Abstract: Based on the observed precipitation and temperature data at 14 meteorological stations of the Huaihe River basin from 1964 to 2007, and the climate projection from 2001 to 2100 by the ECHAM5/MPI-OM model, the streamflow for the basin from 2010 to 2100 were projected under the SRES-A2, -A1B, and-B1 scenarios by applying artificial neural network (ANN) hydrological models. The results show that differences in annual streamflow from 2010 to 2100 are significant under the three scenarios; the streamflow under the SRES-A2 scenario displays a general increasing trend, especially significant from 2051 to 2085, it declines gradually in fluctuation from 2024 to 2037 under the SRES-A1B scenario, and shows no obvious trend under the SRES-B1 scenario. Fluctuations of spring streamflow in 2010-2100 are the smallest in all four seasons, ranging from -15.1% to 18.6% under the three scenarios. Summer average streamflow decreases before the 2040s and increases subsequently whilst the fluctuations are not significant. Autumn streamflow has obvious fluctuations under the SRES-A2 and SRES-A1B scenarios; it drops in the 2060s under the SRES-A2 scenario by 50.6%, which is the maximum decreasing range for all projected seasons. Winter streamflow increases in 2050s under SRES-A1B scenario by 54.7%, which is the maximum increasing range in four seasons.

Key words: Climate change projection, streamflow, projection, Artificial neural networks, 21st century, Huaihe catchment