Semantic change detection aims to localize changed regions and identify the corresponding land-cover transitions from bi-temporal remote sensing images, which is crucial for applications such as urban expansion analysis, disaster assessment, and environmental monitoring. Although vision foundation models such as the Segment Anything Model 2 provide strong visual priors and powerful feature representations, directly transferring them to semantic change detection remains challenging. In particular, the high-frequency details required for precise boundary delineation are often weakened during feature extraction, while the joint optimization of binary change localization and semantic recognition can introduce task interference. To address these challenges, we present FreqSCD, a SAM2-based framework built on a frozen backbone with three task-specific components: a High–Low-Frequency Adapter for frequency-aware feature adaptation, Task-Decoupled Decoding and Semantic Consistency for reducing task interference, and Local Spatial–Semantic Alignment for improving multi-scale feature aggregation. Experiments on the SECOND and Landsat-SCD benchmarks show that FreqSCD achieves strong semantic change detection performance, obtaining an F1 score of 56.72% and a SeK of 24.17% on SECOND, as well as an F1 score of 85.46% and a SeK of 53.76% on Landsat-SCD.

Aim: Charcoal production remains a significant source of energy in sub-Saharan Africa and is increasingly linked to deforestation and land degradation in peri-urban areas. This study aimed to analyze the temporal trends, spatial land cover changes, and socio-environmental impacts of charcoal production in Bamenda I Sub-Division between 2000 and 2023. Methods: Primary data were collected through structured questionnaires administered to 198 respondents across 13 quarters, in addition to key informant interviews and GPS-based field observations. Secondary data included Landsat 5, 7, and 8 satellite imagery from the United States Geological Survey (USGS) for land cover change detection, as well as production records from the Northwest Regional Delegation of Forestry and Wildlife (MINFOF). Analytical methods comprised descriptive statistics, cross-tabulation, and GIS-based spatial analysis. Results: Results indicate that charcoal production in Bamenda I Sub-Division increased from 28,024 bags in 2000 to a peak of approximately 69,000 bags between 2012 and 2015, before declining sharply to 21,563 bags in 2022, likely due to the progressive depletion of accessible wood resources. Over the study period, natural woodland declined from 12.35% to 8.09% of the sub-divisional area, representing a net loss of 34.5%, while built-up areas expanded from 3.58% to 9.77%. Deforestation was rated as high by 50.56% of respondents across 13 quarters. Income generation was identified as the primary motivation for charcoal production (70%), which sustains livelihoods in the 50,000 - 100,000 FCFA/month range for 48% of producers. Conclusion: Charcoal production in Bamenda I Sub-Division has yielded both material livelihood benefits and cumulative geographical damage, manifesting as forest loss, land cover transition, and the erosion of ecosystem services. Recommendation: Policy responses should prioritize afforestation, adoption of improved kiln technologies, promotion of alternative energy sources, and the establishment of regulated forest governance frameworks to mitigate environmental impacts and enhance the sustainability of peri-urban energy systems.

Urban green spaces are increasingly threatened by rapid urban expansion, making their continuous monitoring and prediction essential for sustainable urban management. This study investigates the spatiotemporal dynamics of urban garden landscapes in Shiraz, Iran, by integrating multi-temporal Landsat imagery, GIS analysis, and CA–Markov modeling. Landsat data from 2003, 2013, and 2023 were processed to derive the Normalized Difference Vegetation Index (NDVI), which was classified into four vegetation-density categories to quantify land-cover transitions. A CA–Markov framework implemented in IDRISI TerrSet (Version 20.0) was then employed to simulate spatial dynamics and predict vegetation changes for 2033. Results reveal a significant expansion of non-vegetated areas from 711.93 ha in 2003 to 976.66 ha in 2023, accompanied by a decline in dense vegetation from 403.68 ha to 382.64 ha. Model projections indicate a further reduction in dense vegetation to 239.35 ha by 2033, suggesting ongoing fragmentation of urban green infrastructure driven by development pressures. By combining time-series remote sensing, GIS-based spatial analysis, and predictive modeling, this study provides an integrative framework for detecting, interpreting, and forecasting urban land-cover change. The findings offer evidence-based insights to support sustainable urban planning, green infrastructure protection, and climate-resilient city management in rapidly growing urban environments.

Over the past decades, the combined effects of climate, in terms of thermo-pluviometric variability, and land-cover transformations have profoundly reshaped Mediterranean rural landscapes, altering their ecological balance and geomorphological stability. These pressures are particularly critical in badland environments, fragile erosional systems marked by steep slopes, limited vegetation, and dense networks of gullies and mass-wasting features, often associated with rural landscapes. This study investigates four decades (1984–2023) of vegetation dynamics within the Orcia River catchment in Tuscany (Italy), a UNESCO World Heritage Site containing representative sub-humid badlands, with the aim of explaining these dynamics in relation to climate variability and land-use changes. Monthly Normalized Difference Vegetation Index (NDVI) time series derived from Landsat imagery were analyzed together with MODIS Land Surface Temperature (LST), Global Precipitation Measurement (GPM) data, and ground-based meteorological records. Regional Land Use and Land Cover (LULC) maps were analyzed to evaluate land-cover transitions. Long-term vegetation and climate trends were assessed using the Mann–Kendall test and Sen's slope estimator on NDVI pixels filtered using the Pixel Quality Assessment (PQA) band and on climatic satellite and ground-based datasets (2004–2023). Results revealed a steady and statistically significant greening trend across the study area since 1984, with most pixels showing NDVI increases of about 0.005 units per year. Vegetation changes were mostly insignificant or slightly negative during 1984–2003, whereas a more pronounced increase occurred after 2004 (0.005–0.010 units yr−1, up to 0.015 along river corridors). Satellite and ground-based meteorological station data showed a warming trend, particularly in minimum temperatures, whereas precipitation remained generally stable. About 42% of the area underwent land-cover change, primarily due to forest expansion and a reduction in grazing and sparsely vegetated areas, mostly linked to the badlands surface. NDVI–LULC intersection analyses showed that most NDVI variations coincided with land-use transitions, while NDVI increases in stable areas suggest management-driven change in vegetation cover. This greening process seems to be enhanced by the registered warming trend, as suggested by the positive correlation between MODIS-NDVI and MODIS-LST values within the southern forested region. Our findings demonstrate the synergistic influence of climate warming and land-cover change on Mediterranean landscapes. The general greening process, which leads to a reduction of badland surfaces, appears to be driven by more favourable climate conditions, which enhance vegetation vigour, and by the abandonment of traditional agricultural practices, allowing natural vegetation recovery.

Land cover (LC) change is reshaping terrestrial ecosystems and profoundly impacting sustainable development in Africa, yet the long-term, continental-scale spatiotemporal dynamics of these shifts remain obscured. To address the above issue, this study systematically explores the spatiotemporal dynamics of LC across Africa from 1985 to 2022 by leveraging the fine-resolution remote-sensing-derived GLC_FCS30D LC dataset within a stratified Intensity Analysis framework. To decompose landscape changes into interval, category, and transition levels across five climatic sub-regions of Africa, we systematically evaluate the temporal consistency of land systems. This hierarchical approach disentangles systematic transition pathways from random fluctuations, thereby revealing the distinct regional regimes governing continental transformation of LC. Our results ultimately show a strong LC change acceleration in Africa after 2010, mainly in Southern, Eastern, and Western Africa, which together made up 80 to 90% of the continent's LC dynamics. During the whole study period, shrubland and grassland had the highest gross turnover due to their high bidirectional volatility. Intensity-wise, forest remained inactive even though it was a persistent net loser to crop in East Africa (2010-2020), to shrub in Southern Africa (1990-2022), and to wetland in West Africa during the post-2000 intervals. Wetland had a major change in dynamics from historical growth during 1985-1990 to systematic decline in 2015-2022. Cropland increased by systematically targeting shrubland and grassland, mainly in East Africa. Additionally, the Sahel contributed 40% of continental grassland to bare area transitions, despite some recovery of grassland in the region. These findings show that aggregate net-change metrics obscure the volatility in African LC; therefore, distinct regional regimes such as agricultural expansion and forest degradation necessitate spatially differentiated management strategies.

Monitoring land cover dynamics and understanding vegetation responses to climate change are critical for ecological assessment and management in dryland regions. This study systematically analyzes land cover dynamics, vegetation type transitions, and their climatic drivers across Asian drylands from 2001 to 2022 by integrating MODIS land cover data, TerraClimate climate reanalysis datasets, and the Google Earth Engine (GEE) platform. Using a unified framework that combines land cover dynamic indices, transition probability and transfer matrix analyses, and climate attribution, we quantify spatiotemporal change patterns and identify dominant vegetation transition pathways. The results reveal pronounced land cover changes across Asian drylands over the past two decades, characterized by expansions of grasslands (GRA), savannas (SAV), croplands (CRO), and water, snow, and ice (WSI), alongside contractions of shrublands (SH), mixed forests (MF), permanent wetlands (WET), and barren land (BAR). Land cover transition analysis indicates that the most prominent conversion pathways are from barren land to grasslands and from grasslands to croplands, reflecting the combined influences of climate variability and land use processes. Climate attribution analyses further demonstrate that vegetation dynamics across different stability zones exhibit distinct responses to long-term climate trends, with increasing maximum temperature, soil moisture, and vapor-related variables, together with declining precipitation, drought indices, and surface radiation, jointly shaping vegetation persistence, expansion, or degradation. By integrating long-term multi-source datasets and cloud-based geospatial computing, this study provides a scalable and reproducible framework for assessing land cover change and vegetation stability in arid and semi-arid regions. The findings enhance understanding of dryland ecosystem dynamics under climate change and support large-scale ecological assessment in data-scarce environments.

The coastal mid-Atlantic region of the United States is increasingly vulnerable to soil salinization, primarily driven by sea-level rise and powerful coastal storms, posing a threat to farmland productivity, and ecological stability. However, the spatially heterogeneous nature of salinization across different land covers makes it challenging to monitor their interactions across large areas and longer time periods. To address this gap, we combined remote sensing-based land cover classification with modeled soil salinity data to assess landscape-scale dynamics across the Delmarva Peninsula from 2000 to 2016. Using a Random Forest classifier trained on Continuous Change Detection and Classification (CCDC)-derived synthetic Landsat surface reflectance, we generated gridded land cover datasets for five years (2000, 2002, 2005, 2009, and 2016) to match and compare with the existing Global Soil Salinity Maps. Overall, forests and other vegetation expanded, whereas farmland and bare soil declined. Salinization trend across these land covers is neither uniformly optimistic nor categorically alarming. Our results showed that over 75% of Delmarva remained in the non-saline category in those 5 years, increasing by 1,138km², and extremely saline zones declined by 833km². More than 83% of land cover transitions occurred without changing salinity categories, while 7-11% moved to a lower salinity category. Our findings based on these temporal snapshots reveal fluctuations in salinity across different land covers, underscoring the value of multi-temporal remote sensing for continuous monitoring of salinity-driven land changes.

Mapping land cover, monitoring its changes, and simulating future alterations are essential tasks for sustainable land management. These processes enable accurate assessment of environmental impacts, support informed policymaking, and assist in the planning needed to mitigate risks related to urban expansion, deforestation, and climate change. This study proposes a GeoAI-based framework leveraging Multilayer Perceptron (MLP), a class of Artificial Neural Networks (ANNs), to predict land cover changes in the Aosta Valley region (NW Italy). The model uses Copernicus Earth Observation data, specifically Sentinel-1 and Sentinel-2 imagery, and is trained and validated on land cover maps derived from different time periods previously validated with ground truth data. The objective is to provide a predictive tool capable of simulating potential future landscape configurations, supporting proactive regional land use planning including regulatory constraints under the current land use plan. Model performance is evaluated using accuracy metrics. The land cover classification methodology follows established approaches in the scientific literature, adapted to the specific geomorphological characteristics of the Aosta Valley. To explore and visualize potential future land cover transitions, Sankey and chord diagrams are used in combination with zonal statistics and thematic plots. These provide detailed insights into the intensity, direction, and magnitude of landscape dynamics. Training data were stratified-sampled across the study area, covering a diverse set of land cover classes to ensure robustness and generalization of the MLP model. This GeoAI approach offers a scalable and replicable methodology for anticipating land cover dynamics, identifying vulnerable areas, and informing adaptive environmental management strategies at the regional scale, while simultaneously considering the latest urban planning regulations.

Understanding land-use and land-cover (LULC) dynamics within protected areas (PAs) is fundamental for assessing conservation effectiveness and ecosystem resilience under increasing anthropogenic and climatic pressures. This study examines the spatio-temporal evolution of LULC across Bavaria’s protected areas between 2000 and 2023 by integrating categorical land-cover data, satellite-derived vegetation indices, and environmental drivers. Annual LULC changes were first quantified using MODIS MCD12Q1 land-cover classifications to evaluate class persistence, transitions, and area trajectories and were subsequently interpreted alongside 16-day MODIS NDVI and SAVI composites to assess associated vegetation greening and browning trends. Ecological stability was characterized by using class-level persistence indicators, coefficients of variation (CVs), and linear trend slopes. The results reveal a marked greening signal after 2010, coinciding with pronounced land-cover transitions, including a decline in evergreen needleleaf forests (−480.6 km2; −32.2%) and substantial expansion of deciduous broadleaf forests (+390.8 km2; +106.1%) and grasslands (+275.8 km2; +28.4%), while wetlands experienced a severe contraction (−203.4 km2; −73.7%), indicating heightened hydrological sensitivity within protected ecosystems. Correlation analysis further indicates that anthropogenic pressure, quantified using the human footprint index, remains a dominant driver of change in croplands and urban areas, even within legally protected boundaries. Overall, this study demonstrates that vegetation trends, land-cover transitions, climatic exposure, and human pressure jointly shape ecological stability in protected areas, highlighting the value of an integrated indicator-based framework.

ABSTRACT Sustainable Development Goal (SDG) 15.3.1, ‘Proportion of degraded land,’ assesses land degradation using three core sub-indicators: land productivity, land cover (LC), and soil organic carbon. The LC sub-indicator evaluates global trends through a transition matrix defined in the UNCCD Good Practice Guidance (GPG), classifying land cover changes as degradation, improvement, or stability. However, this approach does not capture intra-class transitions (e.g. changes occurring within the same LC class), potentially overlooking relevant degradation or improvement processes. To address this limitation, we expanded the original 7-class transition matrix by incorporating 44 CORINE Land Cover (CLC) detailed level-three classes, integrating biophysically weighted MODIS NDVI data (2000–2018). This enhanced framework enabled the detection of intra-class changes amounting to 1.06% (artificial), 0.96% (agricultural), 7.32% (natural), and 0.03% (wetland/water bodies). Notably, within the natural class (CLC 3) 8.53% of the area exhibited improvement, 14.9% degradation, and 76.5% stability. The workflow of the analysis using the 44 CORINE CLC detailed level-three classes for the year 2000 and 2018, the monthly NDVI were obtained from Google Earth engine (GEE) and cumulative averages computed, a percentile raking was then used quartiles to split the distribution into four equal parts (25%, 50%, 75%, and 100%) to reduce the number of CLC detailed level-three classes into a LC type. The ranking of the LC types was taken from the good Practice guidance (GPGs).

Human-driven land-use change has significantly altered forest ecosystems and carbon dynamics in mountainous regions. This study aims to quantify land cover transitions and associated forest carbon stocks changes in Bhutan. It also seeks to support the development of a national measurement, reporting, and verification system. Using Landsat-based satellite imagery and object-based image classification techniques, we assessed forest cover transitions, stand structure variations, and forest type changes across temporal intervals. The analysis revealed a consistent increase in agricultural and built-up areas. It also showed a concomitant decline in coniferous forest cover. In particular, agricultural land increased by approximately 0.77 million ha, while coniferous forest decreased by approximately 0.19 million ha over the study period. These changes were driven by both climatic shifts and socio-economic factors. Approximately 57% of Bhutan’s population depends on agriculture. Correspondingly, forest carbon stocks declined from approximately 570 million tC in 1995 to 405 million tC in 2017. This decline was largely attributed to coniferous forest loss and climate-induced mortality. Bhutan has made significant preparations for the implementation of the Warsaw REDD+ framework under the United Nations Framework Convention on Climate Change. These preparations include the establishment of a forest reference emission level for submission. However, challenges remain in detecting small-scale land use changes. Additional challenges include addressing spectral misclassification in mountainous regions. Our study provides a scientific baseline to support national forest monitoring and carbon accounting systems. It also offers policy-relevant insights for achieving Bhutan’s nationally determined contributions and enhancing its carbon sink potential.

Abstract Land Use Land Cover (LULC) changes have significant environmental and socioeconomic implications, necessitating systematic quantification and assessment. This study analyses LULC transitions over two time periods, 1999-2009 and 2009-2019, focusing on key land categories: Forest (FRSD), Agriculture (AGRL), Water Bodies (WATR), Rangeland (RNGE), and Urban Land (URLD). The results indicate substantial land transformations, with agricultural land expansion (637 km², 8.07 annual gain intensity in 2009-2019) occurring at the expense of forest cover (597 km², 1.14 annual gain intensity in 1999-2009). Urban expansion showed a significant increase in intensity, particularly in 2009-2019, with URLD growing by 54.2 km² (9.17annual gain intensity). Hydrological changes were evident, with water body reduction from 135 km² in 1999-2009 to 154 km² in 2009-2019, significantly impacting ecosystem stability. The transition analysis reveals substantial conversions, including 115 km² of agricultural land shifting to forest cover in 2009-2019, while 10.5 km² of water bodies transitioned to forested areas. These findings underscore the dynamic nature of LULC changes driven by anthropogenic and environmental factors, highlighting the need for sustainable land management strategies.

Gabon harbors one of Africa's richest assemblages of non-human primates (NHPs), yet integrated national-scale evidence on their conservation status remains limited. To inform conservation strategies, we conducted the first nationwide assessment integrating habitat dynamics, the geographic distribution of species, and the effectiveness of the protected-area network in the country. We harmonized 300 m land-cover maps (ESA CCI 1992; Copernicus 2022), compiled 481 georeferenced occurrences, and identified concentration areas using kernel density estimation and Getis-Ord Gi* analysis. We quantified land-cover transitions with a per-pixel transition matrix and assessed protected-area capture using Monte Carlo randomization. Ten fully protected species are confirmed, including Gorilla gorilla and Pan troglodytes. Occurrences concentrate mainly in the Ogooué-Ivindo and Haut-Ogooué Provinces; ~10% of the national territory lies above the 90th kernel density percentile (≈26,700 km2), and 1.5% of cells qualify as hotspots at the 99% threshold. Primate records are strongly associated with evergreen broadleaved forests (87.9% of points), which remained persistent from 1992 to 2022 (forest-to-forest = 223,476 km2; 98.13%) with a net decline (-2571.66 km2; -1.19%). Gross losses (4046.58 km2) were mainly attributable to agricultural conversion (68.63%; χ2 = 31,525; p < 0.001). Over 90% of records fall in areas stable across 1992-2022. Protected areas (PAs) captured more occurrences (observed 40.1% vs. expected 18.47%; p < 0.001), yet gaps remain for some taxa (e.g., Allochorocebus solatus, 86% outside PAs). Overall, Gabon retains an extensive core of suitable habitat, but targeted action outside PAs and maintenance of landscape connectivity are needed to secure populations where agricultural expansion and fragmentation are intensifying.

Background The interplay between climate, vegetation and human activity hinders the analysis of fire regimes in socio-ecological systems. With a monsoonal and variable rainfall regime, Argentine Dry Chaco is one of the largest dry forests worldwide and an important agricultural frontier advancing at forest detriment. Overall, fire plays a key role in shaping the landscape and in land management. Aims We describe land cover (LC) transitions and temporal trends in fire frequency, analyze fire-mediated transitions probability, and assess LC and climate influence on fire occurrence in the Argentine Dry Chaco. Methods Employing data generated through remote sensors, we conducted spatially explicit temporal reconstructions of annual fire activity (FA), LC changes and climate conditions from 2000 to 2023. Key results Agriculture and pastures expanded while forests lost ~17% of their initial area; annual FA showed a significant declining trend, with grasslands, pastures and agriculture having the highest burn probability. When fire occurs, forests, shrublands and pastures increase their likelihood of transitioning to agriculture. Fire occurrences within managed lands are decoupled from climatic influences. Conclusions Decrease in FA is driven by a transformation in agricultural systems, moving from traditional to mechanized management. Implications Transdisciplinary approaches are crucial for analyzing the purposes of fire use.

The Western Ghats of India support a diverse assemblage of Ixodid ticks, including multiple vectors of Kyasanur Forest Disease (KFD), yet comprehensive data from forest-fringe landscapes outside recognised outbreak zones remain limited. To address this gap, we undertook a multi-state, cross-sectional survey across 44 forest-fringe villages distributed within seven ecological grids spanning Goa, Maharashtra, Karnataka, Kerala, and Tamil Nadu during peak nymphal activity (December–February). Standardized flagging method yielded 10,350 ticks representing 24 taxa across four genera, including 12 recognised KFD vectors. Haemaphysalis spinigera and Haemaphysalis turturis were the most abundant and widespread species in both historically affected and unaffected districts, indicating that ecological suitability for primary vectors extends substantially beyond the spatial extent of human case reporting. Species richness, Shannon diversity, Simpson diversity, and total abundance exhibited pronounced spatial heterogeneity, with multiple villages in Karnataka and Kerala supporting the highest diversity, while clusters in Goa and Maharashtra exhibited depauperate assemblages dominated by one or two Haemaphysalis species. These community patterns suggest fine-scale variation in habitat suitability within forest-fringe mosaics shaped by land-cover transitions and vegetation structure. Joint Species Distribution Modelling (JSDM) revealed that tick assemblages respond to interacting climatic, habitat, terrain, and temporal gradients. Precipitation, NDVI-derived vegetation structure, land-cover categories, slope-derived terrain metrics, seasonal timing, and inter-annual variation collectively influenced species distributions, demonstrating multidimensional environmental filtering. Distinct and sometimes contrasting predictor–response relationships among taxa highlight niche differentiation within the tick community. Moderate residual correlations among several species pairs further indicate shared habitat preferences or unmeasured ecological processes such as host movement patterns, understory complexity, and microhabitat moisture retention. Together, these findings provide the most geographically extensive ecological assessment of Ixodid ticks in the Western Ghats. The widespread distribution of multiple KFD vectors in districts without confirmed human cases, underscores the urgent need for proactive ecological surveillance, integration of fine-scale environmental data, and landscape-based risk forecasting. Strengthened surveillance frameworks incorporating habitat metrics, host information, and pathogen screening will be critical for anticipating shifts in vector distributions and mitigating emerging tick-borne disease threats in the region.

This study examines the spatiotemporal dynamics of urban expansion and agricultural land loss in the peri-urban areas of Addis Ababa, Ethiopia, from 1985 to 2025, and projects future land-use trajectories to 2055 using Geographic Information Systems (GIS), remote sensing, and hybrid MLP–Markov modeling. Multi-temporal Landsat imagery was classified using supervised Maximum Likelihood Classification and post-classification change detection to quantify historical Land Use and Land Cover (LULC) transitions. Results reveal a dramatic decline in cropland from 32.71% (14,209 ha) in 1985 to 10.28% (4468.8 ha) in 2025 alongside rapid expansion of built-up areas from 12.41% (5392.4 ha) to 68.57% (29,789.19 ha). Transition analysis, combined with key spatial drivers (distance to roads, population density, slope, and elevation), was used to train a Multilayer Perceptron neural network to generate transition potential maps. Integrated MLP–Markov simulations predict that built-up land will increase further to 89%, 90%, and 91% by 2035, 2045, and 2055, respectively, reducing cropland to approximately 5% by mid-century. Classification reliability is supported by high ROC-AUC (0.875) and Kappa (> 0.79) values. The findings highlight severe threats to peri-urban agricultural livelihoods, food security, and ecological stability as urban growth intensifies. The study underscores the urgent need for integrated and enforceable land-use policies that balance urban development with the preservation of agricultural and ecological systems, offering evidence-based insights for sustainable urban planning in rapidly growing cities across the globe.

This study adopts an integrated simulation framework based on Cellular Automata and Artificial Neural Networks (CA-ANN) to model land use and land cover (LULC) transitions driven by urban expansion. By combining machine learning with spatial modeling, the approach enables the forecasting of urban growth dynamics and supports data-driven urban planning. The objective is to assess urban sprawl in the city of Passo Fundo, southern Brazil, using LULC change simulations from 2002 to 2043. Satellite imagery from 2002 to 2023 indicate for supervised classification of three land cover classes-urbanized areas, forests, and non-urbanized areas-alongside key spatial variables, including hypsometry, proximity to water bodies, railways, central business districts, and road networks. These variables served as inputs to the CA-ANN model to simulate future land use scenarios for 2033 and 2043. Results indicate a 45% increase in urbanized areas from 2002 to 2023, with projections reaching 66% growth by 2043, absolute land area expansion. This urban expansion primarily occurs at the expense of agricultural and forest areas, underscoring the risks of landscape fragmentation, biodiversity loss, and pressure on agricultural lands. The findings highlight the urgency of integrating spatial intelligence into sustainable land governance strategies, particularly in regions where urbanization intersects with agribusiness territories and food security systems.

ABSTRACT Rammed earth sites are vital heritage structures with considerable historical and cultural value. However, they face severe degradation due to natural erosion and human activities in arid Northwest China. Current methods for their extraction and monitoring remain limited, particularly in integrating multi-source remote sensing with advanced machine learning for precise localization and predictive analysis. In this study, we aimed to address these gaps by developing an accurate boundary extraction framework for rammed earth sites using multi-source remote-sensing data, and simulating their future spatiotemporal evolution to support proactive conservation. We compared three machine learning approaches: Object-Based Image Analysis combined with Convolutional Neural Networks (OBIA-CNN), Maximum Entropy Model-based Discrete Particle Swarm Optimization (MEDPSO), and U-Net-based semantic segmentation. OBIA-CNN outperforms MEDPSO and U-Net, achieving superior accuracy (OA = 97.46%, Kappa = 0.95) with strong anti-interference and generalization capabilities, effectively minimizing salt-and-pepper noise and preserving structural continuity. While achieving a high recall (0.9731), U-Net exhibited boundary expansion and over-segmentation, limiting its precision in delineating fine archaeological features. We applied the Markov-PLUS model to simulate land-use changes around four representative sites from 2023 to 2056 under natural scenarios, incorporating environmental and socioeconomic drivers. The model indicated critical transitions in land cover that threaten site preservation, enabling the identification of high-risk zones. This study provides an integrated framework that bridges high-precision site extraction with spatiotemporal simulation, offering a scientific basis for the sustainable conservation of rammed earth heritage in arid environments.

Cartographic tools for evaluating land cover transitions: implications for landscape management and almond diversity conservation

A new index for monitoring non-linear trends in passive forest recovery at regional scale