Fig. 4: Effect size of each predict variable for late spring frost (LSF) frequency, FSI, and GDD extracted from linear mixed models.
From: Flowering in the Northern Hemisphere is delayed by frost after leaf-out
In the mixed models, LSF frequency, FSI (the False Spring Index), and GDD (the accumulation of growing degree days) were response variables, while species, region, and period were the predictors, with species and sites as random intercepts. All records (81,093 in total) used in these models were in the presence of LSF, including 33,413 records during 1950–1980 and 47,680 records during 1981–2021. LSF frequency, FSI, and GDD were calculated as the proportion of days with LSF within the period between leaf-out and the last LSF, the days from leaf-out to the last LSF, and the sum of daily mean temperature above 5°C from leaf-out date to the last LSF, respectively. Species, region, and period were categorized into early- and late-leafing species, boreal, temperate, and subtropical region, and period of 1950-1980 and 1981-2021, respectively. The red and blue points indicate negative and positive estimates of predictor effect size, respectively. Error bars represent 95% confidence intervals (CIs). The black vertical dashed line denotes when estimate of effect size equals to zero. The asterisks indicate a significant difference (two-side t-test, p < 0.05) from zero according to the mixed models results. Variables in these mixed models are standardized in order to compare the effect size within and between models. The detailed estimations of the mixed models are provided in Supplementary Data 13. Source data are provided in Source Data Fig. 4.
