论文部分内容阅读
当前小区域的古气候变化研究受模拟资料分辨率和可靠性的严重制约。为了将大区域的气候模拟资料应用到小区域的古气候研究中去,亟待建立有效的降尺度方法。为此以徽鄂地区为例,建立了一个3层BP神经网络拟合模型,利用相关气象要素作为拟合因子,拟合并重建了该地区近千年来1月、7月和年平均的温度和降水序列,通过与观测及模拟资料的对比分析发现,该模型拟合及重建的近千年气候序列有较高的精度和可靠性,能反映小区域气候的年际和年代际变化信号,提高了模拟资料对小区域气候变化的刻画能力。
Paleoclimate changes in the current small area are severely constrained by the resolution and reliability of the simulated data. In order to apply climate simulation data in large areas to paleoclimate research in small areas, an effective downscaling method is urgently needed. Taking Huizhou and Hubei provinces as an example, a 3-layer BP neural network fitting model was established, and the relevant meteorological elements were used as fitting factors to fit and reconstruct the average annual temperature in January, July and in the past 1,000 years in this area And precipitation series. Comparing with the observed and simulated data, it is found that the model has good accuracy and reliability in fitting and reconstructing nearly 1000-year climatic series, which can reflect interannual and interdecadal variation signals of small-scale regional climate and increase The ability of modeling data to describe climate change in a small area.