Table 2 Comparison of trait map performance with existing products

Author	Resolution (km)	SLA	Leaf N (mass)	Leaf N (area)
This study (COMB)	1	0.63	0.56	0.63
	22	0.65	0.62	0.68
	55	0.63	0.59	0.65
	111	0.59	0.54	0.60
	222	0.52	0.46	0.59
This study (CIT)	1	0.53	0.49	0.55
	22	0.45	0.49	0.53
	55	0.44	0.47	0.54
	111	0.42	0.44	0.52
	222	0.41	0.37	0.51
van Bodegom et al.⁸³	55	0.33	-	-
	111	0.23	-	-
	222	0.24	-	-
Boonman et al.⁸⁰	55	0.43	0.11	0.44
	111	0.37	0.08	0.42
	222	0.37	0.12	0.41
Butler et al.⁵⁴	55	0.29	0.20	0.39
	111	0.36	0.29	0.37
	222	0.29	0.25	0.33
Madani et al.⁸¹	55	0.10	-	-
	111	0.25	-	-
	222	0.25	-	-
Moreno et al.⁴⁵	1	0.38	0.26	-
	22	0.38	0.31	-
	55	0.38	0.12	-
	111	0.40	0.21	-
	222	0.44	0.21	-
Schiller et al.⁴⁹	55	0.47	0.38	0.50
	111	0.38	0.35	0.46
	222	0.39	0.31	0.49
Vallicrosa et al.⁸²	1	-	0.29	-
	22	-	0.37	-
	55	-	0.20	-
	111	-	0.33	-
	222	-	0.30	-
Wolf et al.²⁸	22	0.41	0.31	0.38
	55	0.42	0.27	0.41
	111	0.32	0.29	0.34
	222	0.31	0.36	0.37

Pearson’s correlation coefficient r of each trait map in relation to spatially-independent gridded sPlot community-weighted means (CWM) unused in model training at different resolutions (1–222 km²). Bold values represent the highest correlation for each trait-resolution pair, while italicized values represent the second-highest. Models incorporating a combination of citizen science and vegetation survey data (COMB) had the strongest correspondence for all traits and all resolutions, and models incorporating citizen science data only (CIT) exhibited the second-strongest correlations for most trait-resolution pairs. SCI models marginally outperformed COMB models on average but demonstrated consistently lower mean coefficients of variation and are therefore not shown here (Fig. 4a, b and Table S2). All trait values were transformed using a Yeo-Johnson power transform with the same parameters determined in the original transformation of the sPlot CWMs.

Quick links

Search