Comments on “Twentieth-Century Trends of Arctic Precipitation from Observational Data and a Climate Model Simulation”

Abstract

Recently Kattsov and Walsh (2000, hereafter KW) investigated Arctic precipitation changes during the twentieth century. They compared observationally based precipitation data over the Northern Hemisphere subpolar and polar region with climate model output derived from a single long-term ECHAM4/T42 experiment forced with observed SST data. In addition to the SST forcing the model run is also subject to increasing greenhouse gas concentrations according to historical measurements. With regard to the area-averaged annual sums of total preciptation, a clear positive trend is common to both datasets, observed and simulated, particularly during the first half of the twentieth century. Assuming that the model’s ability to reproduce the observed time series is not due to chance, this result suggests that SST and/or CO2 changes may be responsible for this increase in Arctic precipitation. However, the authors point to some uncertainty factors, mainly concerning the accuracy of the observational data (cf. Cullather et al. 2000) and some unrealistic features in the Global Ice and Sea Surface Temperature data version 2.2 (GISST2.2) sea ice data as lower boundary condition of the model experiment. Here we want to present some additional results based on Monte Carlo experiments with the same model to pick up another aspect that generally limits the reliability of intercomparison studies between observations and single model simulations, namely the discrimination of externally induced and internal chaotic variability components. Time–space variations of observed climatological variables are composed of natural variability induced by internal stochastic processes as well as external factors like the solar cycle, vulcanic eruptions, etc. (Fig. 1). A potential external either natural or anthropogenic climate change signal is hardly distinguishable from noise since natural fluctuations of the real climate system are largely unknown and cannot be separated without additional assumptions. Similar considerations ap-