Fig 1.
A: Boxplots of the distribution of individuals for each species, highlighting the median value. B: Scatter plot of the mean vs. variance for individuals by species, and regression line to check how they fit Taylor’s Law.
Table 1.
Feature importance for the Random Forest model with the ABIOTIC set of variables.
Table 2.
Prediction errors for spatial application.
Fig 2.
Prediction errors with a two-step Random Forest Regressor.
A: Relative Squared Error distributions for 100 random choices of training/testing sets, vertical lines set at median values. B: Root Mean Square Error distributions for the same collection of predictors.
Fig 3.
Prediction errors by species using a two-step Random Forest Regressor.
A: Relative Squared Error distributions for 100 random choices of training/testing sets. B: Root Mean Square Error distributions for the same collection of predictors. See Table C in S1 Text for species acronyms.
Fig 4.
Prediction errors by individuals.
Each dot is the value of where y is the recorded value of abundance and
the regression prediction. There are 37260 predictions for each run. A: Error values for a run of the two-step model with Random Forest. B: Error values for a run of the abiotic model with Random Forest.
Table 3.
Prediction errors splitting by year and using Random Forest.