Low winter precipitation, but not warm autumns and springs, threatens mountain butterflies in middle-high mountains

Low-elevation mountains represent unique model systems to study species endangered by climate warming, such as subalpine and alpine species of butterflies. We aimed to test the effect of climate variables experienced by Erebia butterflies during their development on adult abundances and phenology, targeting the key climate factors determining the population dynamics of mountain insects. We analysed data from a long-term monitoring of adults of two subalpine and alpine butterfly species, Erebia epiphron and E. sudetica (Nymphalidae: Satyrinae) in the Jeseník Mts and Krkonoše Mts (Czech Republic). Our data revealed consistent patterns in their responses to climatic conditions. Lower precipitation (i.e., less snow cover) experienced by overwintering larvae decreases subsequent adult abundances. Conversely, warmer autumns and warmer and drier springs during the active larval phase increase adult abundances and lead to earlier onset and extended duration of the flight season. The population trends of these mountain butterflies are stable or even increasing. On the background of generally increasing temperatures within the mountain ranges, population stability indicates dynamic equilibrium of positive and detrimental consequences of climate warming among different life history stages. These contradictory effects warn against simplistic predictions of climate change consequences on mountain species based only on predicted increases in average temperature. Microclimate variability may facilitate the survival of mountain insect populations, however the availability of suitable habitats will strongly depend on the management of mountain grasslands.


Supplementary methods
Monitoring of adult abundances in the Jeseník Mts (J) and Krkonoše Mts (K) was done by counting butterflies along three transects in J and four transects in K. J1 (length 2800 m, altitude 1325-1465 m, centre coordinates N 50°2.948', E 17°13.359') follows the main ridge covered by Nardus grasslands; J2 (570 m, 1370-1400 m, N 50°3.874', E 17°14.678') crosses, in addition to Nardus grasslands, some Pinus mugo shrubs; and J3 (800 m, 1330-1492 m, N 50°4.770', E 17°13.880') ascends perpendicularly to the highest summit, from the tall-herb timber belt formations to Nardus grasslands. K1 (960 m, 1120-1270 m, N 50°41'27.5" E 15°38'39.5") and K2 (2400 m, 1270-1340 m, N 50°42'05.2" E 15°39'52.8") are located at cultural grasslands in the mountain zone; whereas K3 (1200 m, 1340-1360 m, N 50°42'42.3" E 15°40'33.5") and K4 (2500 m, 1360-1510 m, N 50°43'19.2" E 15°41'14.6") cross the subalpine habitats. Table S1. (provided as a separate file) Summary data on the abundance monitoring and the climate variables are included: Walks = the number of transect walks completed during the year, Individuals_recorded = the total number of individuals observed during the year at the transect, Interval = the dates of the start and end of the field sampling. The fit of the generalised additive models (GAM) fitted for data from each transect and year separately is described: GAM_explained_dev = % deviance explained, GAM_df = the number of degrees of freedom describing the complexity of the fitted curve, GAM_F = the F statistic value, GAM_P = the Pvalue. The fitted GAM was used to estimate the abundance index and descriptors of phenology: PAI = the population abundance index, Onset = the estimated onset of the flight period, Duration = the estimated duration of the flight period. The rest of the variables describe the climate fluctuations based on data from the available weather stations. See the Methods in the main text for detailed explanation of the variables.

Population recruitment curves
We monitored adults of Erebia sudetica and Erebia epiphron along three permanent transects in the Jeseník Mts (J1-J3) and four transects in the Krkonoše Mts (K1-K4) (see Figure 1 in the main text). Figures S1-S11 show the daily numbers of individuals of the two Erebia species in individual

Correlations among climate variables
Figures S12-S15 show Pearson's correlation coefficients among all pairs of climate variables calculated separately for the Jeseník Mts ( Fig. S12 and S13) and the Krkonoše Mts, where we used data from two weather stations -one at a lower altitude (Fig. S14) and another at a higher altitude (Fig. S15). In addition, we visualised the relationships among the climate variables also using PCA, with the variables scaled to unit variance (Fig. S16). The explanation of the variables is provided in the Methods in the main text.      Table 1 in the main text).

Temporal trends
We tested the dependence of the climate variables on the year separately for data from the Jeseník Mts (complete data including the 1990s and recent data from the period 2009-2020) and the Krkonoše Mts, where we used data from two weather stations -one at a lower altitude and another at a higher altitude. As expected, we detected several significant temporal trends in climate variables over the entire study period in the complete dataset from the Jeseník Mts, including measurements from the 1990s, but most of these trends were not significant over the short recent period of 2009-2020 (Table S2).
Figures S18 and S19 show the temporal trend, or the lack thereof, in the phenology of Erebia sudetica and Erebia epiphron in the two mountain ranges. We estimated the Onset and Duration of the flight period separately for each transect (see Methods in the main text). We tested the relationship between the Onset or Duration of the flight period (both log-transformed) and the year using generalised linear models (GLM) with a Gaussian error distribution. We used AIC to compare the fit of a model with transect identity and the year as predictors to a model with transect identity only. In most cases, the simpler model was better supported. Hence, there was no evidence of a shift in the Onset or Duration of the flight period in time over the duration of our study, with the exception of E. epiphron in the Jeseník Mts with data from all years, including the 1990s included in the analysis (see also Table 2 in the main text).