Comparison of future uncertainty of AOGCM-TAR and AOGCM-AR4 models in the projection of runoff basin

Increased concentration of greenhouse gases is expected to alter the radiative balance of atmosphere, causing increases in temperature and changes in precipitation patterns and other climatic variables. These changes are associated with the changes in hydrological systems globally and at the basin scale. These changes include: precipitation patterns and extremes; the amount and generation of river flow; the frequency and intensity of flood and drought.
At present, coupled Atmospheric -Oceanic General Circulation Models (AOGCMs) are the most frequently used models for projection of different climatic change scenarios. These scenarios can finally simulate other changes, such as water resources changes in the future. The basis of these models consists of describing the physical processes taking place in the climate system and the dynamics of climate variables as a function of different internal or external changes. Up to now IPCC has released 4 different versions of AOGCM models including: First Assessment Report models (FAR) in 1990, Second Assessment Report models (SAR) in 1996, Third Assessment Report models (TAR) in 2001 and Fourth Assessment Report models (AR4) in 2007. In this paper we evaluate the uncertainty of using different TAR and AR4 AOGCM models on the projection of runoff of a basin. At first we used temperature and precipitation variables of 7 TAR models including CCSR, CGCM2, CSIRO-MK2, ECHAM4, GFDL-R30, HadCM3, NCAR-DOE PCM and 9 AR4 models including; CCSM3, GCM3, CSIRO Mk3, GFDL CM2.1, GISS E-R, HadCM3, ECHAM5, MIROC-med, PCM for the 2040-2069 periods under A2 emission scenario. The A2 scenario corresponds to pessimistic future with higher population growth, lower GDP growth, and fragmented and slower technological change. A conceptual rainfall-runoff model (SYMHYD) was calibrated and verified for the Gharesu basin in baseline period (1971-2000). SIMHYD simulates daily runoff (Surface runoff and base flow) using daily precipitation and potential evapotranspiration (PET) as input data. Gharesu basin is located in North West of great Karkheh River Basin, west of Iran.
Historical data in this study are daily temperature (T), precipitation (P) and runoff (R) for thirty years period (1971 to 2000). These data were acquired from different sources and stations. Temperature records of Kermanshah synoptic station, Outflow measurements of Qarehbaghestan hydrometric station, and daily precipitation records of eleven rain gauge stations were used in this study.
The climate variables (monthly temperature and precipitation) of 16 AOGCMs were downscaled to Gharesu basin. Downscaling is a procedure that derives local or regional scale information from larger scale data like AOGCM outputs. In this study, change factor downscaling techniques was employed to generate monthly precipitation and temperature values for Gharesu basin scale in future period (2040-2069). Results show that in all months the temperature of the basin will increase by an average of 2.5°C. On the other hand the increasing of temperature simulated by TAR models are more than AR4 models. Both AR4 and TAR models simulate precipitation in a same manner, reduction for winter and spring and increase for autumn.
Finally the ranges of precipitation and temperature change of the period 2040-2069 simulated by both models introduced to SYMHYD rainfall-runoff model and the monthly runoff changes of the basin were simulated for the period 2040-2069 relative to the period 1971-2000. Results show that runoff change of the basin due to AR4 models are less than TAR models for most of the months. On the other hand the runoff will increase in winter by 20-60% and by 20-40% in summer and decrease in autumn up to 40% and up to 60% in spring. Finally it can be concluded that although the number of AR4 models used in this study is more than TAR models, the range of uncertainty of AR4 is less than TAR. The final results showed that the projections of AR4 models are more reliable than TAR models.