Journal of Advances in Modeling Earth Systems最新文献

High performance computing (HPC) architectures have undergone rapid development in recent years. As a result, established software suites face an ever increasing challenge to remain performant on and portable across modern systems. Many of the widely adopted atmospheric modeling codes cannot fully (or in some cases, at all) leverage the acceleration provided by General-Purpose Graphics Processing Units, leaving users of those codes constrained to increasingly limited HPC resources. Energy Research and Forecasting (ERF) is a regional atmospheric modeling code that leverages the latest HPC architectures, whether composed of only Central Processing Units (CPUs) or incorporating GPUs. ERF contains many of the standard discretizations and basic features needed to model general atmospheric dynamics. The modular design of ERF provides a flexible platform for exploring different physics parameterizations and numerical strategies. ERF is built on a state-of-the-art, well-supported, software framework (AMReX) that provides a performance portable interface and ensures ERF's long-term sustainability on next generation computing systems. This paper details the numerical methodology of ERF, presents results for a series of verification/validation cases, and documents ERF's performance on current HPC systems. The roughly 5× speed up of ERF (using GPUs) over Weather Research and Forecasting (CPUs only) for a 3D squall line test case highlights the significance of leveraging GPU acceleration.

Large-scale hydrodynamic models are vital for flood risk assessment and understanding the global water cycle; however, their results can include uncertainties related to spatial resolution. Few studies have evaluated hydrodynamic models across a range of spatial resolutions, with most focusing on a few variables (e.g., discharge) and often neglecting performance at ungauged sites or the role of parameter optimization. We addressed these limitations by comparing Catchment-based Macro-scale Floodplain (CaMa-Flood) model simulations in the Amazon River basin at different spatial resolutions, using the higher resolution as a benchmark in each comparison. We found good inter-resolution performance in simulating discharge and water depth, with coefficients of determination exceeding 0.88 in >80% of locations. The normalized Nash–Sutcliffe efficiencies for discharge and water depth were greater than 0.83 and 0.68, respectively, in more than 75% of locations, suggesting that most locations had consistent hydrodynamics. We detected large discrepancies in discharge between simulations at ∼2.5% of locations due to limited representation of bifurcation flow, floodplain conveyance, and backwater at river confluences in the model. Water depth also differed significantly at ∼3% of locations, mainly at headwaters, due to width bottleneck sections. Flood extent patterns differed minimally between simulations around the main stream and large sub-streams, whereas improvements in the downscaling method are required for small sub-streams. Our results demonstrate the need to improve the representation of bifurcation channels and floodplain parameterization for specific locations, although the general river hydrodynamics patterns were well-captured by computationally efficient moderate-resolution (i.e., 6 arcmin) CaMa-Flood simulations.

Surface turbulent fluxes constitute key energy exchanges in the atmospheric boundary layer (ABL). Accurate prediction of variations in the ABL is essential for agricultural ecology and climate studies. Existing prediction methods include those based on Monin–Obukhov similarity theory (MOST) and machine learning (ML). However, the MOST method requires experimental parameters and empirical equations, while the ML method considerably relies on manual feature extraction. Given the potential of deep learning (DL) in time series prediction, an inverted Transformer (iTransformer) model is employed in this study to predict friction velocity, kinematic sensible heat flux, and kinematic latent heat flux values across different seasons. The iTransformer model encodes the data via transposed encoding, and a multivariate self-attention module is employed to capture the correlations between variables. The feed-forward neural networks leverage these correlations to predict surface turbulent fluxes. Compared with other methods, including the Transformer and ML methods, the iTransformer model can not only improve the prediction correlations but also reduce the errors in surface turbulent fluxes. Moreover, the model can effectively capture the trends in various fluxes within 1 month or even one day. In summary, the iTransformer model can significantly increase the predictive performance for surface turbulent fluxes.

Parameterizations of $��O��\left(��1�-�10��\right)��$km submesoscale mixed layer instabilities in General Circulation Models (GCMs) represent the effects of unresolved vertical buoyancy fluxes (VBF) in the ocean mixed layer. These submesoscale flows interact non-linearly with mesoscale and boundary layer turbulence, and it is challenging to account for all the relevant processes in physics-based parameterizations. In this work, we present a data-driven approach for the submesoscale parameterization, that relies on a Convolutional Neural Network (CNN) trained to predict mixed layer VBF as a function of relevant large-scale variables. The data used for training is given from 12 regions sampled from the global high-resolution MITgcm-LLC4320 simulation. When compared with the baseline of a submesoscale physics-based parameterization, the CNN demonstrates high offline skill across all regions, seasons, and filter scales tested in this study. During seasons when submesoscales are most active, which generally corresponds to winter and spring months, we find that the CNN prediction skill tends to be lower than in summer months. The CNN exhibits a dependency on the large scale strain field, a variable closely related to frontogenesis, which is currently missing from the submesoscale parameterizations in GCMs.

In proximity of the Antarctic ice sheet, oceanic motions with different spatial structures contribute to the transport of warm water onto the continental shelf, fueling submarine ice-shelf melting. Emerging evidence suggests that km-sized ocean submesoscale fronts and eddies significantly impact ice-ocean interactions on timescale of a few hours to a few days, with profound implications for the melting regime. However, ocean models often rely on parameterizations to represent these fine-scale, high-frequency processes because of the coarse resolution. This results in uncertainties in heat transfer mechanisms at the ice-ocean boundary which often leads to conservative estimations of melt rates. Here, we assess the impact of sub-kilometer model resolution on capturing ocean features at different scales in the Amundsen Sea Embayment, West Antarctica, and their impact on the melting regime of local ice shelves. While the simulations do not show substantial differences in large-scale ocean dynamics, we find that using a sub-kilometer resolution is necessary to resolve ocean eddies beneath ice shelves. Furthermore, the ocean kinetic energy using 200 m resolution is approximately three times higher than when using a 1 km resolution. This increase is driven by significantly stronger submesoscale activity, both in the open ocean and beneath the ice shelves, leading to improved model-data agreement, particularly in capturing high melt rates observed at the grounding line.

This study examines how different structural choices in bulk microphysics schemes impact the simulation of warm rain initiation. A single liquid category (SLC) approach prognosing up to four moments of a single drop size distribution (DSD) is compared to the traditional two-category, two-moment approach with separate DSDs for cloud and rain (four total prognostic variables). Different methods for calculating tendencies of the prognostic variables from drop collision-coalescence are also tested: a discretized numerical-integration approach, machine learning via neural networks, lookup tables, and traditional power law fits. Relative to simulations using a bin microphysics model, SLC gives smaller error overall than the two-category approach when numerical integration is used to calculate the collision-coalescence tendencies for both. Replacing the numerical integration with a pre-computed lookup table reduces computational cost with little loss of accuracy. However, using fitted power laws with SLC to represent the collision-coalescence tendencies substantially reduces accuracy and leads to an order of magnitude increase in error. It is also demonstrated that with SLC, reasonably accurate solutions are obtained using only three prognostic moments, while a two-moment SLC scheme leads to substantial error. Overall, both the choice of prognostic moments (e.g., SLC vs. two-category) and method to calculate the collision-coalescence tendencies are important to consider for minimizing errors in bulk schemes. SLC with a sufficiently detailed calculation of the collision-coalescence tendencies provides accurate solutions for a reasonable computational cost, providing a viable alternative to the traditional two-category, two-moment approach for bulk microphysics.

This study presents OceanCastNet (OCN), a machine learning approach for wave forecasting that incorporates wind and wave fields to predict significant wave height, mean wave period, and mean wave direction. We evaluate OCN's performance against the operational ECWAM model using two independent data sets: NDBC buoy and Jason-3 satellite observations. NDBC station validation indicates OCN performs better at 24 stations compared to ECWAM's 10 stations, and Jason-3 satellite validation confirms similar accuracy across 228-hr forecasts. OCN successfully captures wave patterns during extreme weather conditions, demonstrated through Typhoon Goni with prediction errors typically within $��\pm �$0.5 m. The approach also offers computational efficiency advantages. The results suggest that machine learning approaches can achieve performance comparable to conventional wave forecasting systems for operational wave prediction applications.

The skill of numerical weather forecasts strongly depends on the quality of the initial conditions (analyses), which are created by assimilating observations into previous short-range model forecasts. Therefore, it is important to carefully assess the influence of different observations on the analysis. The degrees of freedom for signal (DFS) is a useful metric for quantifying this influence. While DFS has long been used in variational data assimilation (DA) systems, its application in ensemble-based DA systems remains limited. In this study, we propose two novel approaches for estimating the DFS in ensemble-based systems. One approach uses the weighting vector calculated in ensemble transform Kalman filters, while the other uses the innovation vector and observation-space increment vector. We also propose a new strategy for implementing the DFS approaches in the presence of domain localization, which first estimates DFS locally and then aggregates the results to derive a global DFS value for each observation. Our numerical results show that the DFS per observation decreases as the localization radius increases. More generally, the proposed DFS approaches and implementation strategy have the potential to be used in practice to inform the optimization of observation networks and DA systems.

Accurately predicting terrestrial ecosystem responses to climate change over long-timescales is crucial for addressing global challenges. This relies on mechanistic modeling of ecosystem processes through land surface models (LSMs). Despite their importance, LSMs face significant uncertainties due to poorly constrained parameters, especially in carbon cycle predictions. This paper reviews the progress made in using data assimilation (DA) for LSM parameter optimization, focusing on carbon-water-vegetation interactions, as well as discussing the technical challenges faced by the community. These challenges include identifying sensitive model parameters and their prior distributions, characterizing errors due to observation biases and model-data inconsistencies, developing observation operators to interface between the model and the observations, tackling spatial and temporal heterogeneity as well as dealing with large and multiple data sets, and including the spin-up and historical period in the assimilation window. We outline how machine learning (ML) can help address these issues, proposing different avenues for future work that integrate ML and DA to reduce uncertainties in LSMs. We conclude by highlighting future priorities, including the need for international collaborations, to fully leverage the wealth of available Earth observation data sets, harness ML advances, and enhance the predictive capabilities of LSMs.

Data assimilation of observations into full atmospheric states is essential for weather forecast model initialization. Recently, methods for deep generative data assimilation have been proposed which allow for using new input data without retraining the model. They could also dramatically accelerate the costly data assimilation process used in operational regional weather models. Here, in a central US testbed, we demonstrate the viability of Score-based Data Assimilation (SDA) in the context of realistically complex km-scale weather. We train an unconditional diffusion model to generate snapshots of a state-of-the-art km-scale analysis product, the High Resolution Rapid Refresh. Then, using SDA to incorporate sparse weather station data, the model produces maps of precipitation and surface winds. The generated fields display physically plausible structures, such as gust fronts, and sensitivity tests confirm learned physics through multivariate relationships. Preliminary skill analysis shows the approach already outperforms a naive baseline of the High-Resolution Rapid Refresh system itself. By incorporating observations from 40 weather stations, 10% lower RMSEs on left-out stations are attained. Despite some lingering imperfections such as insufficiently disperse ensemble DA estimates, we find the results overall an encouraging proof of concept, and the first at km-scale. It is a ripe time to explore extensions that combine increasingly ambitious regional state generators with an increasing set of in situ, ground-based, and satellite remote sensing data streams.