Several sources of uncertainty such as greenhouse gas emission scenarios, global climate sensitivity, and regional climate change affect impact assessment of climate change on water resources. Since the probability distribution function (PDF) of these sources of uncertainties is unknown, the PDF of runoff cannot be predicted with any accuracy. This research work is aimed to find the most effective sources of uncertainty for evaluation of changes in water resources due to climate change and introduce a method to construct consequent PDF. The approaches were applied to the Zayandeh Rud River in Iran.

To do this, monthly climate data (temperature and precipitation) of the periods 1971-2000 (baseline) and 2010-2099 retrieved from the seven AOGCMs climate datasets including: CCSR NIES, CGCM2, CSIRO MK2, ECHAM4 OPEYC3، GFDL R30, HadCM3 and NCAR DOE PCM. These data include surrounding output cells of the AOGCMs with respect to the position of the study area. In the next step, the baseline cells were downscaled to the study area using IDW (Inverse Distance Weighting) and Krigging approaches and the results were compared with observed data. IDW was selected due to its enough accuracy and simplicity. This method was also applied to downscale AOGCMs data for the period 2010-2099.

To construct climate change scenarios under the extreme SRES emission scenarios, so called B1 and A1FI, Pattern scaling method applied using MAGICC software. Comparison of the seasonal climate change scenarios with the internal climate variability, which were calculated based on 1000 years of the AOGCM Control Runs, revealed significant changes in rainfall and temperature for the study area.

To investigate possible impact of climate change on the water resources of Zayandeh Rud basin and its associated uncertainty, a rainfall-runoff model so called IHACRES was calibrated for the study area. In this regard, the uncertainties due to downscaling methods, green house emission scenarios, and the AOGCM projections (climate change scenarios) were evaluated. The results showed greater role of the AOGCM projections on cumulative frequency distribution (CDF) of the seasonal runoff. To quantify it, Bayesian - Monte Carlo approach was carried out and effect of sampling uncertainties (i.e. number and methods of sampling from AOGCM projections) and arbitrary prior distributions (i.e. uniform, normal, triangular) of AOGCM projections on runoff CDF was investigated. Prior distribution was found as the most effective component on the CDF.

The prior distributions of AOGCMs projections were constructed using two novel weighting methods based on the performance of the AOGCMs in reproducing present-day temperature-precipitation (MOTP (Mean Observed Temperature-Precipitation)) and runoff (MOR (Mean Observed Runoff)). Comparison of the resulted runoff CDFs from MOTP and MOR with REA (reliability ensemble averaging) method revealed better performance of the MOTP.

The most critical and nervous result of this research work was high probability for reduction in water resources of the basin. For instance there is between 40 and 77 % probability for 10% reduction in seasonal yields of the Zayandeh Rud River for said future periods respect to the baseline.