Seasonal Probabilistic Forecast of Tropical Cyclone Activity in the North Indian Ocean (PAP014435)

Código

PAP014435

Autores

J. Diez, Raúl Medina, S. Al-Yayai

Tema

Flood forecasting and early warning systems

Resumo

Coastal flooding risk associated to tropical cyclones (TCs) is nowadays a major concern inlow-lying and populated areas of the North Indian Ocean (NIO). In the past few years there have beendevastating examples (i.e., Gonu or Phet in the Arabian Sea, and Nargis and Layla in the Bengal Bay)that encourages better understanding and predicting the activity of such events. Seasonal forecasting TCactivity would lead to a more efficient coping these phenomena and mitigate its destructive effects.It is well known that the TC genesis is boosted by warm sea surface temperature (SST), low to moderatevertical wind shear and upper level divergence. In this study we focus in the role of the SST as a neededprecondition for TCs intensification with a twofold objective: (a) to define a statistical downscaling modelto establish a relationship between large scale SST configurations and the occurrence probability of TCs;(b) to probabilistically forecast the occurrence rate of TCs in the NIO based on operational SST forecasts.To build the statistical downscaling model, Y=f(X), we apply a combination of the principal componentanalysis and the k-means classification algorithm in order to identify representative patterns of large scaleSST (predictor X) and the associated TC activity in terms of occurrence rate and expected tracks(predictand Y). Once explored the existing relations, we make use of the different runs of the NCEP?soperational forecast, Climate Forecast System v2 (CFSV2). By using different runs we are able toconstruct an ensemble forecast (probabilistic) of the TC activity for the next month/season. The proposedmodel reveals that this approach provides complementary and useful information for a very early TCwarning system.