Extended Data Fig. 1: Model accuracy assessment and extent of interpolation and extrapolation across all terrestrial pixels in 73 global covariate layers.

a, Coefficient of variation (standard deviation as a fraction of the mean predicted value) as a measure of the prediction accuracy of our model. b, Proportional extent of interpolation (purple) versus extrapolation (red) in the univariate space. c, Percentage of pixels that fall within the convex hulls of the first 11 principal component spaces (collectively covering >80% of the sample space variation). d, Percentage of pixels interpolated as a function of the percentage of global environmental conditions covered by the sample set. On the global scale, 86% of the Earth’s pixels have at least 90% of the covariate bands falling within the sampled range of environmental conditions. e, Percentage of pixels falling within the 55 convex hull spaces of the first 11 principal components (collectively explaining >80% of the variation). On the global scale, 62% of the Earth’s pixels fell within 100% of 55 convex hull spaces. f, Percentage of terrestrial pixels falling within the sampled range, per covariate band.