Seasonal predictability of tropical cyclone frequency around Okinawa and Taiwan
―Key roles of the Indian Ocean Dipole―

1. Key Points

• Tropical cyclones are the most costly and deadly natural disasters in Japan, South Korea, Taiwan, the Philippines, and other coastal regions of Southeast Asia. Successful prediction of TC frequency in a season, at least a few months in advance, could help reduce socioeconomic losses through necessary mitigation measures and benefit a range of industries, including insurance, agriculture, and tourism.

• The summer frequency of tropical cyclones around Okinawa and Taiwan was found to be predictable from early May.

• The Indian Ocean Dipole contributed to the predictability.

2. Abstract

We have found that the JAMSTEC/APL (Application Laboratory) seasonal prediction system (SINTEX-F) could predict whether tropical cyclones (including typhoons) will increase in the summer near Okinawa and Taiwan from the beginning of May.

Tropical cyclones are one of the most devastating natural disasters in Japan and Taiwan. If we can predict several months in advance whether tropical cyclones (including typhoons) will become more active in the summer, we can use this information to manage risk in various industries, including property and casualty insurance. It has been known that the Northwest Pacific overall activity of tropical cyclones is seasonally predictable to some extent. However, it has been difficult to provide useful seasonal predictions of which area of the Northwest Pacific tropical cyclone seasonal activity increase/decrease.

In this study, we conducted a large-scale ensemble re-forecast experiments in 1982-2022 by a large-ensemble seasonal prediction system (SINTEX-F), and found that it is possible to predict to some extent whether tropical cyclones will increase or decrease during the summer (June-August) around Okinawa and Taiwan from early May. In the cases of successful predictions, the contribution of the Indian Ocean Dipole Mode phenomenon was shown to be important.

In most conventional seasonal forecasts, the three-month average temperature and precipitation have been used as the forecast target, but in this study, it was shown that the seasonal characteristics of extreme phenomena, such as tropical cyclones, that cause extensive damage can be predicted to some extent. This is very helpful for society because it gives people time to get ready and protect against the damage caused by extreme events.

It has been pointed out that the intensity of typhoons in the low latitudes of the northwest Pacific may increase with the progress of anthropogenic global warming, and it is feared that the damage may become more severe due to overlapping phases with natural climate variations such as El Niño/La Niña events and Indian Ocean Dipole events, which occur every few years. It is an urgent task to improve the accuracy of tropical cyclone seasonal forecasts, as shown in this study, and to develop social application research to mitigate damage by using the forecast information.

The paper is published in npj climate and atmospheric science on April 3, 2025.

3. Background

JAMSTEC/APL seasonal prediction system (SINTEX-F, overview) is based on an ocean-atmosphere-sea ice-land coupled general circulation model developed on the JAMSTEC super-computer “Earth Simulator” (https://www.jamstec.go.jp/es/en/) under the international research collaboration with European researchers. The model codes describe physical processes of the climate systems including ocean-atmosphere-sea ice-land. The initialization scheme uses the satellite surface ocean observations and subsurface ocean observations from the Argo floats, mooring buoys, etc. We generate 108 ensemble members to reduce the prediction uncertainties associated with different initial conditions and perturbed physical schemes. The 108-member system has an advantage in finding a predictable signal against unpredictable atmospheric noise on a seasonal timescale and possible co-variability patterns influencing targets in the ensemble phase space.

Using the SINTEX-F, APL have been providing some pioneering results for seasonal prediction; super El Niño/La Niña events, super Indian Ocean Dipole mode events, an East Asian record-breaking warm winter, East African droughts, and Pakistan floodings, etc. However, these seasonal predictions were mostly limited to monthly average climate indices and three-month averaged temperatures and precipitation fields, and did not provide any information on seasonal characteristics of extreme phenomena that cause extensive damage, such as tropical cyclones (e.g., where typhoons are likely to be located in summer compared to normal years). Obviously, it is important to use weather forecasts, i.e., forecasts of the path of individual typhoons made several days in advance, for disaster prevention. In addition, if we can predict several months in advance where typhoons will be more active during the season than in a normal year, this information can be applied to risk management in various stakeholders, such as damage insurance, agriculture, and tourism. In recent years, it has been pointed out that the intensity of typhoons in the low latitudes of the northwestern Pacific may increase with the anthropogenic global warming. Seasonal prediction study of tropical cyclone activity is becoming increasingly important because it is feared that the damage may be exacerbated by the one-two punch of natural year-to-year climate variability and long-term anthropogenic global warming.

4. Results

By analyzing the re-forecast experiments conducted from early May from 1982 to 2022, it was found that it is possible to predict the summer frequency of tropical cyclones (the number of tropical cyclones during the summer) near Okinawa and Taiwan to some extent (Figure 1).

Figure 1 Correlation coefficient skill for predicting the June-August (JJA) tropical cyclone (TC) frequency from early May of 1982-2022.

In particular, the high TC activity in the summer of 2018 was well predicted. It is found that a positive Indian Ocean Dipole (IOD) played a key role in the predictions by analyzing the inter-member co-variability (Figure 2). The prediction of the TC frequency anomalies in the target region was significantly linked with the prediction of the positive IOD, particularly negative sea surface temperature in the eastern pole: ensemble members that predicted stronger positive IOD tend to predict higher TC frequency in the target region. Although the differences among ensemble members due to large atmospheric internal variability have been considered to be unpredictable noise, the inter-ensemble correlation analyses could support that the teleconnection from the positive IOD commonly appears in ensemble members to some extent. It also suggests that by reducing the uncertainty of the IOD prediction, we could reduce the uncertainty of the TC frequency prediction around Okinawa and Taiwan and increase the signal.

Figure 2 Inter-ensemble member correlation in the 108-member prediction between the TC frequency anomalies in the target area (120°E-130°E and 20°N-30°N) and the sea surface temperature anomalies for JJA of 2018 issued in early May 2018.

Although it is difficult to predict IOD occurrence this year at this stage, it is necessary to continue to pay attention to the possibility. The latest forecasts can be found on the SINTEX-F website.

5. Future Perspective

Beyond the stage of predicting three-month average temperature and precipitation fields, it is an important achievement of this research to show that seasonal predictability of the seasonal characteristics of extreme phenomena that cause extensive damage, such as tropical cyclones, is possible to some extent in specific ocean areas. However, the accuracy of the prediction may not be sufficient to meet the needs of various users. In the future, the prediction accuracy needs to be drastically improved. For example, the SINTEX-F used in this study is a large-ensemble system, but its horizontal resolution is limited to about 110 km, which is insufficient to precisely represent the structure of typhoons. A higher resolution version of the system is essential for improved forecasting. A different approach is to identify in advance the events that are most likely to be forecasted. If the past 41 summers from 1982 to 2022 are treated as equivalent, a simple calculation will yield only relatively low forecast accuracy. However, under certain conditions (e.g., when IOD events occur alone), the forecast appears to be relatively successful. By understanding the process/cause of the anomalous conditions in each case, we should further understand possible conditions that make forecasts more useful than in normal years. As a result, we may be able to support user decision making by providing confidence information with forecast, such as: this summer's forecast is more reliable than normal. Alternatively, it is also important to advance research and development to focus on tropical regions, where seasonal forecasts are relatively easy. In addition, weather forecasting using artificial intelligence (AI) has attracted attention in recent years. The hybrid approach of dynamical models and AI-based models is also attracting attention.

APL will continue to develop a set of tools to generate seasonal forecast information and support stakeholder decision making related to the occurrence of extreme events (called the Extreme Climate Digital Twin).