The overall trend of rapid retreat of Alpine glaciers contains considerable variability of responses at the scale of individual glaciers. As a step towards a regional assessment of glacier state that allows a detailed differentiation of single glaciers, we explore the potential of a self-organizing maps (SOM) algorithm to identify and cluster recurring patterns of thickness change at glaciers in western Austria. Using digital elevation models and glacier inventories for three time periods, we compute the frequency distribution of surface elevation change over the area of each glacier in the data set, for each period. The results of the SOM clustering show a distinct pattern shift over time: From 1969 to 1997, surface elevation change occurred at relatively uniform rates across a given glacier. Since 1997, the distribution of surface elevation change at individual glaciers has been far less uniform, indicating accelerated processes of disintegration. Tracking the evolution of individual glaciers throughout the time periods via the clusters highlights both the broader regional trend as well as glaciers that deviate from this trend, e.g. some very small, high elevation glaciers that have reverted to reduced and more uniform volume loss patterns.

高山冰川快速退缩的总体趋势在单个冰川的规模上包含相当大的响应可变性。作为对允许详细区分单个冰川的冰川状态进行区域评估的一步，我们探索了自组织地图 (SOM) 算法在识别和聚类奥地利西部冰川厚度变化的重复模式方面的潜力。使用三个时间段的数字高程模型和冰川清单，我们计算了每个时期数据集中每个冰川区域的表面高程变化的频率分布。SOM 聚类的结果显示出明显的模式随时间变化：从 1969 年到 1997 年，给定冰川的表面高程变化以相对均匀的速率发生。自 1997 年以来，个别冰川表面高程变化的分布远没有那么均匀，这表明解体过程正在加速。通过集群跟踪单个冰川在整个时间段内的演变，突出了更广泛的区域趋势以及偏离这一趋势的冰川，例如一些非常小的、高海拔的冰川已经恢复到减少和更均匀的体积损失模式。