Fig. 2: The effect of cryogenic land surface processes (LSP) in the biomass models.
From: Cryogenic land surface processes shape vegetation biomass patterns in northern European tundra
a The bar plots show the effect sizes of the observed LSP on estimated AGB after controlling for the effects of other environmental predictors. The lines indicate 95% confidence intervals based on boosted regression trees (BRT) modeling over 500 bootstrap samples. Asterisks indicate statistically significant differences between sites with LSP absent and LSP present (P ≤ 0.001, paired one-sided t-test, n = 500). b Prediction error of the AGB models in terms of root mean squared error (RMSE) between observations and each of the model predictions (calibrated first without LSP [Base model] and then with LSP [LSP model]). The boxplots depict median, 1st and 3rd quartiles, and 95% range of variation over 500 cross-validation rounds. Asterisks indicate statistically significant model improvement (P ≤ 0.01, paired one-sided t-test, n = 500). c Predictors’ relative influence in the above-ground biomass (AGB) models. Here, the LSP includes the summed influences of cryoturbation, solifluction and nivation (numerical results are presented in Supplementary Table 2). The black lines indicate the standard deviation over 500 BRT model runs with bootstrap sampling. GDD growing degree days, WAB water balance, Tfeb February mean air temperature, RAD potential incoming solar radiation, RAC residual autocovariate for the LSP model (i.e., RACLSP), TWI topographic wetness index, Peat peat cover.
