Abstract: Based on two methods of R/S and non-periodic cycle analysis, the values of Hurst parameter H, fractal dimension D and non-cycle average cycle length of temperature and precipitation at time series in Lanzhou City were calculated, and then the variation trends, long-term memory effects and memory cycles of temperature and precipitation at the time series were analyzed. The results show that the values of H and D of average seasonal and annual temperature varied in ranges of 0.5-1 and 1-1.5 respectively, there was a fractal structure, and average seasonal and annual temperature was in a continuous increase trend. Average cycle lengths of average annual and spring, summer, autumn and winter temperature at time series were 9 years, 4 years, 8 years, 5 years and 6 years respectively. Precipitation in spring and winter will continuously decrease but increase in summer. H and D of average annual and autumn precipitation series with long-term negative correlation will vary in ranges of 0-0.5 and 1.5-2 respectively, the change process has an anti-persistent character, the average annual and autumn precipitation in the future will in an increase trend, average cycle lengths of average annual, spring, summer, autumn and winter precipitation at time series will be 9 years, 7 years, 10 years, 12 years and 13 years respectively. Precipitation in the past will affect that in the future for a long time. The study revealed that the methods of R/S and non-periodic cycle analysis are new, scientific and reliable in predicting regional climate change.

Key words: rescaled range analysis, non-periodic cycle analysis, temperature, precipitation, heat island effect, climate change, trend prediction, Lanzhou