Changes in future rainfall extremes over Northeast Bangladesh: A Bayesian model averaging approach

Abstract

In this paper, we used a Bayesian model averaging (BMA) approach to analyse the changes in rainfall extremes in the periods 2041-2070 and 2071-2099 over northeast Bangladesh as a consequence of climate change. Climate change over this region could potentially impact agricultural production, water resources management, and the overall economy of the country. We used six regional climate models (RCMs) over the Coordinated Regional Downscaling Experiment South Asia domain. We used one medium stabilization scenario (RCP4.5) and one highemission scenario (RCP8.5) for projecting the extreme rainfall indices. A multimodel ensemble mean was generated using the BMA approach. The BMA mean is a weighted average related to each RCM's predictive skill during the training period. Most of the rainfall extremes are expected to increase in both premonsoon (March-May) and monsoon (June-September) seasons in the future compared with baseline . The average pre-monsoon rainfall of the study area is projected to increase by 12.93 and 18.42% under RCP4.5 and 18.18 and 23.85% under RCP8.5 for the periods 2041-2070 and 2071-2099, respectively. The average monsoon rainfall of the study area is projected to increase by 4.96 and 2.27% under RCP4.5 and 6.56 and 6.40% under RCP8.5 for the periods 2041-2070 and 2071-2099, respectively. All the extreme indices except consecutive wet day are expected to change significantly at the 95% confidence level during the pre-monsoon season. The study area will potentially be subjected to more frequent floods in the future both in pre-monsoon and monsoon seasons as a consequence of climate change. Notably, the intensity and the magnitude of flash flooding in the pre-monsoon season are expected to increase more in the future because the increase in extreme indices is more significant during that season. K E Y W O R D SBayesian model averaging, climate change, multi-model ensemble mean, rainfall extremes, regional climate models, representative concentration pathways

Risk Identification of Mountain Torrent Hazard Using Machine Learning and Bayesian Model Averaging Techniques

“…Bayesian Model Average (BMA) is a probabilistic post-processing technique that generates probability predictions by quantifying prediction uncertainty [46]. The Bayesian formula was used to combine the prior distribution with the likelihood function in the BMA method, which aims to obtain the posterior distribution of the predicted object.…”

Calibration Indonesian-Numerical Weather Prediction using Geostatistical Output Perturbation

“…The mean BMA unifies the posterior distribution of all the final trained models. The purpose of BMA is to combine uncertain models to get the best model [9]. EMOS used a single normal distribution, where the mean is an affine function of the ensemble members, and the variance is an affine function of the ensemble variable [10].…”

A weighted ensemble of regional climate projections for exploring the spatiotemporal evolution of multidimensional drought risks in a changing climate

“…Although the AEM simulation reduces the model bias compared to a single climate model, the systematic bias cannot be neglected and would hinder reliable projections of future droughts. An alternative approach of the AEM approach is Bayesian model averaging 4 (BMA), which has been proven to be a promising tool for improving multi-model hydroclimate simulations (Duan and Phillips 2010;Yang et al 2011;Olson et al 2016Olson et al , 2018Zhang et al 2016;Ahmadalipour et al 2018;Shin et al 2019;Basher et al 2020). However, little effort has been directed towards applying BMA to project future drought characteristics (Ahmadalipour et al 2018;Miao et al 2020).…”