Relating West Antarctic Ice Cores to Climate with Artificial Neural Networks

该奖项提供三年的支持，使用广泛的、适应性强的、多参数的方法，使用包括人工神经网络在内的一系列技术来寻求气象条件与雪坑和冰芯记录之间的关系。已有或正在收集的多参数、高分辨率冰芯数据反映了积雪、大气化学、同位素分馏和其他过程，通常具有亚年分辨率。可获得这类数据的南极西部站点将被用作再分析数据产品中反轨迹分析的起点，以确定提供数据流的气象条件。然后，人工神经网络将用于寻找这些气象条件与其产品之间的最佳关系。以前的工作已经证明了再分析产品在确定积雪量方面的价值，在理解冰川化学方面的反轨迹分析的价值，以及在连接气象条件及其产品方面的人工神经网络的价值。初步工作表明，神经网络成功地将再分析产品降尺度为西南极洲自动气象站数据，使特定站点数据的插值能够提高对西南极洲气候近期变化的理解。

0087380AlleyThis award provides three years of support to use a broad, adaptable, multi-parameter approach, using a range of techniques including artificial neural networks to seek the relations between meteorological conditions and the snow pit and ice core records they produce. Multi-parameter, high resolution, ice core data already in hand or now being collected reflect snow accumulation, atmospheric chemistry, isotopic fractionation, and other processes, often with subannual resolution. The West Antarctic sites from which such data are available will be used as starting points for back-trajectory analyses in reanalysis data products to determine the meteorological conditions feeding the data stream. The artificial neural nets will then be used to look for optimal relations between these meteorological conditions and their products. Previous work has demonstrated the value of reanalysis products in determining snow accumulation, of back trajectory analyses in understanding glaciochemistry, and of artificial neural nets in linking meteorological conditions and their products. Preliminary work shows that neural nets are successful in downscaling from reanalysis products to automatic weather station data in West Antarctica, enabling interpolation of site-specific data to improve understanding of recent changes in West Antarctic climate.