Search for: All records

Abstract Applications of machine learning (ML) in atmospheric science have been rapidly growing. To facilitate the development of ML models for tropical cyclone (TC) research, this binary dataset contains a specific customization of the National Center for Environmental Prediction (NCEP)/final analysis (FNL) data, in which key environmental conditions relevant to TC formation are extracted for a range of lead times (0–72 hours) during 1999–2023. The dataset is designed as multi-channel images centered on TC formation locations, with a positive and negative directory structure that can be readily read from any ML applications or common data interface. With its standard structure, this dataset provides users with a unique opportunity to conduct ML application research on TC formation as well as related predictability at different forecast lead times. 

Abstract In this study, the potential changes in tropical cyclone (TC) lifetime in the western North Pacific basin are examined for different future climates. Using homogeneous 9-km-resolution dynamical downscaling with the Weather Research and Forecasting (WRF) Model, we show that TC-averaged lifetime displays insignificant change under both low and high greenhouse gas concentration scenarios. However, more noticeable changes in the tails of TC lifetime statistics are captured in our downscaling simulations, with more frequent long-lived TCs (lifetime of 8–11 days) and less short-lived TCs (lifetime of 3–5 days). Unlike present-day simulations, it is found that the correlation between TC lifetime and the Niño index is relatively weak and insignificant in all future downscaling simulations, thus offering little explanation for these changes in TC lifetime statistics based on El Niño–Southern Oscillation. More detailed analyses of TC track distribution in the western North Pacific basin reveal, nevertheless, a noticeable shift of TC track patterns toward the end of the twenty-first century. Such a change in TC track climatology results in an overall longer duration of TCs over the open ocean, which is consistent across future scenarios and periods examined in this study. This shift in the TC track pattern is ultimately linked to changes in the western North Pacific subtropical high, which retreats to the south during July and to the east during August–September. The results obtained in this study provide new insights into how large-scale circulations can affect TC lifetime in the western North Pacific basin in warmer climates. Significance StatementUsing high-resolution dynamical downscaling with the Weather Research and Forecasting (WRF) Model under low- and high-emission scenarios, this study shows that the basin-averaged tropical cyclone (TC) lifetime in the western North Pacific (WNP) basin has no noticeable change under both warmer climate scenarios, despite an overall increase in TC maximum intensity. However, the tails of the TC lifetime distribution display significant changes, with more long-lived (6–20 days) TCs but less short-lived (3–5 days) TCs in the future. These changes in TC lifetime statistics are caused by the shift of the North Pacific subtropical high, which alters large-scale steering flows and TC track patterns. These results help explain why previous studies on TC lifetime projections have been inconclusive in the WNP basin and provide new insights into how large-scale circulations can modulate TC lifetime in a warmer climate. 

Abstract Exploring new techniques to improve the prediction of tropical cyclone (TC) formation is essential for operational practice. Using convolutional neural networks, this study shows that deep learning can provide a promising capability for predicting TC formation from a given set of large-scale environments at certain forecast lead times. Specifically, two common deep-learning architectures including the residual net (ResNet) and UNet are used to examine TC formation in the Pacific Ocean. With a set of large-scale environments extracted from the NCEP–NCAR reanalysis during 2008–21 as input and the TC labels obtained from the best track data, we show that both ResNet and UNet reach their maximum forecast skill at the 12–18-h forecast lead time. Moreover, both architectures perform best when using a large domain covering most of the Pacific Ocean for input data, as compared to a smaller subdomain in the western Pacific. Given its ability to provide additional information about TC formation location, UNet performs generally worse than ResNet across the accuracy metrics. The deep learning approach in this study presents an alternative way to predict TC formation beyond the traditional vortex-tracking methods in the current numerical weather prediction. Significance StatementThis study presents a new approach for predicting tropical cyclone (TC) formation based on deep learning (DL). Using two common DL architectures in visualization research and a set of large-scale environments in the Pacific Ocean extracted from the reanalysis data, we show that DL has an optimal capability of predicting TC formation at the 12–18-h lead time. Examining the DL performance for different domain sizes shows that the use of a large domain size for input data can help capture some far-field information needed for predicting TCG. The DL approach in this study demonstrates an alternative way to predict or detect TC formation beyond the traditional vortex-tracking methods used in the current numerical weather prediction. 

Proper representations of stochastic processes in tropical cyclone (TC) models are critical for capturing TC intensity variability in real-time applications. In this study, three different stochastic parameterization methods, namely, random initial conditions, random parameters, and random forcing, are used to examine TC intensity variation and uncertainties. It is shown that random forcing produces the largest variability of TC intensity at the maximum intensity equilibrium and the fastest intensity error growth during TC rapid intensification using a fidelity-reduced dynamical model and a cloud-resolving model (CM1). In contrast, the random initial condition tends to be more effective during the early stage of TC development but becomes less significant at the mature stage. For the random parameter method, it is found that this approach depends sensitively on how the model parameters are randomized. Specifically, randomizing model parameters at the initial time appears to produce much larger effects on TC intensity variability and error growth compared to randomizing model parameters every model time step, regardless of how large the random noise amplitude is. These results highlight the importance of choosing a random representation scheme to capture proper TC intensity variability in practical applications. 

Abstract This study examines the variability of tropical cyclone (TC) intensity associated with stochastic forcings at the maximum potential intensity (PI) equilibrium. By representing TC intensity as an Itô diffusion process in the framework of TC-scale dynamics, we show from both theoretical and numerical analyses that there exists an invariant intensity distribution whose variance is proportional to the variances of stochastic forcings. This result provides further evidence that TC dynamics possess an intrinsic variability that prevents the TC absolute intensity errors in numerical models from being reduced below an arbitrarily small threshold. Analysis of the invariant intensity distribution at the PI limit reveals also that the stochastic forcing component associated with tangential wind and the warm-core anomaly in the TC central region have the largest contribution to TC intensity variability. These results suggest that future development of stochastic representation in TC models should focus on the tangential wind and thermodynamic structure to capture proper TC intensity random fluctuations.