Fig 1.
The map was created using freely obtained shapefiles from the Ethiopian Mapping Agency.
Table 1.
Summary of plot inventory data from studied plantation and natural forests.
Table 2.
Distribution of harvested trees across tree size used in biomass estimation models.
Table 3.
Model forms used to predict tree biomass.
Table 4.
Tree species sampled and their mean of dendrometric properties for biomass modelling.
Table 5.
Multispecies models developed in this study using different combinations of tree variables.
Table 6.
Variable coefficients and validation metrics of cross-validated models with training set.
Table 7.
Predictive accuracy of the viable multispecies models on the test datasets.
Fig 2.
Relationships between predicted and observed aboveground biomass of the harvested trees.
The labelled M2-PF (Fig 2a) and M2-NF (Fig 2d) refer to the best-performing AGB models developed in this study for plantation and natural forests, respectively. Pan-tropical models are represented using the first authors’ name: Brown = Brown [18]; Chave = Chave et al. [17]. In Fig 2a–c, the text “PF” indicates the AGB of harvested sample trees (n = 69) from Wef-Washa plantation forest used in testing model performance, whilst “NF” in Fig 2d–f signifies those from the natural forest (n = 58). rMPE (%) is the relative mean prediction error produced in the estimation of AGB. The diagonal lines show a 1:1 relation. Positive and negative rMPE values indicate under- and overestimation of the AGB, respectively. The p-value for rMPE was significantly different from zero at the 95% confidence interval when P < 0.05.
Fig 3.
Bias in the AGB estimates of individual plots resulted from the aggregation of species in multispecies models.
The x-axis labels M1–7 correspond to multispecies models developed in this study using the dataset of plantation forest (Group I) as shown in Table 5. The remaining letters represent pan-tropical models: BR = Brown [18] and CH = Chave et al. [17]. RE (%) refers to relative errors in %. Orange dots represent AGB estimates from multispecies models that fell within the 95% confidence intervals of species-specific model estimates. Whereas, black dots indicate models whose estimates lie outside this confidence interval.
Table 8.
Bias in across-plot average AGB estimates using data from 40 plots of plantation forest.
Fig 4.
The mean relative error in plot-level AGB estimates resulted from the aggregation of species multispecies models.
Mean relative errors of the plot-level AGB estimates are represented by black dots; error bars show the 95% confidence interval of the mean. The x-axis labels M1–8 correspond to multispecies models developed in this study using the dataset of plantation forest. The remaining letters represent pan-tropical models: Br = Brown [18] and CH = Chave et al. [17].