Global Weather Prediction Model Research Articles

A Hybrid Machine Learning/Physics‐Based Modeling Framework for 2‐Week Extended Prediction of Tropical Cyclones

AbstractPrediction of tropical cyclones (TCs) beyond a week is challenging but of great importance for disaster prevention and mitigation. We propose a hybrid machine learning (ML)/physics‐based modeling framework to extend TC forecasts to 2 weeks. This framework integrates a recently launched ML‐based global weather prediction model (Pangu) and the high‐resolution physics‐based regional weather research and forecasting (WRF) model. The Pangu model shows promise in enhancing the accuracy of predictions for large‐scale circulation and TC tracks, while the high‐resolution WRF model is capable of capturing the core processes underlying TC evolution. To capitalize on the complementary strengths of both the Pangu and WRF models in predicting TCs, the framework comprises three key components: downscaling the Pangu model using the WRF model, adjusting large‐scale circulation through spectral nudging driven by the Pangu model forecasts, and updating sea surface temperature using an ocean mixed‐layer model. These components also ensure the framework's feasibility for real‐time TC forecasting. The prediction skill of the framework has been demonstrated for five long‐lived TCs across various basins from 2018 to 2023. Results indicate that the hybrid ML/physics‐based modeling framework decreased the 2‐week mean TC track and intensity errors by 59% and 32% compared to the global numerical weather prediction models, by 2% and 59% compared to the ERA5‐driven Pangu model, and by 32% and 23% compared to the ERA5‐driven WRF model, respectively. This implies that the framework has great potential to be used for 2‐week extended prediction of TCs.

Dynamical Tests of a Deep Learning Weather Prediction Model

Abstract Global deep learning weather prediction models have recently been shown to produce forecasts that rival those from physics-based models run at operational centers. It is unclear whether these models have encoded atmospheric dynamics or simply pattern matching that produces the smallest forecast error. Answering this question is crucial to establishing the utility of these models as tools for basic science. Here, we subject one such model, Pangu-Weather, to a set of four classical dynamical experiments that do not resemble the model training data. Localized perturbations to the model output and the initial conditions are added to steady time-averaged conditions, to assess the propagation speed and structural evolution of signals away from the local source. Perturbing the model physics by adding a steady tropical heat source results in a classical Matsuno–Gill response near the heating and planetary waves that radiate into the extratropics. A localized disturbance on the winter-averaged North Pacific jet stream produces realistic extratropical cyclones and fronts, including the spontaneous emergence of polar lows. Perturbing the 500-hPa height field alone yields adjustment from a state of rest to one of wind–pressure balance over ∼6 h. Localized subtropical low pressure systems produce Atlantic hurricanes, provided the initial amplitude exceeds about 4 hPa, and setting the initial humidity to zero eliminates hurricane development. We conclude that the model encodes realistic physics in all experiments and suggest that it can be used as a tool for rapidly testing a wide range of hypotheses.

The sensitivity of atmospheric blocking to upstream latent heating – numerical experiments

Abstract. Recent studies have pointed to an important role of latent heating during cloud formation for the dynamics of anticyclonic circulation anomalies such as atmospheric blocking. However, the effect of latent heating on blocking formation and maintenance has not yet been fully elucidated. To explicitly study this cause-and-effect relationship, we perform sensitivity simulations of five selected blocking events with the Integrated Forecast System (IFS) global weather prediction model in which we artificially eliminate latent heating in clouds upstream of the blocking anticyclones. This elimination has substantial effects on the upper-tropospheric circulation in all case studies, but there is also significant case-to-case variability: some blocking systems do not develop at all without upstream latent heating, while for others the amplitude, size, and lifetime of the blocking anticyclones are merely reduced. This strong influence of latent heating on the midlatitude flow is due to the injection of air masses with low potential vorticity (PV) into the upper troposphere in strongly ascending “warm conveyor belt” airstreams and the interaction of the associated divergent outflow with the upper-level PV structure. The important influence of diabatic heating demonstrated with these experiments suggests that the accurate representation of moist processes in ascending airstreams in weather prediction and climate models is crucial for blocking dynamics.

An Array of Ringing Global Free Modes Discovered in Tropical Surface Pressure Data

AbstractWe used newly available ERA5 hourly global data to examine the variations of atmospheric circulation on global scales and high frequencies. The space–time spectrum of surface pressure displays a typical red background spectrum but also a striking number of isolated peaks. Some peaks represent astronomically forced tides, but we show that most peaks are manifestations of the ringing of randomly excited global-scale resonant modes, reminiscent of the tones in a spectrum of a vibrating musical instrument. A few such modes have been tentatively identified in earlier observational investigations, but we demonstrate the existence of a large array of normal mode oscillations with periods as short as 2 h. This is a powerful and uniquely detailed confirmation of the predictions of the theory of global oscillations that has its roots in the work of Laplace two centuries ago. The delineation of the properties of the modes provides valuable diagnostic information about the atmospheric circulation. Notably the amplitudes and widths of the normal mode spectral peaks contain information on the forcing mechanisms and energy dissipation for the modes, and the simulation of these properties for each of the many modes we have identified can serve as tests for global climate and weather prediction models.

The Flexible Ocean and Climate Infrastructure version 1 (FOCI1): mean state and variability

Abstract. A new Earth system model, the Flexible Ocean and Climate Infrastructure (FOCI), is introduced. A first version of FOCI consists of a global high-top atmosphere (European Centre Hamburg general circulation model; ECHAM6.3) and an ocean model (Nucleus for European Modelling of the Ocean v3.6; NEMO3.6) as well as sea-ice (Louvain-la-Neuve sea Ice Model version 2; LIM2) and land surface model components (Jena Scheme for Biosphere Atmosphere Coupling in Hamburg; JSBACH), which are coupled through the OASIS3-MCT software package. FOCI includes a number of optional modules which can be activated depending on the scientific question of interest. In the atmosphere, interactive stratospheric chemistry can be used (ECHAM6-HAMMOZ) to study, for example, the effects of the ozone hole on the climate system. In the ocean, a biogeochemistry model (Model of Oceanic Pelagic Stoichiometry; MOPS) is available to study the global carbon cycle. A unique feature of FOCI is the ability to explicitly resolve mesoscale ocean eddies in specific regions. This is realized in the ocean through nesting; first examples for the Agulhas Current and the Gulf Stream systems are described here. FOCI therefore bridges the gap between coarse-resolution climate models and global high-resolution weather prediction and ocean-only models. It allows to study the evolution of the climate system on regional and seasonal to (multi)decadal scales. The development of FOCI resulted from a combination of the long-standing expertise in ocean and climate modeling in several research units and divisions at the Helmholtz Centre for Ocean Research Kiel (GEOMAR). FOCI will thus be used to complement and interpret long-term observations in the Atlantic, enhance the process understanding of the role of mesoscale oceanic eddies for large-scale oceanic and atmospheric circulation patterns, study feedback mechanisms with stratospheric processes, estimate future ocean acidification, and improve the simulation of the Atlantic Meridional Overturning Circulation changes and their influence on climate, ocean chemistry and biology. In this paper, we present both the scientific vision for the development of FOCI as well as some technical details. This includes a first validation of the different model components using several configurations of FOCI. Results show that the model in its basic configuration runs stably under pre-industrial control as well as under historical forcing and produces a mean climate and variability which compares well with observations, reanalysis products and other climate models. The nested configurations reduce some long-standing biases in climate models and are an important step forward to include the atmospheric response in multidecadal eddy-rich configurations.

Representation of atmospheric blocking in the new global non-hydrostatic weather prediction model ICON

Representation of atmospheric blocking in the new global non-hydrostatic weather prediction model ICON

Resolving Sahelian thunderstorms improves mid-latitude weather forecasts

The summertime West African Sahel has the worldwide highest degree of thunderstorm organisation into long-lived, several hundred-kilometre elongated, fast propagating systems that contribute 90% to the annual rainfall. All current global weather prediction and climate models represent thunderstorms using simplified parameterisation schemes which deteriorates the modelled distribution of rainfall from individual storms and the entire West African monsoon circulation. It is unclear how this misrepresentation of Sahelian convection affects forecasts globally. Our study is the first to demonstrate how a computationally feasible increase of model resolution over West Africa – allowing to avoid convection parameterisation – yields a better representation of organised convection in the Sahel and of moisture within the monsoon system, ultimately improving 5–8-day tropical and mid-latitude weather forecasts. We advocate an operational use of a modelling strategy similar to the one presented here for a cost-effective improvement of global weather prediction and potentially even (sub-)seasonal and climate simulations.

Effect of Evaporating Sea Spray on Heat Fluxes in a Marine Atmospheric Boundary Layer

AbstractA detailed analysis of the evaporating ocean spray effect on the vertical latent and sensible heat fluxes in a marine atmospheric boundary layer (MABL) for different droplet sizes, vertical distributions of air temperature, humidity, and turbulent intensity is presented. For our analysis we have employed a two-temperature nonequilibrium MABL model developed in our previous work. The obtained analytical and numerical solutions show that the latent and total heat fluxes are significantly enhanced by large droplets because these droplets produce steep vertical gradients of moisture and air temperature in a MABL. Small droplets, however, do not noticeably change the total heat flux but rather redistribute the energy between its sensible and latent components. It has been shown that evaporating spray affects the turbulent kinetic energy (thus the intensity of the vertical turbulent transport) mostly mechanically by altering the vertical distribution of the mass density of the air–spray mixture rather than thermodynamically by changing vertical profiles of the air temperature and moisture. Furthermore, we have found that the vertical profiles of heat fluxes are approximately self-similar for a wide range of defining parameters, that is, can be approximately scaled to a reference heat profile for a wide range of vertical distributions of the temperature, humidity, and turbulence intensity. The obtained analytical expressions for the vertical heat fluxes affected by the spray presence enable their quick and efficient calculations. This will allow for the future construction of a computationally efficient spray and accurate parameterization to be used in global weather prediction models.

First Application of the Local Ensemble Tangent Linear Model (LETLM) to a Realistic Model of the Global Atmosphere

Abstract The local ensemble tangent linear model (LETLM) provides an alternative method for creating the tangent linear model (TLM) and adjoint of a nonlinear model that promises to be easier to maintain and more computationally scalable than earlier methods. In this paper, we compare the ability of the LETLM to predict the difference between two nonlinear trajectories of the Navy’s global weather prediction model at low resolution (2.5° at the equator) with that of the TLM currently used in the Navy’s four-dimensional variational (4DVar) data assimilation scheme. When compared to the pair of nonlinear trajectories, the traditional TLM and the LETLM have improved skill relative to persistence everywhere in the atmosphere, except for temperature in the planetary boundary layer. In addition, the LETLM was, on average, more accurate than the traditional TLM (error reductions of about 20% in the troposphere and 10% overall). Sensitivity studies showed that the LETLM was most sensitive to the number of ensemble members, with the performance gradually improving with increased ensemble size up to the maximum size attempted (400). Inclusion of physics in the LETLM ensemble leads to a significantly improved representation of the boundary layer winds (error reductions of up to 50%), in addition to improved winds and temperature in the free troposphere and in the upper stratosphere/lower mesosphere. The computational cost of the LETLM was dominated by the cost of ensemble propagation. However, the LETLM can be precomputed before the 4DVar data assimilation algorithm is executed, leading to a significant computational advantage.

Parallelization and Performance of the NIM Weather Model on CPU, GPU, and MIC Processors

Abstract The design and performance of the Non-Hydrostatic Icosahedral Model (NIM) global weather prediction model is described. NIM is a dynamical core designed to run on central processing unit (CPU), graphics processing unit (GPU), and Many Integrated Core (MIC) processors. It demonstrates efficient parallel performance and scalability to tens of thousands of compute nodes and has been an effective way to make comparisons between traditional CPU and emerging fine-grain processors. The design of the NIM also serves as a useful guide in the fine-grain parallelization of the finite volume cubed (FV3) model recently chosen by the National Weather Service (NWS) to become its next operational global weather prediction model. This paper describes the code structure and parallelization of NIM using standards-compliant open multiprocessing (OpenMP) and open accelerator (OpenACC) directives. NIM uses the directives to support a single, performance-portable code that runs on CPU, GPU, and MIC systems. Performance results are compared for five generations of computer chips including the recently released Intel Knights Landing and NVIDIA Pascal chips. Single and multinode performance and scalability is also shown, along with a cost–benefit comparison based on vendor list prices.

Europeans shine in weather forecasting

The US global model lags the performance of two European competitors in predicting weather up to two weeks ahead.

Wavelet Compression Technique for High-Resolution Global Model Data on an Icosahedral Grid

AbstractModern Earth modeling systems often use high-resolution unstructured grids to discretize their horizontal domains. One of the major challenges in working with these high-resolution models is to efficiently transmit and store large volumes of model data for operational forecasts and for modeling research.A newly developed compression technique is presented that significantly reduces the size of datasets produced by high-resolution global models that are discretized on an icosahedral grid. The compression technique is based on the wavelet transform together with a grid rearrangement algorithm and precision-controlled quantization technology. The grid rearrangement algorithm converts an icosahedral grid to a set of 10 rhombus grids that retain the spatial correlation of model data so that a three-dimensional wavelet transform can be effectively applied. The precision-controlled quantization scheme guarantees specified precision of compressed datasets.The technique is applied to the output of a global weather prediction model, the Flow-Following, Finite-Volume Icosahedral Model (FIM) developed by NOAA’s Earth System Research Laboratory. Experiments show that model data at 30-km resolution can be compressed up to 50:1 without noticeable visual differences; at specified precision requirements, the proposed compression technique achieves better compression compared to a state-of-the-art compression format [Gridded Binary (GRIB) with JPEG 2000 packing option]. In addition, model forecasts initialized with original and compressed initial conditions are compared and assessed. The assessment indicates that it is promising to use the technique to compress model data for those applications demanding high fidelity of compressed datasets.

A Vertically Flow-Following Icosahedral Grid Model for Medium-Range and Seasonal Prediction. Part I: Model Description

Abstract A hydrostatic global weather prediction model based on an icosahedral horizontal grid and a hybrid terrain-following/isentropic vertical coordinate is described. The model is an extension to three spatial dimensions of a previously developed, icosahedral, shallow-water model featuring user-selectable horizontal resolution and employing indirect addressing techniques. The vertical grid is adaptive to maximize the portion of the atmosphere mapped into the isentropic coordinate subdomain. The model, best described as a stacked shallow-water model, is being tested extensively on real-time medium-range forecasts to ready it for possible inclusion in operational multimodel ensembles for medium-range to seasonal prediction.

Massively parallel solvers for elliptic partial differential equations in numerical weather and climate prediction

The demand for substantial increases in the spatial resolution of global weather and climate prediction models makes it necessary to use numerically efficient and highly scalable algorithms to solve the equations of large‐scale atmospheric fluid dynamics. For stability and efficiency reasons, several of the operational forecasting centres, in particular the Met Office and the European Centre for Medium‐Range Weather Forecasts (ECMWF) in the UK, use semi‐implicit semi‐Lagrangian time‐stepping in the dynamical core of the model. The additional burden with this approach is that a three‐dimensional elliptic partial differential equation (PDE) for the pressure correction has to be solved at every model time step and this often constitutes a significant proportion of the time spent in the dynamical core. In global models, this PDE must be solved in a thin spherical shell.To run within tight operational time‐scales, the solver has to be parallelized and there seems to be a (perceived) misconception that elliptic solvers do not scale to large processor counts and hence implicit time‐stepping cannot be used in very high‐resolution global models. After reviewing several methods for solving the elliptic PDE for the pressure correction and their application in atmospheric models, we demonstrate the performance and very good scalability of Krylov subspace solvers and multigrid algorithms for a representative model equation with more than 1010 unknowns on 65 536 cores on the High‐End Computing Terascale Resource (HECToR), the UK's national supercomputer. For this, we tested and optimized solvers from two existing numerical libraries (the Distributed and Unified Numerics Environment (DUNE) and Parallel High Performance Preconditioners (hypre)) and implemented both a conjugate gradient solver and a geometric multigrid algorithm based on a tensor‐product approach, which exploits the strong vertical anisotropy of the discretized equation. We study both weak and strong scalability and compare the absolute solution times for all methods; in contrast to one‐level methods, the multigrid solver is robust with respect to parameter variations.

The gravity wave momentum flux in hydrostatic flow with directional shear over elliptical mountains

Semi-analytical expressions for the momentum flux associated with orographic internal gravity waves, and closed analytical expressions for its divergence, are derived for inviscid, stationary, hydrostatic, directionally-sheared flow over mountains with an elliptical horizontal cross-section. These calculations, obtained using linear theory conjugated with a third-order WKB approximation, are valid for relatively slowly-varying, but otherwise generic wind profiles, and given in a form that is straightforward to implement in drag parametrization schemes. When normalized by the surface drag in the absence of shear, a quantity that is calculated routinely in existing drag parametrizations, the momentum flux becomes independent of the detailed shape of the orography. Unlike linear theory in the Ri→∞ limit, the present calculations account for shear-induced amplification or reduction of the surface drag, and partial absorption of the wave momentum flux at critical levels. Profiles of the normalized momentum fluxes obtained using this model and a linear numerical model without the WKB approximation are evaluated and compared for two idealized wind profiles with directional shear, for different Richardson numbers (Ri). Agreement is found to be excellent for the first wind profile (where one of the wind components varies linearly) down to Ri=0.5, while not so satisfactory, but still showing a large improvement relative to the Ri→∞ limit, for the second wind profile (where the wind turns with height at a constant rate keeping a constant magnitude). These results are complementary, in the Ri⪆O(1) parameter range, to Broad’s generalization of the Eliassen–Palm theorem to 3D flow. They should contribute to improve drag parametrizations used in global weather and climate prediction models.

Evaluation of the Earth Systems Research Laboratory's global Observing System Simulation Experiment system

ABSTRACTAn Observing System Simulation Experiment (OSSE) system has been implemented at the National Oceanographic and Atmospheric Administration Earth Systems Research Laboratory in the US as part of an international Joint OSSE effort. The setup of the OSSE consists of a Nature Run from a 13-month free run of the European Center for Medium-Range Weather Forecasts operational model, synthetic observations developed at the National Centers for Environmental Prediction (NCEP) and the National Aeronautics and Space Administration Global Modelling and Assimilation Office, and an operational version of the NCEP Gridpoint Statistical Interpolation data assimilation and Global Forecast System numerical weather prediction model. Synthetic observations included both conventional observations and the following radiance observations: AIRS, AMSU-A, AMSU-B, HIRS2, HIRS3, MSU, GOES radiance and OSBUV. Calibration was performed by modifying the error added to the conventional synthetic observations to achieve a match between data denial impacts on the analysis state in the OSSE system and in the real data system. Following calibration, the performance of the OSSE system was evaluated in terms of forecast skill scores and impact of observations on forecast fields.

An Investigation of a Commercial Aircraft Encounter with Severe Clear-Air Turbulence over Western Greenland

AbstractThis study presents digital flight data recorder (DFDR) analyses and high-resolution numerical simulations relevant to a severe clear-air turbulence (CAT) encounter over western Greenland by a Boeing 777 aircraft at 10-km elevation at 1305 UTC 25 May 2010. The environmental flow was dominated by an extratropical cyclone to the southeast of the Greenland tip, resulting in easterly flow at all levels. The results of the analyses indicate that the CAT encounter was related to mountain-wave breaking on the western lee (downslope) of the Greenland plateau. The simulations were not of especially high resolution (5-km horizontal grid spacing) by today’s standards, yet the simulation results do produce large-amplitude lee waves and overturning in good agreement with the encounter location as indicated by the DFDR. The success of this and other simulations in reproducing mountain-wave turbulence (MWT) events suggests that operational implementation of high-resolution nonhydrostatic simulation models, possibly an ensemble of models, over MWT-prone areas could produce more reliable forecasts of MWT than are currently available using gravity-wave-drag or MWT-postprocessing algorithms derived from global weather prediction models of relatively coarse scale.

Horizontal grids for global weather and climate prediction models: a review

AbstractA latitude–longitude grid is used by almost all operational atmospheric forecast models, and many research models. However, it is expected that the advantages of a latitude–longitude grid will become outweighed on massively parallel computers by data‐communication bottlenecks. There is therefore renewed interest in quasi‐uniform alternatives. This review surveys and assesses previously proposed horizontal grids for modelling the atmosphere over the sphere. Aspects of numerical accuracy likely to be affected by grid structure are discussed; particular attention is paid to computational modes and grid imprinting. Computational modes are potentially very serious, since they may be excited in realistic applications by boundary conditions, nonlinearity, physical forcing, and data assimilation. The geometry of polyhedra is reviewed due to its relation to numerical degrees of freedom, and hence to numerical wave dispersion and the possible existence of computational modes.All grids proposed to date have known problems or issues that merit further investigation. Orthogonal logically rectangular grids may be generated using conformal maps, but these suffer from singularities and resolution clustering. Resolution clustering may be avoided by using overset grids, but there are potential issues associated with the overlap regions. Alternatively, resolution clustering may be avoided, whilst retaining a logically rectangular grid, by giving up orthogonality; however, existing numerical schemes exploit orthogonality to obtain various properties thought to be important for accuracy, and it is not yet known whether these can also be obtained on non‐orthogonal grids. Quasi‐uniformity and orthogonality can be obtained without resolution clustering or overlaps by using non‐quadrilateral grid cells, such as triangles, or pentagons and hexagons. However, when a staggered placement of variables is used to minimise dispersion errors for fast waves, non‐quadrilateral grids support computational modes.In view of the lack of a single ideal grid, several topics meriting further investigation are identified. Copyright © 2011 Royal Meteorological Society and British Crown Copyright, the Met Office

Downscaling Global Weather Forecast Outputs Using ANN for Flood Prediction

Downscaling global weather prediction model outputs to individual locations or local scales is a common practice for operational weather forecast in order to correct the model outputs at subgrid scales. This paper presents an empirical‐statistical downscaling method for precipitation prediction which uses a feed‐forward multilayer perceptron (MLP) neural network. The MLP architecture was optimized by considering physical bases that determine the circulation of atmospheric variables. Downscaled precipitation was then used as inputs to the super tank model (runoff model) for flood prediction. The case study was conducted for the Thu Bon River Basin, located in Central Vietnam. Study results showed that the precipitation predicted by MLP outperformed that directly obtained from model outputs or downscaled using multiple linear regression. Consequently, flood forecast based on the downscaled precipitation was very encouraging. It has demonstrated as a robust technology, simple to implement, reliable, and universal application for flood prediction through the combination of downscaling model and super tank model.

Midlatitude stratosphere – troposphere exchange as diagnosed by MLS O<sub>3</sub> and MOPITT CO assimilated fields

Abstract. This paper presents a comprehensive characterization of a very deep stratospheric intrusion which occurred over the British Isles on 15 August 2007. The signature of this event is diagnosed using ozonesonde measurements over Lerwick, UK (60.14° N, 1.19° W) and is also well characterized using meteorological analyses from the global operational weather prediction model of Météo-France, ARPEGE. Modelled as well as assimilated fields of both ozone (O3) and carbon monoxide (CO) have been used in order to better document this event. O3 and CO from Aura/MLS and Terra/MOPITT instruments, respectively, are assimilated into the three-dimensional chemical transport model MOCAGE of Météo-France using a variational 3-D-FGAT (First Guess at Appropriate Time) method. The validation of O3 and CO assimilated fields is done using self-consistency diagnostics and by comparison with independent observations such as MOZAIC (O3 and CO), AIRS (CO) and OMI (O3). It particularly shows in the upper troposphere and lower stratosphere region that the assimilated fields are closer to MOZAIC than the free model run. The O3 bias between MOZAIC and the analyses is −11.5 ppbv with a RMS of 22.4 ppbv and a correlation coefficient of 0.93, whereas between MOZAIC and the free model run, the corresponding values are 33 ppbv, 38.5 ppbv and 0.83, respectively. In the same way, for CO, the bias, RMS and correlation coefficient between MOZAIC and the analyses are −3.16 ppbv, 13 ppbv and 0.79, respectively, whereas between MOZAIC and the free model they are 6.3 ppbv, 16.6 ppbv and 0.71, respectively. The paper also presents a demonstration of the capability of O3 and CO assimilated fields to better describe a stratosphere-troposphere exchange (STE) event in comparison with the free run modelled O3 and CO fields. Although the assimilation of MLS data improves the distribution of O3 above the tropopause compared to the free model run, it is not sufficient to reproduce the STE event well. Assimilated MOPITT CO allows a better qualitative description of the stratospheric intrusion event. The MOPITT CO analyses appear more promising than the MLS O3 analyses in terms of their ability to capture a deep STE event. Therefore, the results of this study open the perspectives for using MOPITT CO in the STE studies.