REPRESENTATION OF PRECIPITATION OVER MARITIME CONTINENT IN THE CMIP5 AND CMIP6 MODELS

• Main Author: Marsitha B, Femmy
• Format: Theses
• Language: Indonesia
• Subjects: Geologi, hidrologi & meteorologi
• Online Access: https://digilib.itb.ac.id/gdl/view/68918
• Institution: Institut Teknologi Bandung

In this study, an evaluation of precipitation simulation over Maritime Continent (MC) from Coupled Model Intercomparison Project (CMIPs) i.e., CMIP5 and CMIP6, has been conducted with emphasis on the use of temporal-based metrics, in addition to aggregation-based metrics that have been used in many previous studies, to take into account the non-stationary properties of model errors and uncertainties. The metrics are computed from 12 models of both CMIP5 and CMIP6 with ERA-5 daily reanalysis data as a reference for 1980-2005. Agreements between models and observations are evaluated in terms of temporal and spatial representations. Temporally, the distribution of precipitation errors over MC is evaluated based on the error consistency of the seasonal precipitation distribution. Spatially, the contributions of annual and semi-annual cycles are examined by using harmonic analysis, as well as the climate variabilities including Madden-Julian Oscillation (MJO), Indian Ocean Dipole (IOD) and El Nino Southern Oscillation (ENSO) by using Multivariate Linear Regression (MLR). Overall model representation is assessed by the summation rank method, which sums the ranking of each metric in the first and second metric groups. Results from temporal evaluation show that improvement of precipitation representation over the MC only occurs in 6 out of 12 CMIP6 models. Moreover, spatial representation of key phenomena is better in CMIP5 models if evaluated by individual metrics. However, based on overall spatial metric, the three top-ranked models are GFDL-CM4 (CMIP6), CNRM-CM5 (CMIP5), dan NorESM2-LM (CMIP6), indicating that improved CMIP6 models have more consistent metric performance. These results confirm that CMIP6 models do not necessarily perform better in representing observed rainfall, compared to their CMIP5 counterparts. Furthermore, the non-aggregate statistical metrics used in this study reveal non-stationary characteristics of model errors, which are largely influenced by interannual variability associated with ENSO. Although it has not been thoroughly studied, improved representation in CMIP6 seems to occur when there is an increase in horizontal and/or vertical resolution with moderate modifications in model parameterization. This implies that models from both CMIP5 and CMIP6 can be used for practical analysis of climate projections, while future model development needs to consider wider aspects of model performance such as precipitation representation in various regions.