Geofizika, Vol. 31 No. 2, 2014.
Izvorni znanstveni članak
Verification of WRF rainfall forecasts over India during monsoon 2010: CRA method
Ananda Kumar Das
; India Meteorological Department, Mausam Bhawan, New Delhi, India
Mansi Bhowmick
; The University of Leeds, School of earth and Environment, Leeds, UK
P. K. Kundu
; Jadavpur University, Department of Mathematics, Kolkata, India
S. K. Roy Bhowmik
; India Meteorological Department, Mausam Bhawan, New Delhi, India
Sažetak
The WRF model forecast during monsoon season 2010 has been verified with daily observed gridded rainfall analysis with 0.5° spatial resolution. First- ly, the conventional neighborhood technique has been deployed to calculate common scores like mean error and root mean square error. Along with, widely used two categorical skill scores have been computed for seven different rainfall thresholds. The scores only found the general nature of the model performance and depicted the degradation of forecast accuracy exceeding moderate rainfall category of 7.5 mm. The object oriented Contiguous Rain Area method also has been considered for the verification of rainfall forecasts to gather more informa- tion about model performance. The method similarly has endorsed that the performance of the model degrades along with the increase in rainfall amount. But at the same time, the decomposition of mean square error has pointed out that the maximum error occurred due the shifting of rain object or event in the forecast compared to observation. The volume error contributes less as compared to pattern error in 24 hour forecasts irrespective of rainfall thresholds. But in 48 hour forecasts, their values are comparable and change along with rainfall threshold. During whole monsoon season, all contiguous rain areas in model forecasts have been searched over observed rainfall analyses applying best-fit criteria. For contiguous rain areas below 50 mm more than 70 percent match was found.
Ključne riječi
Hrčak ID:
134938
URI
Datum izdavanja:
31.12.2014.
Posjeta: 1.515 *