论文信息

U.A. Yuguda / Moruff Gbolagade Saka / Ephraim Edward Dishan

FUDMA Journal of Sciences

论文解读

AI 辅助整理
方法
13个1公顷样地实测,异速生长方程估算,结合Landsat 8 NDVI/SAV
发现
平均总生物量29.67 Mg/ha,三指数模型拟合高但样本小、指数共线且缺独立验证。
机会点
可扩展样地数量与重复清查,引入独立验证和不确定性量化,改进高地碳储量遥感估算。

论文摘要

高地在尼日利亚碳核算中的代表性不足。我们估算了贡贝州图拉高地的地上生物量(AGB)、地下生物量(BGB)及生物量碳储量,并考察了其与Landsat衍生植被指数之间的关系。2025年2月至6月,在13个面积各为1公顷、间距1公里的样地中进行了木本植被调查。AGB采用异速生长方程估算,BGB采用0.26的固定根冠比估算,碳含量按0.47的分数计算;凋落物另行报告。利用2020—2024年Landsat 8 OLI影像计算了NDVI、SAVI和ARVI。平均密度为AGB 23.55 Mg ha⁻¹(95% CI 10.27–36.82),BGB 6.12 Mg ha⁻¹,合计29.67 Mg ha⁻¹(13.95 Mg C ha⁻¹),样地间变异较大(0.18–108.62 Mg ha⁻¹)。各样地合计,AGB和BGB分别为306.14和79.60 Mg(181.29 Mg C)。由于BGB是AGB的固定比例,其不携带独立信息;若采用0.56的比率,平均总生物量将升至36.74 Mg ha⁻¹。三指数线性模型对校准样地的调整R²为0.986(AIC = 173.93),但样本量小、指数间近乎共线且缺乏独立验证,限制了对空间预测的信心。模型推导的平均AGB在2020年至2024年间上升了16.99%;这反映的是光谱变化,而非实测的生物量积累或碳固存。该结果是初步基线,表明需要更多样地、重复调查和正式的不确定性评估。

全文

FUDMA Journal of Sciences

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
232

Field-Based Biomass and Carbon Stock Estimation and Landsat-Derived Spatial Patterns in
Tula Highland, Gombe State, Nigeria

*1Yuguda Usman Abubakar, 2Moruff Gbolagade Saka and 2Ephraim Edward Dishan

1Department of Plant Sciences, Gombe State University, Gombe, Gombe State, Nigeria.
2Department of Forestry and Wildlife Management, Modibbo Adama University, Yola, Adamawa State, Nigeria.

*Corresponding authors’ email: uayuguda@gsu.edu.ng
ORCID iD: https://orcid.org/0000-0003-3005-3130

ABSTRACT
Received:
Accepted:
Published:
27 Oct. 2026
10 Nov. 2026
14 Nov. 2026
Highland vegetation is poorly represented in Nigeria’s carbon accounting. We estimated
aboveground (AGB) and belowground biomass (BGB) and biomass carbon stocks in Tula
Highland, Gombe State, and examined their relationship with Landsat-derived vegetation
indices. Woody vegetation was inventoried in 13 one -hectare plots, spaced 1 km apart,
between February and June 2025. AGB was estimated with allometric equations, BGB
with a fixed root -to-shoot ratio of 0.26, and carbon with a fraction of 0.47; litter was
reported separately. Landsat 8 OLI imagery for 2020–2024 was used to compute NDVI,
SAVI and ARVI. Mean densities were 23.55 Mg ha⁻¹ for AGB (95% CI 10.27 –36.82)
and 6.12 Mg ha⁻¹ for BGB, a combined 29.67 Mg ha⁻¹ (13.95 Mg C ha⁻¹), with large
variation among plots (0.18–108.62 Mg ha⁻¹). Summed over the plots, AGB and BGB
were 306.14 and 79.60 Mg (181.29 Mg C). Because BGB is a fixed fraction of AGB, it
carries no independent information; a ratio of 0.56 would raise mean total biomass to
36.74 Mg ha⁻¹. A three-index linear model gave an adjusted R² of 0.986 (AIC = 173.93)
for the calibration plots, but the small sample, near -collinear indices and lack of
independent validation limit confidence in spatial predictions. Model-derived mean AGB
rose by 16.99% between 2020 and 2024; this reflects spectra l change, not measured
biomass accumulation or carbon sequestration. The results are a preliminary baseline and
show the need for more plots, repeated inventories and formal uncertainty assessment.

Keywords:
Aboveground Biomass, Belowground Biomass, Carbon Stock, Landsat 8, Vegetation
Indices, Tula Highland

INTRODUCTION
Aboveground biomass is an Essential Climate Variable
(GCOS, 2016) and a basic input to national greenhouse gas
inventories and REDD+ reporting (Bos et al., 2019; Herold et
al., 2019). Uncertainty in biomass estimates remains a major
constraint on carbon ac counting (Kumar & Mutanga, 2017;
Temesgen et al., 2015). Belowground biomass is measured
less often than aboveground biomass, although roots may
make up 10 –40% of total biomass depending on species
composition and site conditions (Mokany et al., 2006).
Montane and highland vegetation is climate -sensitive and
data-poor, and it is rarely represented in biomass inventories.
Nigeria’s Climate Change Act of 2021 created the National
Council on Climate Change and a legal basis for carbon
markets (Ajayi, 2026; PwC, 2023), which increases the
demand for biomass and carbon data. Such data remain scarce
for non-forest woody and highland vegetation. Tula Highland
in Gombe State combines rock outcrops, mixed montane –
savanna flora and smallholder land use (Jimoh et al., 2022;
Yilangai et al., 2023), and its biomass has not been quantified.
Satellite vegetation indices are widely used to extend plot
measurements over larger areas (Kumar & Mutanga, 2017).
Relationships calibrated in closed -canopy forest or lowland
savanna do not transfer readily to rocky highland terrain,
where bare rock and soil contribute much of the pixel signal,
so Tula Highland needs its own calibration. Three indices
were considered here. NDVI is simple but sensitive to soil
background where canopy cover is sparse (Huete, 1988).
SAVI includes a soil -adjustment factor (Hue te, 1988), and
ARVI uses the blue band to reduce atmospheric effects
(Kaufman & Tanré, 1992). The indices may carry
complementary information, but they are algebraically
related and highly correlated, which complicates regression.
Two limitations shape the interpretation of this kind of work.
First, a standing carbon stock is not carbon sequestration: a
stock is the amount stored at one time, whereas sequestration
is a net annual flux that requires repeated measurement.
Second, a si ngle-date field inventory cannot validate multi -
year change in image -derived estimates, because spectral
change may reflect phenology, herbaceous growth, or
atmospheric and compositing effects rather than change in
woody biomass. This study therefore treats the field inventory
as the measured quantity and the satellite results as
exploratory.
This study aimed to estimate aboveground and belowground
biomass and associated carbon stocks in Tula Highland,
Gombe State, and evaluate the potential of Landsat -derived
vegetation indices for spatial biomass estimation.

MATERIALS AND METHODS
Study Area
Tula Highland lies in Kaltungo Local Government Area,
Gombe State, north -eastern Nigeria (Figure 1), between
latitudes 9°48′N and 9°54′N and longitudes 11°25′E and
11°31′E. It covers approximately 6,381 ha at an elevation of
645 m above sea level. The clim ate is tropical savanna
(Köppen classification; Jimoh et al., 2022). The vegetation is
a light, discontinuous canopy with scattered shrubs and sparse
grass cover, which becomes open and largely grassless where
rock outcrops dominate (Jimoh et al., 2022). Woody species
FUDMA Journal of Sciences (FJS)
ISSN online: 2616-1370
ISSN print: 2645 - 2944
Vol. 10 No. 24, December, 2026, pp 232 – 239
DOI: https://doi.org/10.33003/fjs-2026-1024-6338

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
233
of highest structural importance in the plots were Guiera
senegalensis, Parkia biglobosa, Khaya senegalensis, Lantana
camara, Balanites aegyptiaca and Raphia sudanica ; Ficus
thonningii and Vitex doniana are also characteristic of the
highland, and Panicum maximum occurs in the grass layer
(Jimoh et al., 2022). Local microclimates support relatively
high floristic diversity (Yilangai et al., 2023), and the
highland protects watersheds and supplies fuelwood, fodder
and non-timber products to nearby communities.

Figure 1: Location and extent of Tula Highland vegetation, Kaltungo Local
Government Area, Gombe State, Nigeria (source: GIS Labor atory, Gombe
State University)

Plot Design and Field Measurements
Thirteen plots of 100 m × 100 m (1 ha) were established at
fixed, equal intervals in a systematic design (Ravindranath &
Ostwald, 2008), about 0.2% of the highland area. The number
of plots followed minimum sampling -intensity
recommendations for areas of l ow to moderate tree density
(Rügnitz et al., 2009). Plots were spaced 1 km apart and
measured from February to June 2025. Each plot covers about
11 Landsat 8 pixels. Within each plot, four 20 m × 20 m
subplots at the corners were used for tree enumeration, 5 m ×
5 m subplots nested within them for shrubs, and a 1 m × 1 m
subplot for litter (Bordoloi et al., 2021).
To convert biomass recorded within nested subplots to
standardised per-hectare densities (Mg ha⁻¹), stratum-specific
expansion factors (EF) were applied based on the sampled
surface area within each 1.0 -ha (10,000 m²) macro -plot
(Bordoloi et al., 2021). Fo r mature trees, the four 20 m × 20
m corner subplots sampled a cumulative area of 1,600 m²
(0.16 ha) per plot, so tree biomass was scaled to one hectare
using an expansion factor of
EFtree = 10,000 m² / 1,600 m² = 6.25:
AGBtrees (Mg ha⁻¹) = (Σ AGBtree (kg) × 6.25) / 1,000
(1)
For shrubs and saplings, the four 5 m × 5 m nested subplots
sampled a total area of 100 m² (0.01 ha) per plot, and shrub
biomass was expanded to one hectare using
EFshrub = 10,000 m² / 100 m² = 100.0:
AGBshrubs (Mg ha⁻¹) = (Σ AGBshrub (kg) × 100.0) / 1,000
(2)
For surface litter necromass, four 0.5 m × 0.5 m quadrats
sampled a cumulative area of 1.0 m² (0.0001 ha) per plot, and
oven-dry litter mass was scaled to one hectare using
EFlitter = 10,000 m² / 1.0 m² = 10,000:
Litter necromass (Mg ha⁻¹) = (W dry (g m⁻²) × 10,000) /
1,000,000 = Wdry (g m⁻²) × 0.01 (3)
Diameter at breast height (DBH, measured at 1.3 m) and total
height were recorded for all woody individuals. Trees with
stems of 10 cm DBH or more, and shrubs and saplings with
stems of 1–9 cm DBH. In shrubs and small trees, diameter
was measured near the base or at the point of branching,
whichever was more representative of the stem. Height was
measured with a clinometer.
Litter was collected from 0.5 m × 0.5 m quadrats placed
within the 1 m × 1 m subplots (Asrat et al., 2018). The
material comprised fine necromass, dead wood less than 5 cm
in diameter or 50 cm in length, non -decomposed plant
material, unburned leaves and b ranches, and partly
decomposed litter. Samples were oven -dried and weighed.
Litter is reported separately and is not included in the AGB,
BGB or carbon totals.

Biomass and Carbon Estimation
Tree AGB was estimated from DBH and height with Equation
4, using the coefficients of the multi-species savanna equation
of Henry et al. (2009) (see also Chave et al., 2014): a = 0.051
and b = 0.930, for AGB in kg. Shrub AGB was estimated from
stem diameter with Equation 5 (Conti et al., 2019), with a =
0.0568 and b = 2.584. The equation of Conti et al. (2019) was
used because it is a global shrub model.
AGB = a × (DBH² × H)b (4)
Where DBH is diameter at breast height (cm) and H is total
tree height (m).
AGB = a × Db (5)
Where D is shrub stem diameter (cm).
Aboveground carbon was calculated with a carbon fraction of
0.47 (IPCC, 2006, 2019):
CAGB = AGB × 0.47 (6)
Litter biomass (LB, kg ha⁻¹) was calculated from the dry mass
collected per unit quadrat area:
LB = (W / A) × 10 (7)

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
234
Where W is dry mass of litter (g) and A is quadrat area (m²);
the factor 10 converts g m⁻² to kg ha⁻¹.
BGB was estimated as a fixed fraction of AGB for trees and
shrubs (Equation 8), using a root -to-shoot ratio ( R) of 0.26
(IPCC, 2006; Mokany et al., 2006; Ma et al., 2021). Because
BGB is then a deterministic function of AGB, it carries no
independent belowground information. The ratio is lower
than some IPCC defaults for dry tropical vegetation, and the
effect of a higher value is shown in the Results. Belowground
carbon was derived with the same carbon fraction (Equation
9). Carbon was converted to CO₂ -equivalent by multiplying
by 44/12.
BGB = AGB × R (8)
CBGB = BGB × 0.47 (9)

Remote Sensing Data
Landsat 8 Operational Land Imager (OLI) scenes for 2020 –
2024 (30 m spatial resolution) were downloaded from the
USGS EarthExplorer archive and used to derive NDVI, ARVI
and SAVI datasets to assess vegetation conditions. The
imagery was referenced to WRS -2 Path/Row 186/053, and
the vegetation -index rasters were clipped to the study -area
boundary using the Extract by Mask tool in ArcGIS. The
original scene -level metadata and processing records were
unavailable; exact acquisition dates, annual scene counts,
product processing levels, cloud-cover thresholds and cloud-
masking procedures therefore could not be independently
verified. Bands were stacked in ArcMap, converted from
digital numbers to top -of-atmosphere reflectance with scene
metadata, and corrected for atmospheric scattering by dark
object subtraction. Images were processed at the Zonal
Advanced Space Technology Applications Laboratory,
National Centre for Remote Sensing, Federal University
Kashere, Gombe State, Nigeria. Annual mean temperature
and total precipitation for 2020–2024 were obtained from the
Nigerian Meteorological Agency (NIMET).
NDVI (Rouse et al., 1974; Tucker, 1979), SAVI (Huete,
1988) and ARVI (Kaufman & Tanré, 1992) were computed
for each pixel:
NDVI = (NIR − RED) / (NIR + RED) (10)
SAVI = [(NIR − RED) × (1 + L)] / (NIR + RED + L)
(11)
ARVI = [NIR − (RED − γ (RED − BLUE))] / [NIR + (RED − γ (RED −
BLUE))] (12)
Where NIR, RED and BLUE are reflectance in the near -
infrared, red and blue bands, L is the soil -adjustment factor
(L = 0.5) and γ is the atmospheric correction coefficient (γ =
1). For each index, annual maximum -value composites were
produced by selecting the highest value recorded for each
pixel in a year, which reduces the effects of cloud and
seasonal fluctuation. Five-year mean vegetation-index values
(2020–2024) were paired with the 2025 field inventory data
for biomass calibration. The imagery was used to represent
longer-term spectral conditions rather than contemporaneous
ground observations, and the temporal mismatch was
recognised as a potential source of uncertainty in the
biomass–vegetation index relationships.

Statistical Analysis
Candidate models were multiple linear regressions of plot
AGB on all seven non -empty combinations of NDVI, SAVI
and ARVI, fitted by ordinary least squares. Each combination
was fitted to AGB, log -transformed AGB and inverse -
transformed AGB, giving 21 model s. Models were ranked
with Akaike’s information criterion (AIC = 2k − 2ln L, where
k is the number of estimated parameters and L the maximised
likelihood). AIC values were compared only within a
response form, because likelihoods defined on different
response scales are not comparable. Models with ΔAIC below
2 were considered to have similar support (Burnham &
Anderson, 2002). Prediction uncertainty was assessed by
Monte Carlo simulation with 500 runs (Chave et al., 2004;
Yanai et al., 2010). The coefficien t of variation (CV) of the
simulated predictions indicates the stability of the model
under the simulated parameter variation. It does not measure
predictive accuracy. The selected equation was applied to the
annual composites to produce AGB maps for 2020 –2024.
Climate and spectral rasters were co -registered to a common
grid, and zonal statistics were calculated over the study
polygon (Tomlin, 1990). Pearson correlations between annual
mean indices and annual climate used n = 5 years.

RESULTS AND DISCUSSION
Biomass and Carbon Stocks
Plot values are given in Table 1 and summary statistics in
Table 2. Mean AGB across the 13 plots was 23.55 Mg ha⁻¹
(SD 21.97; 95% confidence interval 10.27–36.82) and mean
BGB was 6.12 Mg ha⁻¹ (SD 5.71), giving a mean total
biomass of 29.67 Mg ha⁻¹ (SD 27. 68; 95% CI 12.94–46.40).
Mean carbon density was 13.95 Mg C ha⁻¹ (11.07
aboveground and 2.88 belowground), equivalent to 51.18 Mg
CO₂-e ha⁻¹. Summed over the 13 sampled hectares, AGB was
306.14 Mg and BGB 79.60 Mg, holding 143.88 and 37.41 Mg
C respectively (181.29 Mg C, or 665.35 Mg CO₂ -e, in total).
These totals refer to the sampled plots only and are not
densities or estimates for the whole highland. Mean litter
biomass was 9,050.83 ± 997.03 kg ha⁻¹ (SD 4.15 and SE 1.31)
and is not included in the totals.

Table 1: Aboveground and Belowground Biomass and Carbon Stocks in the 13 Plots, Tula Highland. Plot Values are
Per Hectare (Mg ha⁻¹)
Plot AGB AGC AGC as CO₂-e BGB BGC BGC as CO₂-e Total biomass
TH 1 21.20 9.97 36.57 5.51 2.59 9.51 26.72
TH 2 18.31 8.61 31.58 4.76 2.24 8.21 23.07
TH 3 12.79 6.01 22.06 3.32 1.56 5.73 16.11
TH 4 36.47 17.14 62.91 9.48 4.46 16.36 45.95
TH 5 26.31 12.37 45.38 6.84 3.22 11.80 33.15
TH 6 25.60 12.03 44.17 6.66 3.13 11.48 32.26
TH 7 0.91 0.43 1.56 0.24 0.11 0.41 1.14
TH 8 33.88 15.92 58.44 8.81 4.14 15.20 42.69
TH 9 19.73 9.27 34.04 5.13 2.41 8.85 24.86
TH 10 0.14 0.07 0.24 0.04 0.02 0.06 0.18
TH 11 5.16 2.42 8.90 1.34 0.63 2.31 6.50
TH 12 86.20 40.52 148.69 22.41 10.53 38.66 108.62

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
235
Plot AGB AGC AGC as CO₂-e BGB BGC BGC as CO₂-e Total biomass
TH 13 19.43 9.13 33.51 5.05 2.37 8.71 24.48
Sum (Mg) 306.14 143.88 528.06 79.60 37.41 137.29 385.73
Mean (Mg ha⁻¹) 23.55 11.07 40.62 6.12 2.88 10.56 29.67
AGB, aboveground biomass; AGC, aboveground carbon; BGB, belowground biomass; BGC, belowground carbon; CO₂ -e,
CO₂-equivalent. Sums and means were calculated from unrounded values and may differ slightly from the sums of the rounded
entries.

Table 2: Sums and Mean Densities of Biomass and Carbon across the 13 Plots
Quantity Sum across 13 plots Mean (per ha) SD 95% CI of mean
AGB 306.14 Mg 23.55 21.97 10.27–36.82
AGC 143.88 Mg C 11.07 10.32 4.83–17.31
BGB 79.60 Mg 6.12 5.71 2.67–9.57
BGC 37.41 Mg C 2.88 2.68 1.26–4.50
Total biomass 385.73 Mg 29.67 27.68 12.94–46.40
Total carbon 181.29 Mg C 13.95 13.01 6.08–21.81
Total CO₂-equivalent 665.35 Mg CO₂-e 51.18 47.74 22.33–80.03
Mean and SD are in Mg ha⁻¹ (carbon in Mg C ha⁻¹; CO₂ -equivalent in Mg CO₂ -e ha⁻¹). Confidence intervals use the t
distribution with 12 degrees of freedom.

Biomass varied about 600 -fold among plots, from 0.18 Mg
ha⁻¹ (TH 10) to 108.62 Mg ha⁻¹ (TH 12), and the coefficient
of variation was 93%. TH 12, where Parkia biglobosa, Khaya
senegalensis and Raphia sudanica were structurally
important, held 28% of the summed AGB. TH 10 and TH 7
(1.14 Mg ha⁻¹) were sparse, rock-dominated plots. The dense
stand in TH 12 may reflect a moisture -favoured micro-site,
but soil moisture and topography were not measured. The
wide confidence interval shows that the 13 -plot mean is an
imprecise estimate for the highland.
Mean total biomass (29.67 Mg ha⁻¹) is lower than the shrub -
savanna (45.29 Mg ha⁻¹) and riparian -forest (94.58 Mg ha⁻¹)
values reported for the Sudan savanna of Benin (Chabi et al.,
2016), although the upper limit of the present confidence
interval (46.40 Mg ha⁻¹) reaches the shrub -savanna value.
These are standing stocks. They do not show that the highland
is removing CO₂ from the atmosphere, which would require
repeated measurement of biomass change.

Belowground Biomass
BGB made up 20.6% of total biomass (79.60 of 385.73 Mg).
This follows directly from the fixed ratio of 0.26 and is not a
property of the vegetation. The spatial pattern of BGB
therefore duplicates that of AGB, and BGB cannot be
validated independently of i t. Shallow, rocky soils often
promote greater root allocation (Mokany et al., 2006; Ma et
al., 2021), so a uniform ratio may underestimate belowground
storage on exposed plots. Using R = 0.56, the IPCC (2006)
default for tropical dry forest with AGB below 125 Mg ha⁻¹,
would raise mean BGB from 6.12 to 13.19 Mg ha⁻¹ and mean
total biomass from 29.67 to 36.74 Mg ha⁻¹. Excavation-based
root allometry for dominant highland species is needed to
narrow this uncertainty.

Vegetation Indices and Climate
Mean annual NDVI, SAVI and ARVI over 2020–2024 were
0.22, 0.33 and 0.14, low values consistent with a
discontinuous canopy and exposed rock and soil. Interannual
variability was small (SD 0.04, 0.06 and 0.05). All three
indices peaked in 2023 (NDVI 0.242, SAVI 0.359, ARVI
0.169), the year with the highest temperature and lowest
precipitation. A peak in the driest year is unlikely to indicate
woody growth. Herbaceous or crop greening, composite
selection and atmospheric contamination of the maximum -
value composite could contribute, and none was tested.
The indices correlated strongly and positively with annual
mean temperature (r = 0.969–0.996; Table 3) and negatively,
but not significantly, with precipitation (r = −0.593 to
−0.692). With five annual observations these coefficients are
unstable, and temperature and greenness may simply share a
trend over the period. They show association only and do not
show that temperature controls greenness.

Table 3: Pearson Correlations between Annual Mean Vegetation Indices and Annual Climate, Tula Highland, 2020 –
2024
Index Climate variable n r p Remarks
ARVI Precipitation 5 −0.593 0.292 Not significant
ARVI Temperature 5 0.969 0.006 Significant
NDVI Precipitation 5 −0.687 0.201 Not significant
NDVI Temperature 5 0.994 0.001 Significant
SAVI Precipitation 5 −0.692 0.195 Not significant
SAVI Temperature 5 0.996 < 0.001 Significant
n, number of years. Source: Landsat 8 analysis and climate data.

Correlations between raster -predicted AGB and the three
indices exceeded 0.90 in every year (NDVI and SAVI
0.9405–0.9589; ARVI 0.9038 –0.9191; 20,000 randomly
sampled pixels per year). These values are expected, because
predicted AGB is a linear function of the same indices, and
they give no independent evidence of model validity. NDVI
and SAVI correlations were identical to four decimal places
in every year, which shows that the two indices are almost
perfectly collinear.

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
236
Model Performance and Selection
Twenty-one models were fitted (Table 4). For untransformed
AGB, the three-index model had the lowest AIC (173.93) and
the highest adjusted R² (0.986), followed by NDVI + SAVI
(AIC 175.86; adjusted R² 0.982) and NDVI alone (AIC
191.53; adjusted R² 0.931). The ΔAIC between the first two
models is 1.93, below the threshold of 2, so NDVI + SAVI
has similar support; NDVI alone was clearly weaker (ΔAIC
= 17.60). Inverse -transformed models explained the least
variance (best adjusted R² 0.476), and log -transformed
models were intermediate (best adjusted R² 0.834, for the
three-index model). Rankings in Table 4 are given within
each response form because AIC is not comparable across
them. The result differs from the regional pattern reported for
single-index models (Yuguda et al., 2026).

Table 4: Fit Statistics of the Candidate Models for Plot AGB and its Transformations, Tula Highland
Response Predictors Adjusted R² AIC Rank within response
AGB NDVI + SAVI + ARVI 0.9857 173.93 1
AGB NDVI + SAVI 0.9823 175.86 2
AGB NDVI 0.9306 191.53 3
AGB NDVI + ARVI 0.9276 192.77 4
AGB ARVI 0.8749 198.60 5
AGB SAVI + ARVI 0.8610 200.60 6
AGB SAVI 0.8513 200.67 7
Log AGB NDVI + SAVI + ARVI 0.8336 31.66 1
Log AGB SAVI + ARVI 0.8075 32.82 2
Log AGB NDVI + SAVI 0.7914 33.78 3
Log AGB SAVI 0.7419 35.60 4
Log AGB NDVI 0.6640 38.77 5
Log AGB ARVI 0.6410 39.56 6
Log AGB NDVI + ARVI 0.6266 40.77 7
Inverse AGB SAVI + ARVI 0.4704 −62.48 1
Inverse AGB NDVI + SAVI + ARVI 0.4761 −62.02 2
Inverse AGB NDVI + SAVI 0.3397 −59.83 3
Inverse AGB SAVI 0.2817 −59.56 4
Inverse AGB NDVI 0.2197 −58.56 5
Inverse AGB ARVI 0.1846 −58.04 6
Inverse AGB NDVI + ARVI 0.1514 −56.82 7
Statistics refer to the calibration plots. Source: field survey (2025) and Landsat 8 analysis.

The selected equation (Table 5) has large coefficients of
opposite sign for NDVI (+106,527) and SAVI (−63,095).
This is typical of collinear predictors: the individual
coefficients are unstable and cannot be read as separate
effects of each index. The gain from NDVI alone (adjusted R²
0.931) to the three -index model (0.986) comes from small
differences between nearly collinear predictors on 13 plots
and may reflect noise as much as added information. The
three-index model should not be regarded as superior until
cross-validation supports it. Complementary information
from the three indices, as suggested for heterogeneous terrain
(Chen et al., 2025; Mu hammad et al., 2024; Zhang et al.,
2023), was not demonstrated here.

Table 5: Selected Vegetation-Index Model for AGB, Tula Highland
Response Predictors Equation Adjusted
R²
AIC Simulation CV
(%)
AGB NDVI + SAVI +
ARVI
AGB = −2,556.13 + 106,527.47 × NDVI −
63,095.02 × SAVI + 14,609.84 × ARVI
0.986 173.93 3.32
Adjusted R² and AIC refer to the calibration plots. Source: field survey (2025) and Landsat 8 analysis.

Table 6: Monte Carlo Summary (500 Runs) for the Untransformed AGB Models, Tula Highland
Predictors Mean simulated AGB SD Lower 95% Upper 95% CV (%)
NDVI 2,384.46 174.01 2,050.42 2,719.55 7.30
SAVI 2,390.93 254.54 1,896.54 2,878.22 10.65
ARVI 2,402.38 238.65 1,939.99 2,868.34 9.93
NDVI + SAVI 2,387.26 85.79 2,218.23 2,556.58 3.59
NDVI + ARVI 2,384.57 172.51 2,041.43 2,710.97 7.23
SAVI + ARVI 2,396.75 241.56 1,920.73 2,871.51 10.08
NDVI + SAVI + ARVI 2,389.78 79.38 2,235.51 2,548.47 3.32
Values in kg ha⁻¹. Source: Monte Carlo analysis.

The three-index model had the lowest simulation CV (3.32%;
Table 6). This CV shows that predictions are stable under the
simulated input variation. It does not show that predictions
are accurate, and a model fitted closely to 13 plots can be
stable and still biased out of sample. Log-transformed models
had CVs of 21.7 –32.6%. Inverse -transformed models had
CVs above 1,000% and 95% intervals that included negative
biomass, and they were unsuitable for prediction, although

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
237
inverse transformation performs well in some studies (Poudel
et al., 2023). Model selection should therefore consider
stability and plausibility in addition to AIC and adjusted R².

Validation
Fit statistics describe agreement with the calibration plots and
are not measures of predictive accuracy. With 13 plots and
three correlated predictors, overfitting is likely. Individual
plots strongly influence the fit; TH 12 alone holds 28% of the
summed AGB and should be tested as an influential
observation. Residual plots (Figure 2) show no obvious fan -
shaped pattern, but with this sample they have little power to
detect departures from constant variance.

Figure 2: Diagnostic plots for the NDVI + SAVI + ARVI regression: observed against predicted AGB,
residuals against fitted values, and normal Q–Q plot of residuals

Biomass in these plots is positively skewed, with two near -
zero plots and one very large plot, yet untransformed AGB
fitted better than the transformed forms. With only 13 plots
this may reflect limited sampling rather than a true site
characteristic. Stra tifying the highland by rock cover and
fitting separate relationships for open-rock and wooded areas
would probably give more transferable models.

Spatial and Interannual Patterns of Predicted AGB
Applying the selected equation to the annual composites gave
mean AGB of 1,963.65 kg ha⁻¹ in 2020 and 2,297.29 kg ha⁻¹
in 2024, an increase of 333.64 kg ha⁻¹ (16.99%; Figure 3).
Annual means ranged from 1,680.85 to 3,142.49 kg ha⁻¹
(mean 2,169.27; SD 590.45; CV 0.272). Interannual variation
was therefore larger than the 2020–2024 difference, and the
highest annual mean coi ncided with the anomalous 2023
index peak. The change between two endpoint years is
sensitive to the years chosen, and no trend test was applied to
the five annual values.

Figure 3: Predicted AGB (kg ha⁻¹) in Tula Highland for 2020 –2024, change between 2020 and 2024, and the Monte
Carlo mean, standard deviation and CV for 2024

An increase of this kind can arise from woody regeneration,
from phenological or herbaceous variation, or from land -use
change such as a shift from cropping to fallow or tree -based
systems (Biah et al., 2023; Wang et al., 2024). The high
interannual CV and the peak in the driest year suggest that at
least part of the signal is not woody growth. Confirming
durable biomass change requires repeated field inventories,
or structural data from very -high-resolution imagery or
LiDAR. The result is therefore reporte d as a model -derived
spectral pattern. It is not evidence of carbon sequestration,
and no projection beyond 2024 is made.

Limitations
The 13 plots cover about 0.2% of the highland and vary
widely (CV 93%). Plots were placed systematically and not
stratified by rock cover, so the mean may not represent the
whole area. The allometric equations were developed in other
regions and may not su it Raphia sudanica , a palm. BGB
follows from a fixed ratio, and litter and soil carbon are
excluded from the totals. Plot-to-pixel matching is imprecise,
because each plot covers about 11 Landsat pixels, many of
them mixed rock and vegetation, and top -of-atmosphere
reflectance with dark object subtraction is less robust than
surface-reflectance products. Finally, the field data come

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
238
from one year (2025) and the imagery precedes it, so image -
based change cannot be validated.

Conclusion
Field inventory of 13 one -hectare plots in Tula Highland
showed large spatial variation in aboveground and
belowground biomass. Mean densities were 23.55 Mg ha⁻¹
aboveground and 6.12 Mg ha⁻¹ belowground (combined
29.67 Mg ha⁻¹, or 13.95 Mg C ha⁻¹), and the summed values
across the plots were 306.14 and 79.60 Mg. A regression on
NDVI, SAVI and ARVI fitted the calibration plots closely
(adjusted R² = 0.986), but the small sample, collinear
predictors and limited independent validation constrain
confidence in spatial predictions. The 2020–2024 changes in
Landsat-derived AGB are preliminary model -derived
patterns and not confirmed carbon sequestration. More field
plots, repeated inventories and a formal uncertainty
assessment are needed before the model can supp ort carbon
accounting.

REFERENCES
Ajayi, O. L. P. (2026). Operationalisation of the Nigerian
carbon market. Olaniwun Ajayi Insights.
https://www.olaniwunajayi.net/blog/operationalisation-of-
the-nigerian-carbon-market/

Asrat, Z., Taddese, H., Ørka, H. O., Gobakken, T., Burud, I.,
& Næsset, E. (2018). Estimation of forest area and canopy
cover based on visual interpretation of satellite images in
Ethiopia. Land, 7(3), Article 92.

Biah, I., Azihou, A. F., Guendehou, S., & Sinsin, B. (2024).
Land use/land cover change and carbon footprint in tropical
ecosystems in Benin, West Africa. Trees, Forests and People,
15, Article 100488.

Bordoloi, R., Das, B., Tripathi, O. P., Sahoo, U. K., Nath, A.
J., Deb, S., Devi, N. B., & Gupta, A. (2021). Satellite data
derived spatiotemporal variation of biomass carbon stocks
and their relation with vegetation type in tropical forests of
Assam in northeast India. Ecological Indicators, 123, Article
107341. https://doi.org/10.1016/j.ecolind.2021.107341

Bos, A. B., De Sy, V., Duchelle, A. E., Herold, M., & Martius,
C. (2019). Global data and tools for local forest cover loss and
REDD+ performance assessment: Gaps, needs and possible
approaches. Environmental Science & Policy, 92, 357–367.

Burnham, K. P., & Anderson, D. R. (2002). Model selection
and multimodel inference: A practical information -theoretic
approach (2nd ed.). Springer.

Chabi, A., Lautenbach, S., Tondoh, J. E., Orekan, V. O. A.,
Kyei-Baffour, N., & Blaes, X. (2016). The relevance of using
in situ carbon stocks and satellite indices to assess
aboveground carbon stock of tropical forest ecosystems.
iForest, 9, 606–615.

Chave, J., Condit, R., Aguilar, S., Hernandez, A., Lao, S., &
Perez, R. (2004). Error propagation and scaling for tropical
forest biomass estimates. Philosophical Transactions of the
Royal Society B: Biological Sciences, 359(1443), 409–420.

Chave, J., Réjou-Méchain, M., Búrquez, A., Chidumayo, E.,
Colgan, M. S., Delitti, W. B. C., Duque, A., Eid, T.,
Fearnside, P. M., Goodman, R. C., Henry, M., Martínez -
Yrízar, A., Mugasha, W. A., Muller -Landau, H. C.,
Mencuccini, M., Nelson, B. W., Ngomanda, A., Nogueira, E.
M., Ortiz -Malavassi, E., Vieilledent, G. (2014). Improved
allometric models to estimate the aboveground biomass of
tropical trees. Global Change Biology, 20(10), 3177 –3190.
https://doi.org/10.1111/gcb.12629

Chen, C., He, G., Fang, H., Shi, L., Zhuang, Y., Ding, Z.,
Guo, J., Yue, X., Yang, K., & Xi, W. (2025). A study on the
remote sensing estimation and spatiotemporal distribution
patterns of aboveground biomass in savanna grasslands of the
Yuanmou dry -hot va lley. Frontiers in Plant Science, 16,
Article 1648539.

Conti, G., Enrico, L., Casanoves, F., & Díaz, S. (2019). Shrub
biomass estimation in the semiarid Chaco forest: A
contribution to the understanding of South American dryland
carbon sinks. Journal of Arid Environments, 166, 51–60.

Global Climate Observing System. (2016). The global
observing system for climate: Implementation needs (GCOS-
200). World Meteorological Organization.

Henry, M., Tittonell, P., Manlay, R. J., Bernoux, M.,
Albrecht, A., & Vanlauwe, B. (2009). Biodiversity, carbon
stocks and sequestration potential in aboveground biomass in
smallholder farming systems of western Kenya. Agriculture,
Ecosystems & Environment , 129(1 –3), 238 –252.
https://doi.org/10.1016/j.agee.2008.09.006

Herold, M., Carter, S., Avitabile, V., Espejo, A. B.,
Jonckheere, I., Lucas, R., McRoberts, R. E., Næsset, E.,
Nightingale, J., Petersen, R., Reiche, J., Romijn, E.,
Rosenqvist, Å., Rozendaal, D. M. A., Seifert, F. M., Sanz, M.
J., & De Sy, V. (2019). The role and need for space -based
forest biomass -related measurements in environmental
management and policy. Surveys in Geophysics, 40, 757 –
778. https://doi.org/10.1007/s10712-019-09510-6

Huete, A. R. (1988). A soil-adjusted vegetation index (SAVI).
Remote Sensing of Environment, 25(3), 295 –309.
https://doi.org/10.1016/0034-4257(88)90106-X

Intergovernmental Panel on Climate Change. (2006). 2006
IPCC guidelines for national greenhouse gas inventories (H.
S. Eggleston, L. Buendia, K. Miwa, T. Ngara, & K. Tanabe,
Eds.). Institute for Global Environmental Strategies.

Intergovernmental Panel on Climate Change. (2019). 2019
refinement to the 2006 IPCC guidelines for national
greenhouse gas inventories (E. Calvo Buendia, K. Tanabe, A.
Kranjc, J. Baasansuren, M. Fukuda, S. Ngarize, A. Osako, Y.
Pyrozhenko, P. Shermanau, & S. Federici, Eds.). IPCC.

Jimoh, A. I., Hussaini, H., Aliyu, J., Sadiq, F. K., & Kefas, P.
K. (2022). Assessment of the impacts of toposequence on soil
properties and quality in Tula, Gombe State, Nigeria.
Agricultura Tropica et Subtropica, 55(1), 99–107.

Kaufman, Y. J., & Tanré, D. (1992). Atmospherically
resistant vegetation index (ARVI) for EOS -MODIS. IEEE
Transactions on Geoscience and Remote Sensing, 30(2),
261–270. https://doi.org/10.1109/36.134076

Kumar, L., & Mutanga, O. (2017). Remote sensing of above-
ground biomass. Remote Sensing, 9(9), Article 935.
https://doi.org/10.3390/rs9090935

Field-Based Biomass and Carbon Stock Estima… Yuguda et al., FJS
FUDMA Journal of Sciences (FJS) Vol. 10 No. 24, December, 2026, pp 232 – 239
239
©2026 This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0
International license viewed via https://creativecommons.org/licenses/by/4.0/ which permits unrestricted use,
distribution, and reproduction in any medium, provided the original work is cited appropriately.

Ma, H., Mo, L., Crowther, T. W., Maynard, D. S., van den
Hoogen, J., Stocker, B. D., Terrer, C., & Zohner, C. M.
(2021). The global distribution and environmental drivers of
aboveground versus belowground plant biomass. Nature
Ecology & Evolution, 5, 1110 –1122.
https://doi.org/10.1038/s41559-021-01485-1

Mokany, K., Raison, R. J., & Prokushkin, A. S. (2006).
Critical analysis of root:shoot ratios in terrestrial biomes.
Global Change Biology, 12(1), 84 –96.
https://doi.org/10.1111/j.1365-2486.2005.001043.x

Muhammad, B., Rehman, A. U., Mumtaz, F., Qun, Y., &
Zhongkui, J. (2024). Estimation of above -ground biomass in
dry temperate forests using Sentinel -2 data and random
forest: A case study of the Swat area of Pakistan. Frontiers in
Environmental Science, 12, Article 1448648.

Poudel, A., Shrestha, H. L., Mahat, N., Sharma, G., Aryal, S.,
Kalakheti, R., & Lamsal, B. (2023). Modeling and mapping
of aboveground biomass and carbon stock using Sentinel -2
imagery in Chure region, Nepal. International Journal of
Forestry Research, 2023, Article 5553957.

PwC. (2023). Nigeria’s Climate Change Act: Implications for
business and investment. PricewaterhouseCoopers Nigeria.

Ravindranath, N. H., & Ostwald, M. (2008). Carbon
inventory methods: Handbook for greenhouse gas inventory,
carbon mitigation and roundwood production projects.
Springer.

Rouse, J. W., Haas, R. H., Schell, J. A., & Deering, D. W.
(1974). Monitoring vegetation systems in the Great Plains
with ERTS. In Proceedings of the Third Earth Resources
Technology Satellite -1 Symposium (Vol. 1, pp. 309 –317).
NASA.

Rügnitz, M. T., Chacón, M. L., & Porro, R. (2009). Guía para
la determinación de carbono en pequeñas propiedades rurales.
ICRAF.

Temesgen, H., Affleck, D., Poudel, K., Gray, A., & Sessions,
J. (2015). A review of the challenges and opportunities in
estimating above ground forest biomass using tree -level
models. Scandinavian Journal of Forest Research, 30(4),
326–335.

Tomlin, C. D. (1990). Geographic information systems and
cartographic modeling. Prentice Hall.

Tucker, C. J. (1979). Red and photographic infrared linear
combinations for monitoring vegetation. Remote Sensing of
Environment, 8(2), 127 –150. https://doi.org/10.1016/0034-
4257(79)90013-0

Wang, M., Ciais, P., Fensholt, R., Brandt, M., Tao, S., Li, W.,
Fan, L., Frappart, F., Sun, R., Li, X., Liu, X., Wang, H., Cui,
T., Xing, Z., Zhao, Z., & Wigneron, J. -P. (2024). Satellite
observed aboveground carbon dynamics in Africa during
2003–2021. Rem ote Sensing of Environment, 301, Article
113927.

Yanai, R. D., Battles, J. J., Richardson, A. D., Blodgett, C. A.,
Wood, D. M., & Rastetter, E. B. (2010). Estimating
uncertainty in ecosystem budget calculations. Ecosystems,
13, 239–248.

Yilangai, R. M., Abalaka, J., Nsor, C. A., Babale, A., Karau,
S. D., & Ivande, S. (2023). Effect of disturbance on bird
feeding guilds in a West African dry forest. African Journal
of Ecology, 61(2), 461–468.

Yuguda, U. A., Saka, M. G., Dishan, E. E., Abba, H. M., &
Bello, M. B. (2026). Spatial patterns and temporal dynamics
of aboveground-belowground biomass and carbon stocks in
Kanawa Forest Reserve, Gombe State, Nigeria. Bima Journal
of Science and Technolog y, 10(3).
https://doi.org/10.64290/bima.v10i3.1651

Zhang, Y., Liang, S., & Yang, L. (2019). A review of regional
and global gridded forest biomass datasets. Remote Sensing,
11(23), Article 2744.

论文原文与出版信息

FUDMA Journal of Sciences
查看论文原文

继续探索