Evaluation of precipitation forecasts of WRF model for daily heavy rain in Qazvin Province during 2002-2011
Subject Areas : Marine Physics
F.
Arkian
1
(Meteorology Dept., Marine Science and Technology Faculty, Islamic Azad University Tehran North Branch)
N.
Mashatan
2
(Meteorology Dept., Marine Science and Technology Faculty, Islamic Azad University Tehran North Branch)
P. S.
Katiraie Borojerdi
3
(Meteorology Dept., Marine Science and Technology Faculty, Islamic Azad University Tehran North Branch)
E.
Mirzaei Haji Baghlo
4
(Iranian Meteorological Organization, Tehran)
Keywords: Qazvin Province, WRF, Verification of precipitation forecast,
Abstract :
In this study, the precipitation forecasts by WRF model for daily pervasive heavy rains in Qazvin province was evaluated. For this purpose, 30 cases of the heavy and pervasive rains in Qazvin province during (2002-2011) with two different configurations (KFMYJ, GDMYJ) schemes by the WRF model at intervals of 24, 48 and 72 hours, have been simulated. Because of various heights and different climates of Qazvin and considering average rainfall, Qazvin was divided into five precipitation regions containing plain, plain margins, submontane, and mountains of Northeast and Southwest. Then with the two methods, point and areal, simulated rainfall and the corresponding observed values were evaluated. According to evaluated results, for GDMYJ configuration, root mean square error and bias multiple, with respective values of about 8.7 and 1.7 for 24-hour rainfall simulation are better than 48 and 72 hours. Also correlation coefficient between observations and model simulations of precipitation in the submontane region in comparison with other regions has slightly higher accuracy of approximately 0.5. In general, the fraction of systematic error to total error is very low and it is more due to random errors. Considering the rainfall threshold ≥10 mm and the 2×2 contingency tables for the occurrence or absence of precipitation, skill scores were calculated for the model in Qazvin region. The results showed that the model skill in predicting precipitation in 24, 48 and 72 hours on the threshold has acceptable accuracy and on average about 71% of cases were correctly predicted.
