Detection and interpretation of recent and historical streamflow alterations caused by river damming and hydropower production in the Adige and Inn river basins using continuous, discrete and multiresolution wavelet analysis

In alpine catchments, fluctuations of river discharge occur at multiple temporal scales due to natural (e.g., snow and glacier melt) and anthropogenic factors (e.g., reservoir and hydropower operation). In the last century, several modifications of the natural flow regime have been observed, originating in the complex interplay between construction of dams for hydropower production, changes in water management policies and climate. In this work, we show that multiresolution analysis via discrete wavelet transform and continuous wavelet transform allow identifying both old breakpoints caused by dam and hydropower plant construction as well as more recent breakpoints caused by changes in hydropower plant operation. According to the detected breakpoints we can differentiate a pre-impact period (natural flow regime), post-impact period (characterized by river damming and hydropower facilities operation) and a last period after the recent detected breakpoints (caused by changes in water management). Moreover, by analyzing streamflow time series and their log transform it is possible to highlight low flow alterations and distinguish the causes of the breakpoints, since they affect different temporal scales. The change point analysis is performed using different mother wavelets: Haar, Daubechies 2 (db2), and Daubechies 20 (db20). Considering the results obtained with the Haar mother wavelet: a total of 14 historical breakpoints in 13 streamflow time series belonging to the Adige and Inn catchments were identified and 93% correspond to metadata associated to river damming and hydropower production; a total of 5 recent breakpoints were identified, which correspond to metadata associated to river damming and hydropower production. In case of the log transform of the streamflow: 32 historical breakpoints were identified and 100% confirmed by metadata; 20 recent breakpoints were detected and 83% correspond to metadata. The uncertainty in the temporal location is investigated computing the mean difference between detected breakpoint temporal location and metadata and is generally smaller than 1 year. This work shows that change point analysis performed using wavelet transform allows us to differentiate and correctly locate in time most of change points in streamflow variability that can be associated to river damming and hydropower production.