Hyun-Cheol Kim

Arctic sea ice radiative forcing remains incompletely quantified because of persistent observational gaps in ice surface temperature (IST) during polar night, when temperature-driven radiative effects are most active in the surface energy balance. The mid-2000 s Arctic transition further raises the question of whether same-season coupling between atmospheric circulation and the radiative response has persisted or weakened. We address these issues by reconstructing gap-free IST fields for 1988–2023 using automated machine learning to gap-fill MODIS observations, validated against ice mass balance buoys ( R = 0.86 ; RMSE = 4.33 K; bias = − 2.35 K). This enables year-round assessment of sea ice radiative forcing (SIRF). Temperature-driven and albedo-driven SIRF exhibit full-period trends of 0.29 ± 0.02 and 0.13 ± 0.01 W m − 2 yr − 1 respectively; in seasonally consistent windows winter temperature-SIRF ( 0.40 – 0.55 W m − 2 yr − 1 ) is comparable to summer albedo-SIRF, indicating that temperature-driven forcing is not negligible in the Arctic radiative balance. Change-point analysis identifies a statistically significant transition around 2004–2005 (Pettitt test, p < 0.05 under an AR(1) Monte Carlo reference), marking a shift to predominantly positive SIRF anomalies. Empirical orthogonal function analysis documents a reorganization of the dominant DJF sea level pressure mode across this transition, with the leading mode’s variance fraction increasing from 52.5% to 62.8%. Multiple regression of SIRF anomalies onto the leading PCs shows a spatial contraction of the statistically significant atmosphere–SIRF coupling: pre-2005 coupling ( β ∗ ≈ ± 0.2 to ± 0.6 ) is retained over the Barents–Kara–Laptev–East Greenland sector but largely disappears over the Pacific-sector and Central Arctic after 2005. This sector retains positive SIRF anomalies despite the post-2005 shift toward anticyclonic SLP—a regional co-occurrence we term feedback entrenchment.

Sea Ice Concentration (SIC) plays an important role in the Antarctic polynyas, especially where biological interactions through changes in chlorophyll‐a concentration (CHL) phenology are concerned. To specifically understand how much SIC change affects CHL in the Ross Sea polynya (RSP), the regime shift index (RSI) was applied to the SIC anomaly, as per Rudionov (2004), https://doi.org/10.1029/2004gl019448 . This resulted in three principal temporal regimes: Regime 1 (R1), 1997–2004; Regime 2 (R2), 2004–2016; and Regime 3 (R3), 2016–2019. Based on reconstructed daily CHL observations from remote sensing and application of a random forest algorithm with 10 physical predictors, our analysis revealed that (a) earlier sea ice melt onset corresponds to an earlier CHL peak across all regimes; (b) SIC changes explain approximately 10%–12.5% of CHL variation in the RSP: a first‐ever estimation of SIC influence in the RSP; and (c) although CHL is significantly influenced by SIC changes, the relative contributions of the 10 geophysical predictors vary across the three regimes, with photosynthetically active radiation consistently emerging as the dominant driver. This result may explain why CHL has been observed to decrease despite the unprecedented decrease in Antarctic SIC since 2016.

Adélie penguins (Pygoscelis adeliae) are key indicators of the Antarctic ecosystem, and remotely sensed guano stains offer an efficient means of estimating breeding abundance. Landsat medium‐resolution (MR) imagery enables long‐term monitoring through its archive. However, the effect of spatial resolution on detection accuracy and abundance estimates remains unquantified. This study evaluated resolution‐dependent errors using unmanned aerial vehicle (UAV; 0.1 m), very high‐resolution (VHR; 1.24 m, WorldView‐4), and MR (30 m, Landsat 8) imagery acquired concurrently over the Cape Hallett colony in November 2018. UAV‐based delineations were used as reference data to assess guano detection accuracy and to perform Monte Carlo and Poisson‐based simulations. VHR imagery achieved high accuracy (F1 = 0.73; omission = 21%), whereas MR imagery showed large commission (59%) and omission (21%) errors. Detection errors rose nonlinearly with coarser resolution, reaching 97% at 30 m. Abundance estimates were sensitive to apparent density; using the Antarctic mean (0.34 nests m ⁻² ) sometimes gave lower errors than the UAV‐based value (0.44 nests m ⁻² ). These results provide a quantitative framework linking spatial resolution, detection error, and estimation uncertainty, emphasizing the need for UAV‐based validation and colony‐specific calibration for reliable long‐term monitoring of Antarctic penguin populations.

This study presents a methodological framework for developing a high-precision digital surface model (DSM) dataset based on complementary unmanned aerial vehicle (UAV) light detection and ranging (LiDAR) and satellite-derived data for Russell Glacier terminus, Greenland. Field surveys were conducted across four annual campaigns (2022-2025), acquiring UAV LiDAR data that were subsequently cross-referenced with three satellite-derived DSMs, ASTER GDEM V3, WorldDEM, and Arctic- DEM Mosaic v4.1. Vertical rectification using inverse distance weighting interpolation with six ground control points was applied to all datasets, achieving a remarkable 94.1% improvement in accuracy, reducing mean root mean square error (RMSE) from 19.05 to 1.12 m. WorldDEM demonstrated the most substantial improvement (98.6% reduction), while post-correction ArcticDEM achieved the highest accuracy (0.20 m RMSE). UAV LiDAR maintained centimeter-level precision (0.51 m RMSE). Spatiotemporal analysis revealed significant morpho-dynamic changes, including proglacial lake expansion, moraine evolution, and surface elevation variations. Over 20 years (2000-2025), terminus-region elevation differences reached to 20 m. The vertical rectification methodology demonstrates the effectiveness of employing multi-sensor datasets derived from complementary platforms to overcome individual sensor limitations. This dataset supports glacier mass balance research, dynamics investigations, and climate change impact assessments. The dataset is publicly available through the Korea Polar Data Center (KPDC).

This study presents a deep-learning (D2S) model to generate historical vertically polarized (TBv) and horizontally polarized (TBh) brightness temperatures (TBs) at the Soil Moisture Active-Passive (SMAP) 1.4 GHz channels using TBv and TBs polarization difference from the Defense Meteorological Satellite Program (DMSP) satellite series' 19.35 GHz channels. The proposed D2S model employed the ResNet9-based generator and PatchGAN-based discriminator, with a compositve loss function synergizing mean squared error and L1 losses, similar to the Pix2Pix architecture. Training, validation, and testing were conducted using paired datasets of polarized TBs from SMAP (1.41 GHz) and DMSP (19.35 GHz) collected between July 2015 and June 2022, followed by application to historical data spanning from September 1987 to June 2015. The D2S model demonstrated robust performance in generating both SMAP TBv and TBh, achieving test set average correlation coefficients of 0.994 and 0.992, biases of 0.150 K and -0.259 K, root-mean-square errors of 6.463 K and 9.011 K, and mean absolute errors of 3.883 K and 5.518 K, respectively. Furthermore, this study showed reasonable agreement between the D2S-generated SMAP TBs and SMOS TBs and provided preliminary results of thin sea ice thickness retrievals, which were further used for evaluating the limitations of the D2S approach in specific sea ice regions. Consequently, the D2S model can provide empirically consistent SMAP-like polarized TB estimates over the Arctic, including periods prior to the L-band operational era. The D2S framework offers significant potential to enhance the reconstruction of historical Arctic sea ice conditions and to support climate change studies by providing surrogate L-band–equivalent information extending back to 1987.

This study introduces case adaptive detection models for detecting asbestos-containing building materials employing short-wave infrared hyperspectral images. The models are optimized for indoor and outdoor conditions, which could be applied to asbestos detection in interior and exterior building materials. Samples were collected from the most probable types and forms of building materials for both asbestos-containing materials (ACMs) and non-asbestos-containing materials (non-ACMs) to represent real-world conditions, utilizing the largest and most diverse dataset to date (11 types, 4 directions, and 4 forms). X-ray diffraction (XRD), scanning electron microscopy (SEM), and energy dispersive spectroscopy (EDS) analyses demonstrated that chrysotile was the main asbestos mineral in the most of ACMs, as evidenced by characteristic chrysotile diffraction peaks, fibrous serpentine morphology, and coincident Mg-Si elemental distributions. The spectral characteristics of building materials are manifested by mineral components for ACMs and non-ACMs, where asbestos absorption (Mg-OH) identifies ACMs. The indoor RF model identified critical absorption features at 2299-2350nm, 1923-1967nm, and 2366-2426nm, corresponding to Mg-OH, OH, and CO₃²⁻ absorptions, achieving 95.3% accuracy. The outdoor RF model incorporated additional Fe-OH and Al-OH bands, achieving 96.5% accuracy. Real-world validation at three asbestos abatement sites and one large-scale site demonstrated practical applicability including scalability and computational efficiency with overall accuracies of 92.1% for indoor and 92.5% for outdoor conditions. Comparative analysis with existing benchmark methodologies confirmed superior performance, with our models significantly outperforming previous approaches. The models successfully detected ACMs across various material conditions, scales and effectively verified complete removal post-abatement with high computational speed. This approach delivers immediate practical value for compliance verification and quality assurance in real-world abatement scenarios and large-scale surveys, providing objective documentation for regulatory compliance while substantially reducing costs compared to conventional destructive sampling methods.

This study presents a field-applicable detection models for host rocks of naturally occurring asbestos (NOA) through hyperspectral imaging and machine learning. We developed detection models using the most comprehensive dataset to date: 104 rock samples spanning 27 distinct lithologies, including 5 NOA-associated and 22 non-NOA rock types representing prevalent lithospheric formations. Sample characterization integrated visual inspection, X-ray fluorescence (XRF), X-ray diffraction (XRD), and spectral analyses, revealing distinctive signatures in NOA-associated rocks. High MgO content emerged as a crucial indicator for asbestos formation. NOA-carbonate spectral characteristics were primarily controlled by calcite, dolomite, and tremolite asbestos, while NOA-ultramafic spectral responses were governed by chrysotile, lizardite, olivine, and pyroxene. Traditional absorption-based methods (SAM, MSRFF) demonstrated poor performance due to their reliance on endmember spectra and inability to manage complex spectral variability. Machine and deep learning approaches showed superior performance by effectively leveraging key spectral bands: MgOH and CO₃²⁻ for NOA-carbonate detection and MgOH and FeOH for NOA-ultramafic identification. The 3D CNN model achieved superior performance (93.6% accuracy, Kappa coefficient: 0.88), excelling in discriminating subtle mineralogical variations through complex spatial-spectral pattern recognition. Field validation using 2 NOA and 3 non-NOA outcrops confirmed the model's reliability (89.2%-99.6% accuracy). Given the fact that the model utilizes mineral absorptions while excluding atmospheric bands and has been field-tested, the model can be applied to NOA surveys in various contexts, including construction projects and underground development.

This study presents a Dual-Attention SuperResolution Generative Adversarial Network (DA-SRGAN) framework to enhance the spatial resolution of passive microwave (PMW) brightness temperature (TB) data from 12.5/25 km to 3.125/6.25 km. By integrating channel and spatial attention mechanisms into the SRGAN architecture, the proposed model prioritizes critical low- and mid-frequency features inherent to PMW measurements, effectively suppressing noise while preserving radiometric accuracy and geophysical structures such as atmospheric fronts, precipitation bands, and coastlines. Quantitative evaluations demonstrate DA-SRGAN's superiority over baseline models (SRCNN, SRGAN) across metrics including MAE, RMSE, SSIM, and LPIPS, with cosine similarities exceeding 0.995 in histogram-based distribution comparisons. Cross-sensor validation with AMSR2 89 GHz data demonstrates the model's ability to maintain statistical relationship between sensors at enhanced spatial resolution despite hardware and frequency differences. Spatio-temporal consistency analysis across diverse environmental conditions reveals robust performance with RMSE values consistently below 1K across all surface types and seasons, demonstrating temporal stability and regional consistency. Exploratory assessment at finer resolution (3.125/6.25 km) confirms effective spatial detail enhancement while preserving statistical distribution integrity. These advancements position DA-SRGAN as a transformative tool for bridging resolution gaps in PMW satellite remote sensing, offering improved observational fidelity for environmental applications.