Enhanced climate projections over Sindh, Pakistan: a bayesian model averaging ensemble methodology

Climate change poses formidable challenges, particularly for the regions which are among the most vulnerable to these alterations. Sindh Province of Pakistan is the region of focus of this study situated in the south-east of Pakistan. The study aims to focus on model selection (General Circulation Model = GCM) and developing ensemble climate projections for the three key climate variables, rainfall, maximum and minimum temperature. For model selection, we employed Bayesian Model Averaging (BMA), utilizing observed data as the response variable and GCM outputs as predictors, for the duration of 1985–2014. For model ranking, posterior inclusion probability (PIP) is used and models with the higher PIP are preferred. For ensemble projections, BMA model is trained over the baseline period (1985–2014) and evaluated its performance both graphically and numerically. Subsequently, the trained BMA model is employed to project climate data for the target location spanning the years 2015 to 2044, under two distinct socioeconomic pathways: SSP2-4.5 and SSP5-8.5. For projections, BMA utilized the corresponding PIP as weight for each GCM, a weighting scheme established during the training period. Regarding model selection, top 5 models which performed better are selected for each variable. To get insight about climate change, the daily projected data was compared with baseline data on monthly and seasonal basis. The results indicate an upward trend in minimum and maximum temperatures during 2015–2044. Maximum raise in maximum and minimum temperature is 1.3^oC and 0.5^oC, respectively. In addition, winter season witnessed a higher increase in temperature. On the other hand, the projected precipitation has mixed trend under both SSPs. The information from this study are invaluable for policymakers, planners, and relevant stakeholders to make informed decisions against the changing climate.