Extreme heat is commonly defined using climatological thresholds, but such definitions do not directly capture when daily activity begins to weaken under thermal stress. This study estimates a mobility-based behavioral benchmark for heat and examines whether post-benchmark mobility decline varies across neighborhood socioeconomic contexts in the contiguous United States. Using anonymized census tract-day mobility data from 2018 to 2024, linked to daily weather conditions, we analyze warm-season mobility responses to temperature with high-dimensional fixed effects, spline Poisson models, and piecewise Poisson threshold models. The pooled national mobility-temperature relationship is nonlinear. Predicted mobility rises through a moderate warm range, peaks near 26.7°C, and then declines at higher temperatures. In the pooled piecewise model, the benchmark turning point is estimated at approximately 23.0°C, with an uncertainty band of 22.0–26.0°C. Income-stratified models show that heterogeneity is expressed more clearly in post-benchmark slope magnitude than in benchmark location, with the steepest post-benchmark decline appearing in the lowest-income quartile, while higher-income quartiles show more moderate declines. Tract-level maps of post-benchmark heat exposure and modeled mobility burden further show that larger modeled burdens appear across parts of the South and Southwest, although substantial within-region variation remains. These results suggest that behavioral disruption under heat reflects thermal exposure, local context, and unequal neighborhood-level capacity to sustain routine activity once temperatures move beyond a broadly shared benchmark. This approach complements climatological heat definitions with an empirically estimated mobility-sensitive threshold.
With the surging adoption of deep learning models, multi-source remote sensing imagery has been increasingly explored to offer a data-driven approach for large-scale flood event analysis and mapping. However, DL-based methods usually suffer from a scarcity of high-quality labeled datasets used for effective model training. Accurately capturing imagery of actual flood-affected areas from massive, dispersed satellite data archives remains significantly challenging under traditional spatiotemporal retrieval paradigms. This challenge arises from insufficient knowledge of flood event semantics, which leads to imprecise spatiotemporal retrieval ranges and difficulty in discriminating flooded areas from permanent water bodies. To address these issues, we propose a semantic-augmented multimodal remote sensing big data retrieval method for analyzing large-scale flood events. It employs a two-phase retrieval strategy: semantic-augmented metadata-based pre-screening across distributed data archives, which refines flood event semantics from the Global Flood Database and Internet news, and DL-enabled multimodal image retrieval for further identification of flood-related imagery, jointly leveraging flood-event-augmented textual semantics and visual features from optical and SAR imagery. Experiments on MFED show that the method achieves an F1-score of 0.965; semantic augmentation improves retrieval accuracy by 11.5%, and multimodal retrieval outperforms single-modality approaches by 28.2%. Across 15 global regions, it reaches 91.2% Top-5 accuracy with a second-level latency.
Accurate and timely crop type mapping offers a wide range of benefits, from supporting annual crop inventories and agricultural monitoring to enabling sustainable resource management and a deeper understanding of landscape dynamics. This study proposes a practical framework for mid-season parcel-level crop classification utilizing models derived from past Earth Observation (EO) and training data, eliminating the necessity for additional training data collection for future inference. Utilizing Sentinel-1 SAR and Sentinel-2 optical data with a 21-day interval from April 1 to August 4, major crop types were mapped in the province of Quebec during the growing season spanning 5 years (2020–2024). Finally, the most accurate trained models were used to predict the crop types of 2025 to ensure the accuracy and transferability of the trained models. This study took into account several important crops, such as corn, soybeans, wheat, and hay. Two machine learning models and four deep learning models, including Transformers, were utilized for multitemporal crop mapping. Additionally, crop mapping was performed using Google’s AlphaEarth annual dataset and ML algorithms, and the results were compared with Sentinel-1 and Sentinel-2 data. Our findings indicated that the Transformers outperformed other models when using Sentinel-1 and Sentinel-2 data, achieving an overall accuracy of 0.97 on the validation data and 0.93 on the test data. Using Google’s AlphaEarth data, RF delivered a comparable accuracy, with overall accuracies of 0.96 and 0.92 for the validation and test data, respectively. The generated crop type maps for unseen 2025 data in this study show the practicality of the proposed framework to perform mid-season parcel-based crop mapping without the need for additional in-year field data.
The increasing frequency and intensity of floods pose substantial risks to residential areas, infrastructure, and economic systems, highlighting the need for reliable flood mapping to support emergency response and damage assessment. Even though synthetic aperture radar (SAR) facilitates all-weather flood monitoring, speckle effects and intricate surface backscattering mechanisms make precise flood delineation challenging. To address these, we propose a hybrid flood-mapping framework centered on a Mixture-of-Experts Spatially-Aware Gated Enhancement U-Net (SAGE-UNet) for post-event segmentation using Sentinel-1 SAR imagery. This architecture incorporates polarization-specific denoising to minimize the impact of speckle effects and a spatially aware bottleneck to improve the representation of contextual features. To further reduce false alarms and missed detections, particularly in vegetated regions, we introduce a lightweight bi-temporal refinement module that combines discrete wavelet transform (DWT)-based change detection with context-guided morphological reconstruction using normalized difference vegetation index (NDVI)-derived vegetation constraints and hazard-related spatial priors. The framework was evaluated on two major European flood events (Hannover, Germany, December 2023; Valencia, Spain, October 2024). The experimental results demonstrated that the standalone SAGE-UNet achieved IoU improvements of 1.7% and 2.3% over the strongest CNN- and transformer-based baselines on the Hannover and Valencia datasets, respectively. After bi-temporal refinement, the complete framework achieved relative IoU improvements of 7.8% and 15.8%, over the strongest competing methods on the Hannover and Valencia datasets, respectively. The refinement also improved spatial consistency by reducing false alarms and missed detections and better delineating flooding beneath vegetation. LULC analysis also confirmed the framework’s utility for flood exposure assessment, identifying agricultural areas and infrastructure within floodplains as the most affected classes. These results demonstrate the framework’s potential for operational SAR-based flood mapping and impact assessment.
The impacts of the modifiable areal unit problem (MAUP) on the analysis results have long been documented in empirical studies of many disciplines, in particular, the diversified appearance and degrees of sensitivities of geographical variables and observations. Despite diligent efforts, there remains epistemological insufficiency in expressing sensitivities to the MAUP in a unified and consistent manner, compromising the clarity and reliability of the analysis results. To address this issue, a conceptual and methodological framework has been proposed to systematically define and quantify the sensitivities of a geographical observation to different aspects of the MAUP in a probability-based approach, exemplified by demographic, socio-economic, and environmental datasets. The proposed framework aims to provide a theoretically grounded foundation for unambiguous communication and more in-depth investigations of MAUP-related sensitivity analyses, promoting more rigorous and replicable geographical research involving spatially aggregated data.
Solid Earth tides reflect the Earth’s elastic response to the tidal forces of the Moon and Sun. Studying these tides reveals Earth’s internal structure, response to ocean loading, and the relationship between tides and earthquakes. Solid Earth tides are also essential for understanding sea level changes and establishing international reference frames. Since the time of Lord Kelvin, scientists have generally accepted that the tidal forces of the Moon and the Sun induce periodic deformations on the solid Earth’s surface, which can be simulated using a superposition of ellipsoids. However, quantitative attempts to determine the geometric parameters of this ellipsoid, such as the lengths of the semi-major and semi-minor axes and the flattening ratio, have not been realized. This study aims to introduce and optimize a previously proposed high-precision solid Earth tide rotational ellipsoid geometric (Geo) model, which accurately characterizes solid Earth tidal displacements by imposing two rotational ellipsoids corresponding to the Moon and the Sun. Using data from four superconducting gravimeter (SG) stations, we optimized the Geo model’s parameters. Compared to mainstream models (Etideload and Solid/pyTMD), the Geo model demonstrated root mean square error, amplitude difference, phase lag difference and vector difference: 1.08 cm, 0.17 cm, 1.77° and 0.22 cm, respectively, versus 2.74 cm, 0.34 cm, 4.87° and 0.47 cm for Etideload, and 3.14 cm, 0.31 cm, 4.25° and 0.40 cm for Solid/pyTMD. The cotidal charts in areas with a denser distribution of SG stations (2°E–22°E，40°N–60°N) show that the results align with previous analyses, demonstrating the effectiveness of the Geo model in simulating solid Earth tide.
Leaf Area Index (LAI) serves as a key biophysical parameter for characterizing vegetation canopy structure and ecosystem functions. To address the absence of LAI products for the Fengyun-3B (FY-3B) satellite and the limitations of current satellite LAI products, this study proposes an LAI retrieval framework integrating a high-quality benchmark library with machine learning from Fengyun-3B Visible and Infra-Red Radiometer (VIRR) Data. Under rigorous quality control, we constructed a long-term, high-quality LAI benchmark dataset covering Asia by spatiotemporally fusing and screening the MODIS and GEOV2 products. Random Forest, XGBoost, and MLP regression models were trained and optimized for specific vegetation types, generating an 8-day composite 0.01º LAI product for Asia from 2011 to 2020. Validation against in-situ measurements, MODIS, and GEOV2 LAI products indicates that: (1) the accuracy of the FY-3B LAI product (R = 0.581, RMSE = 1.307) showed improved performance compared to MODIS (R = 0.484, RMSE = 1.831) and GEOV2 (R = 0.391, RMSE = 1.920); and (2) the product achieves seamless spatiotemporal coverage. This dataset provides robust support for ecosystem monitoring in Asia and contributes to the construction of a diversified, multisource, synergistic global satellite observation system. This dataset is publicly available at https://doi.org/10.5281/zenodo.18217405.
Universal access to adequate housing is crucial for preventing diseases and improving human well-being, particularly in flood-affected areas. However, existing studies and data are limited in providing comprehensive estimates of housing adequacy across Africa on fine spatial and temporal scales. In this study, we combined satellite imagery, household survey data, and deep learning to produce high-resolution geospatial estimates of access to adequate housing from 2005 to 2020. We find that the adequate housing index (AHI) in Africa increased from 0.47 (95% CI: 0.47–0.48) to 0.51 (0.51–0.52) between 2005 and 2020, yet the number of people living in inadequate housing (528.9, 525.1–533.2 million) has increased by 36.5% across Africa over the study period. This reflects that rapid population growth outpaced the gains in overall housing adequacy. More than 47.4% of people living in inadequate housing were in floodplains such as the basins of the Niger, Nile, and Congo Rivers, and more people were exposed to high-risk flood areas. Our results demonstrate that the developed approach can fill local-level housing data gaps and identify priority areas in Africa. They provide a benchmark for measuring housing changes in flood-risk areas and help to develop mitigation strategies.
Deep semantic interpretation of high-resolution remote sensing (RS) imagery is critical for refined Earth system monitoring; however, current data-driven approaches are impeded by the "semantic gap" inherent in existing datasets, which typically feature flat taxonomies, coarse categories, and insufficient fine-grained semantic descriptions. To mitigate these limitations, this paper presents LuoJia-FG, a large-scale, attribute-driven, and hierarchical fine-grained land-cover dataset tailored for next-generation vision-language models. The dataset comprises 119,619 multimodal triplets based on Gaofen-2, Ziyuan-3, and aerial imagery, with spatial resolutions ranging from 0.5 to 2.0 m. Furthermore, LuoJia-FG is distinguished by three core innovations. First, it establishes a deep three-level hierarchical taxonomy expanding from 8 primary classes to 52 secondary and 95 tertiary fine-grained classes, offering unprecedented semantic granularity. Second, it bridges the pixel-knowledge gap through a novel attribute-coding system that programmatically generates rich, attribute-driven text descriptions based on physical properties such as phenology and canopy density. Third, to demonstrate the utility of these multimodal annotations, we propose the CLIP-guided Hierarchical Classification Network (CLIP-HCNet) as a robust benchmark. This framework effectively leverages the dataset's attribute-driven text descriptions as semantic priors to resolve visual ambiguity among spectrally similar fine-grained categories. Experiments verify that LuoJia-FG constitutes a challenging testbed and that incorporating attribute-driven semantic priors significantly enhances hierarchical classification accuracy, opening new avenues for text-guided geospatial understanding. The LuoJia-FG benchmark dataset is publicly available at , and the source code is available at .
Fine particulate matter (PM2.5) pollution is a major environmental risk factor contributing to multiple diseases and premature mortality. Although numerous studies have examined the health effects of PM2.5, a critical data gap remains in the availability of high-resolution, openly accessible, gridded datasets that attribute premature deaths to long-term exposure to PM2.5. Here, we constructed a high-resolution, long-term dataset of premature deaths attributable to PM2.5 in China, from 2000 to 2019, integrating multiple PM2.5 sources and two epidemiological models: the Integrated Exposure-Response (IER) model and the Global Exposure Mortality Model (GEMM). The dataset incorporates spatially explicit population distributions, baseline mortality rates, and stratification by age and sex, covering populations aged 25 years and above, and includes major disease categories, such as ischemic heart disease, stroke, chronic obstructive pulmonary disease, lung cancer, lower respiratory infections, and other non-communicable diseases. This study outlines the analytical workflow used to generate premature death estimates and documents the key data characteristics, processing assumptions, validation procedures, and sources of uncertainty. The comprehensive dataset provides a quantitative foundation for nationwide health risk evaluation, enabling policy-relevant analyses of population vulnerability and exposure patterns.
To address physical distortion and "black-box" bottlenecks in Total Column Water Vapor (TCWV) modeling, this study develops a physics-constrained interpretable spatio-temporal attention network (ISTAN). By integrating multi-head self-attention mechanisms with bidirectional long short-term memory (Bi-LSTM) networks, this architecture achieves deep decoupling of large-scale spatial teleconnections and local evolutionary dynamics, while explicitly embedding the vertically integrated water vapor continuity equation as a strong inductive bias. Evaluation results based on multi-source data indicate that ISTAN demonstrates exceptional seasonal stability, achieving a peak spring R-2 of 0.931 (MAE: 1.842 kg/m(2)), representing a 9.79%-16.96% accuracy improvement over PINN and U-Net models. Summer R-2 remains robust at 0.922, and validation across ten typical climatic zones confirms high consistency with GNSS measurements (R-2 > 0.90), effectively overcoming scale-smoothing effects. The physics-constraint mechanism significantly reduced Wasserstein distance from 2.76 to 0.76, correcting non-physical biases in the ITCZ, Amazon Basin, and Tibetan Plateau. Attribution analysis confirms that the model has internalized nonlinear thermodynamic scaling laws following the Clausius-Clapeyron physical principle. Trend predictions for 2025-2029 reveal a phased accelerated TCWV rise driven by global warming, with significant wetting potential in Australia and northern North America. This work provides a robust technical foundation for high-fidelity and interpretable TCWV prediction modeling.
The degradation of croplands due to traditional tillage practices is a significant challenge for sustainable agriculture. Crop stubble retention (CSR), as a key conservation tillage method, is crucial for evaluating cropland sustainability, and the efficient acquisition of CSR category information directly affects tillage strategy optimization. However, existing studies often overlook the acquisition of fine CSR category information. This study proposes a method for CSR category recognition based on spatiotemporal cubes and multi-modal remote sensing data. Taking Lishu County, Jilin Province, China, as an example, we constructed a unified standard spatiotemporal dataset using the Raster Dataset Clean & Reconstitution Multi-Grid (RDCRMG) architecture and developed a spatiotemporal cube-based computation framework. Combined with the Random Forest model, we evaluated the effectiveness of the method by comparing single-modal, multi-modal, and optimal percentile-featured multi-modal remote sensing data. The proposed method achieved an overall accuracy of 86.1% and a Kappa coefficient of 0.79 when using multi-modal data with optimal percentile features. Our results demonstrate that this method significantly improves the accuracy of CSR category recognition, providing precise spatial information and technical support for conservation tillage and sustainable cropland management.
Significant changes in snow, glaciers, and the water system across High Mountain Asia (HMA) under global warming have intensified regional water system instabilities. Accurate mapping of the spatial distribution and connectivity of rivers and lakes is essential for understanding and managing the regional water balance and ecosystems. However, existing surface-water datasets lack clear distinctions and connectivity information between flowing rivers and lakes. In this study, three vector datasets over the HMA were developed-Rivers, Lakes, and River-Lake spatial Connectivity-based on the European Space Agency's (ESA) WorldCover data and the 2020 Global Surface Water dataset from the European Commission's Joint Research Centre (JRC). Rivers and lakes were extracted separately using Geographic Information System tools for all basins in the HMA and incorporated into a surface water time series. The analysis results showed that the total area of rivers and lakes in the HMA increased by 23.5% from 2000 to 2019, with lakes accounting for 78.3% of this expansion, while river areas expanded by 40.8%. Glacier-fed rivers and lakes contributed 75.5% of the total water area increase, although non-glacier-fed waters exhibited higher relative growth (44.8%) than glacier-fed ones (20.4%). In the endorheic basins, a consistent annual increase in lake area from 2000 to 2019 was observed: the area of hydrologically connected lakes and rivers expanded from 20,782 km2 to 25,396 km2. In contrast, the area of the isolated lakes grew from 9,550 km(2) to 12,970 km(2). Statistical analysis revealed a significant correlation between the lake area expansion and precipitation. The datasets developed in this study provide the fundamental basis for analyzing changes in rivers and lakes in the HMA, along with their connectivity, supporting regional water resource management, ecological conservation, and sustainable agricultural and pastoral development.
Antarctic coastal snow algal blooms are ecologically important, but remain poorly documented at continental scales because of logistical constraints on field observations. Here, we present a high-resolution, multi-season Sentinel-2 dataset of Antarctic coastal snow algal blooms covering the western Antarctic coastal margin 2018-2025. The dataset was generated using all available Sentinel-2 Level-2A observations acquired during the austral summer within a 100 km inland coastal buffer, processed through a hybrid workflow combining cloud-based preprocessing on Google Earth Engine with parallelized server-side analysis. Seasonal median composites at 10 m spatial resolution were used to calculate the chlorophyll-sensitive index (I-B4) for green snow algae detection. I-B4 values were converted into algal cell concentration estimates using a published field-calibrated model, followed by spectral, spatial, and concentration-based filtering to remove false positives. To our knowledge, this study produced the first spatially continuous seven-year Sentinel-2-derived trend dataset for Antarctic coastal snow algae by quantifying pixel-level temporal trends using a modified Mann-Kendall test. The open dataset includes seasonal maps of algal presence (I-B4), estimated cell concentration, and robust trend metrics. This resource supports studies of cryosphere-biosphere interactions, albedo feedback, and climate sensitivity. The dataset is publicly available in ScienceDB (https://doi.org/10.57760/sciencedb.34802).
Current global Outgoing Longwave Radiation (OLR) products often face a trade-off between spatial resolution and temporal frequency, constraining their utility in monitoring rapid weather dynamics and fine-scale climate processes. To bridge this gap, this study presents a generation and comprehensive validation of the global Long-term Earth System spatiotemporally Seamless Radiation budget OLR dataset (LessRad OLR, ). Derived from over two decades of MODIS observations (March 2000-May 2024), the dataset was generated using an optimized framework that integrates scene-dependent instantaneous retrievals with physically constrained temporal upscaling. The resulting product provides global coverage at an unprecedentedly high resolution (0.05 degrees/h). Comprehensive validation against the CERES Energy Balanced and Filled (EBAF) product confirms robust climatological consistency, with a global monthly Root Mean Square Error (RMSE) of approximately 2.4 W m(-2). Crucially, high-frequency evaluations reveal robust performance at hourly (RMSE similar to 5.65 W m(-2)) and daily (RMSE similar to 4.33 W m(-2)) scales. Furthermore, the 0.05 degrees resolution effectively resolves the fine-scale structures of extreme weather events, which are typically obscured in coarser datasets. This multi-decadal, high-resolution climate data record serves as a valuable resource for investigating diurnal land-atmosphere interactions and validating high-resolution climate models in the era of Big Earth Data.
Efficient management of Pole-like objects (PLOs) is important in urban environments. Deep learning methods offer potential improvements but depend extensively on large training datasets for high accuracy. This paper presents a novel one-shot learning framework for PLO recognition. Our method requires only a single support sample from each category for effective classification, capitalizing on the prior knowledge that PLOs have very similar features. At the core of our approach is OnePole, a non-parametric framework designed to distinguish different PLOs from other road assets while avoiding the large training datasets typically needed by deep learning models. Our experimental results demonstrate that the proposed method achieves an accuracy of 72.1% as measured by Intersection over Union (IoU) on PLOs, a performance comparable to that of state-of-the-art fully supervised deep learning methods. Our method achieves this with only a single labelled object per category, substantially enhancing the efficiency of detecting PLOs in point cloud datasets. This not only illustrates the effectiveness of our approach but also represents a significant advancement in reducing the reliance on extensive labelled data, which is a common challenge in fully supervised methods.
Polders have supported grain production in China for centuries and are typically low-lying farmlands protected by embankments and hydraulic infrastructures. This study extracts farmland use intensity (FUI) in Dongting Lake polders and quantifies their three-dimensional changes and static topographic exposure using multi-temporal remote sensing data. In 2023, single- and double-cropping rice dominated summer cultivation with planting areas of 1,922.75 km2 and 1,469.98 km2 in the Dongting Lake polders, while rapeseed and winter wheat were the main winter crops with 952.34 km2 and 663.30 km2, respectively. Spatial and temporal variations in FUI were mainly driven by the conversion between double- and single-cropping rice. The FUI declined during 1990-2005, increased during 2005-2015, and declined again from 2015 to 2023, resulting in an overall downward trend from 1990 to 2023. The northern boundary of double-cropping rice shifted southward by 133.2 km over 1990-2023, contributing to the decline in FUI. This study also quantifies the vertical distribution of FUI under hypothetical reference lake-level scenarios, indicating that approximately 50% and 95% of the cultivated area lies below 31 m and 36 m, respectively. These results reflect static topographic exposure based solely on elevation comparisons, and should not be interpreted as actual inundation or flood-related outcomes.
Digital twins (DTs) have evolved from domain-specific simulation tools into integrative cyber–physical–social infrastructures that reshape how complex systems are observed, modeled, and governed. Rather than treating DTs as digital replicas, this paper conceptualizes them as dynamic epistemic architectures that integrate observation, physics-based modeling, AI, and decision processes through persistent bidirectional exchange. Drawing on a systematic review of 251 papers (from 449 screened abstracts within 22,434 publications, supplemented by foundational literature), we examine how this transition is enabled by advances in sensing, scalable computing, data assimilation, uncertainty quantification, and AI–physics integration. We argue that the defining feature of mature DTs is not replication fidelity alone, but their capacity to support uncertainty-aware, scenario-driven decision-making across scales, from engineered components to Earth system processes. DTs increasingly operate as system-of-systems (“DT-of-DTs”), requiring interoperable architectures, machine-readable metadata, and standardized trust frameworks to enable composability across organizational boundaries. This review synthesizes technological enablers, cross-domain applications, and emerging governance challenges, and identifies key research priorities in multiscale modeling, probabilistic inference, interoperability standards, certification pathways, and human-centered design. We conclude that the next generation of DTs will depend less on isolated technological advances and more on disciplined integration across physics, data science, computation, domain science, decision science, governance science and institutional practice. In this trajectory, DTs are not static models but evolving infrastructures for adaptive, resilient, and evidence-based decision support in complex socio-technical systems.
Floods are among the most destructive natural hazards with growing social consequences. However, global flood research has exhibited pronounced regional disparities in thematic focus and scholarly attention, reinforcing unequal risk perceptions. This study analyzes 64,989 flood-related publications and uses large language models to extract structured information for synthesis. Over the past four decades, flood research has shifted from hydrological process modeling to integrated risk governance and climate adaptation. A high-risk-low-attention mismatch also prevails, with research output correlating more strongly with socioeconomic conditions than risk. Most developed countries attract disproportionately high research attention relative to risk, whereas 53.1% of the countries exhibit a risk-attention mismatch. The highly exposed regions in Africa and South America remain underrepresented and methodologically fragmented. Global research on flood types exhibits pronounced spatial differentiation, reflecting the combined influence of climatic and geographic settings, as well as political, historical, and institutional contexts. In some countries, such as Egypt and Iraq, research agendas are excessively concentrated on a single flood type, with limited investigations of compound floods. Furthermore, climate change has exposed historically low-risk arid and semi-arid regions to emerging flood threats, potentially exacerbating such mismatches. Overall, this study highlights the urgency of targeted investment and collaboration in high-risk, but low-attention regions.
Remotely sensed data are increasingly used with model-based inference to estimate forest population characteristics (e.g., areal means). However, at-sensor radiance affected by measurement errors has inherent variations over time, space and spectrum, violating the stationary data assumption in regression-based applications. These errors bias relationships with forest variables by attenuating regression coefficients towards zero; heteroscedasticity, often observed in models of forest variables, could aggravate the biased coefficients. We proposed SIMEX-WLS, an errors-in-variables (EIV) modeling approach that corrects coefficient attenuation by incorporating both measurement errors and non-constant residual variances, to rectify the downweighed contribution of remotely sensed data. Validation using Landsat 8 Collection 2 Level-2 data yielded four relevant conclusions: (1) SIMEX-WLS also corrected the attenuated variances of regression coefficients along with the coefficients themselves; (2) in model-based inference, the attenuation correction achieved an 11% reduction in the variance of population mean estimates; (3) pixel-level RMSE may not be an appropriate metric for evaluating the precision of population characteristics; and (4) across samples of varied sizes, attenuation correction consistently improved the estimation precision. These findings explicitly emphasize the significance of incorporating remote sensing measurement errors and validating models for population-level estimation rather than relying solely on pixel-level prediction or mapping.
