Abstract. Accurate quantification of surface mass balance (SMB) in the Antarctic interior underpins ice sheet mass budget assessments and ice core interpretation. Stake measurements, however, systematically underestimate SMB because firn densification causes surface lowering unrelated to mass change. Here, we simulate firn compaction with a firn densification model and correct stake records from 2008-2024 at Dome Argus (Dome A), East Antarctica, thereby refining SMB estimates and their spatial variability. The mean annual corrected SMB is 24.14 kg m-2 yr-1, 8.8 % higher than the uncorrected value (22.19 kg m-2 yr-1). Over the stake array, the RACMO2.4p1 regional model yields lower and more spatially uniform SMB (17.50 kg m-2 yr-1). Using automatic weather station observations, we estimate annual sublimation of 2.34 mm w.e yr-1. and hoar deposition of 0.87 mm w.e. yr-1, indicating that the net vapor flux is equivalent to 5.7 % of the total mass input. This framework reduces densification induced bias in stake-derived SMB, provides an observational benchmark for evaluating regional climate models, and supports accurate dating of ice core climate records from Dome A.

The continental ice sheet, particularly that of Antarctica, is recognized as a critical component of the global climate system, with its subglacial topography understood to exert a fundamental control on ice dynamics and stability. However, the sparse and uneven distribution of measurement-derived data across the continent presents substantial challenges to achieving high-resolution and reliable bed topography reconstructions. Although a range of reconstruction approaches have been proposed to generate detailed terrain, their performance remains constrained in regions with large spacing between radar survey lines, often resulting in artifacts or structural inconsistencies. To address this, a novel deep learning framework integrating terrain inpainting and super-resolution strategies is proposed, termed InpaintSR. Large-scale terrain structures are first recovered through low-resolution inpainting guided by geomorphological priors and spatial continuity, and then refined through a super-resolution process to enhance local details via high-resolution texture learning. When validated in the Gamburtsev Subglacial Mountains and Princess Elizabeth Land, InpaintSR is demonstrated to preserve geomorphic integrity and reconstruct fine-scale landforms more effectively than existing approaches. By adopting this method, a 250m resolution subglacial digital elevation model of PEL, the highest-resolution product yet developed for the region, was generated. With this product, the irregular morphology of Qilin Subglacial Lake is clearly delineated, and a series of previously unmapped canyon systems is newly revealed.

The internal ice layers and ice-bedrock interfaces are critical indicators for understanding ice sheet evolution, englacial dynamics, and climate change impacts. Ice-penetrating radar (IPR) delivers high-resolution imaging of the englacial structures; however, low contrast of internal ice layers and ambiguous layer boundaries in IPR profiles challenge the concurrent extraction of these features. Traditional amplitude-based methods often struggle to efficiently process large-scale profiles or maintain accuracy under complex stratigraphic conditions. Therefore, we propose a U-Net-based pipeline that integrates IPR instantaneous phase analysis to simultaneously extract ice layers and ice-bedrock interfaces from the IPR data at Dome A, Antarctica. To enhance performance, a discriminator first classifies IPR profiles based on layer roughness differences. To address the scarcity of annotated data, we design a synthetic data generation method using morphological profile analysis and introduce an innovative post-processing operation to enhance the continuity of the extracted layers. Compared with existing methods widely used in layer extraction, such as U2-Net and DeeplabV3+, our results not only maintain high accuracy in identifying deep ice-bedrock interfaces but also improve the detection of low-contrast internal ice layers. The proposed method establishes an IPR-attribute-driven system for the automatic processing of IPR data, improving the interpretation accuracy of complex englacial environments by calibrating with ice core records, and provides a tentative technical reference for analyzing the ice flow patterns of the Antarctic Ice Sheet.

Mapping structural boundaries beneath ice-covered and remote polar regions remains a challenge for marine geophysics, where traditional methods are sparse. Gravity-derived quantities, such as gravity anomaly $\Delta g$, vertical gravity gradient $\Gamma_{z z}$, and tensor invariant ratio $I$, provide practical means for boundary detection independent of surface morphology. This study evaluates whether satellite gravity attributes alone can delineate geological boundaries in Antarctic subglacial settings and compares them to surface morphometric edges derived from a digital elevation model (DEM). The gravity attributes were processed using gradient-based edge detection and morphological filtering to extract continuous boundary curves. Results show that $\Gamma_{z z}$ and $I$ delineate sharp mass-contrast boundaries more coherently than $\Delta g$, exhibiting strong spatial correspondence with DEM-derived boundaries. This highlights the complementary value of gravity observations for under-ice structural interpretation, suggesting that gravity-based boundary detection is feasible in areas with sufficient density contrast.

Abstract Understanding the coastal zone of the Antarctic Ice Sheet (AIS), where it interacts with the Southern Ocean and warmer air masses, is crucial for predicting Antarctica's influence on the global climate and sea level. This region has multiple tipping mechanisms that could trigger large, rapid, and potentially irreversible changes in the AIS, the Southern Ocean and their global connections in the coming centuries. The AIS remains the largest source of uncertainty in future sea‐level projections. Bed topography beneath the ice shelves and the coastal ice sheet is not yet well documented, and is a major source of this uncertainty. This review assesses current knowledge of the coastal zone and highlights methods to investigate it, including aerogeophysical surveys, ground‐ and ship‐based measurements, satellite observations, and computer modeling. An ensemble analysis of published bed topography data sets identifies significant data gaps and their regional distribution, framed in the context of current ice‐sheet behavior and potential instability. We propose scientific priorities and guidelines for future aerogeophysical surveys, advocating for a comprehensive, coordinated international effort to build a next‐generation data set of Antarctic bed properties. Such an initiative would significantly advance understanding of the role of coastal processes in ice‐sheet dynamics, reducing uncertainties in sea‐level rise projections and improving predictions of future ocean and climate changes.

In the above article [1], we correct the following texts that contain inaccurate descriptions and numerical errors: (1) Page 6 (Section III B): In the fourth paragraph of the left column the last sentence “...... $\sigma_{\frac{dh}{dt}_{Grid} }$ , is obtained through an error propagation from ......” is revised to “...... $\sigma_{\frac{dh}{dt}_{Grid} }$ , is estimated as the root mean square of ...... .”(2)Page 9 (Section IV D): In the last paragraph of the left column: the sentence “...... volumetric losses in the basin, $-81.9~\pm$ 0.8 km3 yr-1......, with an increase of $-14.5 \pm 1$ km3 yr-1” is revised to “...... volumetric losses in the basin, $-81.9~\pm$ 3.5 km3 yr-1......, with an increase of $-14.5~\pm$ 3.6 km3 yr-1.”(3)Page 12 (Section Appendices C): In the third paragraph of the right column the sentence “......, which are mostly under 0.5 cm yr-1” is revised to “......, which are mostly under 1.0 cm yr-1.”(4)Page 12 (Figure 15): The figure contained numerical errors and is revised. The black area in the lower-left corner represents the open ocean region.

Supraglacial lake depth is a key variable for quantifying surface meltwater storage and assessing ice-shelf stability, yet spatially continuous and reliable bathymetric information remains difficult to obtain in polar regions because in situ measurements are scarce and optical imagery cannot directly provide water depth. This study develops an integrated framework for supraglacial lake identification and bathymetry retrieval by combining ICESat-2 ATL03 photon-counting lidar data with Sentinel-2 multispectral imagery. ICESat-2 lake photons were used to constrain lake-region extraction from Sentinel-2 imagery, and the photon-derived along-track depths were corrected for scattering and refraction before being converted into Sentinel-2 pixel-level depth labels. Based on these labels, four retrieval models were constructed and evaluated, including an empirical model, CatBoost, a convolutional neural network (CNN), and a residual dense network (RDN). CatBoost generated initial depth estimates, while CNN and RDN further incorporated the CatBoost-derived depth prior and Sentinel-2 multispectral features for pixel-level depth prediction. Experiments over four investigated supraglacial lakes showed that RDN achieved the best average performance across the investigated lakes, with mean R2, RMSE, and MAE values of 0.927, 0.187 m, and 0.144 m, respectively. For the investigated lakes, the integration of ICESat-2 and Sentinel-2 extended discrete along-track reference-depth observations to spatially continuous bathymetry maps. Because the training and validation samples were obtained from different spatial blocks within the same four lake scenes, the reported performance primarily reflects within-lake spatial generalization under the investigated conditions, and transferability to unseen lakes remains to be evaluated. These maps may provide inputs for future lake-volume estimation and ice-shelf hydrological analyses, while their applicability to lakes with different morphological and optical conditions requires further evaluation.

This study investigates the impacts of two types of El Ni & ntilde;o-Eastern Pacific (EP) and Central Pacific (CP)-on extreme precipitation occurrences across Antarctica over the 1979-2023 period. While the main analyses focus on the austral cold (May-October) and warm (November-April) seasons, the study also examines the seasonal variations in extreme precipitation across all four austral seasons-summer (DJF), autumn (MAM), winter (JJA), and spring (SON)-to facilitate comparison with previous studies. Analysing data from a global reanalysis (ERA5), a regional climate model (RACMO2) and a general circulation model (CESM2), the study highlights significant seasonal and spatial differences in extreme precipitation occurrence associated with EP and CP El Ni & ntilde;o events across Antarctica. EP El Ni & ntilde;o exhibits a broader and more substantial influence on the magnitude and spatial extent of extreme precipitation anomalies compared to CP events, with both types showing their most pronounced impacts during the cold season. In that season, both EP and CP El Ni & ntilde;o events are associated with positive extreme precipitation anomalies over the Ross Sea and Marie Byrd Land and negative anomalies over the Bellingshausen Sea and the Antarctic Peninsula. EP El Ni & ntilde;o also produced marked negative anomalies over East Antarctica (60 degrees-90 degrees E) and positive anomalies over Dronning Maud Land and the southern Atlantic Ocean, whereas such anomalies are minimal or absent during CP events. In the warm season, EP El Ni & ntilde;o generates a dipole pattern, with positive anomalies over the Amundsen and Ross Seas and negative anomalies over the Weddell Sea and Dronning Maud Land, a structure that is significantly weakened for CP events. These differences are attributed to distinct upper-tropospheric wavetrains, driven by positive tropical Pacific and Indian sea-surface-temperature (SST) anomalies, which modulate water vapour flux and vertical velocity across the Antarctic continent.

Understanding the relationships between bed roughness and glaciological and geological characteristics is crucial for predicting the dynamics and stability of the Antarctic Ice Sheet (AIS), which further informs climate change projections and sea-level rise estimates. The links between bed roughness and ice flow, basal temperature, and sediment distribution in the Princess Elizabeth Land (PEL) region of East Antarctica are intricate and poorly understood. However, with bed elevation data collected by airborne ice-penetrating radar in seven seasons we are able to derive and analyze bed roughness in both spectral and spatial domains in PEL. The results show pronounced spatial heterogeneity in bed roughness, with distinct regions exhibiting either "smooth" or "rough" bed conditions. Comparative analysis between spectral and spatial domain roughness parameters indicates relative consistency across the region, suggesting their reliability as indicators of bed conditions. We explore the interrelations between bed roughness and basal temperature, ice flow, and sediment distribution, demonstrating the interdependencies among these glaciological characteristics in PEL and, by this demonstration, likely elsewhere. Our study underscores the importance of bed roughness analysis in understanding subglacial conditions and processes critical to ice sheet dynamics and stability.

Princess Elizabeth Land (PEL) has a complex subglacial lake system and rugged topography that significantly affect the stability of the Antarctic Ice Sheet and global sea-level change. Gravity fields provide essential constraints for investigating the structure and mass balance of ice sheets. However, existing gravity observations in PEL have low spatial resolution, measurement accuracy, and areal coverage. We employed high-precision airborne gravity data acquired during the 32nd and 33rd Chinese National Antarctic Research Expedition (CHINARE 32, 2015/16; CHINARE 33, 2016/17) and data from the global gravity field model XGM2019e_2159 to construct a high-resolution regional gravity field model (SKIce-Grav 2025) of PEL using the spherical radial basis function (SRBF). The numerical stability and physical consistency of the model were substantially improved by minimizing edge effects and using the diagonal standard deviation of the cofactor matrix. The SKIce-Grav 2025 model yielded a significantly lower residual standard deviation (3.29 mGal) than XGM2019e_2159 (10.74 mGal), with an average deviation from ground measurements of 2.47 mGal, representing a marked improvement over XGM2019e_2159 (14.02 mGal). These findings demonstrate that the SRBF-based data fusion approach offers superior stability and precision in reconstructing the polar gravity field, providing a robust framework for resolving subglacial topographic features and improving our understanding of Antarctic sub-ice geological structures.

Melt ponds are typical surface features of Arctic sea ice during the melting season. Their formation significantly reduces ice surface albedo, intensifying ice-albedo feedback. Evaluating the interannual variability and spatial evolution of Pan-Arctic melt pond fraction (MPF), as well as its relationship with atmospheric circulation, is crucial for understanding Arctic sea ice melting mechanisms. In this study, based on the MODIS 8-day composite surface reflectance product (MOD09A1), we improved upon the extant three-surface-class linear spectral unmixing (LSM) by introducing a bare ice endmember and optimizing band combinations. Through comparison with reference data, the incorporation of bare ice reduced the mean bias/RMSE of MPF retrievals from 8.69%/ 11.12% to 3.84%/8.15%, thereby allowing for the reconstruction of the Pan-Arctic LSM-MPF dataset spanning May to August from 2000 to 2023. An analysis revealed that the negative-phase North Atlantic Oscillation (NAO) during June-July and negativephase Arctic Oscillation (AO) during July-August significantly influences the interannual variability and evolution of Pan-Arctic MPF. In June, the negative-phase NAO induced a high-pressure anomaly over Greenland, facilitating the transport of warm and humid moisture masses from the North Atlantic into the Western Arctic, resulting in a marked increase in MPF in the Northwestern Central Arctic (NWCA) region. In July, the highpressure center gradually shifts northward, and the atmospheric circulation pattern transitions from a negative-phase NAO to a negative-phase AO. This shift facilitates the further transportation of moisture into the Eastern Arctic, extending significant MPF increases from the NWCA to the Northeastern Central Arctic (NECA) regions. This process intensifies the ice-albedo feedback and enhances radiative energy absorption by sea ice. By August, the influence of the negative-phase NAO weakens, while the negative-phase AO sustains anticyclonic circulation over the central Arctic, which favors moisture retention. Consequently, MPF continues to increase in areas such as the NWCA, delaying the refreezing of melt ponds. The LSM-MPF dataset developed in this study is suited for large-scale climate process analysis and holds promise for advancing our understanding of sea-ice-atmosphere interactions.

Accurate and continuous extraction of ice layer structures from airborne ice sounding radar data is critical for analyzing subglacial hydrothermal environments to support key applications such as geothermal heat flux mapping, subglacial hydrology delineation, ice flow dynamics prediction, ice temperature profile reconstruction, and Antarctic expedition efficiency. However, existing algorithms often suffer from missed detections, false positives, and poor layer continuity-particularly in complex datasets with dense internal layers or noise interference-limiting their applicability in regional-scale subglacial research. To address these limitations, this article proposes a fully automatic ice layer extraction algorithm: It employs a 2-D multiscale curvelet transform (CT) for local orientation extraction and denoising, integrates a weight-optimized minimum spanning tree (MST) to construct a robust layer-connection framework, and supplements these with turn-back point detection and sublayer merging for independent layer segmentation. Validated on field data from the CHINARE 32 expedition and compared against state-of-the-art methods on a standardized CReSIS dataset, the algorithm demonstrates superior performance in comprehensive evaluations. It achieves higher layer detection completeness, better continuity, and improved alignment with actual stratigraphy compared to existing approaches while effectively eliminating hallucinated layers. The proposed method maintains robust performance across diverse ice conditions without requiring manual intervention, establishing an effective bridge between local feature detection and global optimization in radar data processing.

Heat flow is an important parameter reflecting the thermal structure of the Earth's interior and a key indicator of geothermal resource potential. Mexico lies at the tectonic intersection of the Pacific Ring of Fire and holds potential for geothermal energy exploration. Existing studies primarily use traditional interpolation methods, magnetic data inversion, or similarity methods for regional heat flow estimation. However, these approaches are limited by sparse and unevenly distributed data as well as crustal complexities, resulting in difficulties in capturing detailed spatial heterogeneity of heat flow. To accurately predict the surface heat flow distribution in the Mexico region, this study proposes an integrated framework combining two-stage feature selection and a cascaded machine learning model. First, a compact and informative feature set was constructed using a two-stage selection process that integrates correlation analysis and modeling-based evaluation. The selected features include lithosphere-asthenosphere boundary (LAB) depth, gravity mean curvature, 230 km shear wave velocity, 300 km shear wave velocity, topography (elevation), distance to ridge, and distance to volcano. Subsequently, an initial heat flow prediction was generated using a Random Forest (RF) model based on this feature set. Finally, both the initial prediction and the selected features were jointly input into an Extreme Gradient Boosting (XGBoost) model for further learning, resulting in a more accurate heat flow estimation. A 0.5 degrees x 0.5 degrees resolution heat flow distribution map of Mexico was generated based on key geological features. The map reveals that Mexico's heat flow exhibits a spatial distribution characterized by high values in the west and low values in the east, with extreme values located along volcanic belts. High heat flow concentrates in central-southern Mexico and the Gulf of California coast; low values occur along the south Pacific coast, southern Gulf of Mexico, and Yucatan Peninsula.

Subglacial Lake Qilin (SLQ), in the center of Princess Elizabeth Land (PEL), East Antarctica, is a potential drilling target for detecting extreme-life and studying ice sheet evolution, due to its tectonic origin, stability, thick sediments and isolation by 3600 m thick ice. Prior to drilling, an ideal site is needed to meet scientific goals, requiring characterization of water circulation and subglacial hydrology-related basal melting/refreezing. This study quantifies SLQ's basal melt rate using an optimized 1D steady-state thermodynamic model and multi-source remote sensing/geophysical data. The model improves accuracy via dynamic thermal boundaries and a temperature gradient correction. Results show the lake center has a high annual mean basal melt rate of 2.195 mm a-1, increasing northward. Sensitivity analysis indicates geothermal heat flux has the most significant effect on melt rate, compared to ice thickness and temperature gradient.

The Antarctica is surrounded by sea ice in winter, while at the end of summer, sea ice hundreds of kilometers wide can remain near the coast. However, in the Ross Sea section, an ocean channel usually forms in the melt season, connecting the Ross Sea Ice Shelf front with the Southern Ocean. This phenomenon has received little attention. In December, the Ross Sea Ice Shelf Polynya expands under the prevailing southerlies, which facilitates sea ice melting by pushing sunlight-heated ocean water northward. In January, the Ice Strip to the north of the Ross Sea Ice Shelf breaks, and an ocean channel appears around the 180 degrees date line. The key factor of prevailing southerlies to the north of the Ross Sea Ice Shelf in December is generally correlated with the Southern Annular Mode. This study underlines the role of Antarctic katabatic winds during sea ice melt season.

The rapid mass loss of the Antarctic Ice Sheet (AIS), with severe consequences such as sea-level rise, underlies the importance of understanding ice-sheet behavior into the future. Previous studies have identified subglacial volcanoes in Antarctica using various observational and geophysical methods, and have shown that subglacial volcanism can modify subglacial conditions and influence ice-sheet dynamics. However, the distribution, structure, and morphology of Antarctic subglacial volcanoes have not yet been systematically documented across the continent, despite this being a potentially valuable boundary condition for ice-sheet modelling, because volcanic morphology—including edifice relief, shape, and slope—can regulate basal meltwater production and routing, thereby shaping subglacial hydrology and modulating basal traction and ice-flow dynamics, with implications for AIS stability. This study presents a continent-wide reference inventory of Antarctic subglacial volcanic candidates, in which we compile and quantify their key morphological characteristics to explore how volcanic relief and edifice shape may influence basal melt, subglacial hydrology, and, in turn, the dynamics of the AIS and its global sea-level implications. The inventory includes 207 subglacial volcanic edifices that have been interpreted as volcanic in previous studies and for which we can constrain geographic locations. From this compilation, we analyze their distribution, morphological characteristics, and classification in order to provide key geometric and topographic constraints for future studies of their potential interactions with the AIS. This effort enables a novel understanding of Antarctic subglacial volcanism and also provides a needed reference for subglacial volcanoes to support further critical research concerning the evolution of Antarctica's great ice sheets.

Surface meltwater on Antarctic ice shelves commonly occurs as supraglacial lakes, ponded-water patches and meltwater channels, and its storage and drainage can modify ice-shelf surface hydrology, increase surface loading and promote hydrofracture where water is routed into crevassed regions. Previous studies have mapped surface meltwater on the Amery Ice Shelf (AmIS) using satellite remote sensing and machine-learning approaches, but they have mainly focused on selected years, local drainage systems, individual melt events or algorithm evaluation. Due to the diagnostic value of January peak surface-meltwater extent for assessing AmIS surface hydrology and ice-shelf stability, a long-term, high-resolution and consistently processed dataset is needed for reliable interannual comparison. Here, we present a high-resolution January surface meltwater dataset for the AmIS from 2005 to 2026. The cross-sensor dataset was generated from Landsat 7 ETM+, Landsat 8 OLI and Sentinel-2 MSI imagery using a weakly supervised segmentation framework with source-adaptive components trained using NDWI-derived labels. The dataset includes annual January surface meltwater extent maps, annual surface meltwater area statistics and a multi-year surface meltwater occurrence-count product. All raster products are provided on a common 10 m product grid to support spatial comparison across image sources and years. For Landsat-derived products, this grid is used for spatial alignment, while the source Landsat multispectral bands retain their native 30 m spatial resolution. Technical validation showed high pixel-level agreement, with an overall accuracy of 0.9882, precision of 0.9507, recall of 0.9086, F1 score of 0.9291, intersection over union of 0.8677 and kappa coefficient of 0.9227. Compared with threshold-based methods, conventional machine-learning classifiers and multiple deep-learning models, our method achieved the highest F1 score, IoU and kappa. Component-wise ablation experiments and representative visual and boundary comparisons indicated improved delineation of narrow channels, small ponded-water features and continuous lake margins. The 2005–2026 dataset shows pronounced interannual variability in January surface meltwater extent, with relatively extensive surface meltwater coverage in 2005–2006, 2014, 2017, 2019 and 2025 and reduced coverage in 2007, 2011, 2021 and 2023. The occurrence-count product identifies recurrent January surface meltwater mainly across the western and central AmIS, with additional clusters along the southeastern ice-shelf margin. Recurrent surface meltwater is more concentrated near rock outcrops, varies across surface-elevation bands and shows spatial correspondence with MEaSUREs ice-velocity fields and ice-flow direction patterns. The dataset can be combined with runoff, surface mass balance, crevasse fields and MEaSUREs ice-velocity products to support studies of peak-season surface hydrology, recurrent surface meltwater pathways and stability-relevant surface meltwater settings on the AmIS.

In East Antarctica, the largest thinning rates are observed at Totten Glacier in recent years. Hydrologic activity of the three active subglacial lakes (Totten1, Totten2, and Wilkes1) located on Totten Glacier may affect the ice sheet mass balance in the region on time scales of decades. In this article, we utilized laser altimetry data (ICESat and ICESat-2) and radar altimetry data (CryoSat-2 and Sentinel-3) to establish a 20-year time series of surface ice sheet elevation changes for three subglacial lakes, employing different least squares fitting methods, and this analysis aimed to study their hydrological activities. Additionally, we combined REMA and BedMachine data to acquire the subglacial drainage pathways in the region, analyzing the hydrological connections among the three subglacial lakes. The results indicate that Totten1 and Totten2 exhibited frequent inflow and outflow throughout the observation period, with periodic characteristics in lake activities. From 2003 to 2009, Wilkes1 showed an ascending trend in surface ice sheet elevation, followed by a relatively stable state. The characteristics of lake activities changes and subglacial drainage pathways indicate connections among these three subglacial lakes. This article highlights that CryoSat-2 and Sentinel-3 radar data can fill the gaps between ICESat and ICESat-2 data. Furthermore, ICESat-2 laser altimetry data not only extend the records of subglacial lake activities but also capture more densely and accurately resolved spatial details. The integration of these four altimeters proves effective for long-term monitoring of active subglacial hydrological activities.

Heat flow is a key parameter for revealing the Earth's internal structure and the distribution of geothermal resources, playing a crucial role in geoscientific research and resource assessment. Current heat flow prediction typically relies on statistical regression or machine learning methods, which model and estimate heat flow by uncovering its relationships with geological and geophysical features. However, the existing studies still face two core challenges: first, key modeling features generally suffer from low spatial resolution at the global scale, which limits the resolution of heat flow predictions; second, current machine learning methods often exhibit insufficient feature extraction and generalization capabilities under limited data conditions. To address these issues, various high-resolution geological feature data were systematically integrated and analyzed, and a geographic-climatic proximity feature combination (GeoClimaProx) was proposed to balance effectiveness with accessibility. In addition, the tabular prior-data fitted network (TabPFN), an advanced tabular foundation model characterized by strong feature extraction and generalization capabilities, is introduced to enhance the performance of continental heat flow prediction. To validate the effectiveness of the proposed approach, comparative experiments were conducted at multiple spatial resolutions. The results demonstrate that GeoClimaProx outperforms traditional feature combinations, contributing to improved accuracy and spatial resolution in heat flow prediction, and TabPFN exhibits superior generalization ability and higher predictive accuracy across various spatial resolutions and training data sizes. GeoClimaProx and TabPFN provide a novel technical pathway for accurate heat flow prediction using only easily accessible surface observations and tectonic distance features. Based on them, the first 0.2 degrees resolution heat flow map covering global continental regions is constructed, improving the spatial resolution of heat flow modeling, which means the beginning of fine-scale modeling of continental heat flow at the global scale. The code, datasets, and resulting data products are publicly available at https://github.com/zhang152267/GCHF to facilitate reproducibility.