Table 1.

Coefficient of determination (R²), Akaike Information Criterion (AIC_c), difference in AIC_c value between the respective model and the ´best´ model (ΔAIC_c), root mean square error (RMSE) and Moran´s I of linear (lm) and linear mixed effects models (lme) used to predict stand structural complexity index (SSCI) based on climate and soil variables.

Model	Predictor variables	r²	AICc	Δ AICc	RMSE	Moran´s I
lm	MAP + prec. seasonality	0.89	48.02	0.00	0.62	−0.006
lm	MAP + temp. during growing season	0.83	57.79	9.77	0.79	−0.051
lm	MAP-PET + MAT	0.81	60.22	12.20	0.84	−0.087
lm	Field capacity + nitrogen + MAP	0.83	61.28	13.26	0.78	−0.052
lm	Prec. seasonality	0.77	63.19	15.17	0.89	−0.116
lm	MAP + field capacity	0.77	63.49	15.47	0.91	−0.056
lm	Prec. seasonality + field capacity	0.77	63.67	15.65	0.91	−0.084
lm	MAP-PET	0.72	64.36	16.34	1.00	−0.150
lm	MAP + Solar radiation	0.76	64.55	16.53	0.93	−0.095
lm	Field capacity + nitrogen	0.74	66.19	18.17	0.96	0.026
lme	Field capacity + nitrogen	0.78	68.15	20.13	0.74	0.026
lm	MAP	0.66	68.18	20.16	1.10	−0.070
lm	Field capacity	0.61	71.10	23.08	1.18	−0.011
lme	Field capacity	0.74	71.66	23.64	0.82	−0.011
lme	Prec. Seasonality + field capacity	0.76	72.16	24.14	0.89	−0.084
lme	MAP	0.85	74.82	26.80	0.54	−0.070
lme	MAP + field capacity	0.86	75.16	27.14	0.57	−0.056
lme	Prec. seasonality	0.75	76.15	28.13	0.80	−0.116
lme	MAP-PET	0.83	76.88	28.86	0.61	−0.150
lme	MAP + prec. seasonality	0.89	77.19	29.17	0.56	−0.006
lm	Nitrogen	0.45	77.99	29.97	1.41	−0.064
lme	Field capacity + nitrogen + MAP	0.85	78.15	30.13	0.59	−0.052
lme	MAP-PET + temp. during growing season	0.89	78.41	30.39	0.49	−0.051
lm	Solar radiation + nitrogen	0.45	81.14	33.12	1.41	−0.064
lme	MAP-PET + MAT	0.56	81.90	33.88	0.55	−0.087
lm	Solar radiation	0.29	83.10	35.08	1.60	−0.022
lme	Solar radiation	0.84	86.49	38.47	0.69	−0.022
lme	Solar radiation + nitrogen	0.84	89.59	41.57	0.69	−0.064

MAP mean annual precipitation, PET potential evapotranspiration, MAP–PET water balance (mm), MAT mean annual temperature.

Field capacity and soil nitrogen content were log-transformed. Biome was used as random effect in linear mixed effects models. Each model was significant at p < 0.01.