Eusebi, Ryan, Gabriel A Vecchi, Ching-Yao Lai, and Mingjing Tong, January 2024: Realistic tropical cyclone wind and pressure fields can be reconstructed from sparse data using deep learning. Communications Earth and Environment, 5, 8, DOI:10.1038/s43247-023-01144-2. Abstract
Tropical cyclones are responsible for large-scale loss of life and property1,2,3,4, motivating accurate risk assessment and forecasting. These objectives require accurate reconstructions of storms’ wind and pressure fields which assimilate real-time observations5,6,7,8,9, but current methods used for these reconstructions remain computationally expensive and limited10. Here, we show that a physics-informed neural network11,12 can be a promising and computationally efficient algorithm for tropical cyclone data assimilation. Using synthetic training data sparsely sampled from hurricanes simulated in a forecast model, a physics-informed neural network is able to reconstruct full realistic 2- and 3-dimensional wind and pressure fields which capture key features of the cyclone. We also demonstrate how a set of sparse, real-time observations, can be used to accurately reconstruct Hurricane Ida. Our results highlight how recent advances in deep learning can augment data assimilation schemes. The methods are also general and can be applied to other flow problems.
Tropical cyclone (TC) intensity forecasting poses challenges due to complex dynamical processes and data inadequacies during model initialization. This paper describes efforts to improve TC intensity prediction in the Geophysical Fluid Dynamics Laboratory (GFDL) System for High-resolution prediction on Earth-to-Local Domains (SHiELD) model by implementing a Vortex Initialization (VI) technique. The GFDL SHiELD model, relying on the Global Forecast System (GFS) analysis for initialization, faces deficiencies in initial TC structure and intensity. The VI method involves adjusting the TC vortex inherited from the GFS analysis and merging it back into the environment at the observed location, enhancing the analyzed representation of storm structure. We made modifications to the VI package implemented in the operational Hurricane Analysis and Forecast System, including handling initial condition data, reducing input domain size, and improving storm intensity enhancement. Experiments using the T-SHiELD configuration demonstrate that using VI significantly improves the representation of initial TC intensity and size, enhancing TC predictions, particularly in storm intensity and outer wind forecasts within the first 48 h.
A subseasonal-to-seasonal (S2S) prediction system was recently developed using the GFDL Seamless System for Prediction and Earth System Research (SPEAR) global coupled model. Based on 20-yr hindcast results (2000–19), the boreal wintertime (November–April) Madden–Julian oscillation (MJO) prediction skill is revealed to reach 30 days measured before the anomaly correlation coefficient of the real-time multivariate (RMM) index drops to 0.5. However, when the MJO is partitioned into four distinct propagation patterns, the prediction range extends to 38, 31, and 31 days for the fast-propagating, slow-propagating, and jumping MJO patterns, respectively, but falls to 23 days for the standing MJO. A further improvement of MJO prediction requires attention to the standing MJO given its large gap with its potential predictability (38 days). The slow-propagating MJO detours southward when traversing the Maritime Continent (MC), and confronts the MC prediction barrier in the model, while the fast-propagating MJO moves across the central MC without this prediction barrier. The MJO diversity is modulated by stratospheric quasi-biennial oscillation (QBO): the standing (slow-propagating) MJO coincides with significant westerly (easterly) phases of QBO, partially explaining the contrasting MJO prediction skill between these two QBO phases. The SPEAR model shows its capability, beyond the propagation, in predicting their initiation for different types of MJO along with discrete precursory convection anomalies. The SPEAR model skillfully predicts the observed distinct teleconnections over the North Pacific and North America related to the standing, jumping, and fast-propagating MJO, but not the slow-propagating MJO. These findings highlight the complexities and challenges of incorporating MJO prediction into the operational prediction of meteorological variables.
We describe the third version of the Geophysical Fluid Dynamics Laboratory cloud microphysics scheme (GFDL MP v3) implemented in the System for High-resolution prediction on Earth-to-Local Domains (SHiELD). Compared to the GFDL MP v2, the GFDL MP v3 is entirely reorganized, optimized, and modularized into functions. The particle size distribution (PSD) of all hydrometeor categories is redefined to better mimic observations, and the cloud droplet number concentration (CDNC) is calculated from the Modern-Era Retrospective analysis for Research and Applications, version 2 (MERRA2) aerosol data. In addition, the GFDL MP has been redesigned so all processes use the redefined PSD to ensure overall consistency and easily permit the introduction of new PSDs and microphysical processes. A year's worth of global 13-km, 10-day weather forecasts were performed with the new GFDL MP. Compared to the GFDL MP v2, the GFDL MP v3 significantly improves SHiELD's predictions of geopotential height, air temperature, and specific humidity in the Troposphere, as well as the high, middle and total cloud fractions and the liquid water path. The predictions are improved even further by the use of reanalysis aerosol data to calculate CDNC, and also by using the more realistic PSD available in GFDL MP v3. However, the upgrade of the GFDL MP shows little impact on the precipitation prediction. Degradations caused by the new scheme are discussed and provide a guide for future GFDL MP development.
We present the System for High‐resolution prediction on Earth‐to‐Local Domains (SHiELD), an atmosphere model developed by the Geophysical Fluid Dynamics Laboratory (GFDL) coupling the nonhydrostatic FV3 Dynamical Core to a physics suite originally taken from the Global Forecast System. SHiELD is designed to demonstrate new capabilities within its components, explore new model applications, and to answer scientific questions through these new functionalities. A variety of configurations are presented, including short‐to‐medium‐range and subseasonal‐to‐seasonal prediction, global‐to‐regional convective‐scale hurricane and contiguous U.S. precipitation forecasts, and global cloud‐resolving modeling. Advances within SHiELD can be seamlessly transitioned into other Unified Forecast System or FV3‐based models, including operational implementations of the Unified Forecast System. Continued development of SHiELD has shown improvement upon existing models. The flagship 13‐km SHiELD demonstrates steadily improved large‐scale prediction skill and precipitation prediction skill. SHiELD and the coarser‐resolution S‐SHiELD demonstrate a superior diurnal cycle compared to existing climate models; the latter also demonstrates 28 days of useful prediction skill for the Madden‐Julian Oscillation. The global‐to‐regional nested configurations T‐SHiELD (tropical Atlantic) and C‐SHiELD (contiguous United States) show significant improvement in hurricane structure from a new tracer advection scheme and promise for medium‐range prediction of convective storms.
Liu, Qingfu, Xuejin Zhang, and Mingjing Tong, et al., September 2020: Vortex Initialization in the NCEP Operational Hurricane Models. Atmosphere, 11(9), DOI:10.3390/atmos11090968. Abstract
This paper describes the vortex initialization (VI) currently used in NCEP operational hurricane models (HWRF and HMON, and possibly HAFS in the future). The VI corrects the background fields for hurricane models: it consists of vortex relocation, and size and intensity corrections. The VI creates an improved background field for the data assimilation and thereby produces an improved analysis for the operational hurricane forecast. The background field after VI can be used as an initial field (as in the HMON model, without data assimilation) or a background field for data assimilation (as in HWRF model).
Motivated by the use of the GFDL microphysics scheme in the FV3GFS, the all-sky radiance assimilation framework has been expanded to include precipitating hydrometeors. Adding precipitating hydrometeors allows the assimilation of precipitation-affected radiance in addition to cloudy radiance. In this upgraded all-sky framework, the five hydrometeors, including cloud liquid water, cloud ice, rain, snow and graupel, are the new control variables, replacing the original cloud water control variable. The Community Radiative Transfer Model (CRTM) was interfaced with the newly added precipitating hydrometeors. Sub-grid cloud variability was considered by using the average cloud overlap scheme. Multiple scattering radiative transfer was activated in the upgraded framework. Radiance observations from the Advanced Microwave Sounding Unit-A (AMSU-A) and the Advanced Technology Microwave Sounder (ATMS) over ocean were assimilated in all-sky approach. This new constructed all-sky framework shows neutral to positive impact on overall forecast skill. Improvement was found in 500 hPa geopotential height forecast in both Northern and Southern Hemispheres. Temperature forecast was also improved at 850 hPa in the Southern Hemisphere and the Tropics.
Zhang, Zhan, and Mingjing Tong, et al., July 2020: The Impact of Stochastic Physics-Based Hybrid GSI/EnKF Data Assimilation on Hurricane Forecasts Using EMC Operational Hurricane Modeling System. Atmosphere, 11(8), 801, DOI:10.3390/atmos11080801. Abstract
The National Oceanic and Atmospheric Administration’s (NOAA) cloud-permitting high-resolution operational Hurricane Weather and Research Forecasting (HWRF) model includes the sophisticated hybrid grid-point statistical interpolation (GSI) and Ensemble Kalman Filter (EnKF) data assimilation (DA) system, which allows assimilating high-resolution aircraft observations in tropical cyclone (TC) inner core regions. In the operational HWRF DA system, the flow-dependent background error covariance matrix is calculated from the HWRF self-cycled 40-member ensemble. This DA system has proved to provide improved initial TC structure and therefore improved TC track and intensity forecasts. However, the uncertainties from the model physics are not taken into account in the FY2017 version of the HWRF DA system. In order to further improve the HWRF DA system, the stochastic physics perturbations are introduced in the HWRF DA, including the cumulus convection scheme, the planetary boundary layer (PBL) scheme, and model surface physics (drag coefficient), for HWRF-based ensembles. This study shows that both TC initial conditions and TC track and intensity forecast skills are improved by adding stochastic model physics in the HWRF self-cycled DA system. It was found that the improvements in the TC initial conditions and forecasts are the results of ensemble spread increases which realistically represent the model background error covariance matrix in HWRF DA. For all 2016 Atlantic storms, the TC track and intensity forecast skills are improved by about ~3% and 6%, respectively, compared to the control experiment. The case study shows that the stochastic physics in HWRF DA is especially helpful for those TCs that have inner-core high-resolution aircraft observations, such as tail Doppler radar (TDR) data.
Lim, A H., J A Jung, S E Nebuda, J M Daniels, W Bresky, Mingjing Tong, and Vijay Tallapragada, February 2019: Tropical Cyclone Forecasts Impact Assessment from the Assimilation of Hourly Visible, Shortwave, and Clear-Air Water Vapor Atmospheric Motion Vectors in HWRF. Weather and Forecasting, 34(1), DOI:10.1175/WAF-D-18-0072.1. Abstract
The assimilation of atmospheric motion vectors (AMVs) provides important wind information to conventional data-lacking oceanic regions, where tropical cyclones spend most of their lifetimes. Three new AMV types, shortwave infrared (SWIR), clear-air water vapor (CAWV), and visible (VIS), are produced hourly by NOAA/NESDIS and are assimilated in operational NWP systems. The new AMV data types are added to the hourly infrared (IR) and cloud-top water vapor (CTWV) AMV data in the 2016 operational version of the HWRF Model. In this study, we update existing quality control (QC) procedures and add new procedures specific to tropical cyclone assimilation. We assess the impact of the three new AMV types on tropical cyclone forecasts by conducting assimilation experiments for 25 Atlantic tropical cyclones from the 2015 and 2016 hurricane seasons. Forecasts are analyzed by considering all tropical cyclones as a group and classifying them into strong/weak storm vortices based on their initial model intensity. Metrics such as track error, intensity error, minimum central pressure error, and storm size are used to assess the data impact from the addition of the three new AMV types. Positive impact is obtained for these metrics, indicating that assimilating SWIR-, CAWV-, and VIS-type AMVs are beneficial for tropical cyclone forecasting. Given the results presented here, the new AMV types were accepted into NOAA/NCEP’s operational HWRF for the 2017 hurricane season.
Tong, Mingjing, et al., December 2018: Impact of Assimilating Aircraft Reconnaissance Observations on Tropical Cyclone Initialization and Prediction using Operational HWRF and GSI Ensemble-Variational Hybrid Data Assimilation. Monthly Weather Review, 146(12), DOI:10.1175/MWR-D-17-0380.1. Abstract
This study evaluates the impact of assimilating high-resolution inner-core reconnaissance observations on tropical cyclone initialization and prediction in the 2013 version of the operational Hurricane Weather Research and Forecasting (HWRF) model. The 2013 HWRF data assimilation system is a GSI-based hybrid ensemble-variational system that in this study uses the Global Data Assimilation System ensemble to estimate flow-dependent background error covariance. Assimilation of inner-core observations improves track forecasts and reduces intensity error after 18-24 h. The positive impact on the intensity forecast is mainly found in weak storms, where inner-core assimilation produces more accurate tropical cyclone structures and reduces positive intensity bias. Despite such positive benefits, there is degradation in short-term intensity forecasts that is attributable to spin-down of strong storms, which has also been seen in other studies.
There are several reasons for the degradation of intense storms. First, a newly-discovered interaction between model biases and the HWRF vortex initialization procedure causes the first-guess wind speed aloft to be too strong in the inner core. The problem worsens for the strongest storms, leading to a poor first-guess fit to observations. Though assimilation of reconnaissance observations results in analyses that better fit the observations, it also causes a negative intensity bias at the surface. In addition, the covariance provided by the NCEP global model is inaccurate for assimilating inner-core observations, and model physics biases result in a mismatch between simulated and observed structure. The model ultimately cannot maintain the analysis structure during the forecast, leading to spin-down.
Bernardet, L, Vijay Tallapragada, S Bao, S Trahan, Y Kwon, Q Liu, Mingjing Tong, M Biswas, T Brown, D Stark, L Carson, R Yablonsky, E Uhlhorn, Sundararaman Gopalakrishnan, X Zhang, and Timothy Marchok, et al., June 2015: Community Support and Transition of Research to Operations for the Hurricane Weather Research and Forecast (HWRF) Model. Bulletin of the American Meteorological Society, 96(6), DOI:10.1175/BAMS-D-13-00093.1. Abstract
The Hurricane Weather Research and Forecasting (HWRF) model is an operational model used to provide numerical guidance in support of tropical cyclone forecasting at the National Hurricane Center. HWRF is a complex multi-component system, consisting of the Weather Research and Forecasting (WRF) atmospheric model coupled to the Princeton Ocean Model for Tropical Cyclones (POM-TC), a sophisticated initialization package including a data assimilation system, and a set of postprocessing and vortex tracking tools. HWRF's development is centralized at the Environmental Modeling Center of NOAA's National Weather Service, but it incorporates contributions from a variety of scientists spread out over several governmental laboratories and academic institutions. This distributed development scenario poses significant challenges: a large number of scientists need to learn how to use the model, operational and research codes need to stay synchronized to avoid divergence, and promising new capabilities need to be tested for operational consideration. This article describes how the Developmental Testbed Center has engaged in the HWRF developmental cycle in the last three years and the services it provides to the community in using and developing HWRF.
