Abstract Understanding the temporal and spatial variability of river discharge of alpine hydrological systems is of particular interest due to their relevance for water uses including water provisioning, hydropower production and touristic activities. Streamflow variability is highly heterogeneous both in time and space due to several reasons, such as a differentiated response to climate change, differences in catchment morphology and geographic location. Therefore, catchment classification for these systems is challenging. A suitable tool to determine the crucial scales of variability of a non-stationary time series is wavelet transform. In this work, we compute the wavelet coherence between fifty selected gauging stations located within the Inn River catchment to classify them by runoff behavior, focusing only on long term variability, between one and eight years scales. This choice allows us to filter out the effect of local meteorological patterns and the effects of hydropower production. In addition, we decompose the streamflow signals in three levels (256 days, three years, six years) using Discrete Wavelet Transform to further understand the detected alterations in the streamflow signal. Three main runoff behaviors (referred as three classes) are found at the yearly scale. Focusing on two-four years scales a loss of coherence between time series located at different elevations becomes significant in the 1980s. Prior to 1980 we detect four different behaviors, while after 1980 we detect eleven different classes. At larger scales the stations are clustered in four classes. Our analysis highlights that catchment classification may depend on the scale of the analyzed signal and it may vary both in time and space. This research contributes to the development of new methods for catchment classification, which is highly relevant for many hydrological applications such as prediction in ungauged basins, model parameterization, understanding the potential impact of environmental and climatic changes, and transferring information from gauged catchments to the ungauged ones.

中文翻译：

---

使用小波分析对阿尔卑斯山长期流量变化的流域内比较和分类

摘要 了解高山水文系统河流流量的时空变异性尤其令人感兴趣，因为它们与水资源利用（包括供水、水电生产和旅游活动）相关。由于多种原因，例如对气候变化的不同响应、流域形态和地理位置的差异，河流流量变异在时间和空间上都具有高度异质性。因此，这些系统的流域分类具有挑战性。确定非平稳时间序列的关键可变性尺度的合适工具是小波变换。在这项工作中，我们计算了位于因河流域内的 50 个选定测量站之间的小波相干性，以根据径流行为对它们进行分类，仅关注长期变化，一年到八年的量表。这种选择使我们能够过滤掉当地气象模式的影响和水电生产的影响。此外，我们使用离散小波变换将水流信号分解为三个级别（256 天、三年、六年），以进一步了解检测到的水流信号变化。在年尺度上发现了三种主要的径流行为（称为三类）。在 2-4 年尺度上，位于不同海拔的时间序列之间的一致性在 1980 年代变得很重要。在 1980 年之前，我们检测到四种不同的行为，而在 1980 年之后，我们检测到十一种不同的类别。在更大的范围内，站点被分为四类。我们的分析强调，流域分类可能取决于分析信号的规模，并且可能在时间和空间上有所不同。这项研究有助于开发流域分类的新方法，这与许多水文应用高度相关，例如未测量流域的预测、模型参数化、了解环境和气候变化的潜在影响以及将信息从测量流域传输到未测量流域。那些。