4 Climate Modeling

4.1 RCM Bias Correction

Linear regression was used to correct the RCM simulation results. The climate correction sequence in the Xiangxi River Basin is from 1991 to 2005, in the Jinghe River Basin is from 1981 to 1987, and in the Zhongzhou River Basin is 2009-2013. The variability of MAE in the three basins is calculated (Fig. 5).
The climate simulation corrected for precipitation and temperature, respectively. As can be seen from Fig. 5, the correction has different effects in the two climatic factors. In general, this method has a better correction on temperature. Among the three basins, Zhongzhou River has the best correction effect, with an average of 83.66%; the correction effect of Xiangxi River ranks the second, with an average of 78.05%; and the correction effect of Jinghe River is worse than the others, with an average of 77.42%. The results on precipitation is not ideal, and the correction effect are achieved just in Xiangxi River and Zhongzhou River, i.e. 17.37%, 2.63%. This is because the RCM simulation sequence and the measured sequence are arranged in the ascending order firstly when the correction method is used. The climate correction is performed on the basis of the climate sequence that disturbs the one-to-one correspondence. Although this method may have negative bias growth in precipitation correction, it has a more accurate forecast effect on extreme weather during the whole forecast period for long-term climate simulation.
At the same time, the results show that the climate simulation accuracy by the four RCMs, i.e., RSM3, HadGEM3_RA, RegCM4 and WRF, has not much different among the three basins. Therefore, in the subsequent part of this paper, only the mean analysis results of the four RCMs are presented.

4.2 Annual Climate Changes

Using RSM3, HadGEM3_RA, RegCM4 and WRF to forecast climate in the period 2021-2050, the forecast is carried out in Xiangxi River, Jinghe River and Zhongzhou River, respectively. And the forecast results are corrected by the above climate bias-correction method. Since the four RCMs have similar climate simulation accuracy, this section takes the average forecast data of the four RCMs as forecast results. Taking the mean historical observation data as baseline, the climate change trend under two emission scenarios was analyzed as follows.
The annual analysis results are shown in Fig. 6. It can be seen that annual average temperature in the three basins will increase in the next 30 years. Under the RCP8.5 emission scenario, the temperature rise trend is more obvious. Temperature rise in Jinghe River is the strongest, and the highest temperature under the RCP8.5 emission scenario appears in 2047, an increase of 43.39%. Compared with Jinghe River, the temperature changes of the others are more gradual. The increase in Xiangxi River does not exceed 30%, and in Zhongzhou River does not exceed 10%. Under the RCP4.5 emission scenario, temperature in Zhongzhou River has a negative growth in a few years.
Future average annual precipitation in the three river basins show different trends. The value in Xiangxi River in the next 30 years shows a significant upward trend. And there was no significant difference in Xiangxi River under the two emission scenarios, all of which are around 16%. The annual average precipitation in Jinghe River shows a great swing change in the next 30 years. The variability under RCP8.5 emission scenario is higher than that under RCP4.5 emission scenario. Therefore, it is more likely that extreme weather will occur in Jinghe River in the next 30 years, and this situation is even more serious under the RCP8.5 emission scenario. Different from the other two basins, the annual average precipitation in Zhongzhou River has a downward trend, with an average decline of 16.27% under RCP4.5 emission scenario and 18.32% under RCP8.5 emission scenario.

4.3 Monthly Climate Changes

Like annual climate changes analysis, based on historical observation data, the climate monthly change trend of the three basins in the next 30 years under two emission scenarios was analyzed. The analysis results are shown in Fig. 7.
As can be seen from the figure, the monthly average temperature in the three basins has an upward trend. The rise in Jinghe River is the most obvious, especially in the summer (June-August) and winter (December-February). And the growth rate is larger under RCP8.5 emission scenario. Therefore, the summer temperature in Jinghe River will be higher than the original, and the “warm winter” phenomenon may continue to occur. Temperature rise in Xiangxi River mainly occurs in the summer and autumn (September-November), while the temperature in spring (March-May) and winter show a downward trend. Therefore, the temperature difference among four seasons in Xiangxi River may be even more different in the future. That means, the four seasons in Xiangxi River may more distinct. The rise in Zhongzhou River mainly occurs in the spring, while in the other three seasons have not obvious change. Therefore, the temperature difference among four seasons in Zhongzhou River may decrease in future.
There is a significant difference in the monthly average precipitation. In the next 30 years, monthly average precipitation in Xiangxi River and Zhongzhou River during the precipitation peak period (July) will increase. Therefore, the probability of extreme weather in these two basins may increase in the future. In particular, the precipitation decreased significantly in April and May when precipitation was less in Zhongzhou River, which would aggravate the occurrence of extreme weather. However, Jinghe River did not show significant changes. Significantly, it was analyzed that the annual average precipitation in the Jinghe River has a swing change. Therefore, no significant change in monthly average precipitation does not indicate that there is no change in future precipitation.