Removing climbers more than doubles tree growth and biomass in degraded tropical forests

Huge areas of tropical forests are degraded, reducing their biodiversity, carbon, and timber value. The recovery of these degraded forests can be significantly inhibited by climbing plants such as lianas. Removal of super-abundant climbers thus represents a restoration action with huge potential for application across the tropics. While experimental studies largely report positive impacts of climber removal on tree growth and biomass accumulation, the efficacy of climber removal varies widely, with high uncertainty as to where and how to apply the technique. Using meta-analytic techniques, we synthesise results from 26 studies to quantify the efficacy of climber removal for promoting tree growth and biomass accumulation. We find that climber removal increases tree growth by 156% and biomass accumulation by 209% compared to untreated forest, and that efficacy remains for at least 19 years. Extrapolating from these results, climber removal could sequester an additional 32 Gigatons of CO2 over 10 years, at low cost, across regrowth and production forests. Our analysis also revealed that climber removal studies are concentrated in the Neotropics (N=22), relative to Africa (N=2) and Asia (N=2), preventing our study from assessing the influence of region on removal efficacy. While we found some evidence that enhancement of tree growth and AGB accumulation varies across disturbance context and removal method, but not across climate, the number and geographical distribution of studies limits the strength of these conclusions. Climber removal could contribute significantly to reducing global carbon emissions and enhancing the timber and biomass stocks of degraded forests, ultimately protecting them from conversion. However, we urgently need to assess the efficacy of removal outside the Neotropics and consider the potential negative consequences of climber removal under drought conditions and for biodiversity.

Methods
Dataset collected from published articles and associated data. Response data (treatment and control) and variance has then been converted into effect size (standardised mean difference) using metafor and compute.es R packages (Viechtbauer, 2010; Del Re, 2013). Missing variances were imputed using linear relationships between variance, mean growth and sample size. Imputation was run 10 times, generating 10 sets of imputed SD and SS for the tree growth and biomass datasets. Climate variables were calculated from the Climate Research Unit (CRU) (Harris et al., 2020), and elevation was calculated from the International Centre for Tropical Agriculture (CIAT), using the site coordinates with a 1 km buffer (Jarvis et al., 2008). Dataset includes calculations from correction to original article.