Carbon sources and sinks are crucial to climate regulation. Megacities, as major emitters, encounter increasing pressure from rapid land-use changes and ecological degradation. These challenges are prominent in mountain–basin transition regions owing to fragmented terrain, ecological sensitivity, and accelerated urbanization. However, the combined effects of climate and landscape patterns on carbon sinks in these regions remain poorly understood. This study investigated the spatiotemporal dynamics (2000−2020) and combined effects of climate and landscape patterns on carbon sinks across production–living–ecological spaces in a typical transition-region megacity. Net ecosystem productivity (NEP) was used as a carbon sink indicator. An interpretable machine learning model, eXtreme Gradient Boosting with Shapley Additive Explanations, was applied to quantify the independent and interactive effects of four key climatic factors (temperature, precipitation, potential evapotranspiration, and sunshine duration) and landscape patterns. Climatic factors exerted stronger effects on NEP than on landscape patterns, with moderate temperatures, sufficient rainfall, and sunshine enhancing sequestration. The key landscape pattern indices were the percentage of landscape, landscape shape index, edge density, and splitting index, with living space being the most sensitive. Climate–landscape interactions were context-dependent: fragmentation and edge density reduced carbon sinks but were potentially mitigated under optimal climate conditions. Improved landscape patterns further enhanced NEP, highlighting the importance of expanding and connecting ecological spaces, using a “large-scattered, small-clustered” layout in production spaces, and integrating green infrastructure in living spaces. These findings inform landscape planning and management to optimize land-use structures and strengthen carbon sinks in megacities under climate change.

Invasive alien plants, particularly trees, are a global problem because invasions have ecological, social, and economic impacts. Due to the rapid rate of tree invasions, continuous monitoring for management is important. Airborne hyperspectral imagery is ideal for monitoring but expensive to acquire, often prohibitively so for continuous monitoring. Conversely, spaceborne multispectral imagery is freely available and often has high temporal resolution, but the trade-off is lower spatial and spectral resolutions. Freely available spaceborne hyperspectral imagery has higher spectral resolution, but lower spatial resolution. In resource-constrained regions, freely available remote sensing imagery is often the only option for monitoring. To address trade-offs between spatial and spectral resolution in invasive alien tree mapping for resource-constrained regions, this study compares four sensors that are freely available in South Africa. These sensors cover a wide range of spatial and spectral resolutions (0.25–60 m and 3–285 bands respectively). The use of data fusion to overcome this trade-off was also explored. We found that all three spaceborne sensors performed similarly, with SPOT6 marginally resulting in the highest accuracy overall and for discriminating among alien taxa. Sentinel-2 exhibited the highest accuracy for discriminating alien taxa from other classes. Notably aerial photography collected as part of government airborne campaigns performed poorly. Data fusion of EMIT with Sentinel-2, i.e. imagery with higher spectral and higher spatial resolutions respectively, led to a marked improvement in accuracy (∼ 5 %). Therefore, in resource-constrained countries, using any of these spaceborne sensors for invasive alien tree mapping may provide robust classification results, but data fusion of sensors with high spatial and spectral resolutions is highly recommended.

Arid ecosystems remain under-mapped at actionable scales despite their ecological importance. Decision makers lack reliable, high-resolution habitat maps in drylands to prioritize protection and target restoration. This research integrates spaceborne hyperspectral imaging from the Environmental Mapping and Analysis Program (EnMAP) with deep learning semantic segmentation models to produce an updated level of habitat classification based on the International Union for Conservation of Nature (IUCN) for part of the Imam Turki bin Abdullah Royal Reserve, Saudi Arabia. Using ground control points and the full EnMAP spectral cube without band selection, U-Net and DeepLabV3+ architectures were each implemented with VGG19 and ResNet-101 encoder backbones, resulting in four model configurations for comparative evaluation. Among the tested models, U-Net with a VGG19 backbone achieved the highest performance, attaining an F1 score of 0.90, demonstrating superior capability for habitat mapping in arid environments. The resulting map segmented the dendritic wadi network, Rawdat depressions, and sand-plateau contacts with sharp boundaries. Aligned with the IUCN habitat scheme, Rawdat, seasonal vegetation-bearing desert carbonate sinkholes, are introduced as a new IUCN habitat class. The map produces verifiable indicators relevant to Sustainable Development Goals 13 (Climate Action) and 15 (Life on Land), supporting protection, restoration targeting, and monitoring. Therefore, the proposed deep learning hyperspectral framework can be applied to other arid and semi-arid regions worldwide to upgrade terrestrial ecosystem mapping and conservation planning. • EnMAP with U-Net yield precise arid habitat map using all 224 hyperspectral bands. • U-Net-VGG19 surpassed both DeepLabV3+ variants and U-Net-ResNet101 (F1 = 0.90). • Rawdat is introduced as a distinct, IUCN-aligned Level-3 habitat class. • Map captures Wadis, Rawdat depressions, Sand-Plateau contacts with sharp edges.