Non‐destructive allometric estimates of above‐ground and below‐ground biomass of high‐mountain vegetation in the Andes

Abstract

AbstractAimStudies that monitor high‐mountain vegetation, such as paramo grasslands in the Andes, lack non‐destructive biomass estimation methods. We aimed to develop and apply allometric models for above‐ground, below‐ground and total biomass of paramo plants.LocationThe paramo of southern Colombia between 1°09′N and 077°50′W, at 3,400 and 3,700 m a.s.l.MethodsWe established 61 1‐m2 plots at random locations, excluding disturbed, inaccessible and peat bog areas. We measured heights and basal diameters of all vascular plants in these plots and classified them into seven growth forms. Near each plot, we sampled the biomass from plants of abundant genera, after having measured their height and basal diameter. Hence, we measured the biomass of 476 plants (allometric set). For each growth form we applied power‐law functions to develop allometric models of biomass against basal diameter, height, height x basal diameter and height × basal area. The best models were selected using AICc weights. Using the observed and predicted plant biomass of the allometric set we calculated absolute percentage errors using cross‐validation. The biomass of a plot was estimated by summing the predicted biomass of all plants in a plot. Confidence limits around these sums were calculated by bootstrapping.ResultsFor groups of <20 plants the biomass predictions yielded large (>15%) errors. Applying groups that resembled the 1‐m2 plots in density and composition, the errors for above‐ground and total biomass estimates were <15%. Across all plots, we obtained an above‐ground, below‐ground and total plot biomass of 329 ± 190, 743 ± 486 and 1011 ± 627 g/m2 (mean ± SD), respectively. These values were within the range of biomass estimates obtained destructively in the tropical Andes.ConclusionsIn new applications, if target vegetation samples are similar regarding growth forms and genera to our allometric set, their biomass might be predicted applying our equations, provided they contain at least 50–100 plants. In other situations, we would recommend gathering additional biomass measurements from local plants to evaluate new regression equations.