hazards. Timely mapping these events on satellite images remains challenging due to complex topography and spatial heterogeneity, while pixel-based deep learning often suffers from high costs and noise sensitivity. Here, we propose an open-source object-based Landslide Residual Graph Attention Network (LANDS-ResGAT) for landslide segmentation in rapidly transforming mountains. Superpixel-based segmentation generates spatial units, from which spectral and DEM-derived topographic features and terrain metrics are extracted. A spatial object graph encodes contextual relationships via centroid distances and edge attributes. LANDS-ResGAT’s 8-layer HybridResNet, integrating SAGEConv and GATv2Conv with residual connections, learns robust structural and contextual patterns. Evaluated on six Himalayan subregions with 3 m/pixel resolution PlanetScope satellite imagery, it achieves F1-scores above 0.92, outperforming Swin Transformer and U-Net++ while reducing inference time to under 3.5 min per scene. This scalable, interpretable approach can be adopted for any other high-mountain region globally to support real-time risk assessment and climate-sensitive geohazard monitoring.

Glacier-like forms (GLFs) are one subtype of glacial features found on the Martian surface. They are located within the mid-latitudes of Mars (30-60 degrees) in both hemispheres. These features having formed within the Amazonian period during a period of higher obliquity than Mars' is at today which allowed for the preferential accumulation of icy material in the mid-latitudes. While previous studies have investigated the geographic controls on GLF formation, their former extent, and their former dynamics (Souness, et. al., 2012; Brough, et. al. 2016, 2019), the boundary conditions under which GLFs formed remain poorly constrained, particularly on a local-scale.Our primary aim is to improve our understanding of how Martian GLFs formed and evolved with respect to their climactic and geomorphological setting using terrestrial rock glaciers as analogues. As there is still ongoing debate as to the formation dynamics of rock glaciers on Earth, be they permafrost-derived or derived from debris-covered glaciers, with the issue being that both start points can adequately describe the end-state of palaeo rock glaciers, we need to take an approach which acknowledges this issue of equifinality. Bayesian inversion is one such method that can do this. We start with the assumption that these GLFs represent permafrost-derived ice bodies where ground-temperature is a key boundary-condition for their formation. With this method, we use observed glacier geomorphology to reconstruct the former extent, volume, and thickness of the GLF to compute a posterior probability distribution for ground temperatures that are physically consistent with the reconstructed geometry of the palaeo glacier. We also consider near-surface air temperature as a secondary factor in accumulation feasibility. Here we present our ongoing work in this effort. We manually demarcated the geomorphological constraints of multiple GLFs on Mars within GIS software based on identifiable geomorphology within the orthorectified imagery that mark the former maximum extent of the glacier, and extract morphometric data using the georeferenced HiRISE DEM. We then used the perfect-plasticity approximation to reconstruct palaeo ice-thicknesses and volume of the palaeo glacier. These morphometrics are then compared with modelled outputs for glacier deformation, employing Bayesian logic to constrain a boundary range of long-term mean ground temperature that would be compatible to produce the reconstructed glacier morphology. We also investigate several terrestrial rock glaciers in order to assess the accuracy and validity of our approach against measurable analogue examples, which further enables us to compare the dynamics of terrestrial and Martian glaciers.References:Brough, Stephen, Bryn Hubbard, and Alun Hubbard. 2016. “Former Extent of Glacier-Like Forms on Mars.”, Icarus 274 (August): 37–49. https://doi.org/10.1016/j.icarus.2016.03.006.Brough, S., Hubbard, B., & Hubbard, A. (2019, 02). Area and volume of mid latitude glacier-like forms on mars. Earth and Planetary Science Letters, 507 , 10–20. Retrieved from https://linkinghub.elsevier.com/retrieve/pii/S0012821X18306903 doi: 10.1016/j.epsl.2018.11.031Souness, Colin, Bryn Hubbard, Ralph E. Milliken, and Duncan Quincey. 2012. “An Inventory and Population-Scale Analysis of Martian Glacier-Like Forms.” Icarus 217 (1): 243–55. https://doi.org/10.1016/j.icarus.2011.10.020.

Previous studies have mapped end-of-season snowlines (ESS) on glaciers from satellite imagery to find their snowline altitudes (SLA) to use as a proxy for the glacier equilibrium line altitude (ELA). A line is traced along the boundary between snow and ice, then, from a digital elevation model (DEM), elevation values are extracted at regular intervals along the line. The average elevation of these points is taken to be the SLA. While this approach would be advantageous, as it offers a solution to measuring glacier ELAs in remote regions, it is prone to an oversampling bias. Where snow cover is patchy, for example, in shaded areas or where avalanching has occurred, a greater length of line is mapped in order to follow the snow-ice boundary than is required for smoother segments. This is regardless of whether the region contributes a larger area of snow cover or not. Consequently, SLA calculations are prone to oversampling from areas of irregular snow cover. Even when the ESS is mapped accurately and precisely, the SLA value may differ significantly from the true ELA. This poster investigates alternative methods of calculating the SLA from mapped ESSs to reduce bias towards patchy and irregular areas of snow cover.

Urban expansion is an inherently complex and nonlinear process shaped by heterogeneous spatial patterns, temporal dynamics, and stochastic uncertainties. To model such complexity, this study presents a novel hybrid framework that integrates logistic function-based urban proportion prediction with an adaptive Monte Carlo simulation under a cellular automata (CA) structure. The logistic function captures the nonlinear, sigmoidal growth trend of urban saturation at a microscale, while the adaptive Monte Carlo introduces controlled randomness based on local growth rates and probabilistic stratification. This dual mechanism enhances the system's capability to simulate emergent urban dynamics governed by local interactions and macro constraints. Empirical validation in Changsha, China, shows that the proposed model achieves over 5 % improvement in F1score accuracy compared to traditional CA-Markov models. Moreover, it reveals spatially differentiated urban transformation pathways and enables city-scale prediction constrained by dynamically evolving local capacities. The model features emphasized adaptability and transferability. This interdisciplinary approach demonstrates strong potential for understanding and forecasting complex urban systems from a nonlinear and data-driven modelling perspective.

During the past decade, a plethora of research articles have considerably advanced our understanding of the genesis, evolution, and hazard potential of the Himalayan glacial lakes. However, given the rapidly changing regional climate and land-use practices that exacerbate the risks, a research and policy shift towards adaptive management of glacial lakes, ranging from regulated siphoning to multistakeholder participation, is urgently required. In this perspective piece, we propose a Strengths, Weaknesses, Opportunities, and Threats (SWOT) framework, to highlight the key focus areas for future glacial lake outburst flood (GLOF) research and policymaking. The suggested approach is of proactive nature in influencing researchers and policymakers, and aims at integrating local socioeconomic and cultural contexts, while also assessing the hazard perception of communities. This also means that while the focus of this article is the Himalayan region, the ideas and the SWOT framework proposed here can easily be adapted for any other geographical region globally.

Growing global population, changing climate, and shrinking land resources demand for quicker, efficient, and more accurate methods of mapping and monitoring vegetation cover in remote sensing datasets. Many deep learning-based methods have been widely applied for semantic segmentation tasks in remote sensing images of vegetated environments. However, most existing models are pixel-based, which introduces challenges such as high time consumption, cumbersome implementation, and limited scalability. This paper presents the SAGRNet model, a Graph Convolutional Neural Network (GCN) that incorporates sampling aggregation and self-attention mechanisms, while leveraging the ResNet residual network structure. A key innovation of SAGRNet is its ability to fuse features extracted through diverse algorithms, enabling comprehensive representation and enhanced classification performance. The SAGRNet model demonstrates superior performance over leading pixel-based neural networks, such as U-Net++ and DeepLabV3, in terms of both time efficiency and accuracy in vegetation image classification tasks. We achieved an overall mapping accuracy of similar to 90 % using SAGRNet, compared to similar to 87% and similar to 85% from U-Net++ and DeepLabV3, respectively. Additionally, it offers more convenience in data processing. Furthermore, the model significantly outperforms cutting-edge graph-based convolutional networks, including Graph U-Net (achieved overall accuracy similar to 65%) and TGNN (achieved overall accuracy similar to 75%), showcasing exceptional generalization capability and classification accuracy. This paper provides a comprehensive analysis of the various processing aspects of this object-based GCN for vegetation mapping and emphasizes its significant potential for practical use. The model's versatility can also be expanded to other image processing domains, offering unprecedented possibilities of information extraction from satellite imagery. The code for practical application experiment is available at https://github.com/baoling123/GCN-remote-sensing-classification.git.

Changing climate is enhancing the occurrence and intensity of natural disasters, profoundly impacting human lives, livelihoods, infrastructure, and economic growth. Modelling and prediction of deadly high-mountain slope failure hazards such as snow, ice, and rock avalanches have always been challenging. Current in-situ sensor-based approaches for slope failure predictions of hanging glaciers and rock faces are quite limited in their spatial continuity and extent and there is also a research gap on linking the pre-collapse slope movements with subsequent avalanche runouts. Earth observation datasets can offer a viable alternative for quantifying and monitoring pre-collapse dynamics at larger spatial scales. For the catastrophic 2021 rock-ice collapse in Chamoli, India, several studies had reported some anomalous movements weeks-to-months prior to the collapse. However, we need more analyses to understand how common such pre-collapse anomalous movements are before we can even start considering investigating them as potential precursors for effective avalanche predictions. To fill this research gap, using satellite remote sensing datasets and digital elevation models, we investigated several high-mountain slope failure events (e.g., Piz Scerscen in 2024, Piz Cengalo Bondo in 2017) of varying magnitudes and nature (i.e., rockfall, rock-ice avalanche, and ice avalanche) in different topographical and climate settings. While we were able to quantify pre-collapse dynamics for these events, we also observed variations in the occurrence and magnitude of anomalous movements prior to the events. These preliminary findings are encouraging and the future research and results from such analyses can bridge the knowledge gap on the detection and modelling capabilities, ultimately enhancing resilience to mountain hazards.

Agricultural and ecological land use (AELU) classification using remote-sensing technology is crucial for understanding agro-environmental processes. However, traditional pixel-level classification methods often struggle with accuracy and classification issues when applied to diverse AELU landscapes, owing to significant seasonal dependence of spectral characteristics of AELU pixels. Similarly, pixel-based deep learning algorithms face challenges such as overfitting, high computational demands, and limited generalization ability for large-scale AELU classification. This study explores the application of object-based classification to mapping AELU types using remote sensing data. Compared to pixel-based methods, object-based classification groups pixels into spatially coherent units, allowing the extraction of aggregated spectral and texture features. This not only reduces salt-and-pepper noise and enhances classification stability but also improves computational efficiency - making it well-suited for large-scale, seasonally dynamic landscapes. Methodologically, this study introduces several feature attribute calculation methods for participating in spatial statistical analysis based on object classification. Various band combinations of Sentinel-2 satellite multispectral data were evaluated across seasonal periods to gauge classification accuracy. Findings underscore the efficacy of integrating all bands or selectively using bands 11 and 12, consistently achieving high F-scores (similar to 0.8). However, variability across AELU types and seasons emphasizes the need for optimized band selection and consideration of temporal influences. The study also highlights the significant improvement in accuracy through spectral difference calculations, particularly beneficial for challenging classifications like bog and heath. Looking forward, future research should focus on refining band selection strategies, leveraging advanced machine learning techniques to enhance classification robustness, and exploring the impacts of climate change on spectral signatures. Addressing these aspects will advance the utility of object-based classification for broader applications in agricultural monitoring and ecological assessment.

While algorithms have been created for land usage in urban settings, there have been few investigations into the extraction of urban footprint(UF). To address this research gap, the study employs several widely used image classification method classified into three categories to evaluate their segmentation capabilities for extracting UF across eight cities.The results indicate that pixel-based methods only excel in clear urban environments, and their overall accuracy is not consistently high. RF and SVM perform well but lack stability in object-based UF extraction, influenced by feature selection and classifier performance. Deep learning enhances feature extraction but requires powerful computing and faces challenges with complex urban layouts. SAM excels in medium-sized urban areas but falters in intricate layouts. Integrating traditional and deep learning methods optimizes UF extraction, balancing accuracy and processing efficiency. Future research should focus on adapting algorithms for diverse urban landscapes to enhance UF extraction accuracy and applicability.

Rapid urbanization and land-use changes are placing immense pressure on resources, infrastructure, and environmental sustainability. To address these, accurate urban simulation models are essential for sustainable development and governance. Among them, Cellular Automata (CA) models have become key tools for predicting urban expansion, optimizing land-use planning, and supporting data-driven decision-making. This review provides a comprehensive examination of the development of urban cellular automata (UCA) models, presenting a new framework to enhance individual UCA sub-modules within the context of emerging technologies, sustainable environments, and public governance. By addressing gaps in prior UCA modelling reviews—particularly in the integration and optimization of UCA sub-module technologies—this framework is designed to simplify UCA model understanding and development. We systematically review pioneering case studies, deconstruct current UCA operational processes, and explore modern technologies, such as big data and artificial intelligence, to optimize these sub-modules further. We discuss current limitations within UCA models and propose future pathways, emphasizing the necessity of comprehensive analyses for effective UCA simulations. Proposed solutions include strengthening our understanding of urban growth mechanisms, examining spatial positioning and temporal evolution dynamics, and enhancing urban geographic simulations with deep learning techniques to support sustainable transitions in public governance. These improvements offer data-driven decision support for environmental management, advancing policies that foster sustainable urban development.

Urban vegetation is critical for mitigating summer heat, but previous studies have largely relied on static greenness metrics, leaving a gap in understanding of how vegetation phenology (its seasonal life cycle) regulates urban temperatures on a global scale. Here, we utilize interpretable machine learning and satellite data to quantify the influence of key phenological metrics, encompassing growth intensity (e.g., peak greenness), timing (e.g., start of season), and duration, on summer Land Surface Temperature (LST) across 24 major global cities. We found that: 1) a significant temporal mismatch exists in over 80% of cities, with vegetation green-up lagging seasonal surface warming by 50–100 days, creating a window of thermal vulnerability; 2) seasonal accumulation of Enhanced Vegetation Index (EVI) provides stable, linear cooling, whereas Maximum EVI (MEV) and EVI amplitude (EA) exhibit a distinct threshold effect, with their cooling benefits diminishing or even reversing beyond a critical point; 3) vegetation's cooling effect changes with context, delivering roughly 25% greater cooling in the top 10% of temperature extremes compared to moderate conditions; and 4) in certain contexts, vegetation's cooling effect is observationally weakened or even offset when it is masked by the dominant influence of a positively correlated warming factor, such as high elevation. These findings provide mechanistic evidence that simply increasing green cover is insufficient; future urban heat mitigation must shift to "phenology-aware" designs that synchronize vegetation's life cycle with seasonal heat peaks to achieve maximum cooling benefits.

Impact craters, resulting from the collision of meteorites, asteroids, or comets with planetary surfaces, manifest as circular-elliptical depressions with diverse sizes and shapes influenced by various factors. These morphological features play a crucial role in planetary exploration, offering insights into the geological composition and structure of celestial bodies. Beyond their scientific importance, craters may also hold valuable natural resources, such as frozen water in the Moon's permanently shadowed craters. Furthermore, understanding craters’ spatial distribution is pivotal for terrain-relative navigation and for selecting future landing sites.Manual crater mapping through visual inspection is an impractical and laborious process, often unattainable for large-scale investigations. Moreover, manual crater mapping is susceptible to human errors and biases, leading to potential disagreements of up to 40%. In order to tackle these issues, semi-automatic crater detection algorithms (CDA) have been developed to mitigate human biases, and to enable large-scale and real-time crater detection and mapping.The majority of CDAs’ are based on machine learning (ML) and data-driven methods. ML-based CDAs’ are trained in a supervised manner using specific datasets that were manually labelled. Because of that, existing ML-based CDAs’ are constrained to specific data types according to the type of their training data. This makes current ML-based CDAs’ unstable and un-practical, since applying an ML scheme to a different type of data requires acquiring and labelling a new training set, and subsequently use it to train a new ML scheme, or fine-tune an already existing one.In this study, we describe a universal approach [1] for crater identification based on Segment Anything Model (SAM), a foundational computer vision and image segmentation model developed by META [2]. SAM was trained with over 1 billion masks, and is capable to segment various data types (e.g., photos, DEM, spectra, gravity) from different celestial bodies (e.g., Moon, Mars) and measurement setups. The segmentation output undergoes further classification into crater and no-crater based on geometric indices assessing circular and elliptical attributes of the investigated mask. The proposed framework is proven effective across different datasets from various planetary bodies and measurement configurations. The outcomes of this study underlines the potential of foundational segmentation models in planetary science. Foundational models tuned for planetary data can provide universal classifiers contributing towards an automatic scheme for identifying, detecting and mapping various morphological and geological targets in different celestial bodies. References[1] Giannakis, I., Bhardwaj, A., Sam, L., Leontidis, G., (2024). A Flexible Deep Learning Crater Detection Scheme Using Segment Anything Model (SAM), Icarus, 2024.[2] Kirillov, A., et al. (2024). Segment Anything, arXiv:2304.02643

Glacial lake outburst floods (GLOFs) are natural catastrophic events that pose a growing threat to mountain communities worldwide. Despite extensive research, hazard mapping, and risk modelling, these events continue to cause large-scale destruction to the downstream communities and infrastructure. This persistent vulnerability stems from weak policy enforcement, inadequate early warning systems, and poor community preparedness. While the threat of GLOF is increasing, their sudden and destructive nature raises a critical concern—are frontline communities truly aware and resilient enough to cope with such disasters? The recent South Lhonak GLOF in Sikkim, underscores the urgency of this question, highlighting the growing disconnect between scientific understanding, policy implementation and community preparedness. To address this, we propose the GLOF-WATCH approach—an integrated, globally coordinated, watershed-based monitoring system for continuous glacial lake observation and GLOF risk assessment. This study also highlights the key vulnerabilities in community preparedness and the failure of structural measures, outlining critical steps to bridge the gap between scientific research and on-ground resilience to build GLOF-resilient communities across high-mountains globally before the next disaster strikes.

Unplanned urban sprawl can cause environmental degradation, infrastructural overload, and a diminished quality of life. Consequently, accurate urban expansion prediction models are vital for guiding sustainable city development and aligning policy decisions with long-term community needs. Current urban expansion simulation model often struggles with rigid neighbourhood definitions and static factor weighting in conventional Cellular Automata (CA) models, causing inaccuracies in capturing non-linear, multi-scale dynamics. To address these limitations, we propose an enhanced U-Net++-based framework that adaptively integrates attention mechanisms and autoregressive CA simulation. Tested on Changsha city's multi-source spatiotemporal data from 2010 to 2022, our model achieved an Overall Accuracy of 0.87 and a Figure of Merit of 0.90, surpassing traditional methods in both quantitative metrics and visual coherence. By leveraging multi-scale convolutional features and attention-driven factor weighting, this approach effectively addresses the challenges of multi-scale and spatiotemporal heterogeneity. Furthermore, its end-to-end design allows for adaptive optimization of the entire prediction process, yielding more coherent and robust urban expansion forecasts. Moving forward, integrating wider socio-political drivers and policy constraints can further enhance the model's practicality for sustainable urban development, ensuring that city growth management aligns with environmental goals and community well-being. This modelling framework is applicable to predict any urban setting globally.