PREDICTABILITY OF THE SUB-SEASONAL TO SEASONAL (S2S) SCALE RAINFALL IN INDONESIA

• Main Author: Hanif Damayanti, Rosi
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/85283
• Institution: Institut Teknologi Bandung

Sub-Seasonal to Seasonal (S2S) prediction models have been developed but not operationally run by National Hydrometeorological Services (NHMS) around the world due to the still limited forcast skill. However, there has not been comprehensive evaluation of S2S rainfall prediction in Indonesia. In this study, the skill of the ECMWF (European Centre for Medium-Range Weather Forecasts) S2S prediction model in Indonesia is evaluated using three commonly used metrics i.e., Root Mean Squared Error (RMSE), bias, and correlation, with National Centers for Environmental Prediction- National Center for Atmospheric Research (NCEPNCAR) Reanalysis I data to represent observations. The results show that the skill evolution of rainfall, as well as 850 hPa wind, prediction over Indonesia drops sharply in the second week. Although the results are consistent with those of global and entire tropical domains, it is also found that the prediction skill shows strong seasonal and variable dependance with higher predictability of 850 hPa zonal during March-April-May (MAM) and June-July-August (JJA). On the other hand, the skill 850 hPa meridional wind prediction shows higher skill during MAM and September-October-November (SON), whereas the skill of rainfall prediction is better than the others during December-January-February (DJF). Further examination of the 850 hPa winds and rainfall as potential source of predictability is conducted by deseasonalizing and apllying Empirical Orthogonal Function (EOF) analysis to the time series data. The results indicate that there is no single dominant pattern of S2S rainfall variability. Nevertheless, some largest EOFs are examined in more detail. Lag-correlation analyses between the principal components (PC) of rainfall and 850 hPa winds show that eastward-propagating Madden-Julian Oscillation (MJO), and west-ward propagating equatorial Rossby waves might serve as the sources of S2S predictability. Moreover, the largest percentage of rainfall EOF is associated with eastward Madden-Julian Oscillation (MJO) propagation. A case study of rainfall event during May 27 to July 14, 2015 shows that 850 hPa zonal wind is the most relevant predictor for the rainfall pattern, but the predictability of the zonal wind is found to be weak during active MJO period. Another case of heavy rainfall event of January 29, 2020 has also been studied, which exhibits the influence of westward equatorial Rossby (ER) wave propagation. In this case, rainfall seems to be the best predictor for high wave-number ER wave. However, the predictability of the ER wave is also weak around the time of convective event. Results of this study indicate that S2S rainfall prediction in Indonesia is still a challenging problem. As it is found in previous studies, this study confirms that the largest S2S rainfall variability is influenced by MJO phenomenon. In addition, more sporadic convective events can be attributed to high wave-number ER wave propagation. In either case, the skill of S2S prediction model is still limited in representing both MJO and ER wave propagation during active convective phases. Therefore, application of S2S prediction in Indonesia may require more advanced post-processing of model output to improve its accuracy and skill. This study has focused on deterministic skill and has not addressed the probabilistic skill of the S2S forecast.