<p>Human activities and natural processes are the main drivers of the spatio-temporal variability of thermal regime. Despite a few local studies on the thermal regime variability, regional assessments are scarce in the scientific literature. However, regional assessments allow tracing systematic human-induced changes emerging from some types of anthropogenic structures like dams or ponds and identifying the locations of highly influenced reaches.</p><p>In the current study, we propose a framework to detect the influence of dams and ponds on stream temperature. We use observational data from 526 evenly distributed hourly stream temperature stations in the Loire River catchment, France (110,000 km<sup>2</sup>). The data consist of unbalanced time series of natural and altered thermal regimes that contain at least 80 summer days from 2000&#8211;2018. By comparing time series of observed stream temperature and air temperature, we define five indicators to distinguish different patterns of thermal regime. Three of them are based on weekly stream-air temperature linear regressions (slope; intercept; and coefficient of determination). The remaining two indicators compare monthly air and stream temperature regime: 1) the proportion of times stream temperature is greater than air temperature from March&#8211;October (&#8220;frequency&#8221;), and 2) the lag time between the annual peak in air temperature and annual peak in stream temperature (&#8220;shift&#8221;).</p><p>K-means clustering partitioned stations into three clusters: 1) pond-like, 2) dam-like 3) and natural, with 164, 37, and 316 stations, respectively. Supporting this cluster analysis, 93% of stations in pond-like cluster have upstream ponds, and 55% of stations in dam-like cluster have upstream large dams. Pond-like stations have the greatest slope between weekly stream and air temperatures (slope = 0.4) and have stream temperatures greater than air temperatures more frequently (68%) than other clusters. In contrast, dam-like stations have the lowest correlations between weekly stream and air temperatures (mean R<sup>2</sup>=0.3, compared to 0.7 for the other two clusters). Dam-like stations also exhibit the largest shifts in stream thermal regime relative to air temperature (mean shift = 30 days). Impounded runoff index (IRI), the ratio of reservoir volume to annual discharge, best explaines variability within the dam-like cluster. For pond-like stations, catchment areas and mean upstream ponded surface area best explain the within-cluster variability, particularly for the frequency indicator, although this relationship is sensitive to interannual air temperature regime.</p><p>These findings support modelers in quantifying the downstream impacts of different types of anthropogenic structures and managers in surveying and monitoring stream networks through identification of critical reaches.</p>
Supplementary material to "Regional, multi-decadal analysis reveals that stream temperature increases faster than air temperature"