Uncertainty analysis of rainfall projections (case study: Bojnourd and Mashhad synoptic gauge station)

Despite recent progress in developing reliable climate models, the different uncertainties inherent in climate change projections. Climate can change due to a number of anthropogenic and natural factors in spatial and temporal large scales. Therefore, a successful application of a climate parameters simulation in applied water research strongly depends on uncertainty analysis of model output. Here we present a detailed and quantitative uncertainty assessment of rainfall for first future epoch (2011-2040) and second future epoch (2040-2070), based on the projections of wide range of rainfall projections resulting from the factorial combination of four emission scenarios, five GCMs and two downscaling methods (LARS-WG æ SDSM) in Bojnourd and Mashhad synoptic stations. This enabled us to decompose the uncertainty in the ensemble of projections using Box-whisker plot and Bootstrapping method. The uncertainty in precipitation change in response to the general circulation model (GCM) from HadCM3, NCPCM, CNCM3, GFCM2, CGCM3, SRES emission scenarios (A1B, A2, B1, and B2) and two downscaling method (SDSM and LARS-WG) was investigated in two future epochs. In this study, we evaluate the impact of uncertainty in climate change projections on the future precipitation by Box-whisker plots and Bootstrap technique. In the first step, the outliers were excluded by box-and-whisker plots. In the next step the precipitation projected which is reported by ten different scenarios, is then a vector of about 6000 bootstrap replications (500 per model), from which we take the 2.5th and 97.5th percentiles to calculate the range containing 95% of projected estimates. The GCM models show wide variation in their results, particularly for Bojnourd precipitation forecasting. According to Box-whisker graph in Bojnourd synoptic station (BSS), the projected precipitations by CGCM3 and HadCM3 in first and second epoch fall under the 2.5th and 97.5th percentiles. In Mashhad synoptic station (MSS) some scenarios projected precipitation significantly different from other scenarios which were belonging to CGCM3 in January and March and GFCM3 in summer months. On the basis of these results, it is clear that both stations will experience an increase in precipitation for epoch1 and epoch2, with the largest increase found for epoch2. In the next step confidence interval estimation by the bootstrap method is investigated for the uncertainty quantification of precipitation projections using the random sampling method. In BSS the confidence interval band is large in all month except in August and October. It is interesting that for MSS, the range in GCM predictions is relatively small for all seasons except in spring. This means that the uncertainty in climate predictions is considerably smaller for these months. All GCM and downscaling outputs are inherently uncertain because no model can ever fully describe physical systems. Most studies in the literature on the climate change projection do not capture the full range of plausible future climate variation, making their findings seem more precise than they actually are, and as a result making them less credible among climate scientists and potentially misleading for policymakers. We feel that the methodological approach presented here addresses a fundamental shortcoming in the past research. We show that failing to account for climate uncertainty lead to a false sense of confidence about the likely future impacts of climate change, when in fact impacts are actually far less certain.