A Catboost-based Model for Intensity Detection of Tropical Cyclones over the Western North Pacific Based on Satellite Cloud Images

A Catboost-based intelligent tropical cyclone (TC) intensity-detecting model is built to quantify the intensity of TCs over the Western North Pacific (WNP) with the cloud-top brightness temperature (CTBT) data of Fengyun-2F (FY-2F) and Fengyun-2G (FY-2G) and the best-track data of the China Meteorological Administration (CMA-BST) in recent years (2015-2018). Catboost-based model is featured with the greedy strategy of combination, the ordering principle in optimizing the possible gradient bias and prediction shift problems, and the oblivious tree in fast scoring. Compared with the previous studies based on the pure convolutional neural network (CNN) models, the Catboost-based model exhibits better skills in detecting TC intensity with the root mean square error (RMSE) of 3.74 m s^-1. Besides of the three mentioned model features, there are also two reasons on model design. On one hand, the Catboost-based model uses the method of introducing prior physical factors (e.g., the structure and shape of the cloud, deep convections and background fields) into its training process, on the other hand, the Catboost-based model expands the dataset size from 2342 to 13471 samples by hourly interpolation of the original dataset. Furthermore, this paper investigates the errors of the model in detecting different categories of TC intensity. The results show that the deep learning-based TC intensity-detecting model proposed in this paper has systematic biases, namely, overestimation (underestimation) of intensities in TC which are weaker (stronger) than typhoon level, and the errors of the model in detecting weaker (stronger) TCs are smaller (larger). This implies that more factors than the CTBT should be included to further reduce the errors in detecting strong TCs.