The Rothamsted Carbon (RothC) model with its multi soil carbon pools is widely used to estimate soil organic carbon (SOC) change in response to climate and/or land use change. Many of these pools are conceptual and it is a challenge to correctly parameterize them. Taking Irish temperate grasslands as an example, we study the SOC fractionation procedure of Zimmermann et al. (2007) to partition the measured SOC into the pools required in RothC. This was done with the aim of predicting SOC change in response to climate change. We found good correlation between the measured and modelled values for the pools of BIO (microbial biomass) and HUM (humified organic matter), but poor correlation for the pools of DPM (decomposable plant material) and RPM (resistant plant material). The measured carbon pools more reasonably reflected the real environmental conditions than the modelled. Because of the fast decomposition rate and short term simulation (only 40 years), the RPM pool controlled the trends in the future SOC change. The difference in the trends of the predicted total SOC between using measured and modelled carbon pools (to initialize RothC) rapidly increased in the initial years and slowly decreased thereafter. In order to limit this difference to 1% within the first 3 years (the turnover period for RPM), the difference between the measured and modelled RPM pool should be constrained to be less than 10%. In response to higher temperature and, drier summers and wetter winters, RothC predicted a decrease in the SOC of Irish grasslands. (C) 2010 Elsevier B.V. All rights reserved.