Accurate and seamless reconstruction of column-averaged dry-air CO2 mole fraction (XCO ${}_{2}$ ) at high spatial and temporal resolution remains challenging because satellite observations are sparse, unevenly distributed, and frequently interrupted by clouds and retrieval-related limitations. In this study, we propose Trans-XCO2, a spatiotemporal deep learning (DL) framework for reconstructing global daily XCO2 at 0.05° resolution during 2021–2024 by integrating transport background fields with multisource environmental predictors. The framework incorporates instantaneous environmental covariates and short-term atmospheric memory to represent temporal transport effects, and uses a wavelet-enhanced Transformer (WET) to improve multiscale spatial representation. A dual-resolution decoding strategy is further introduced to preserve large-scale background structures while refining fine-scale spatial details. The model is trained using data from 2021 to 2023 and evaluated through strict temporal extrapolation in 2024. Independent validation against measurements from the Total Carbon Column Observing Network (TCCON) shows strong agreement, with an R2 of 0.93 and an RMSE of 1.04 ppm. Compared with coarse-resolution background products, Trans-XCO2 achieves lower reconstruction errors while better preserving spatial gradients. The reconstructed dataset reveals a persistent global XCO2 growth rate of approximately 2.36 ppm yr−1 during 2021–2024, with similar temporal trends over land and ocean. The resulting seamless high-resolution XCO2 dataset provides valuable support for atmospheric carbon monitoring, satellite-model data fusion, and carbon cycle analysis.
4DVarGen is proposed, a 4DVar-inspired generative framework for reconstructing sea surface variable fields at eddy-resolving scales from sparse remote-sensing observations that establishes a mathematical equivalence between 4DVar and an observation-guided denoising process.
Land surface models (LSMs), such as the Community Land Model version 5 (CLM5), represent complex vegetation processes; however, systematic biases persist between modeled and observed leaf area index (LAI) because of parameter uncertainties and knowledge gaps. The high computational cost of CLM5 is a barrier to extensive sensitivity analyses and ensemble simulations at the global scale. This study addresses these limitations by training an evidential deep neural network (EDNN) emulator on a 500‐member CLM5 perturbed parameter ensemble generated via Latin hypercube sampling of 32 key plant physiological parameters. The EDNN employs cyclic temporal encoding to preserve seasonal periodicity and to predict LAI anomalies, thereby emphasizing variability. It also quantifies predictive uncertainty by jointly learning aleatoric and epistemic components in a single forward pass, yielding well‐calibrated probabilistic outputs without requiring computationally intensive ensembles. Across the contiguous United States, the EDNN reproduces CLM5‐simulated LAI with a median R2≈0.8 ${R}^{2}\approx 0.8$ on held‐out members and years while requiring substantially less computation. The EDNN emulator captures seasonal cycles, interannual variability in LAI, and uncertainties (aleatoric and epistemic) in a single pass. The sensitivity analysis highlights photosynthetic capacity and the leaf carbon‐to‐nitrogen ratio as dominant controls on LAI variability, with seasonal shifts in their influence reflecting phenological dynamics. These capabilities enable comprehensive parameter‐sensitivity studies and more efficient calibration and tuning of land surface models by supporting rapid probabilistic forecasting and adaptive model refinement, thereby paving the way for scalable Earth‐system modeling, robust parameter exploration, and uncertainty‐informed projections of land‐surface processes.
Accurate representation of atmospheric moisture is essential for reliable weather forecasting, particularly for small-scale convective systems and extreme events. However, determining high-resolution water vapor (WV) fields remains challenging. Conventional global navigation satellite system (GNSS) troposphere tomography reconstructs 4-D atmospheric wet refractivity fields but is limited by sparse and uneven ray paths, an ill-conditioned coefficient matrix, and an ill-posed inverse problem. Stabilization through constraints and regularization may introduce biases, while the low probability of ray–ray intersections in the lowest tropospheric layers reduces observational influence, causing some regions to depend more on background models than observations. To address these limitations, DeepTomo, to the best of the authors’ knowledge, the first artificial intelligence (AI)-based 4-D GNSS troposphere tomography is introduced as an explainable physics-informed deep learning approach that combines hybrid observational constraints with spatiotemporal learning. Beyond tomographic reconstruction, DeepTomo is conceived as an AI-based assimilation of GNSS observations into ERA5 fields; it integrates a 3-D convolutional neural network (CNN) with residual learning and attention mechanisms and employs a hybrid physics-informed loss function that combines GNSS-derived zenith wet delay (ZWD) with radio occultation (RO) and radiosonde refractivity profiles to correct the ERA5 background toward observational constraints. By learning spatiotemporal relationships between observations and background fields, DeepTomo refines wet refractivity estimates and enables physically consistent reconstruction even in voxels with limited observations. Trained and validated over a dense GNSS network in coastal California using a six-month dataset and evaluated against radiosonde and GNSS-derived ZWD data, DeepTomo performs strongly during the extreme weather event of Hurricane Hilary, a tropical cyclone (TC), in August 2023. Compared with conventional voxel-based tomography, it reduces the root mean square error (RMSE) by up to 64.85% during the TC and 41.62% overall, capturing large moisture variability. Explainable AI (XAI) analysis reveals dynamic spatial attention to regions of enhanced variability. A preliminary sensitivity analysis using GraphCast forecasts shows that the moisture corrections introduced by DeepTomo correspond to short-range forecast errors. This analysis provides an initial indication that DeepTomo, by producing physically consistent GNSS-constrained moisture analyses from ERA5 background fields, has the potential to improve initial conditions and forecast performance in next-generation AI weather forecasting systems, such as GraphCast, bridging GNSS observations with AI-based forecast initialization.
High-resolution measurements of near-surface optical turbulence over the Tibetan Plateau are essential for optical propagation studies, astronomical site characterization, and adaptive optics applications. In this study, a multi-level tethered UAV system was deployed in Ngari Prefecture, China, to obtain synchronous in situ observations of atmospheric optical turbulence from 10 to 190 m above ground. Using measurements at 10–90 m as inputs, we developed a physics-informed reconstruction framework to estimate Cn2 profiles at 110–190 m, thereby approximately doubling the vertical range covered relative to the directly used observations. The framework decomposes the turbulence structure into an equilibrium component describing the large-scale vertical profile and a residual component representing departures from equilibrium. The equilibrium structure is modeled using a dynamically fitted power-law relationship, while a residual-learning module captures short-term variability associated with evolving thermal and dynamical processes. On the independent test period from the same field campaign, the reconstructed profiles achieve an average RMSE below 0.6 in lg(Cn2) and an average R2 above 0.75, outperforming empirical extrapolation and representative data-driven baselines. The proposed framework provides reliable estimates of upper-layer optical turbulence under daytime and transition-period conditions over the Ngari Plateau.
Ocean dissolved oxygen (DO) is a key indicator for assessing the health of marine ecosystems. However, constrained by sparse observational data and insufficient spatiotemporal resolution, traditional reconstruction methods struggle to accurately characterize its nonlinear spatial distribution. This paper proposes a conditional denoising diffusion probabilistic model (CD-DDPM) for ocean dissolved oxygen super-resolution reconstruction. Taking the South China Sea as the study area, we perform a reconstruction task from low-resolution (one-degree by one-degree) to high-resolution (quarter-degree by quarter-degree) using the Copernicus Marine Environment Monitoring Service (CMEMS) global reanalysis product. The model employs U-Net as the denoising network, integrating temperature and salinity (as auxiliary fields) with dissolved oxygen to form a multi-channel input. Additionally, a loss function is introduced to exclude interference from land pixels, and an exponential moving average strategy is adopted to stabilize the training process. Experimental results on the test set show that CD-DDPM achieves a Pearson correlation coefficient of 0.9532, significantly outperforming SRCNN. From the perspectives of PSNR, RMSE, and SSIM, our method also exhibits superior performance compared with traditional optimal interpolation and baseline deep learning models. Ablation experiments further demonstrate that removing both temperature and salinity auxiliary fields increases root mean square error by 98 percent and decreases peak signal-to-noise ratio by 6.72 dB, confirming the irreplaceable constraint effect of multivariate physical information on reconstruction accuracy. This study is the first to apply diffusion models to ocean dissolved oxygen super-resolution reconstruction, providing a new technical pathway for generating high-resolution dissolved oxygen data under sparse observation conditions.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
