Tropical Cyclone Intensity Probabilistic Forecasting System Based on Deep Learning

Table 1

A detailed list of domain expert features constructed from the best path data is shown.

No.	CLIPER features	Brief explanation
1–10	, ,	Persistence factors
11–20	, , ,
21–24	, , ,
25–28	, , ,	Climatology factors
29–32	, , ,
33–36	, , ,
37–41	, , , ,	Domain knowledge inspired by SHIPS
42–45	, , ,
46–50	, , , ,
51–55	, , , ,
56–65	, , , , , , , ,
66–70	, , , ,
71–75	, , , ,	Wind speed and pressure features
76–80	, , , ,
81–85	, , , ,
86–90	, , , ,
91–95	, , , ,
96	Month

The features are subscripted with numbers, for example, indicates the barometric pressure value 6 hours ago.