Groundwater exploration in Enugu Urban, southeastern Nigeria, is constrained by complex subsurface geology within the Anambra Basin. This study investigates the spatial controls on groundwater occurrence by combining gravity-derived sediment thickness and rock density models with Vertical Electrical Sounding (VES) data and in-situ well information. Gravimetric modelling reveals sediment thicknesses ranging from approximately 2 to 3 km, with thicker sedimentary sections occurring in structurally depressed zones and thinner sections associated with uplifted basement areas. Rock density values (2.648-2.689 g cm^-3) indicate heterogeneous lithological conditions that influence groundwater storage and flow. VES results show significant resistivity variation (3.8-987.6 Ωm), reflecting a predominantly shale-dominated stratigraphy with localized weathered and fractured zones that may serve as potential aquifers. However, extensive low-resistivity layers primarily correspond to clay-rich, low-permeability units, thereby limiting groundwater productivity. Analysis of 50 in-situ wells indicates that the majority are shallow (4.5-50 m) and highly seasonal, with over half drying shortly after the rainy season, confirming limited near-surface storage and strong dependence on rainfall recharge. Integration of geophysical and well data enabled the delineation of groundwater potential zones, with more favourable conditions associated with moderate sediment thickness, lower density contrasts, and structurally controlled features at depth. The results demonstrate that groundwater occurrence is highly heterogeneous and primarily controlled by lithological variability and structural complexity. This study highlights the effectiveness of incorporating geophysical remote sensing approaches for hydrogeological characterization and provides a basis for targeted groundwater development in complex sedimentary environments.

Climate change poses a significant threat to cultural heritage worldwide, manifesting through sea-level rise, increased humidity, extreme weather events, and shifting land dynamics. This review explores the potential of advanced remote sensing technologies, particularly Light Detection and Ranging (LiDAR) and Synthetic Aperture Radar (SAR), for climate-adaptive heritage monitoring. LiDAR provides high-resolution topographic and structural data, enabling accurate 3D reconstruction and vulnerability assessment, while SAR offers all-weather, day-and-night imaging for deformation monitoring and early detection of structural changes. Together, these technologies provide a robust, non-invasive framework for large-scale and continuous heritage observation. This Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA)-based review synthesizes recent advances in LiDAR and SAR applications for cultural heritage conservation, with emphasis on case studies from climate-vulnerable regions. It highlights interdisciplinary approaches integrating geospatial analytics, machine learning, and historical data to strengthen conservation strategies. The synthesis reveals a shift toward multi-sensor and multi-scale monitoring approaches, underscoring the complementary roles of LiDAR and SAR. Demonstrative case studies illustrate the empirical contributions: terrestrial laser scanning of the 18th-century Dutch House in Patna achieved sub-2 cm 3D model accuracy (Root Mean Squared Error [RMSE] = 1.3 cm) relative to independent total-station measurements, establishing a baseline for structural health monitoring, while a preliminary Persistent Scatterer Interferometry (PSInSAR) workflow over Shergarh Fort serves as a methodological proof-of-concept for SAR-based deformation screening. The novelty of this review lies in bridging methodological advances with practical implementation, particularly for data-constrained environments in the Global South, and in advancing a formalized, tiered integrated LiDAR-SAR framework for climate-adaptive heritage monitoring.

Wildfires are complex disasters that extend beyond vegetation combustion, producing cascading impacts on the atmosphere, soils, hydrology, ecosystems, and human society. Their risks and impacts are being amplified by climate change, highlighting the need to understand and monitor wildfires not as isolated events but as full-cycle phenomena that encompass pre-fire, active-fire, and post-fire stages. This study systematically reviews wildfire research published over the past decade that has used satellite remote sensing and artificial intelligence (AI) and synthesizes trends in data use and methodology from a lifecycle perspective. Relevant literature was retrieved from the Scopus database using search terms related to wildfire, satellite observations, and AI. The selected studies were classified into pre-fire, active-fire, and post-fire stages. They were compared in terms of study region, input data, reference data, input representation, and algorithmic approach. The results show that relevant studies have increased rapidly in recent years, with clear stage-specific differences in research objectives. Pre-fire studies mainly focus on wildfire occurrence prediction and risk assessment. Active-fire studies emphasize active fire and smoke detection, as well as the monitoring of fire fronts and spread dynamics. Post-fire studies concentrate on burned-area mapping, burn severity assessment, damage evaluation, and vegetation recovery analysis. Geographically, the literature is strongly concentrated in North America and Asia, with a substantial proportion of studies focused on a small number of countries, including the United States, China, Australia, Canada, Brazil, and South Korea. Although single-country studies still dominate, multi-regional and global-scale studies have increased in recent years. In terms of data use, optical satellite imagery remains the most widely used source, while the integration of thermal infrared, synthetic aperture radar, meteorological, topographic, and land-cover data has become increasingly common. Methodologically, machine learning and deep learning approaches increasingly supplement or replace conventional index-based and rule-based methods, with model architectures and input representations differing by wildfire stage and research objective. Nevertheless, regional bias in training data, data imbalance, limited model interpretability, and restricted cross-regional generalizability remain major challenges. Future wildfire monitoring research should therefore move toward integrated lifecycle frameworks that combine multi-source data fusion, regional scalability, physically informed learning, and interpretable AI.

GOCI-II was launched on the 19th of February in 2020 as the successor of Geostationary Ocean Color Imager-I (GOCI-I) onboard the Geo-Kompsat 2B satellite with a 10-year mission lifetime. GOCI-II captures the local area 10 times a day at hourly intervals with 250 m spatial resolution, along with the full-disk imaging over the Asia-Oceania region, selectively once a day. GOCI-II has passed the halfway point of its mission lifetime and produces 26 official L2 products with improved algorithms compared to the initial ones, which were developed using simulated data prior to launch. Here, the current status of the algorithm development and improvement for GOCI-II ocean color products is discussed, including atmospheric correction, chlorophyll-a, colored dissolved organic matter, total suspended matter, diffuse attenuation coefficient, and photosynthetically available radiation. We are also mentioning the GOCI-II data processing system, GOCI-II Ground System (G2GS), and the distribution system for the public data service.

This study aims to classify inland water bodies using deep learning-based semantic segmentation models, namely U-Net and DeepLabV3+, based on Sentinel-2 imagery, and to comparatively analyze their decision-making processes through explainable artificial intelligence (XAI) techniques, including SHapley Additive exPlanations (SHAP) and Gradient-weighted Class Activation Mapping (Grad-CAM). The study areas comprised two representative artificial reservoirs on the Korean Peninsula, Andong Lake and Daecheong Lake. Five spectral bands from Sentinel-2 imagery-Blue, Green, Red, Near-Infrared (NIR), and Shortwave Infrared (SWIR)-were used as input features. Both models achieved high segmentation performance: U-Net attained a validation Intersection over Union (IoU) of 0.9655 and an F1-score of 0.9824, while DeepLabV3+ achieved higher validation accuracy (IoU 0.9895, F1-score 0.9947) and a modest, though not statistically significant, scene-level accuracy advantage on both inference scenes. Under the zero-baseline Gradient SHAP setting adopted here, computed on the pre-sigmoid water-class logit, the Blue band received the largest mean attribution in both models and at both sites, whereas the ordering of the remaining bands varied across configurations. We note that attribution magnitude reflects marginal contribution relative to the chosen baseline and does not, by itself, quantify spectral separability. In contrast, Grad-CAM results showed that U-Net produced sharply localized activations along water boundaries, whereas DeepLabV3+ exhibited more spatially distributed activations, likely due to its Atrous Spatial Pyramid Pooling (ASPP) architecture. Overall, this study reports the quantitative performance of two deep learning models for satellite-based water body segmentation and provides an attribution-based comparative analysis of their decision behavior-consistent Blue-band dominance but architecture-dependent spatial activation-under controlled, single-acquisition case-study conditions, suggesting preliminary considerations for selecting between accuracy-oriented and boundary-oriented models.

Recent climate change has intensified wintertime convective snowbands over the Yellow Sea, causing severe localized heavy snowfall and socio-economic damage in coastal Korea. Current weather radar has observational limitations (radius, elevation angle) for these events, especially for systems originating from distant seas, making proactive hazardous weather monitoring challenging. Existing proxy radar research predominantly focuses on summertime precipitation. To address this, we developed a wintertime proxy radar model using Geo-Kompsat-2A (GK2A) satellite data and time-series deep learning. The optimal model was selected through qualitative and quantitative evaluations, using 30-minute time-series input data consisting of three 10-minute-interval images, without the Multi-Task Learning (MTL) application. This model achieved an accuracy of Probability Of Detection (POD) 0.599, False Alarm Rate (FAR) 0.390, Critical Success Index (CSI) 0.435, Pearson Correlation Coefficient (CC) 0.605, bias -0.532 dBZ, and Root Mean Squared Error (RMSE) 3.334 dBZ, partially mitigating the typical surface cooling/snow cover issue in winter satellite data. Furthermore, a sliding-window approach was utilized to generate data for a larger area, with a mean and 3/4 patch overlap yielding the most continuous results. This developed proxy radar was able to generate early-stage snowfall information from distant seas that was previously undetectable by conventional radar. This research demonstrates the promising role of wintertime proxy radar data as an auxiliary forecasting tool for proactive detection of hazardous weather and damage mitigation in radar-deficient regions.

Burned area is defined as the spatial extent damaged by wildfire, where changes occur in the land surface and vegetation. This extent can be effectively observed by a satellite, which can cover large areas in a short time. In general, satellite-based burned area can be derived from temporal changes in vegetation indices such as the Normalized Difference Vegetation Index (NDVI), obtained from polar-orbiting optical satellites. However, data gaps from clouds and aerosols in optical satellite observations can make it difficult to obtain sufficient time series for burned area detection. In this study, we proposed a framework for evaluating the applicability of Geostationary Ocean Color Imager-II (GOCI-II)time series of observations, which provide higher temporal resolution than polar-orbiting satellites, for burned area detection to overcome this limitation. The framework used time series from daily compositing as the main input and time series from noon-single for comparison. Burned area was detected from both time series using a procedure adapted from the burned area mapping algorithm used to generate the MCD64A1 product and the results were compared to evaluate the effect of daily compositing when using geostationary satellite data. In commonly detected burned area, daily compositing showed a lower mean interquartile temporal range (IQR^*)than noon-single, 5.56 vs. 6.85, and higher temporal separability (S^*), 6.89 vs. 6.16. The proportion of pixels with burn-date differences within ±1 day relative to active fire detection date was also higher, 0.458 vs. 0.340.In comparison with the standard MCD64A1 product, the spatial agreement of the Dice coefficient (DC) reached 0.8534 at a burned fraction (τ) of 0.50, where the GOCI-II burned area was 107,286 ha, about 6.2% larger than the MCD64A1 product. These results suggest that high temporal resolution of GOCI-II can contribute to temporal stability in burned area detection and high spatial agreement with existing satellite-based burned area products, demonstrating the applicability of geostationary satellite data for burned area detection.

Particulate organic carbon (POC) monitoring in dynamic coastal environments is essential for estimating Blue Carbon fluxes and understanding local biogeochemical cycles. Conventional satellite algorithms developed for open-ocean (Case-1) waters frequently underperform in optically complex coastal environments because of competing optical signals from Colored Dissolved Organic Matter (CDOM) and suspended particles. To address this gap, we present a preliminary effort to evaluate and locally optimize POC retrieval algorithms for the Geostationary Ocean Color Imager-II (GOCI-II) in the coastal waters of Jeju Island, Korea. Previously published GOCI-II POC products in Korean waters have applied global open-ocean models without regional tuning. Using a structured diagnostic methodology, we first evaluated three standard blue-to-green band ratio algorithms (412/555, 443/555, and 490/555 nm) using in-situ radiometric and biogeochemical data from 23 stations in the Korea Strait. All three algorithms performed poorly (R² ≤ 0.093), indicating that high CDOM concentrations decouple blue reflectance from particulate scattering. A cross-sensor comparison with Visible Infrared Imaging Radiometer Suite (VIIRS) (R² = 0.617, N = 2,675 coincident pixels) provided evidence of GOCI-II's radiometric integrity, suggesting that the failures were algorithmic rather than instrumental. We then evaluated all 15 GOCI-II band-ratio combinations spanning 412-745 nm under the same Stramski et al. (2008) power-law framework. The analysis revealed a clear spectral hierarchy: Red-near-infrared (NIR) ratios consistently outperformed all shorter-wavelength alternatives, with the 680/745 nm combination yielding the best performance (R² = 0.2422, root mean square error (RMSE) = 31.69 mg m^-3, mean bias error (MBE) = -2.82 mg m^-3, p = 0.0171). Although the explained variance of the best-performing model remains modest, the substantial and consistent improvement over blue-green algorithms across multiple Red-NIR pairs supports the conclusion that the Red-NIR domain is more appropriate for POC retrieval in this CDOM-rich environment. The proposed 680/745 nm relationship should be regarded as a preliminary regional candidate that requires further validation against an expanded multi-seasonal in-situ database before operational deployment.