Aboveground Biomass Models for Common Woody Species of Lowland Forest in Borana Woodland, Southern Ethiopia

Abstract

Aboveground biomass models are useful for assessing vegetation conditions and providing valuable information on the availability of ecosystem goods and services, including carbon stock and forest/rangeland products. This study aimed to develop aboveground biomass estimation models for the common woody species found in Borana woodland. Multispecies and species-specific models for aboveground biomass were developed using 114 destructively sampled trees representing five species. The dendrometric variables selected as predictors of the trees’ aboveground dry biomass for both multispecies and species-specific models were diameter at breast height, tree height, wood basic density (ρ), crown area (ca) and crown diameter (cd). The distribution of biomass across trees’ aboveground components was estimated using destructively sampled trees. Most tree biomass is allocated to branches, followed by the stems. The tree diameter, wood basic density, and crown diameter were significant predictors in generic and species-specific biomass models across all tree components. Incorporating wood basic density into the model significantly improved prediction accuracy, while tree height had a minimal effect on biomass estimation. The stem and twig biomasses were the highest and least predictable plant parts, respectively. Compared with the existing models, our newly developed models significantly reduced prediction errors, reinforcing the importance of location-specific models for accurate biomass estimation. Hence, this study fills the geographic and ecological gaps by developing models tailored with the unique conditions of the Borana lowland forest. The accuracy of species-specific biomass models varied among tree species, indicating the need for species-specific models that account for variations in growth architecture, ecological factors, and bioclimatic conditions.

1. Introduction

The estimation of carbon stock sequestered in each ecosystem is a function of plant biomass and reflects the amount of CO₂ stored in each vegetation [1,2,3]. A reliable estimate of biomass is crucial for monitoring forest conditions and reporting changes in carbon stocks [4]. Biomass estimations are a function of applied methods and models [2]. Biomass can be calculated by the direct method of destructively harvesting all the trees [5] or the indirect method of using an allometric model [3] and a remote sensing approach [6,7]. The direct method of cutting down trees is costly and time-consuming [2], whereas indirect methods using an allometric model are time-efficient and less expensive [1]. Therefore, different allometric models must be developed for species and forest types to obtain accurate biomass and carbon stock estimates in forests [8].

In Ethiopia, some allometric equations were developed for different trees/shrubs and forests in the northern [4,9,10], northwestern [11,12], central [13], and southern [14,15] regions of the country. All the above-mentioned studies were focused on species-specific biomass models, except for a few cases [4,10,11]. Additionally, wood basic density is rarely used in aboveground biomass prediction in Ethiopia, except for a few studies by [4,10].

The specific biomass prediction models provide a proxy estimate of carbon sequestered in each ecosystem, improve the estimates of carbon stocks in the region, and support the implementation of policies and mechanisms designed to mitigate climate change, like REDD+ [16]. The general multispecies pan-tropical models developed by [17] have been used for the aboveground biomass estimation of different forest types in Africa. However, [4] proves the applicability limitation of the model in Ethiopia. Similarly, [10] tested general multispecies biomass models developed in the drylands of northern Ethiopia and reported significant bias and under-predictions.

Most of the previous studies developed biomass estimation models for different areas, raising uncertainty for the application of the models beyond the area [4,14,17,18,19]. The variability in environmental conditions, such as soil type, climate, and vegetation structure across regions, makes it difficult to generalize the applicability of allometric models. Particularly, the unique ecological conditions of the Borana woodland, which are known for low elevation, dry soil, arid climates, erratic rainfall, and diverse tree species, necessitate the need for developing area-tailored allometric models for accurate carbon stock estimation and vegetation monitoring. Existing allometric models developed for highland areas and/or temperate environments may not account for the stressors and adaptive features of the Borana dryland ecosystem. Thus, developing area-specific allometric models for Borana is crucial for improving the reliability of biomass estimates and ensuring better management of the woodland’s carbon stocks.

The previous biomass models developed for Ethiopian drylands have focused primarily on exclosures and grasslands. Notably, [14,18] developed species-specific allometric models for woody species found in the Borana grassland. However, these studies are limited in scope, as they focus primarily on grassland ecosystems, which differ significantly from the dryland afromontane forests. The species composition, environmental conditions, and vegetation structure in grasslands and afromontane forests are different, which makes the applicability of these grassland-based models less reliable for afromontane forest ecosystems. Although belowground biomass is a critical component of total biomass and is important in carbon stock assessments, it can alternatively be estimated using root-to-shoot ratios derived from other similar ecosystems. Therefore, this study was limited to only aboveground biomass due to the complexity and difficulty of excavating root systems in the rocky mountain of a dryland environment. Hence, our study emphasized aboveground biomass models for tree species of dryland afromontane forests in Borana woodland, located in southern Ethiopia.

Accordingly, the main objective of this study was to develop the aboveground allometric biomass equations for the lowland forest in the Borana woodland of Ethiopia. Specifically, the study aimed to (i) compare the patterns of biomass partitioning to aboveground biomass components within and between tree species; (ii) develop multispecies and species-specific allometric biomass models for predicting the stem, branch, twig, and total tree aboveground biomasses; and (iii) evaluate the prediction performances of previously published multispecies biomass estimation models.

2. Materials and Methods

2.1. Site Description

This study was undertaken in the Gamadu forest in the Borana woodland of Oromia Regional State (hereafter Oromia) in southern Ethiopia (Figure 1). The forest is located at 04°06′ north and 038°06′ east. The study site was selected systematically based on the diversity of agroclimatic zones in the forest cover, with the Gamadu forest block in Borana woodland encompassing the vast area and featuring a range of diverse agroclimatic zones [20].

The main soil texture of the Borana area is 53% red sandy loam soil, 30% black clay and volcanic light-colored silty clay, and 17% silt and vertisols [21]. In Borana, the common land use and land cover types are dense and natural indigenous tree species (woodland), areas dominated by indigenous grass species (grassland), areas neither covered by vegetation nor used for crop production (bare land), and the cultivated areas [22]. Vegetation in the Gamadu forest is dominated by drought-adapted woody species, including Juniperus procera Hochst. ex Endl., Pittosporum viridiflorm Sims, Euclea divinorum Hiern, Teclea nobilis Delile, Olea europaea L., Oncoba spinosa Forssk., Pappea capensis Benth., Commiphora africana (A. Rich.) Engl., Vangueria apiculate K. Schum., and Rytigynia neglecta (Hiern) Robyns, covered with native grass species and herbs. The tree canopy of this forest is typically open to moderately closed, with understory vegetation, which is influenced by grazing intensity and seasonal rainfall. The climate of the Borana woodland is arid to semi-arid with a mean annual precipitation of 580 mm [23]. The area has a bimodal rainfall pattern, with the peak rainy season from March to May and small rain from September to November [23]. Based on data from the nearby weather station, the area receives an average monthly rainfall of 4–5 mm between August and September, up to 121 mm in April, and the average monthly temperature is 17–20 °C.

2.2. Data Collection

For measuring standing trees and shrubs (hereafter referred to as ‘trees’), 24 main plots with 10 m × 10 m were designed, following [24]. First, we crossed the forest by using the narrow road that went through the forest from the foot of the mountain to the top edge, which we utilized as the main transect. While walking along this route, we established three starting points representing low, medium, and high forest disturbance levels. The distance between each site was determined based on variations in the visually detected degree of disturbances.

The degree of forest disturbance level was determined visually by observing canopy cover, density of the trees, dominance of tree stumps, and presence of animal dung and paths, as also used by [25]. Hence, the forest location with a high extent of canopy cover, dense tree stands, and fewer stumps and animal paths was taken as the site of a low disturbance level. Similarly, places with moderate canopy cover, medium tree density, and noticeable signs of some tree stumps and animal paths were considered as sites of medium disturbance. The location with sparse canopy cover, low tree density, frequent tree stumps, and clear signs of animal paths and dung observed was taken as a high-disturbance site. This classification approach was also used in previous ecological studies in Ethiopia [25,26,27].

From the starting points of each site, we made a 1 km transect line to the right and left of the road. In this way, we created six transects from the three starting points, three to the left and three to the right of the road. By taking four main plots, each measuring 100 m², along each transect, we made 24 main plots, each spaced by 200–300 m based on access to gully passages (Figure 2).

Within each main plot, the diameter at stump height (dsh) of all trees was measured at 0.3 m above the ground. Trees with a dsh ≥ 2.5 cm were counted, their species identified, their diameter at breast height (dbh) measured at 1.3 m, and their basal area calculated. Accordingly, the type of plant species included in the destructive sample for developing allometric biomass equations was determined based on the total basal area per hectare of respective species and the relative dominance of the species.

2.2.1. Tree Sampling

The individual tree species in the destructive sample for developing biomass models were selected based on the total basal area per hectare. As the basal area is the circular cross-sectional area of a tree stem at breast height, this was determined from the circle area, which is πr² [28]. In forest inventory, we usually measure the diameter of a tree, rather than the radius [29], using dbh/2 instead of radius.

$$BA=\pi { \left({\displaystyle \frac{dbh}{2}}\right)}^2$$

where BA = basal area (m²), dbh = diameter at breast height, and π (pi) = 3.14.

The basal area is calculated at the stand level as the total of the values for each live tree, usually expressed as square meters per hectare and commonly measured in centimeters.

$$BA \left(\mathrm{i}\mathrm{n} {\mathrm{c}\mathrm{m}}^2\right)=3.14\times {\displaystyle \frac{{dbh}^2}{4}} \mathrm{and} BA \left(\mathrm{i}\mathrm{n} {\mathrm{m}}^2\right)=0.785\times {\displaystyle \frac{{dbh}^2}{10000}}=0.00007854{dbh}^2$$

The total basal area coverage for each species recorded in the forest was considered to select species to be included in developing allometric biomass models for estimating the biomass of the given forest. The top ten plant species with the highest basal area per ha were Juniperus procera Endl., Pittosporum viridiflorm, Euclea divinorum Hiern., Teclea nobilis Del., Olea europaea Subsp. cuspidata (Wall. Ex G. Don) Cif., Oncoba spinosa Forsk., Pappea capensis, Commiphora africana (A. Rich) Engl., Vangueria apiculate, and Rytigynia neglecta (Hiern) Robyns (

Table A1. Summary of tree species characteristics in the study area.

S/N	Scientific Name	Family Name	Number of Trees per ha	Mean dsh	Mean dbh	Total BA per ha	Mean BA per Tree
1	Juniperus procera Endl.	Cupressaceae	188	41.76	34.82	27.454	0.146
2	Pittosporum viridiflorm	Pittosporaceae	163	31.51	24.55	9.213	0.0565
3	Euclea divinorum Hiern.	Ebenaceae	933	6.98	5.75	2.902	0.0031
4	Teclea nobilis Del.	Rutaceae	750	6.85	6.08	2.887	0.0038
5	Olea europaea Subsp.	Oleaceae	200	13.45	9.81	1.879	0.0094
6	Oncoba spinosa Forsk.	Flacourtiaceae	258	9.35	8.38	1.799	0.007
7	Pappea capensis	Sapindaceae	129	17.80	11.29	1.594	0.0124
8	Commiphora africana (A. Rich) Engl.	Burseraceae	167	6.23	6.27	0.626	0.0037
9	Vangueria apiculata	Rubiaceae	104	7.00	6.42	0.386	0.0037
10	Rytigynia neglecta (Hiern) Robyns	Rubiaceae	133	6.63	5.19	0.358	0.0027
11	Hiddi qaalluu	Hiddi qaalluu	104	7.10	4.51	0.219	0.0021
12	Maytenus senegalensis (Lam.) Exell	Celastraceae	100	5.11	4.09	0.156	0.0016
13	Ruttya fruticosa Lindau	Acanthaceae	67	4.75	3.71	0.085	0.0013
14	Commiphora kua (R.Br.ex Royle) Vollesen	Burseraceae	71	4.70	3.49	0.082	0.0012
15	Clausena anisata (Willd.) Benth.	Rutaceae	54	3.95	3.58	0.059	0.0011
16	Buuxxee	Buuxxee	21	5.26	5.66	0.055	0.0026
17	Buddleja polystachya	Scrophulariaceae	79	3.14	2.58	0.046	0.0006
18	Acokanthera schimperi (A. DC.) Schweinf	Apocynaceae	33	4.01	3.26	0.031	0.0009
19	Secamone punctulata Decne.	Asclepiadaceae	42	3.73	2.88	0.030	0.0007
20	Carissa edulis Vahl	Apocynaceae	29	3.63	3.01	0.023	0.0008
Total			3625

dsh: diameter at stump height, dbh: diameter at breast height; Basal Area (BA): Cross-sectional area of a tree trunk measured at a specific height (m²). Italicized species names follow scientific nomenclature conventions.

in Appendix A). However, the large plant species with a mean basal area per tree (BA/tree) greater than 0.01 m² were excluded from destructive sampling due to the prohibition of felling large trees from the studied forests. Hence, despite their higher total basal area per ha, Juniperus procera, Pittosporum viridiflorm, and Pappea capensis were excluded. Additionally, these species with big trees have a higher demand for wood products and are mostly cut by the local community.

On the other hand, plant species with a total basal area of less than 0.5 m²/ha were also excluded from the sample, because such tree species hold a small fraction of aboveground biomass in forests. Accordingly, five tree species, including Euclea divinorum, Teclea nobilis, Olea europaea, Oncoba spinosa, and Commiphora africana, were included in developing the allometric model.

Among the destructively sampled species, Olea europaea had the highest tree diameter (dbh = 17 cm, dsh = 28.5 cm), wood basic density (ρ = 1.24 g/cm³), and total aboveground biomass (Tagb = 120.89 kg) (

Table A2. Dendrometric variables and dry biomass of trees.

Species	Statistic	dsh (cm)	dbh (cm)	th (m)	cd (m)	ca (m2)	ρ (g/cm3)	Sb	Bb	Tb	Tagb
Commiphora africana (A. Rich) Engl.	Mean	6.13	6.17	3.28	1.89	3.92	0.37	2.1	3.03	1.35	6.48
	Sd	2.5	2.94	1.47	1.21	4.77	0.06	1.87	3.53	1.09	6.25
	Min	2.5	2.0	1.65	0.25	0.05	0.28	0.33	0.30	0.25	0.88
	Max	11.2	13.0	6.2	4.4	15.2	0.52	6.73	16.4	3.97	26.45
Euclea divinorum Hiern.	Mean	7.33	6.05	4.38	2.15	5.51	0.75	4.52	5.64	1.80	11.95
	Sd	2.82	2.72	1.12	1.58	8.53	0.07	1.91	2.01	0.62	4.36
	Min	2.7	1.7	2.55	0.55	0.24	0.62	1.50	1.95	0.27	3.72
	Max	15.0	14.0	6.85	6.4	32.15	0.85	8.91	11.08	2.91	22.27
Olea europaea Subsp. Cuspidata	Mean	12.25	10.68	4.72	4.06	12.73	0.97	16.17	30.28	8.63	55.38
	Sd	5.78	4.63	1.64	0.87	4.59	0.16	10.95	24.8	6.36	39.62
	Min	2.5	2.3	1.14	2.35	4.34	0.71	1.65	1.52	0.76	3.92
	Max	28.5	17.0	7.2	5.2	18.85	1.24	40.14	71.08	21.17	120.89
Oncoba spinosa Forsk.	Mean	9.48	8.37	4.92	3.25	10.62	0.65	13.55	20.42	6.50	40.47
	Sd	5.43	4.55	1.87	1.78	9.18	0.05	7.16	10.11	4.01	20.26
	Min	3.0	2.3	1.65	0.9	0.64	0.56	4.02	7.85	1.92	17.97
	Max	17.0	16.0	7.4	5.55	24.18	0.72	23.22	35.91	14.93	72.89
Teclea nobilis Del.	Mean	8.1	7.73	5.44	2.84	7.34	0.82	10.93	15.72	7.85	34.5
	Sd	3.56	4.16	1.78	1.17	5.34	0.05	5.14	8.30	3.80	16.70
	Min	2.5	2.0	2.85	0.7	0.38	0.75	3.0	5.04	1.90	9.93
	Max	15.5	15.0	8.45	5.05	20.02	0.9	21.06	32.33	15.13	68.52
Aggregate of Five Species	Mean	8.32	7.49	4.49	2.7	7.43	0.7	8.47	13.2	4.7	26.4
	Sd	4.37	4.0	1.69	1.54	7.39	0.21	7.74	14.77	4.61	26.15
	Min	2.5	1.7	1.14	0.25	0.05	0.28	0.33	0.3	0.25	0.88
	Max	28.5	17.0	8.45	6.4	32.15	1.24	40.14	71.08	21.17	120.89

dsh: Diameter at stump height (cm); dbh: Diameter at breast height (cm); th: Tree height (m); cd: Crown diameter (m); ca: Crown area (m²); ρ: Wood basic density (g/cm³); Sb: Dry biomass of stem; Bb: Dry biomass of branches; Tb: Dry biomass of twig (leaves plus small branches); Tagb: Total aboveground dry biomass of tree.

in Appendix A). On the other hand, the tallest recorded tree is from the Teclea nobilis species, with 8.45 m, and Euclea divinorum is a species with the widest crown area among the samples.

2.2.2. Tree Measurements

Following the identification of tree species to be included in the sample and determination of the sample size, the diameter at stump height, diameter at breast height, tree height, and crown diameters of the selected individual trees were measured. The crown diameter was used to calculate the crown area of the trees [10].

$$ca=\pi {\displaystyle \frac{cd1 \times cd2}{4}}$$

where ca is the crown area, cd₁ is the widest crown diameter, and cd₂ is its perpendicular crown diameter.

For trees with many stems from the same root at a height of below 1.3 m, the dbh was calculated [30,31].

$${\mathit{dbh}}_c=\sqrt{{\displaystyle {\sum }_{i=1}^n}\left({\left({dbh}_i\right)}^2\right)}$$

where dbh_c is the computed diameter at breast height, dbh_i is the diameter at breast height of individual stems extending from the same root, and i, is the number of stems measured.

The selected trees were cut down by chainsaw and subdivided into stems, branches, and twigs with leaves. Leaves, which might have dropped down during cutting, were carefully collected manually and weighed with the twig component of the trees. All components were immediately weighed separately in the field using a weight balance scale, and the fresh weight of each section was recorded.

Sub-samples were collected from each divided tree component of the harvested plants and weighed for their fresh weight using a digital balance with an accuracy of 0.01 g. The dry weight was determined in the laboratory. Accordingly, from the stem, branch, and twig-leaves components, the sample biomass of 300 g fresh weight was taken by three replications. Stem subsamples were taken at 25%, 50%, and 75% of the total stem length. Similarly, the subsample was taken from large, medium, and small branch sizes. A subsample of the twig-leaves component was taken from a homogeneous mixed sample of the part. An average of three subsamples was used for dry weight determination.

The stem and branch samples were oven-dried at 105 °C, and twig-leaves samples at 70 °C for 48 h, and weighed in the laboratory. The moisture content percentage and dry weight of the stem, branch, and twig leaves were determined. The ratio of sub-sample dry weight to fresh weight was used to convert component fresh weight to dry weight. The total dry weight of each biomass component of each tree was calculated as follows:

$$\begin{array}{c}\mathrm{Total} \mathrm{d}\mathrm{r}\mathrm{y} \mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{e}\mathrm{a}\mathrm{c}\mathrm{h} \mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{n}\mathrm{t}\hfill \\ ={\displaystyle \frac{\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e} \mathrm{d}\mathrm{r}\mathrm{y} \mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{n}\mathrm{t}}{\mathrm{s}\mathrm{u}\mathrm{b}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e} \mathrm{f}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{h} \mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{n}\mathrm{t}}}\times \mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l} \mathrm{f}\mathrm{r}\mathrm{e}\mathrm{s}\mathrm{h} \mathrm{w}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{o}\mathrm{f} \mathrm{c}\mathrm{o}\mathrm{m}\mathrm{p}\mathrm{o}\mathrm{n}\mathrm{e}\mathrm{n}\mathrm{t}\hfill \end{array}$$

The total dry biomass for each sampled plant was obtained by adding the dry biomass of the stem, branches, and twig leaves.

2.3. Data Analysis

A nonlinear regression model was used to develop the biomass allometric model. The dependent variable was tree biomass. For predicting tree biomass, diameter at stump height (dsh at 30 cm), diameter at breast height (dbh at 1.30 m), tree height (th), crown area (ca), crown diameter (cd), and wood basic density (ρ) were used. The relationships between biomass and the predictor variables were determined using Pearson’s correlation before modeling. The correlation matrix plot and correlation coefficients were used to determine the linear relationship of aboveground biomass with predictor variables. Accordingly, different candidate biomass models were generated and evaluated to predict the total aboveground biomass, stem biomass, branch biomass, and twig biomass. All statistical analyses were performed using STATA software version 14.

2.3.1. Linearity Test

The relationships between biomass and each independent variable were examined using a scatterplot. This helps to obtain an insight into the nature of candidate models to be examined for fitness. The linearity test indicates (Figure 3) a nonlinear relationship between aboveground biomass and dendrometric variables.

2.3.2. Allometric Equations Development

Following the demonstrated nonlinear relationship between the dry biomass of the trees and the dendrometric variables, a nonlinear exponential form was used to find the best model fit for the data. Accordingly, the nonlinear version of this power function model, which is frequently employed in allometric modeling studies [10,11,32,33,34,35] was fitted using a logarithmic transformation of the biomass and tree biometric data.

$$ln\left(Y\right)=\alpha +ln\left(X\right){\beta }^{\prime }$$

where Y is the dry biomass, X is the vector of predictor dendrometric variables, α is the constant, and β is the vector of regression parameters to be estimated.

Total aboveground dry biomasses (i.e., Tagb—total aboveground biomass, Sb—stem biomass, Bb—branch biomass, and Tb—twig biomass) and the probable independent variables were correlated using Pearson’s formula. The tree aboveground dry biomass was significantly correlated with dsh, dbh, th, cd, ca, and ρ (

Table A3. Correlation between the dry biomass of trees and dendrometry variable of trees.

	Tagb	Sb	Bb	Tb
dsh	0.7399 *	0.7419 *	0.7224 *	0.634 *
dbh	0.8208 *	0.8247 *	0.7938 *	0.7349 *
th	0.5106 *	0.532 *	0.4562 *	0.5371 *
cd	0.5991 *	0.5881 *	0.5565 *	0.6209 *
ca	0.5607 *	0.5466 *	0.5261 *	0.5731 *
wbd	0.6288 *	0.623 *	0.6092 *	0.5633 *

Note: Tagb—Total aboveground dry biomass of tree; Sb—dry biomass of stem; Bb—dry biomass of branches; Tb—dry biomass of twig (leaves plus small branches); dbh—diameter at breast height (at 1.3 m); dsh—diameter at stump height (at 0.3 m); ca—crown area; cd—crown diameter; th—tree height; wbd—wood basic density of the stem; and *—the correlation coefficient statistically significant at p < 0.001.

in Appendix A).

Tree diameter at breast height had a strong positive and significant correlation with total aboveground dry biomass. This result agrees with previous studies [8,10,17,33,36]. Wood basic density possesses a moderate positive correlation with total aboveground biomass. This result is consistent with several studies [10,37], but contrasts with [17,35,36], who claim that wood density has no significant correlation with the dry biomass of trees. The dsh and ca were eliminated from the model to reduce the issue of multicollinearity.

Considering its strong link with aboveground biomass [8,17,36], different candidate model forms were fitted using dbh as the single predictor and combined with a stepwise inclusion of height, cd, and ρ.

Accordingly, in developing allometric equations, we have divided the model forms into four groups (A, B, C, D) (

Table A4. Model groups.

GROUP A	GROUP B
MA1: ln(Tagb) = α + β1ln(dbh) MA2: ln(Tagb) = α + β1ln(dbh) + β2ln(th) MA3: ln(Tagb) = α + β1ln(dbh) + β2ln(cd) MA4: ln(Tagb) = α + β1ln(dbh) + β2ln(wbd) MA5: ln(Tagb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) MA6: ln(Tagb) = α + β1ln(dbh) + β2ln(th) + β3ln(wbd) MA7: ln(Tagb) = α + β1ln(dbh) + β2ln(cd) + β3ln(wbd) MA8: ln(Tagb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) + β4ln(wbd)	MB1: ln(Sb) = α + β1ln(dbh) MB2: ln(Sb) = α + β1ln(dbh) + β2ln(th) MB3: ln(Sb) = α + β1ln(dbh) + β2ln(cd) MB4: ln(Sb) = α + β1ln(dbh) + β2ln(wbd) MB5: ln(Sb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) MB6: ln(Sb) = α + β1ln(dbh) + β2ln(th) + β3ln(wbd) MB7: ln(Sb) = α + β1ln(dbh) + β2ln(cd) + β3ln(wbd) MB8: ln(Sb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) + β4ln(wbd)
GROUP C	GROUP D
MC1: ln(Bb) = α + β1ln(dbh) MC2: ln(Bb) = α + β1ln(dbh) + β2ln(th) MC3: ln(Bb) = α + β1ln(dbh) + β2ln(cd) MC4: ln(Bb) = α + β1ln(dbh) + β2ln(wbd) MC5: ln(Bb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) MC6: ln(Bb) = α + β1ln(dbh) + β2ln(th) + β3ln(wbd) MC7: ln(Bb) = α + β1ln(dbh) + β2ln(cd) + β3ln(wbd) MC8: ln(Bb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) + β4ln(wbd)	MD1: ln(Tb) = α + β1ln(dbh) MD2: ln(Tb) = α + β1ln(dbh) + β2ln(th) MD3: ln(Tb) = α + β1ln(dbh) + β2ln(cd) MD4: ln(Tb) = α + β1ln(dbh) + β2ln(wbd) MD5: ln(Tb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) MD6: ln(Tb) = α + β1ln(dbh) + β2ln(th) + β3ln(wbd) MD7: ln(Tb) = α + β1ln(dbh) + β2ln(cd) + β3ln(wbd) MD8: ln(Tb) = α + β1ln(dbh) + β2ln(th) + β3ln(cd) + β4ln(wbd)

Tagb—Total aboveground dry biomass of tree; Sb—dry biomass of stem; Bb—dry biomass of branches; Tb—dry biomass of twig (leaves plus small branches); dbh—diameter at breast height; th—tree height; cd—crown diameter; wbd—wood basic density. α is a constant of the model while β₁, β₂, β₃ and β₄ are parameters (regression coefficients) of the model. Tagb = Sb + Bb + Tb.

in Appendix A). Group A represents models tested for total aboveground biomass, and groups B, C, and D include models for the biomass of tree compartments, specifically stem, branch, and twig, respectively. A total of 32 model forms, with 8 models in each group, were examined for the accuracy of dry biomass estimation.

The goodness of fit for the biomass equations was analyzed using linear regression models. The post-estimation goodness of fit of each candidate model was assessed using the p-value, adjusted coefficient of determination ($R_{adjusted}^2$), Akaike’s information criterion (AIC), mean prediction error, and relative prediction bias (%).

Adjusted R-squared is an adjustment for the coefficient of determination that considers the number of variables in a data set and is computed as follows:

$$R_{adjusted}^2=1-\left[\left({\displaystyle \frac{\left({1-R}^2\right)\times \left(n-1\right)}{n-p-1}}\right)\right]$$

The Akaike information criterion (AIC) is also an estimator of prediction errors to determine the relative quality of statistical models for given data [38]. AIC is the expected relative distance between the better-fitted and the other candidate model. The general formula to determine AIC is as follows:

AIC = −2lnL + 2k

where lnL = maximized log-likelihood and k = number of parameters estimated.

$$MPE={\displaystyle \frac{1}{n}} \sum _{i=1}^n|Observed biomass-Predicted biomass|$$

$$rBias(\%)={\displaystyle \frac{MPE}{Mean of Observed biomass}}\times 100$$

Accordingly, the model with the best prediction was selected among candidate models based on significant coefficients (p ≤ 0.05), higher adjusted R², and smaller AIC, MPE, and rBias.

Additionally, as the biomass estimation was made by reversing the logarithmical values into the original units, the transformation might cause a systematic bias. Hence, the bias was adjusted by a correction factor to modify the underestimated biomass. Accordingly, the CF was computed for each selected model using the respective mean square error of the regression.

This CF was used when the back transformation of data to power biomass function was required, at which the calculated CF of the model was then multiplied by the biomass estimated by the model. Hence, the linear model of the form [ln(Y) = α + βln(X)] was then back transformed to a power function in its actual application to determine biomass (Y), as follows:

y = Exp [α + β₁₂₃ln(x₁₂₃)] × CF

3. Results

3.1. Distribution of Biomass Along the Tree Compartments

The proportion of tree biomass across the components shows that branches accounted for the largest aboveground biomass, followed by the stem and twigs for the five destructively sampled species (

Table 1. Biomass distribution (mean, range) of selected tree species.

Plant Species	n	Stem Biomass	Branch Biomass	Twig Biomass
Commiphora africana (A. Rich) Engl.	25	(2.1, 0.3–6.7)	(3, 0.3–16.4)	(1.4, 0.3–4)
Euclea divinorum Hiern.	31	(4.5, 1.5–8.9)	(5.6, 1.9–11)	(1.8, 0.3–2.9)
Olea europaea Subsp. cuspidata	18	(16.2, 1.7–40)	(30.3, 1.5–71.1)	(8.6, 0.8–21.2)
Oncoba spinosa Forsk.	17	(13.6, 4–23.2)	(20.4, 7.8–35.9)	(6.5, 1.9–14.9)
Teclea nobilis Del.	23	(10.9, 3–21.1)	(15.2, 5–32.3)	(7.9, 1.9–15.1)

Note: n: Sample size. The mean biomass values represent the proportion of total biomass distributed among stem, branch, and twig components. The range indicates variability in biomass allocation among sampled trees.

). Biomass distribution within components across different tree species was shown by ranges. Hence, among the tree components, the branch biomass was found to have the most widely distributed biomass allocation between species. This indicates that the branches are the main plant parts accountable for biomass variability across different plant species. Similarly, stems are found to have the second-highest biomass variability among plant components across species, while twigs have the smallest biomass variability across the species.

3.2. Biomass Models

A comparison of the overall goodness of fit of biomass models for different biomass components shows that the model fitted for stem and twig biomasses had the best and weakest relative values of prediction performance indicators. The model, which combines wood basic density, diameter at stump height, and crown diameter, demonstrated the best prediction performance for estimating total aboveground, stem, and branch biomasses (

Table 2. Model expressions and parameter estimates for biomass prediction.

Model	Model Expression	n	α	β1	β2	β3	CF	adj. R2	MPE	rBias (%)	AIC
MA7	AGB = Exp [α + β1ln(dbh) + β2ln(cd) + β3ln(ρ)] × CF	114	1.703	0.790	0.300	1.560	1.136	0.78	−1.01	3.82	172.0
MB7	Sb = Exp [α + β1ln(dbh) + β2ln(cd) + β3ln(ρ)] × CF	114	0.663	0.764	0.294	1.548	1.111	0.81	−0.38	4.47	150.6
MC7	Bb = Exp [α + β1ln(dbh) + β2ln(cd) + β3ln(ρ)] ×CF	114	0.897	0.871	0.225	1.692	1.181	0.75	−0.44	3.33	202.1
MD5	Tb = Exp [α + β1ln(dbh) + β2ln(cd) + β3ln(ρ)] × CF	114	−0.160	0.643	0.469	1.216	1.274	0.57	−0.19	4.04	244.9

AGB: Aboveground biomass; CF: Correction factor; adj R²: Adjusted R²; MPE: Mean prediction error; rBias: Relative bias; AIC: Akaike information criterion; ln: Natural logarithm; dbh: Diameter at breast height; cd: Crown diameter; ρ: Wood basic density; parameter estimates: α (intercept), β₁, β₂, and β₃ (coefficients).

). Likewise, despite its relatively poor prediction power, the twig’s dry biomass model was predicted by a combination of tree diameter, tree height, and crown diameter.

The coefficients of determination (R²_adjusted) and relative bias of selected multispecies species models for the estimation of tree biomass structure ranged from 0.57 to 0.81 and 3.33% to 4.47%, respectively (Table 2). When the biomass model accuracy among the tree compartments was compared, the stem biomass was the most predictable part, and the twig biomass was the least predictable (

Table 3. Generic biomass prediction models for aboveground Components.

Biomass Component	Model Expression
Total Aboveground Biomass (AGB)	AGB = Exp [1.703 + 0.790ln(dbh) + 0.300ln(cd) + 1.560ln(ρ)] × 1.136
Stem Biomass (Sb)	Sb = Exp [0.663 + 0.764ln(dbh) + 0.294ln(cd) + 1.548ln(ρ)] × 1.111
Branch Biomass (Bb)	Bb = Exp [0.897 + 0.871ln(dbh) + 0.225ln(cd) + 1.692ln(ρ)] × 1.181
Twig Biomass (Tb)	Tb = Exp [−0.160 + 0.643ln(dbh) + 0.469ln(cd) + 1.216ln(ρ)] × 1.195

Note: AGB: Aboveground dry biomass; Sb: Stem dry biomass; Bb: Branch dry biomass; Tb: Twig dry biomass; Exp: Exponential function; ln: Natural logarithm; dbh: Diameter at breast height; cd: Crown diameter; ρ: Wood basic density.

). The coefficient of determination for the models used to determine the biomass of individual species ranged from 0.71 to 0.96 (

Table 4. Biomass prediction models for selected species.

Species	Model	Model Expression	α	β1	β2	CF	R2adj	MPE	rBias (%)	AIC
C. africana	MA3Co	AGB = Exp [α + β1ln(dsh) + β2ln(cd)] × CF	−1.764	1.769	0.465	1.0146	0.964	0.09	1.39	−14.6
E. divinorum	MA1Eu	AGB = Exp [α + β1ln(dsh)] × CF	0.785	0.847	—	1.022	0.716	−0.07	0.58	−6.9
O. europaea	MA1Ol	AGB = Exp [α + β1ln(dbh)] × CF	−0.187	1.694	—	1.097	0.848	−0.38	0.07	1.26
O. spinosa	MA2On	AGB = Exp [α + β1ln(dbh) + β2ln(th)] × CF	1.911	0.436	0.544	1.014	0.897	0.10	0.25	−0.62
T. nobilis	MA3Te	AGB = Exp [α + β1ln(dbh) + β2ln(cd)] × CF	1.779	0.764	0.204	1.012	0.919	−0.18	0.52	−18.2

AGB: Aboveground biomass, CF: Correction factor; R²adj: Adjusted coefficient of determination; MPE: Mean prediction error; rBias: Relative bias; AIC: Akaike information criterion; ln: Natural logarithm; dsh: Diameter at stump height; dbh: Diameter at breast height; cd: Crown diameter; th: Tree height; parameter estimates: α (intercept), β₁ and β₂ (coefficients).

).

3.3. Predictive Performance of Previously Developed Models

We have assessed the prediction power of previously reported general multispecies models using the data collected from our study site (

Table 5. Comparison of allometric models for aboveground biomass estimation.

Model Type	Reference	Model Equation	N	Observed Mean	Predicted Mean	MPE	rBias (%)
Multispecies	This study (MA7)	AGB = Exp [1.703 + 0.79ln(dbh) + 0.3ln(cd) + 1.56ln(ρ)] × 1.136	114	26.41	27.42	−1.01	3.82
Multispecies	[11]	AGB = Exp [−1.726 + 2.030ln(DBH) + 0.616ln(H) + 0.915ln(ρ)] × 1.029	114	26.41	32.28	−5.87	22.22
Multispecies	[10]	AGB = 0.350 × dbh0.864 × ca0.171 × ρ0.485	114	26.41	23.59	2.82	10.69
Multispecies	[15]	AGB = Exp (−1.58 + 1.197ln(dbh) + 0.818ln(th) + 0.321ln(ca)) × 1.08	114	26.41	19.05	7.35	27.84
Multispecies	[4]	AGB = 0.3102 × ds1.51554 × cw0.6453	114	26.41	18.00	8.41	31.84
Multispecies	[39]	AGB = 0.2451 × (DSH2 × H)0.7038	114	26.41	16.38	10.03	37.98
Multispecies	[17]	AGB = 0.0673 × (D2 × H × WD)0.976	114	26.41	18.00	8.41	31.84
Multispecies	[3]	AGB = 0.0763 × DBH2.2046 × H0.4918	114	26.41	20.17	6.24	23.62

N: Sample size; AGB: Aboveground biomass; MPE: Mean prediction error; rBias: Relative bias (%); ln: Natural logarithm; dbh: Diameter at breast height; cd: Crown diameter; ρ: Wood density; DSH: Diameter at stump height; H: Tree height; ca: Crown area; WD: Wood density; cw: Crown width.

). For this purpose, seven generic models were selected, among which five models from the Ethiopia sample trees [4,10,11,15,39], one model from trees of other African forests [3], and one also from a tree dataset of global pantropical forests [17] were selected. Among the four multispecies models of Ethiopian forests, two were the models developed using sample trees from dryland forests [4,10], which were similar to the ecology of our study area. These selected generic multispecies models were tested by comparing the values of their mean prediction error [40] and relative bias (rBias). The result shows that all tested previous multispecies models [3,4,15,17,39] produced high prediction bias on our data and highly underestimated the biomass of the study area, except [10]. It is not surprising that our model fits the tree and shrub species biomass of the study area and was found to have the least prediction error and the best estimate for aboveground biomass of the studied forest.

4. Discussion

4.1. Biomass Partition in Tree Components

The study documents that more than half of the tree biomass was allocated to branches, followed by the stem biomass. This result is consistent with [32], which reported that branches contributed two-thirds of the dry aboveground biomass of trees. Other studies, such as [4,41,42], document that branches account for the largest portion of a tree’s biomass. However, [33,43] reported contrasting results where their findings show that the largest fraction of trees’ aboveground biomass was allocated to the stem as compared with branches and leaves. These differences in the allocation of biomass along tree components might be due to the ecological responses of woody plants to environmental factors [32]. Similarly, the biomass allocation can also be determined by factors like climatic conditions and land-use systems.

In arid or semi-arid regions affected by limited rainfall and recurrent droughts, trees may prioritize branching by increasing their canopy size to enhance water retention through lateral growth. Additionally, in open woodlands, characterized by high disturbance, the allocation of greater biomass in branches than stem is common. This is because the trees pruned by grazing and other human activities remain shorter and expand their canopy as a survival strategy. In contrast, in our study area of pastoral land use system, grazing pressure can potentially make trees develop more lateral growth near the ground, which could lead to more biomass being invested in branches than stem. On the other hand, in temperate dense forests where competition for light is intense, trees often allocate more biomass to stems to achieve greater height and access more sunlight. In such an ecosystem, with limited human disturbance, stem biomass often dominates because of the vertical growth strategy adopted by trees to outcompete neighboring vegetation. This could explain why studies like [33,43] find that the stem often make up the largest fraction of the biomass in more temperate or moist environments.

Regarding the variability of biomass across different species, the highest biomass variability was found within branches as compared with stems and twigs. This could be because the size and number of branches vary significantly across the tree species. Similarly, the result shows that the biomass of twigs is less variable compared with stems and branches because twigs are typically smaller, less complex, and more uniform in structure across tree species.

4.2. Multispecies Biomass Estimation Models

The model combining the three-predictor variables produced the greatest possible model fit for biomass estimation of total biomass and biomass of tree compartments (stem, branch, and twig). For the tree biomass estimations, the model with the combination of tree diameter, wood basic density, and crown diameter produced the greatest possible model fit. This result is similar to that found by [10], where more than 95% of the variability in the aboveground biomass of trees was explained by the tree diameter, wood basic density, and crown area as predictors.

Importantly, tree diameter remains an essential predictor of trees’ aboveground biomass. In line with this, [44] documented that 94%–98% of the variation in the aboveground biomass of major woody species in Ethiopia was explained by diameter as a predictor variable. Other studies, such as [9,13,17,30,35,39,45,46], emphasize the importance of tree diameter in dry biomass estimation.

Moreover, our model demonstrates the importance of adding wood basic density to the biomass model in significantly improving its prediction performance when compared with the inclusion of tree height and crown diameter. However, the addition of tree height to the biomass model was found to be statistically not significant (p < 0.05) for total aboveground biomass and biomass of stem, branch, and twig. For instance, in models 3 and 7 of group A (Supplementary Table S1), with the addition of wood basic density to the model, the adjusted R² for total aboveground biomass increased from 0.58 to 0.78, the AIC decreased from 247.5 to 172, and the MSE decreased from 0.5 to 0.25.

This finding also agrees with that of [47], which states that diameter at breast height is a good predictor for aboveground biomass and that its accuracy increases with the addition of wood basic density and/or crown diameter. Our data demonstrates that the inclusion of total tree height in the model increases the prediction bias of the biomass model and that this variable does not significantly affect the trees’ aboveground biomass. Other studies [10,48] also report that all models that included tree height as a predictor variable had non-significant model parameters. Likewise, [47] documents that the biomass model’s goodness of fit increased by the inclusion of wood basic density data, while no noticeable effect was observed by the addition of height data. More specifically, [49] estimated that adding tree height to the model resulted in a 5% drop in allometric model accuracy. However, this result contrasts with the findings of [11,15,50,51], in which tree height was a preferable predictive variable to wood basic density in estimating aboveground biomass.

On the other hand, [43] concludes that the inclusion of tree height improves the accuracy of the tree biomass model in the wet zone, where tree height is a more important factor in determining biomass. However, for the moist climatic zone, they observed that specific wood density increases model accuracy rather than including tree height. This indicates that, in regions with abundant moisture, the structural characteristics of wood (wood basic density) play a more critical role in biomass estimation than tree height. Additionally, other studies, such as [52,53], document that the relationship between tree diameter and height in estimating the biomass was affected by various environmental factors. More specifically, [50] concludes that the variation of the relationship among predictive biometric variables of trees depends directly on bioclimatic variables. This may explain some of the contradictory findings on the importance of various tree metric variables (i.e., th and ρ) in improving the accuracy of biomass estimating models. Hence, this leads to the realization that the significance of the dendrometric variables in predicting a tree’s aboveground biomass varies depending on the ecological and climatic variations.

This study demonstrates that the prediction accuracy of allometric models is different among biomass components of the tree. The stem biomass was the most predicted plant component, whereas twig biomass was the most unpredictable. This result corroborates the finding by previous studies [11,32,34,54,55,56,57], in which the biomass variation of leaves and tiny branches is less explainable by tree metric variables as compared with that of stems and woody components. Nevertheless, this outcome goes against the claims made by [14,15] that the total aboveground and branch biomasses were predicted more accurately than the stem biomass. According to [2,58], species differences, plant growth conditions, and site characteristics like stand density and availability of light, water, and soil nutrients could all be contributing factors to the variations in predictability in biomass of plant components.

4.3. Species-Specific Biomass Estimation Models

The dominant tree variables in the biomass equations of the five studied species were tree diameter, crown diameter, and wood basic density. Other studies, including [10,57], also used a combination of diameter, crown diameter, and wood basic density to construct the best-fit allometric equations. However, other studies, such as [59], have found the diameter at breast height and total height as the most frequent parameters across their models.

The accuracy of the aboveground biomass models varied among tree species. For example, the total aboveground biomass for Commiphora africana was the most predicted dry biomass. A similar adjusted R² of 0.93 was reported by [14], while [15] reports an adjusted R² of 0.47 for the same species. Differences in the predictability level of the same species might be due to the differences in growth architecture and growth stages [47]. The result shows that the combination of tree diameter and crown diameter was the main predictor for the biomass of all tree partitions of Commiphora africana. In line with this, [57] finds diameter at stump height and crown width as the main explanatory variables of dry biomass in Commiphora species. In their generic model, [4] also finds the best-fit equation by combining the diameter at stump height with crown width.

The aboveground biomass of Olea europaea was explained by tree diameter as a single predictor variable. This is consistent with the findings of [36] in biomass models for Olea europaea, which show that the AGB had a substantial correlation with dbh but not with wood basic density or tree height individually. Other studies, such as those of [13,46,50], also corroborate the significance of diameter in predicting tree biomass.

The total aboveground biomass and stem biomass of Oncoba spinosa were better predicted by the model with combinations of the tree height and diameter, but the dry biomasses of its branches and twigs were better explained when the tree diameter and crown diameter were combined. This result shows that, in Oncoba spinosa species, the twig biomass was more predictable than other biomass compartments (i.e., stem and branch). This result is consistent with the findings of [14,15] and contrasts with those documented by [11,57].

The total aboveground biomass of Teclea nobilis was best explained by the combination of tree and crown diameter. Similarly, this result shows that a combination of tree diameter and height typically determines the variation in the dry biomass of the stem. The branch biomass of Teclea nobilis was found to be best predicted by the allometric model with tree diameter and height. Despite the equation with tree diameter, crown diameter, and wood basic density best fit biomass of the twig, the regression constant of the equation was not statistically significant (p > 0.05). This shows the lesser predictability of twig (leaves and tiny branches) biomass than the biomass of the stem and big branches of the tree.

4.4. Performances of Existing Multispecies Models

The model developed by [10] resulted in relatively low differences between mean values of observed and predicted dry weight of aboveground biomass, with the lowest relative prediction bias and relative squared error. The lower prediction error of the model observed by [10] is a result of the model being developed using the sample trees from the dryland afromontane forest, which was found to be similar to the ecology of our study area.

On the contrary, this outcome shows that the general multispecies models developed by [4,39] have proved to underpredict the total aboveground biomass of our data. Like our finding, [10] also confirms the higher prediction bias of [4,39], which significantly underestimated their data by 40.3% and 29.4%, respectively.

Per our data, the commonly used pantropical model of [17] also underpredicted the aboveground biomass. The large prediction errors of the “pan-tropical” model by [17] are documented in several studies [4,60,61]. The multispecies model of [15] also produced a high prediction error on our dataset. This was due to the differences in tree species composition and ecology of the study sites, as their model relied on the majority of sample trees from a single species (i.e., acacia species) in the grassland area, while our sample of trees was from five different plant species in the dryland afromontane forest.

Generally, the high prediction error recorded for the majority of the existing generic allometric equations confirmed the concerns raised by previous studies [4,14,15,17,19] regarding the application of their models outside their study areas. Therefore, the newly developed multispecies model, which has a mean prediction error of 1.01 kg and relative prediction bias of 3.82%, was found to be the model with the best prediction performance to estimate the aboveground biomass of the Borena woodland. This shows the importance and reliability of location-specific biomass estimation models, which is also emphasized by earlier studies [11,39].

5. Conclusions

Multispecies aboveground tree biomass models were developed for Borana woodland, using 114 sample trees from five dominant species in the Gamadu afromontane forest. According to the biomass distribution along the tree compartments of the five species, the branch holds the highest fraction of the aboveground biomass. In generic and species-specific allometric models, tree diameter, wood basic density, and crown diameter were the most significant predictors for biomass across all tree components. The addition of wood basic density to the model significantly improved prediction accuracy, while tree height was found to have a minimal effect on biomass estimation. This further supports the findings of other studies that have emphasized the importance of tree diameter and wood density over tree height for accurate biomass predictions.

The models fitted for the biomass of tree compartments were found to have better statistical prediction than the models of total aboveground biomass of the trees. Among the biomass of tree compartments, the stem and twig biomasses were found to be the most and least predictable plant parts, respectively.

The result affirms that species-specific models possess better prediction performance with minimum errors than the multispecies models. Moreover, the study revealed that there are varying prediction accuracies for different species, with some species, such as Commiphora africana, showing strong predictive power while others, like Euclea divinorum, were poorly explained by dendrometric variables. The results highlight the need for species-specific models that account for variations in growth architecture, ecological factors, and bioclimatic conditions.

Testing existing multispecies biomass models revealed that some models performed well for the study area while others significantly underestimated biomass, confirming concerns about the applicability of generic models across different regions. The newly developed multispecies model, tailored to the study area, demonstrated the best prediction performance, reinforcing the importance of location-specific models for accurate biomass estimation. Hence, this study emphasizes the significance of ecological and species-specific factors in biomass estimation and underscores the importance of developing and applying regionally calibrated models to improve the accuracy of aboveground biomass predictions in different forest ecosystems.