Modification of a linear regression-based multi-model super-ensemble technique and its application in forecasting of wave height during extreme weather conditions

Figures & data

Figure 1. Tracks of cyclones Phailin (red), Hudhud (magenta) and Nilofar (green). The positions at 00UTC are marked along the track. The locations of buoys at Gopalpur, Visakhapatnam (VSK), Pondicherry and AD06 are marked in blue.

Table 1. Description of different wave models.

Model	Domain of study	Wind	Resolution
SW1	Indian Ocean, 30°–120° E, −60°–30° N	ECMWF	30°–120°E, 60°S–0° (resolution 110 km), AS and BOB offshore (resolution 45 km), coastal area (resolution 8–15 km)
SW2	Indian Ocean, 30°–120° E, −60°–30° N	ECMWF	Same as SW1
SW3	Indian Ocean, 30°-120° E, −60°–30° N	Corrected ECMWF wind (spatially using wind observations and wind forecast from India Meteorological Department (IMD))	Same as SW1
WWIII	Global, 0°–359° E, −80°–70° N	NCMRWF	Indian Ocean (24°–125° E, −60°–32° N, 50 km), AS (42°–79° E, −5°–31° N, 25 km), BOB (75°–100° E, −5°–23° N, 25 km), coastal grids (East coast(76°–95° E, 5°–23° N, 5 km) and West coast (65°–79° E, 5°–24° N, 5 km)
SWAN	Pondicherry,79.5°–80.5°E, 11.2°–12.2°N AS 60°–79°E, 5°–26°N BOB 76°–95°E, 5°–25°N	ECMWF	250 m
			12 km
			12 km

Table 2. RMSE (in °C) between observation and ensemble members and model predictions of SST for March, April and May with January initial condition.

Period of Prediction	Different models and ensemble techniques	RMSE
Prediction for March (1982–2009)	MEM	0.4
	KEM	0.44
	M01	0.65
	M05	0.73
	M09	0.9
	M14	0.62
	M17	0.51
	M21	0.94
Prediction for April (1982–2009)	MEM	0.28
	KEM	0.4
	M01	0.36
	M05	0.5
	M09	0.63
	M14	0.42
	M17	0.43
	M21	0.53
Prediction for May (1982–2009)	MEM	0.32
	KEM	0.4
	M01	0.48
	M05	0.83
	M09	0.51
	M14	0.54
	M17	0.64
	M21	0.9

Figure 2. RMSE (in second) of ensemble and individual model predictions of wave period during SWM of 2014 at Pondicherry.

Table 3. RMSE (in metre) and Scatter Index of ensemble and individual model predictions of significant wave height during the southwest monsoon of 2014 and during Nilofar, Hudhud and Phailin cyclones.

Period of prediction	Different models and ensemble techniques	RMSE	Scatter Index
Nilofar (AD06)	MEM	0.26	0.11
	KEM	0.3	0.12
	SWAN	0.31	0.13
	WWIII	0.75	0.3
	SW1	0.63	0.26
	SW2	0.37	0.14
	SW3	0.43	0.17
Hudhud (Vizag)	MEM	0.57	0.17
	KEM	0.91	0.28
	SWAN	1.73	0.54
	WWIII	2.2	0.68
	SW1	0.95	0.3
	SW2	0.78	0.24
	SW3	0.62	0.2
Phailin (Gopalpur)	MEM	0.6	0.21
	KEM	0.72	0.25
	SWAN	0.75	0.26
	WWIII	1.23	0.43
	SW1	0.85	0.3
	SW2	0.9	0.32
SW monsoon (2014) (Pondicherry)	MEM	0.12	0.17
	KEM	0.15	0.22
	SWAN	0.19	0.28
	WWIII	0.26	0.38
	SW1	0.2	0.3

Figure 3. Significant wave height (in metre) is plotted for different model outputs, observation and ensemble predictions at the time of Nilofar.

Figure 4. Significant wave height (in metre) is plotted for different model outputs, observation and ensemble predictions at the time of Hudhud.

Figure 5. Significant wave height (in metre) is plotted for different model outputs, observation and ensemble predictions at the time of Phailin.