Electronic supplementary material from Historical citizen science to understand and predict climate-driven trout decline

Figure S1. Annual average daily mean temperatures in Spain for the period 1850-2005. Represented values are anomalies (in °C) from the mean values from 1961 to 1990, as reported by Brunet et al. (2007). Squares and whiskers are mean and standard error values for the periods 1850-1870, 1900-1920 and 1990-2005. A and B mark the significant differences among periods found by a Tukey HSD post-hoc test (P< 0.001). The temporal temperature trend is represented by a weighted average line with a 10-year period Brunet M et al. 2007 Temporal and spatial temperature variability and change over Spain during 1850–2005. J. Geophys. Res. 112, D12117.; Figure S2. Response curves of thermic suitability for trout to temperature variables using data from 1850 and 2000. The values shown are average suitability for temperature categories of 0.5 °C intervals, using 2000 temperature as a reference. Annual mean, July maximum and January minimum have 32, 36 and 35 categories, respectively. The curves shown correspond to those modelling techniques that accurately predicted trout distribution in both periods (i.e., AUC > 0.7).GLM: Generalized Linear Models; GBM: Generalized Boosting Models; GAM: Generalized Additive Models; ANN: Artificial Neural Networks; FDA: Flexible Discriminant Analysis; MARS: Multivariate Adaptive Regression Splines; and RF: Random Forest. Classification Tree Analysis (CTA) and Surface Range Envelope (SRE) were also run but are not shown in the figure for not reaching AUC thresholds.; Figure S3. Extinction probability as determined by trout prevalence in 10×10km cells. Proportion of extinction cells (trout cited in the 1850s but not in the 2000s) among the different categories of trout prevalence defined for cells with information from at least 3 localities from the Madoz (1850s data). Low: trout present in 1 to 50% of the localities (n= 48); High: trout present in 51 to 90% of the localities (n= 178); Widespread: trout present in more than 90% of the localities (n= 439).; Figure S4. Changes in temperatures in Spain between the 1850s and 2000s climatic scenarios. See methods for a description of the development of climatic scenarios; Figure S5. Relationship between the predicted and observed changes in thermal suitability for trout between the 1850s and 2000s. Predicted change was calculated as the suitability difference between the projection of the 1850s model to the 2000s thermal scenario (2000s predictions) and the 1850s estimates, while observed change is the difference between the suitability estimates of the 2000s and the 1850s models.