Storm Surge Peak and Timing Forecasts for the Pearl River Estuary, 2000-2023

Ruolan Xiong developed a machine learning model for storm surge forecasting using data from 42 tropical cyclones between 2000 and 2023. The dataset likely contains model outputs and inputs, including tropical cyclone parameters and observed surge data from nine stations in the Pearl River Estuary. It was last updated on 2026-05-19 and is shared under a CC-BY-4.0 license.

Use Cases

• Benchmarking new storm surge forecasting models based on the described GRU and SVR hybrid approach.
• Analyzing the relationship between tropical cyclone parameters and surge timing errors based on the described feature selection.
• Training models for multi-station joint forecasting based on the embedded spatial information between stations and TC tracks.
• Evaluating iterative forecasting schemes for reducing peak underestimation in sequential prediction tasks.

Strengths

• Model results are validated using leave-one-out cross-validation on 42 tropical cyclones.
• The study reports specific performance improvements, including a 27% reduction in Peak Error at 12–18h lead times.
• Data covers a 23-year time span (2000-2023) from nine monitoring stations.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small (46.4 KB), indicating limited raw observational data is included.

Provenance

Source

Ruolan Xiong

Collection Method

Data likely originates from storm surge observations and tropical cyclone best-track data, processed through a custom GRU and SVR model.

Time Range

2000-2023

Freshness

Last updated 2026-05-19 04:20:50; freshness should be verified.

Geography

Pearl River Estuary, China