Elevation use over time
Changes in elevation were analysed by applying a “segmented” (or “broken-line”) regression model with record elevation as a response, and the year as a predictor. We obtained the best-fit breakpoint value with the davies.testfunction of the ’segmented’ library in R (Muggeo 2003). Additional break-points were tested by applying the davies.test to each time frame previously detected. Linear regressions were performed on the two obtained intervals (namely to the left and to right of the breakpoint value), a t-test was applied to compare the two regression slopes. For these analyses, we included observations older than 1949 in order to include data evenly distributed over time. Only for B. konradinia different analysis was used: as the few records are unevenly distributed over time, we grouped them in three ranges:“1960s” from 1961-1963, “1980s-1990s” from 1984 to 1998, “2010s-2020” from 2011 to 2020. To calculate the uphill shift, we considered the difference in 25% quantile elevation values of the oldest 40 observations in the recent and the older time group, although we calculated the differences also with other quantiles too for comparison.