Abstract: The authors examine meteorological observation data over the Tibetan Plateau (TP) during the passed 50 years. The Plateau was divided into five temperature and precipitation subareas using the method of self-organizing feature maps. For each subarea, mean generating function-optimal subset regression was applied to predict climatic variations in the future 5 years. The results indicate that there is no obvious trend in precipitation for the TP as a whole, except southeastern Qinghai and eastern Tibet, where a significant decreasing trend is found, and annual fluctuations of precipitation are violent. However, temperature of the TP exhibits an increasing tendency, with the exception of the southeastern part.

Key words: Tibetan Plateau, mean generating function, optimal subset regression, SOFM

摘要： 以青藏高原78个站点50年的逐年降水和温度数据为基础, 使用 SOFM 人工神经网络模型对高原的降水和温度变化进行了分区, 并采用均生函数-最优子集回归( MGF-OSR) 预测模型对青藏高原的降水和温度进行了5年情景的预测。预测结果表明:总体而言, 今后 5 年青藏高原的降水年际波动较大, 并没有显著的趋势;但青海东南和西藏东部部分地区有明显的减少。青藏高原的总体温度变化增加趋势显著, 仅高原东南部明显降温。

关键词: 青藏高原, 均生函数, 最优子集回归, 人工神经网络