Figure 2 Metric-based rolling window indicators estimated for dormice body temperature (black lines) and air temperature (yellow lines) separated for each alternative state (activity and hibernation). (a and b) activity state and hibernation state for the female; (c and d) activity state and hibernation state for the male. Red lines show the Gaussian filtering of the time series. Panels 1, 2 and 3 show autocorrelation at lag-1 (AR(1)), standard deviation and skewness respectively, estimated within sliding windows of 20% the size of the time series. Yellow lines show the indicators for air temperature. The Kendall τ indicates the strength of the trend in the indicators for body temperatures.