论文部分内容阅读
基于贝叶斯神经网络,构建了资料匮乏地区的径流降尺度模型,模拟了叶尔羌河卡群站月平均径流,与BP神经网络的结果进行了对比,验证了BNN的优越性,并结合CMIP5三种气候模式GFDL_ESM2G,GFDL_ESM2M及MIROC5的RCP 4.5,RCP 6.0,RCP 8.5三种情景,对未来3个时段(2020年代,2050年代,2080年代)卡群站月平均径流进行了预测,并定量计算了预测的不确定性区间,研究表明:贝叶斯神经网络降尺度模型可以较好地捕捉叶尔羌河的径流特征,即相关系数达到0.9以上,效率系数达到0.8,且模拟效果比ANN较优;未来情景下,叶尔羌河流域受气温升高影响,3个时段年径流均呈现增加的趋势,增加幅度分别为75%~92%,83%~110%,88%~127%,其中RCP8.5情景下的径流增加幅度比其他情景较明显;不同月份径流存在不同程度的增加趋势,其中5-8月份变化趋势相对较明显。
Based on the Bayesian neural network, a runoff downscaling model was constructed for the lack of data and the monthly average runoff of the Yarkand River simulator was simulated. Compared with the results of BP neural network, the superiority of BNN was verified. Combining with CMIP5 Climate model GFDL_ESM2G, GFDL_ESM2M and MIROC5 RCP 4.5, RCP 6.0, RCP 8.5 three scenarios, the next three periods (2020, 2050, 2080) monthly average runoff of the card group station has been predicted and quantitative calculations of the forecast The results show that the Bayesian neural network downscaling model can capture the runoff characteristics of the Yarkant River better, ie the correlation coefficient reaches above 0.9 and the efficiency coefficient reaches 0.8, and the simulation result is better than ANN. In the future scenario . The annual runoff in the Yarkant River basin was increased by 75% -92%, 83% -110% and 88% -127%, respectively. The runoff in the RCP8.5 scenario was affected by the increase of air temperature. The increase range is more obvious than other scenarios. The runoff in different months has an increasing tendency in varying degrees, of which the trend of change in May-August is relatively obvious.