Research interest

My research mainly focuses on the impacts and the track forecast of tropical cyclones (TCs) over East Asia.


Impacts of tropical cyclones

Hydrological impact
In addition to the East Asian Summer Monsoon, TCs are another major contributor to summer rainfall over East Asia, especially the coastal region. The variability of both TC-induced rainfall and monsoonal rainfall strongly modulates the hydrological cycle and related climate risk in the region. This study analyzes the contribution of TCs to the variability of summertime rainfall in the years of strong versus weak Changma/Baiu (the monsoon season in Korea/Japan). The possible large-scale driver of TC- and non-TC-induced rainfall variability is also examined.

Differences in TC and non-TC rainfall between strong and weak Changma/Baiu years

Air pollution
Numerous studies confirm that a TC near Luzon Strait or Taiwan can create favorable conditions for regional air pollutant episodes (light northerly for pollutant transport, subsidence unfavorable for pollutant dispersion, and increased insolation for enhanced photochemical reaction). These TC-induced episodes can cause higher pollutant concentrations than the others. Our study verifies the impact of TCs on local air quality in summer on a climate scale by investigating the influence of the change in the prevailing TC track over the western North Pacific.

How typhoon induce air pollution episode


Related article:


Extratropical transition

An example of tropical cyclone and extratropical cyclone

When TCs move to the mid-latitudes, they oftentimes undergo extratropical transition (ET) by which they change their structure. Upon transforming into extratropical cyclones (ETCs), they tend to impact larger areas and thus larger populations. In our study, we used a state-of-the-art Earth system model to examine the changes in global ET activities under increased concentrations of CO2 in the atmosphere, with a focus on the destructiveness of the transitioned storms.

Also read this blog post


Related article:


Machine learning models for tropical cyclone track forecast

Physical models have long been the tool for making weather forecasts. Owing to the advancement of computer resources and the availability of tremendous amount of data, machine learning algorithms for weather forecast has gained popularity in recent years. I developed two machine learning models for improving TC track forecast in the medium range (forecast lead time >5 days).

The first model (track-pattern-based model) applies the fuzzy c-means clustering method on historical TC tracks similar to an operational 5-day forecast track; makes predictions based on each track pattern; and combines the forecast for each track pattern to form the final forecast. The second model uses different classes of artificial neural networks in concert to correct the TC track predicted by a global weather forecast model (GEFS). The neural network model is improved by including a shortcut connection, feature selection, and hyperparameter tuning. The performance of the deep learning model is evaluated using accuracy, association, and skill.

Schematic of the track-pattern-based model and deep learning model


Related article: