Tropical secondary forests (SFs) cover large extents of human modified landscapes and can help restore ecosystems at low costs and high ecological effectiveness. Recovery rates, however, vary with environmental conditions and are negatively affected by anthropogenic impacts. Hence, to be able to assess and monitor the ecological integrity of SF, it is important to identify indicators that have a consistent response to degradation (are predictable) and can be applied across a region (are generalizable).

In this study we aimed (i) to identify the relative importance of environmental conditions and anthropogenic impacts for SF structure, diversity and function, and (ii) to identify the best indicators for assessing and monitoring SF across the Amazon region. We consider as best indicators those which variation is mainly described by anthropogenic impacts and are weakly affected by site and environmental conditions.

We used information from 135 vegetation plots from 12 sites across the Brazilian Amazon, with ages ranging from one to 35 years after abandonment. We fitted general linear mixed models to select (based on the lowest AIC) the best models for each indicator of vegetation structure (basal area, maximum DBH, Gini index, and stem density), diversity (Hill numbers q0, q1, q2), and function (aboveground biomass). We assess the fixed effects of the following drivers: land use and cover (forest cover, fire frequency, number of clearances, land use type: pasture, shifting cultivation and “mix”), soil and topography (CEC, CFV, SOC, N, pH, silt, clay, sand, HAND), and climatic conditions (MAP, MAT, CWD, dry season temperature, dry season precipitation). Site was included as a random factor.

The selected best models had marginal R2 higher than 0.53 (apart from Hill q0 with R2=0.24), and R2 of the random effects between 0.06 and 0.14. All models were significantly and strongly affected by age (average standardized effect size of 0.63±0.23), followed by anthropogenic impact (0.32±0.10) and environmental factors (0.19±0.08).

We ranked the models in descending order of (i) lowest dependency on site location (represented by the variation explained by random effects), (ii) highest effect size of anthropogenic impact variables and (iii) lowest effect size of environmental variables. Accordingly, the most promising indicators were the Gini index of structural heterogeneity and the species diversity hill numbers q1 and q2. Absolute values of these indicators can be derived for each successional stage (initial, intermediate and advanced) and used to assess the ecological integrity of SF across the Amazon region