Fig. 3: The sensible heat difference between forests and non-forest and its relationship with cloud effects.

a Potential effects of forests on cloud cover from MODIS and MSG data (duplicates from Fig. 1a, b). b The sensible heat differences between forests and non-forest (ΔH) estimated from satellite data⁴, Community Land Model (CLM), and c Twenty-eight paired forest and non-forest flux sites (including one Amazon pair not shown on the map, see Supplementary Table 1). The three circles marked in a and b denote the locations of cloud inhibition which correspond to the negative ΔH in Amazon, Central Africa, and Southeast US. The connection lines with a dot in panel c indicate the location of flux tower clusters where multiple flux pairs are close in distance. d The relationship between sensible heat (ΔH) and cloud differences between forest and non-forest (ΔCloud) at paired flux towers. The cloud effects at paired flux site locations are extracted from ΔCloud aggregated to 1° based on MODIS data. The line is fitted by geometric mean regression⁸⁰. The Spearman’s correlation coefficient (ρ), which is a nonparametric measure of rank correlation, is calculated by the spearmanr function of scipy.stats module in Python with its p value (p) determined by a two-tail t-test.