使用岭回归对哈萨克斯坦月平均气温的统计降尺度研究

哈萨克斯坦是世界最大的内陆国家,拥有典型的大陆性气候和多样的地理环境及生态系统,同时哈萨克斯坦的自然环境和人类社会对于气候变化这一全球性问题是敏感的、脆弱的,需要运用科学的研究方法应对气候变化的挑战。通常,区域或局地尺度的气候变化影响研究需要对气候模式输出或再分析资料进行降尺度以获得更细分辨率的气候资料。近年来,大量验证统计降尺度方法在各个地区能力的研究见诸文献,然而在哈萨克斯坦地区验证统计降尺度方法的研究非常少见。本文使用了岭回归的方法对哈萨克斯坦地区11个气象站点1960~2009年的月平均气温进行了统计降尺度研究。结果显示,使用前30年数据和岭回归模型建立大尺度预报因子和观测资料的统计关系可以较好地预测后20年的月平均气温,预测能力在各站各月均有不同程度的差异,地形复杂的站点预测效果较差,夏季预测结果好于冬季;此外,将哈萨克斯坦地区平均来看则与观测数据相吻合。

Statistical Downscaling of Monthly Mean Temperature for Kazakhstan Using Ridge Regression

Kazakhstan is the largest landlocked country in the world with a typical continental climate, and its natural environment and human society are sensitive and vulnerable to climate change. In climate change impact studies, one widely-used approach to obtain future climate change scenario at regional or local scale is to downscale future climate projection from General Circulation Model (GCM). So far, to our knowledge, no statistical downscaling study has been carried out in Kazakhstan region. In this study, the authors explored and validated the ability of a statistical downscaling model that is based on ridge regression to predict monthly mean temperatureat of 11 stations in Kazakhstan from NCEP/ NCAR monthly mean reanalysis. The 30-year dataset for the period from 1960 to 1989 was used to train the downscaling model and the next 20-year data for the period of 1990-2009 was used for validation of the downscaling model. The result shows that despite certain disagreements with observations at several stations, the ridge regression model generally is able to reasonably reproduce monthly mean temperature over Kazakhstan region. The authors also find that the performance of the ridge regression model is better in the summer than in the winter and better in flat terrain areas than in complex terrain areas.