Evaluation of GRAPES-TCM rainfall forec ast for China landfalling tropical cyclone in 2006.

Up to the present, little verification has been performed for rainfall predictio ns from numerical forecasts of landfalling tropical cyclones. Using the output f rom the operational run of GRAPESTCM in 2006, this paper evaluated capability of the predicted rainfall for landfalling TC of this system. As comparison, the predicted rainfall from TRaP(Tropical Rainfall Potential) and a climatology met hod based on analogical tropical cyclone（Typhoon Analog Prediction Technique, TAPT） were analyzed simultaneously. Several measures of forecast quality were u sed to evaluate the predicted rainfall from these runs, using daily rain gauge d ata as ground truth. The overall quality was measured by the scattering plot bet ween observed and predicted rainfall, mean absolute errors and root mean square- errors. Additionally, more traditional precipitation verification scores were calculated including Bias Score, Probability of Detection (POD), False Alarm Ratio (FAR), and Equitable Threaten Score (ETS). The MAE (mean absolute error) of pr edicted rainfall from GRAPES-TCM is 31 mm/d which is much larger than that f rom TRaP(20 mm/d) and TAPT(17 mm/d). At the same time, GRAPES-TCM exhibited a higher pattern correlation with observations than TRaP and TAPT. The POD of predicted rainfall from GRAPES-TCM is also higher than that of TRaP and climatol ogy method while the FAR of these 3 methods are in the same grade. In addition, GRAPES-TCM got much higher ETS for every rainfall threshold than the other two methods. For the stronger sever storm (100 mm/d), ETS of the predicted rainfal l from TRaP and climatology is 0. The correlation between 6-h forecasted precip itation and the observation also shows that GRAPESTCM has a better performance than the other two methods in each period (00-06, 06-12, 12-18, 18-24 hour). Overall, GRAPES-TCM could forecast the precipitation pattern well while TC makes landfall and it has better performance on server precipitation forecast.