Intercomparison and Evaluation of Reanalysis Datasets in Representing Tropical Cyclones over the South Indian Ocean

Tropical cyclones (TCs) in the South Indian Ocean are among the most destructive hydrometeorological hazards affecting surrounding regions, including Indonesia. Reliable representation of their tracks, intensity, and structure is therefore essential for improving regional risk assessment and advancing studies of climate change. Atmospheric reanalysis datasets are widely used to investigate TC climatology and environmental conditions because they provide dynamically consistent atmospheric fields with long temporal coverage. However, previous studies have shown that reanalysis products often differ substantially in their ability to reproduce TC characteristics such as storm track, intensity, and structural evolution, with the best-performing datasets varying across ocean basins. Despite these advances, a systematic multi-dataset evaluation of tropical cyclone representation in the South Indian Ocean using a consistent and objective tracking framework remains limited. This basin is characterized by relatively sparse observational coverage, which may further influence the performance of different reanalysis products. As a result, the reliability of commonly used reanalysis datasets for representing TC characteristics in this region is still not well understood.

This study evaluates the capability of several widely used reanalysis datasets—ERA5, JRA-3Q, MERRA-2, and NCEP—to represent tropical cyclones over the South Indian Ocean. Tropical cyclone tracks are objectively detected using the CNRM tropical cyclone tracking scheme, which identifies candidate vortices based on relative vorticity, sea-level pressure minima, warm-core temperature anomalies, and wind structure before linking them into coherent trajectories. The detected tracks are paired with the International Best Track Archive for Climate Stewardship (IBTrACS) dataset to assess their consistency with observed cyclone evolution. The analysis focuses on the domain 30°E–120°E and 0°–40°S and spans multiple decades of TC activity.  Dataset performance is evaluated using statistical skill metrics, including bias, root-mean-square error (RMSE), Probability of Detection (POD), False Alarm Rate (FAR), and the Critical Success Index (CSI), focusing on storm position, maximum wind speed, minimum sea-level pressure, life cycle, and peak intensity timing.

Preliminary results reveal notable differences among the datasets. JRA-3Q shows the closest agreement with IBTrACS in representing storm position and minimum sea-level pressure, outperforming other reanalysis data sets. Nevertheless, all reanalysis datasets substantially underestimate maximum wind speed, indicating persistent limitations in representing tropical cyclone intensity. These findings highlight the strengths and limitations of current reanalysis products in the South Indian Ocean and provide guidance for selecting appropriate datasets for cyclone climatology and process-based studies.