Tropical Cyclone Track Forecasts Research Articles

Uncertainty forecasting system for tropical cyclone tracks based on conformal prediction

Tropical cyclone tracks are one of the most critical factors in tropical cyclone forecasting, but there is an inherent uncertainty in their forecasts, but there are no relevant machine learning methods to carry out uncertainty studies. This study proposes an uncertainty forecasting system using machine learning models within a conformal forecasting framework, aiming to provide reliable forecast regions to improve the ability of decision makers to prepare for and respond to potential hazards. This study jointly examines the path forecasting performance of 10 major machine learning models and 10 major conformal forecasting methods through a comparative study. The research work models forecast timescales of 6, 12 and 24 h, and the study covers hurricanes from 1975 to 2021. The experimental results show that the deterministic forecast performance of the model is comparable to the skill of the operated benchmark model, demonstrating that the machine learning model possesses forecasting skill, while also providing tight uncertainty intervals about the path forecast. The method has high prediction accuracy and reliability and is expected to be widely used in the field of tropical cyclone track forecasting and risk communication in the future.

A Potential Vorticity Diagnosis of Tropical Cyclone Track Forecast Errors

AbstractTropical cyclone (TC) track forecasting provides essential guidance for coastal communities. However, track forecast errors still occur, highlighting the need for continued research into error sources. Piecewise potential vorticity (PV) inversion is used systematically to quantitatively diagnose errors in track forecasts in four models during the 2017 Atlantic hurricane season. The deep layer mean steering flow (DLMSF) provides a sufficient proxy for hurricane movement, and DLMSF errors are correlated with TC track errors. Analysis of track forecasts for Hurricanes Harvey, Irma, and Maria reveals that their track errors are attributed to steering errors caused by misrepresentations of specific pressure systems. Harvey's westward track error in the GFS resulted from zonal wind errors from the Continental High, while Irma's northward track error in the SHiELD gfsIC resulted from meridional wind errors in the Bermuda High and Continental High. Maria's southward track error in the IFS resulted from meridional wind errors in the Bermuda High and a misrepresentation of Jose to Maria's northwest. The mean absolute error of the DLMSF shows that the Bermuda High contributed the most to steering flow errors in the cases examined. Our results show that piecewise PV inversion can identify the sources of biases in TC track forecasts. The correction of these biases may lead to improved track forecasts. Quantitative diagnostics presented here provide useful information for future model development.

Development of Interpretable Probability Ellipse in Tropical Cyclone Track Forecasts Using Multiple Operational Ensemble Prediction Systems

AbstractMost tropical cyclone (TC) forecasting centers have implemented a probabilistic circle to represent track uncertainty at a specified lead time. Recent studies suggest that probability ellipses constructed from ensemble prediction systems can convey the anisotropy of track predictability. In this study, a new probability ellipse model is developed to interpret the extent of forward speed and heading uncertainties in ensemble forecasts by selecting an equal proportion of members in the along‐ and cross‐track directions. This method is validated using the 2019–2021 western North Pacific (WNP) TC track forecasts from the ensemble predictions of the European Centre for Medium‐Range Weather Forecasts, the United States National Centers for Environmental Prediction, and the Korea Meteorological Administration. When the proportion of ensemble members in the ellipse is set to 70%, more than one‐half (50.0%–73.6%) of the forecasts, depending on the lead time, indicate reduced area compared with that of the circle. The mean areas of the probability ellipses are 4.9%, 7.0%, 10.0%, and 11.5% smaller than those of the circle in 48‐, 72‐, 96‐, and 120‐hr forecasts, respectively. The forward speed shows greater uncertainty than the heading, as evidenced by the along‐track radii being larger than the cross‐track counterpart in ∼60% of the samples, regardless of the lead time. In addition, the regional distribution of the along‐track/cross‐track ratio in the probability ellipses can explain the dominant direction of the track error in a particular location. The proposed probability ellipse shows potential for application in operational TC track predictions.

Comparison of different global ensemble prediction systems for tropical cyclone intensity forecasting

AbstractMany meteorological centers have operationally implemented global model‐based ensemble prediction systems (GEPSs), making tropical cyclone (TC) forecasts from these systems available. The relatively low resolution of these GEPSs means that limits previous studies primarily focused on TC track forecasting. However, recent GEPS upgrades mean that TC intensity predictions from GEPSs are now also becoming of interest. This study focuses on the verification and comparison of the latest generation of GEPSs for TC intensity forecasts, particularly during the rapid intensification (RI) period over the western North Pacific (WP), eastern North Pacific (EP), and North Atlantic (NA) basins in 2021–2022. On average, the National Centers for Environmental Prediction (NCEP) GEPS performed best in predicting both TC intensity and RI across all three basins. Nevertheless, the exact timing of RI remains highly uncertain for these GEPS, indicating significant limitations in using GEPSs to forecast RI.

Unusual tracks: Statistical, controlling factors and model prediction

The progress of research and forecast techniques for tropical cyclone (TC) unusual tracks (UTs) in recent years is reviewed. A major research focus has been understanding which processes contribute to the evolution of the TC and steering flow over time, especially the reasons for the sharp changes in TC motion over a short period of time. When TCs are located in the vicinity of monsoon gyres, TC track forecast become more difficult to forecast due to the complex interaction between the TCs and the gyres. Moreover, the convection and latent heat can also feed back into the synoptic-scale features and in turn modify the steering flow. In this report, two cases with UTs are examined, along with an assessment of numerical model forecasts. Advances in numerical modelling and in particular the development of ensemble forecasting systems have proved beneficial in the prediction of such TCs. There are still great challenges in operational track forecasts and warnings, such as the initial TC track forecast, which is based on a poor pre-genesis analysis, TC track forecasts during interaction between two or more TCs and track predictions after landfall. Recently, artificial intelligence (AI) methods such as machine learning or deep learning have been widely applied in the field of TC forecasting. For TC track forecasting, a more effective method of center location is obtained by combining data from various sources and fully exploring the potential of AI, which provides more possibilities for improving TC prediction.

Using the Orthogonal Conditional Nonlinear Optimal Perturbations Approach to Address the Uncertainties of Tropical Cyclone Track Forecasts Generated by the WRF Model

Abstract The orthogonal conditional nonlinear optimal perturbations (O-CNOPs) approach for measuring initial uncertainties is applied to the Weather Research and Forecasting (WRF) Model to provide skillful forecasts of tropical cyclone (TC) tracks. The hindcasts for 10 TCs selected from 2005 to 2020 show that the ensembles generated by the O-CNOPs have a greater probability of capturing the true TC tracks, and the corresponding ensemble forecasts significantly outperform the forecasts made by the singular vectors, bred vectors, and random perturbations in terms of both deterministic and probabilistic skills. In particular, for two unusual TCs, Megi (2010) and Tembin (2012), the ensembles generated by the O-CNOPs successfully reproduce the sharp northward-turning track in the former and the counterclockwise loop track in the latter, while the ensembles generated by the other methods fail to do so. Moreover, additional attempts are performed on the real-time forecasts of TCs In-Fa (2021) and Hinnamnor (2022), and it is shown that O-CNOPs are very useful for improving the accuracy of real-time TC track forecasts. Therefore, O-CNOPs, together with the WRF Model, could provide a new platform for the ensemble forecasting of TC tracks with much higher skill.

A probabilistic track model for tropical cyclone risk assessment using multitask learning

Tropical cyclone (TC) track forecasting is critical for wind risk assessment. This work proposes a novel probabilistic TC track forecasting model based on mixture density network (MDN) and multitask learning (MTL). The existing NN-based probabilistic TC track prediction models focus on directly modeling the distribution of the future TC positions. Multitask learning has been shown to boost the performance of single tasks when the tasks are relevant. This work divides the probabilistic track prediction task into two sub-tasks: a deterministic prediction of the future TC position and a probabilistic prediction of the residual between the deterministic prediction and the actual TC location. The MDN is employed to realize the probabilistic prediction task. Since the target values of the MDN in this work are the residuals, which depend on the prediction result of the deterministic task, a novel training method is developed to train the MTL model properly. The proposed model is tested against statistical and other learning-based models on historical TC data. The results show that the proposed model outperforms other models in making probabilistic predictions. This approach advances TC track forecasting by integrating MDN and MTL, showing promise in enhancing probabilistic predictions and improving disaster preparedness.

The influence of large-scale climate modes on tropical cyclone tracks in the southwest Pacific

Tropical cyclones (TCs) impact the economy, properties, lives and infrastructure of island nations and territories of the southwest Pacific (SWP), accounting for three in four regional disasters each year. To increase the resilience of the SWP to the destructive impacts of TCs, improved TC track forecasts are needed since a high degree of uncertainty exists around the likely path a TC will take in this region post-formation. This requires better comprehension of the factors contributing to TC track variability occurring at different timescales. Therefore, we examine the modulating impact of key Indo-Pacific climate drivers: the El Niño-Southern Oscillation (ENSO), Interdecadal Pacific Oscillation (IPO), Southern Annular Mode (SAM) and the Indian Ocean Dipole (IOD), on SWP TC track variability. We present new insights into the spatial (i.e. prevailing trajectories) and temporal (i.e. track length, average speed and duration) components of TC tracks, being modulated by both individual and combined climate modes. Overall, TC tracks tend to shift northeast during El Niño, IPO positive, IOD east positive and/or SAM negative phases (with a southwest shift observed during the opposite climate phases). Further, we show that when two of these climate modes are in their positive phase (e.g. El Niño with the positive phases of IPO or eastern pole of IOD and SAM), TC track length and average speed are enhanced. However, for cases where either one (e.g. El Niño/negative phase of IPO and IOD east) or two (La Niña/negative phase of IPO, IOD east and SAM) climate modes were in the negative phase, an increase in TC track duration was observed. The findings of this study may be used to improve TC forecasting and better quantify TC-related risks.

Tropical Cyclone Forecasts in the DIMOSIC Project—Medium‐Range Forecast Models With Common Initial Conditions

AbstractThe tropical cyclone (TC) forecast skill of the eight global medium‐range forecast models which are participating in the DIfferent Models, Same Initial Conditions project is investigated in this study. Each model was used to generate 10‐day forecasts from the same initial conditions provided by the European Centre for Medium‐Range Weather Forecasts. There are a total of 123 initial dates spanning in one year from June 2018 to June 2019 at 3‐day intervals. The TC track and intensity forecasts are evaluated against the best track data set. TC‐related precipitation and tropical cyclogenesis forecasts are also compared to explore the differences and similarities of TC forecasts across the models. This comparison of TC forecasts allows model developers in different centers to benchmark their model against other models, with the impact of the initial condition quality removed. The verifications reveal that most models show slow‐moving and right‐of‐track biases in their TC track forecasts. Also, a common dry bias in TC‐related precipitation indicates a general deficiency in TC intensity and convection in the models which should be related to insufficient model resolution. These findings provide important references for future model developments.

A Multiresolution Ensemble Hybrid 4DEnVar with Variable Ensemble Sizes to Improve Global and Tropical Cyclone Track Numerical Prediction

Abstract The current global operational four-dimensional ensemble-variational (4DEnVar) data assimilation (DA) system at NCEP adopts a background ensemble at a reduced resolution, which restricts the range of spatial scales that the ensemble background error covariance can resolve. A prior study developed a multiresolution ensemble 4DEnVar method and determined that this approach can provide a comparable forecast to an approach using solely high-resolution members, while substantially reducing the computational cost. This study further develops the multiresolution ensemble 4DEnVar approach to allow for a flexible number of low- and high-resolution ensemble members as well as varying localization length scales between the high- and low-resolution ensembles. Three 4DEnVar experiments with the same computational costs are compared. The first experiment has an 80-member high-resolution background ensemble with single-scale optimally tuned localization (SR-High). The second and third experiments utilize the multiresolution background ensembles. One has 130 low-resolution and 40 high-resolution members (MR170) while the other has 180 low-resolution members and 24 high-resolution members (MR204). Both multiresolution ensemble experiments utilize differing localization radii with ensemble resolution. Despite having the same costs, both MR170 and MR204 improves global forecasts and decreases tropical cyclone track errors for up to 5 days’ lead time compared to SR-High. Improvements are most apparent in larger-scale features, such as jet streams and the environmental steering flow of tropical cyclones. Additionally, MR170 outperforms MR204 in terms of global and tropical cyclone track forecasts, demonstrating the value of both increasing sampling at large scales and retaining substantial information at small scales.

Forecasting tropical cyclone tracks in the northwestern Pacific based on a deep-learning model

Abstract. Tropical cyclones (TCs) are one of the most severe meteorological disasters, making rapid and accurate track forecasts crucial for disaster prevention and mitigation. Because TC tracks are affected by various factors (the steering flow, the thermal structure of the underlying surface, and the atmospheric circulation), their trajectories present highly complex nonlinear behavior. Deep learning has many advantages in simulating nonlinear systems. In this paper, based on deep-learning technology, we explore the movement of TCs in the northwestern Pacific from 1979 to 2021, divided into training (1979–2014), validation (2015–2018), and test sets (2019–2021), and we create 6–72 h TC track forecasts. Only historical trajectory data are used as input for evaluating the forecasts of the following three recurrent neural networks utilized: recurrent neural network (RNN), long short-term memory (LSTM), and gated recurrent unit (GRU) models. The GRU approach performed best; to further improve forecast accuracy, a model combining GRU and a convolutional neural network (CNN) called GRU_CNN is proposed to capture the characteristics that vary with time. By adding reanalysis data of the steering flow, sea surface temperatures, and geopotential height around the cyclone, we can extract sufficient information on the historical trajectory features and three-dimensional spatial features. The results show that GRU_CNN outperforms other deep-learning models without CNN layers. Furthermore, by analyzing three additional environmental factors through control experiments, it can be concluded that the historical steering flow of TCs plays a key role, especially for short-term predictions within 24 h, while sea surface temperatures and geopotential height can gradually improve the 24–72 h forecast. The average distance errors at 6 and 12 h are 17.22 and 43.90 km, respectively. Compared with the 6 and 12 h forecast results (27.57 and 59.09 km) of the Central Meteorological Observatory, the model proposed herein is suitable for short-term forecasting of TC tracks.

Evaluation of Western North Pacific Typhoon Track Forecasts in Global and Regional Models during the 2021 Typhoon Season

The track forecasts of tropical cyclones (TC) in the western North Pacific (WNP) basin during 2021 typhoon season with five global models and four regional models are evaluated here. The results show that the average direct position errors (DPEs) of the global and regional models are approximately 80, 150, 200, 300, and 400 km at 24 h, 48 h, 72 h, 96 h, and 120 h lead-times, respectively. The European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (ECMWF-IFS) achieved the best track forecast performance at each lead among the five global models. Among the four regional models, The China Meteorological Administration Tropical Regional Atmosphere Model for the South China Sea (CMA-TRAMS) attained the smallest DPEs within a 72 h lead, while The Hurricane Weather Research and Forecasting (HWRF) achieved the best track forecast performance at 96 h and 120 h leads. Most of the models produced an obvious westward systematic bias on track forecast from a 24 h to a 120 h lead. Further correlation and cluster analyses indicate that initial TC intensity and size and environmental steering flow can be regarded as good predictors for TC DPEs. TCs with a stronger initial intensity, a bigger initial size, and a larger environmental steering flow in general attain a smaller DPE, and the improvements may go up to 36% at short lead-time.

Numerical tests for tropical cyclone track prediction by the global WRF model

This work use the global WRF model containing quadruply nesting with which the highest resolution reached 10 km to simulate the typhoons landed on the coast of China in 2011. The model is driven by the reanalysis data fnl with the resolution of 1° x 1°. The study assess the feasibility and applicability of the global WRF model in the 1–7 days prediction of Tropical Cyclone (TC) track by comparing it with the regional WRF model containing the same setting (physical scheme, dynamical frame, model resolution and nesting grid domain). The global model obtain a similar forecast accuracy to the regional model in 1–7 days, with a difference less than 50 km. The forecast accuracy of the global model for 1, 2, 3, 4, 5, 6 and 7 days is about 70 km, 120 km, 180 km, 240 km, 320 km, 400 km, and 500 km, respectively. The reason of the significant TC track errors in the forecast more than 3 or 4 days is analyzed, it is due to the poor representation of typhoon and its steering flow under the situation of binary typhoon system. The study show that the global WRF model can be exploited to proceed the high resolution TC simulation and make the TC track forecast up to 7 days but not in the case of multiple typhoon.

The Effect of Assimilating AMSU-A Radiance Data from Satellites and Large-Scale Flows from GFS on Improving Tropical Cyclone Track Forecast

This study aimed to investigate the effect of assimilating either AMSU-A radiance data from satellites, large-scale flows from the Global Forecast System (GFS), or both together, on improving the track forecast of tropical cyclone (TC). The scale-selective data assimilation (SSDA) approach was employed for the assimilation of large-scale GFS flows, while the conventional 3D variational data assimilation (3DVAR) method was used for that of AMSU-A radiance data. The results show that assimilating either AMSU-A radiance data or large-scale GFS flows has a significant improvement on TC track forecast, but the improvement occurs within the first 72 h and after 72 h, respectively. When assimilating both AMSU-A radiance data and large-scale GFS flows, the forecast can take advantage of both data and thus lead to the smallest 5-day mean errors of the track forecast. These results are instructive to future operational TC track forecasting.

A Novel Tropical Cyclone Track Forecast Model Based on Attention Mechanism

Tropical cyclones are one of the most powerful and destructive weather systems on Earth. Accurately forecasting the landing time, location and moving paths of tropical cyclones are of great significance to mitigate the huge disasters it produces. However, with the continuous accumulation of meteorological monitoring data and the application of multi-source data, traditional tropical cyclone track forecasting methods face many challenges in forecasting accuracy. Recently, deep learning methods have proven capable of learning spatial and temporal features from massive datasets. In this paper, we propose a new spatiotemporal deep learning model for tropical cyclone track forecasting, which adopts spatial location and multiple meteorological factors to forecast the tracks of tropical cyclones. The model proposes a multi-layer ConvGRU to extract the nonlinear spatial features of tropical cyclones, while Spatial and Channel Attention Mechanism (CBAM) is adopted to overcome the large-scale problem of high response isobaric surface affecting the tropical cyclones. Meanwhile, this model utilizes a Deep and Cross framework to combine the traditional CNN model with the multi-ConvGRU model. Experiments were conducted on the China Meteorological Administration Tropical Cyclone Best Track Dataset (CMA) from 2000 to 2020, and the EAR-Interim dataset provided by the European Centre for Medium-Range Weather Forecasts (ECMWF). The experimental results show that the proposed model is superior to the deep learning tropical cyclone forecasting methods.

Impacts of Assimilating Additional Reconnaissance Data on Operational GFS Tropical Cyclone Forecasts

Abstract This study reviews the recent addition of dropwindsonde wind data near the tropical cyclone (TC) center as well as the first-time addition of high-density, flight-level reconnaissance observations (HDOBs) into the National Centers for Environmental Prediction (NCEP) Global Forecast System (GFS). The main finding is that the additional data have profound positive impacts on subsequent TC track forecasts. For TCs in the North Atlantic (NATL) basin, statistically significant improvements in track extend through 4–5 days during reconnaissance periods. Further assessment suggests that greater improvements might also be expected at days 6–7. This study also explores the importance of comprehensively assessing data impact. For example, model or data assimilation changes can affect the so-called “early” and “late” versions of the forecast very differently. It is also important to explore different ways to describe the error statistics. In several instances the impacts of the additional data strongly differ depending on whether one examines the mean or median errors. The results demonstrate the tremendous potential for further improving TC forecasts. The data added here were already operationally transmitted and assimilated by other systems at NCEP, and many further improvements likely await with improved use of these and other reconnaissance observations. This demonstrates the need of not only investing in data assimilation improvements, but also enhancements to observational systems in order to reach next-generation hurricane forecasting goals. Significance Statement This study demonstrates that data gathered from reconnaissance missions into tropical cyclones substantially improves tropical cyclone track forecasts.

Hurricane Forecasting: A Novel Multimodal Machine Learning Framework

Abstract This paper describes a novel machine learning (ML) framework for tropical cyclone intensity and track forecasting, combining multiple ML techniques and utilizing diverse data sources. Our multimodal framework, called Hurricast, efficiently combines spatial–temporal data with statistical data by extracting features with deep learning encoder–decoder architectures and predicting with gradient-boosted trees. We evaluate our models in the North Atlantic and eastern Pacific basins in 2016–19 for 24-h lead-time track and intensity forecasts and show they achieve comparable mean absolute error and skill to current operational forecast models while computing in seconds. Furthermore, the inclusion of Hurricast into an operational forecast consensus model could improve upon the National Hurricane Center’s official forecast, thus highlighting the complementary properties with existing approaches. In summary, our work demonstrates that utilizing machine learning techniques to combine different data sources can lead to new opportunities in tropical cyclone forecasting. Significance Statement Machine learning techniques have not been fully explored for improving tropical cyclone movement and intensity changes. This work shows how advanced machine learning techniques combined with routinely available information can be used to improve 24-h tropical cyclone forecasts efficiently. The successes demonstrated for 24-h forecasts provide a recipe for improving predictions for longer lead times, further reducing forecast uncertainties and benefiting society.

Anomaly Format of Atmospheric Governing Equations with Climate as a Reference Atmosphere

To reduce numerical instability and increase forecast accuracy of a numerical weather prediction (NWP) model, one approach is to subtract a reference atmosphere from atmospheric governing equations. In the past, scientists have proposed one-dimensional, two-dimensional, and three-dimensional static (in time) reference atmospheres with respect to temperature and pressure. These three reference atmospheres were first reviewed, and their corresponding perturbation equations were derived. Then, a new four-dimensional (space and time) all-variable (temperature, pressure, wind, moisture, etc.) reference atmosphere was defined using observed climatic states. Unlike the previous three approaches, the perturbations derived from this new method are actual anomalies relative to climate and directly a part of individual weather systems in both structure and strength. By subtracting climatic states, anomaly equations were derived and analyzed. Finally, the benefits and challenges of the anomaly-equation-based NWP model were discussed. Theoretically, an anomaly model should reduce model systematic errors (bias) and should avoid model climate drift to significantly enhance a model’s performance. An example of tropical cyclone track forecasts using the Beta advection model (vorticity) was demonstrated. The separation of model physics into climatic and anomalous physics is a significant challenge if a pure anomaly-equation-based NWP model is desired. Fortunately, a model including both anomaly and climatic equations should work with current full physics. In an anomaly climate mixed model, the anomaly part needs to be predicted and the climate parts are precalculated constants. It is hoped that this study will inspire model developers to explore the approach, which could be a possible new direction in developing next-generation NWP models. A high-resolution reanalysis is also key to the success of this new approach.

Assimilation of multi-channel radiances in mesoscale models with an ensemble technique to improve track forecasts of Tropical cyclones

Assimilation of multi-channel radiances in mesoscale models with an ensemble technique to improve track forecasts of Tropical cyclones

Using a 3D convolutional neural network and gated recurrent unit for tropical cyclone track forecasting

The tropical cyclone (TC) track forecast is an essential task in meteorological operations. An accurate forecast should be based on a comprehensive understanding and description of TCs. A TC has a complex three-dimensional structure, and the surrounding atmosphere is the driving force for its development. Traditional forecasting methods performed relatively well for the TCs with stable moving speed and direction. However, the forecast accuracy still leaves some space to improve. In recent years, machine learning methods that can extract features from a large amount of historical data have been used in meteorological services and have shown excellent performance. To better forecast 6, 12, 18, and 24 h TC tracks in the Western North Pacific, a hybrid optimization model, combining the 3D convolutional neural network (3DCNN), gated recurrent unit (GRU), and smoothing algorithm is designed, which is called smoothed 3D-GRU. The 3DCNN is used to explore the potential relationship between environmental variables and TC movements at different pressure levels. The GRU is used to convert the TC track forecasting problem into a spatio-temporal sequence problem. The smoothing algorithm is used as a post-processing method to suppress unreasonable jumps of the model output. The mean spherical distances (MSDs) of the proposed smoothed 3D-GRU model at four prediction times are 27.89, 52.37, 79.16, and 112.05 km, which are lower than the comparative machine learning-based forecasting algorithms. Compared with the numerical prediction methods, the MSDs of the smoothed 3D-GRU model are lower in most situations. In general, the smoothed 3D-GRU model can provide reliable guidance for the TC trajectory prediction.