Oases are water-dependent landscape systems that support ecological processes and human activities in drylands. However, a global dataset delineating oasis boundaries and spatial distribution is still lacking. We constructed a global oasis vector dataset for the reference year 2020 through standardized visual interpretation of metre- to sub-metre-scale Google Earth imagery, supported by multiple auxiliary datasets. The dataset contains 3,437 oasis patches covering 2,272,595.47 km2. Field-survey validation and random hexagonal-grid validation yielded overall accuracies of 94.88% (95% CI: 93.12–96.20%) and 93.47% (95% CI: 93.02–93.90%), respectively. The results show that Asia contains the largest share of global oasis area, accounting for 65.67% of the total. Global oasis area is concentrated in a small number of countries, with the 10 countries having the largest oasis areas accounting for 66.77% of the global total. The Green Region class occupies 66.00% of the raster-overlay area within the oases, with cropland representing the dominant land-cover class. The dataset provides a reusable spatial baseline for dryland research, with potential applications in sustainability assessment, cross-regional synthesis, land-cover product validation, and oasis dynamics monitoring.
Sand dams are small water-harvesting structures that store water within the sandy matrix of ephemeral riverbeds. They are widely promoted as nature-based solutions for water security and climate adaptation in drylands. Nonetheless, it remains difficult to systematically assess their environmental and socio-ecological impacts for multiple reasons. This review synthesizes Earth observation and geoinformatics research on sand dams and evaluates how reliably their impacts can be detected from spatial data. A structured literature search and citation tracking identified 11 relevant studies published between 2008 and 2026. Evidence on impacts is dominated by vegetation effects: six studies assess vegetation responses, five report localized greening, higher vegetation vitality, or greater drought resilience, while one finds no average greening. Direct hydrological evidence is less prominent: gravimetric and radar-based approaches could not fully resolve local storage effects, and long-term regional responses remain inconclusive. Geomorphological evidence is limited to one study, while thermal approaches remain largely exploratory. Across these studies, differences in spatial scale, control-site selection, dam maturity, and temporal design constrain comparability and attribution. We develop a methodological taxonomy and propose future research based on time series, Before-After-Control-Impact designs, climatic controls, in-situ validation, multi-sensor integration, automated inventories, and data fusion supported by artificial intelligence.
Rubber is both a strategic material and an essential industrial raw commodity, making accurate yield estimation critically important. Current remote sensing-based physical models require enhancement to converting potential yield into actual yield using limited field samples, while improving parameter applicability and estimation accuracy. Integrating DEM, Sentinel-2 imagery, and meteorological data, this study develops a multi-stage framework that couple the CASA model, SOLARFLUX, a conversion module, and a parameter calibration scheme. Calibrated and validated with field-measured data, the model enables high-resolution dynamic monitoring of dry rubber yield across Hainan Island. Validation demonstrates strong model performance (R2 > 0.95), with relative error (RE) reduced from 21.99 % to 16.62 % after refinement. The model substantially outperforms machine learning approaches. Results reveal a year-on-year decline in total rubber yield from 2021 to 2023, followed by a marked rebound during the 2023-2024. The proposed framework offers a robust and practical solution for data‑scarce regions, and its products provide valuable support for the sustainable development and ecological monitoring of Hainan’s rubber industry.
Vegetation dynamics across the Tibetan Plateau are shifting under climate change, yet our understanding remains hindered by uncertainties among satellite products. We assessed the consistency of seven Leaf Area Index (LAI) products (2001–2018) in capturing mean states and spatiotemporal trends: MODIS, GIMMS, GLOBMAP, GLASS-AVHRR, GLASS-MODIS, HIQ, and GEOV2. Results revealed high agreement on climatological means (Pearson's r > 0.75, p < 0.001), whereas agreement was much lower regarding long-term dynamics (r < 0.6). Only GLOBMAP, GLASS-MODIS, and GEOV2 identified significant greening (0.0059, 0.0052, and 0.0093 m2 m−2 yr−1, p < 0.05), while MODIS, GIMMS, GLASS-AVHRR and HIQ showed non-significant trends. Spatial trend heterogeneities peaked in southeastern forest-dominated regions (<2200 m), which may be related to complex canopy structures. In pixel‑scale analysis, 77% of the vegetated areas lacked consistent and significant trends. Only 0.66% showed unanimous greening across all products, mainly in the northeastern Tibetan Plateau. High consistency occurred in a warm-dry core (10–15°C, 50–150 mm) but weakened toward colder or wetter regimes. These patterns likely reflect sensor disparities, retrieval algorithms, and the extreme environments. These findings provide a comprehensive assessment of the consistency among multiple satellite LAI products and have important implications for monitoring vegetation responses to environmental change.
Reliable view selection is essential for accurate deep multi-view stereo (MVS) reconstruction from satellite imagery. However, existing strategies mainly rely on acquisition metadata or simplified geometric criteria and therefore cannot adequately characterize the heterogeneous reliability of satellite views under complex geometric, radiometric, and temporal conditions. This paper proposes a reliability-aware view selection framework that evaluates candidate source views by jointly considering geometric strength, radiometric consistency, RPC-derived geometric stability, and temporal reliability. Unlike conventional angle-based geometric criteria, the proposed framework characterizes the spatial variation in RPC-derived epipolar orientations across the image domain, thereby providing complementary geometric information for identifying more reliable multi-view configurations. Experiments on the US3D benchmark show that the proposed framework improves reconstruction performance in terms of MAE and PAG1 relative to conventional metadata-driven and simulation-based selection strategies, reducing MAE by up to 9.9% and increasing the Percentage of Accurate Grids at 1 m (PAG1) by up to 8.5%. RMSE exhibits region-dependent behavior, decreasing in JAX but increasing in OMA. Additional evaluations on the MVS3D benchmark and cross-backbone experiments using two satellite MVS architectures further support the transferability of the proposed framework across the evaluated urban environments and reconstruction architectures.
