基于小波变换与神经网络的石羊河流域夏季地温预测模型研究

地温变化在气候反馈效应中起着重要作用， 理解地温及其与影响因素之间的时空关系对预测全球温度变化至关重要。利用1998 - 2017年石羊河流域的逐日常规气象观测资料， 采用小波分析结合BP（Back Propagation）神经网络构建了石羊河流域夏季地温预报模型， 结果表明： 日平均地温预测效果在不同站点均为最佳， 其中预测值和观测值的相关系数均大于0.87， 3 ℃以内的预测概率均大于84%。其中， 民勤地区地温预测效果最好， 预测值和观测值的相关系数达到0.91， 3 ℃以内的预测概率达到86%。日最高地温的预测值与观测值的相关系数高于0.8， 但误差平方和、 标准差较大。永昌地区日最高地温的模拟效果最好， 3 ℃以内的预测概率达到83%。日最低地温的预测与观测值的平均相关系数高于0.66， 3 ℃以内的预报概率高于83%， 但预测值略低。其中， 武威地区日最低地温的预测效果最好， 预测值与观测值的相关系数为0.72， 3 ℃以内的预测概率达到94%。研究成果可为有效弥补干旱、 半干旱区地温观测资料缺失和探讨其与局地气候的关系提供一些参考。

Prediction model of summer land surface temperature in the Shiyang River basin based on the wavelet transform and neural network

The change of ground temperature plays an important role in the climate feedback effect， and understanding the spatio-temporal relationship between ground temperature and its influencing factors is crucial to the prediction of global temperature change. The summer land surface temperature （LST） was constructed by wavelet analysis and BP neural network based on the daily observed data of meteorological stations in the Shiyang River basin from 1998 to 2017. The prediction results and the accuracy are tested. The results show that： （1） The prediction effect of daily average land surface temperature is the best at different stations， and the correlation coefficients between the predicted values and the observed values are both greater than 0.87， and the prediction probability within 3 ℃ is higher than 84%. Among them， Minqin has the best prediction results. The correlation coefficient between predicted and observed values reaches 0.91， and the prediction probability within 3 ℃ is 86%. （2） The prediction results of daily maximum LST in the Shiyang River basin can reflect its variation trend， and the correlation coefficient between the predicted value and the observed value is higher than 0.8. Among them， the simulation effect in Yongchang is the best， and the prediction probability within 3 ℃ is 83%. （3） For the daily minimum LST， the average correlation coefficient between the observed and simulated values is higher than 0.66， but it is little underestimated. The prediction probability of daily lowest LST in the Shiyang River basin at different stations within 3 ℃ is all higher than 83%. Among them， Wuwei has the best forecasting effect， the correlation coefficient between forecasting value and observation value is 0.72， and the forecasting probability within 3 ℃ reaches 94%. This study can provide some references for making up for the lack of ground temperature observation in arid and semi-arid areas and discussing its relationship with local climate.