Quantifying carbon benefits from protecting and restoring tropical forest landuse mosaics in Vanuatu

Abstract

Background

Forest loss and degradation in Vanuatu is a pertinent environmental issue, particularly for rural communities who are inherently reliant on subsistence-based agriculture derived from tropical forest landuse mosaics. Payment for Ecosystem Services (PES) projects provide a mechanism to diversify agricultural livelihoods while reducing deforestation and forest degradation and promoting ecosystem restoration. This study evaluates the Life & Land Project, a PES initiative working with communities at Whitegrass and Middlebush on Tanna, and South River on Erromango. Few primary studies in Vanuatu have quantified forest carbon stocks or evaluated changes in carbon stocks resulting from PES interventions.

Research objectives and methods

These knowledge gaps are addressed by assessing: (a) forest carbon stocks and structural characteristics; (b) the drivers and rates of deforestation and forest degradation, together with pressures affecting forest health, to inform management interventions; and (c) the potential carbon benefits of implementing these interventions using a gain—loss approach.

Results

Results showed high variability in structural composition and carbon stocks, reflective of differences in landuse and environmental pressures. The mature, predominantly primary tropical rainforest at Whitegrass had an average carbon stock of 490.7 ± 32.6 Mg C ha⁻¹, in contrast to the invasive vine-smothered degraded forests of Imanaka Creek (Middlebush) and the Forest Enhancement Zone of South River (107.3 ± 10.7 Mg C ha⁻¹ and 128.2 ± 22.7 Mg C ha⁻¹ respectively). Forest loss at Whitegrass was driven through logging, ad-hoc garden development, and quarry operations at an estimated rate of 3.55 ha⁻¹ year⁻¹, while invasive vines were the main impediment impacting forest health at most sites. To address pressures, a conservation area at Whitegrass (Forest Protection) was established with community developed and enforced by-laws, while Forest Restoration and Improved Forest Management interventions involved undertaking vine clearance, maintenance, and replanting with a mix of agro-forestry and native species (all sites). Overall carbon benefits are estimated at 6,688 Mg CO₂e year⁻¹.

Conclusions

The study helps Vanuatu and other Small Island Developing States, particularly those of the South Pacific, strengthen the platform from which other PES projects could be developed and implemented.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s13021-026-00505-6.

Background

Vanuatu faces a wide range of pertinent environmental issues, of which deforestation, forest degradation and the impacts of invasive plant species are among the most pressing. The country has repeatedly ranked first in the World Risk Index [9, 65], with vulnerability particularly persistent for rural populations and permeating through activities essential for subsistence and wellbeing such as agricultural production. Effects are compounded by underlying societal vulnerability and high susceptibility to natural disasters and climate change [5, 49], with the inherent nexus of impacts leading to degenerative environmental feedback loops [41].

For example, the country has experienced more than a century of commercial forest exploitation beginning with the rapid depletion of sandalwood during the nineteenth century [59]. This was followed by extensive harvesting of other commercially valuable timber species, notably Pacific kauri (Agathis macrophylla) and tamanu (Calophyllum inophyllum), during the twentieth century [38]. Deforestation and degradation of old growth, primary tropical forests through activities such as logging and ad-hoc garden development have disrupted natural ecological processes and enabled invasive plants to proliferate and dominate affected areas. Forest regeneration is typically impeded as seedlings and saplings are smothered, while remaining exposed trees are vulnerable to the effects of windthrow as canopy gaps emerge [53]. Liana overgrowth can also overrun community gardens that rural communities rely on for sustenance, increasing labour-intensive maintenance needs.

To help address such issues, Payment for Ecosystem Service (PES) projects, primarily those focused on developing marketable benefits (e.g. ‘carbon credits’) from improving forest carbon stocks or preventing their loss, have been developed and implemented over the last ~30 years [28]. Successful projects have provided tens of thousands of people in developing countries with supplementary income through payments to design, develop, and conduct projects. If products, such as carbon credits traded on voluntary markets, find consistent and reliable buyers it also provides on-going financial support for project activities and management whilst continuing to improve environmental and societal outcomes.

The current study is associated with the Life & Land Project, a small-scale, community led PES initiative currently being implemented on the islands of Erromango and Tanna (Vanuatu) which incorporates Forest Restoration, Forest Protection, and Improved Forest Management interventions (see Box 1) under the voluntary market’s longest-standing forest carbon standard, Plan Vivo [61]. A key project requirement was to assess forest community structure, carbon stocks and health, and the land-use and land-cover change (LULCC) factors influencing spatial variability across the tropical forest land-use mosaics at each site. These data are critical for informing land-use decision making [12] and developing activity data and high-resolution, locally derived forest inventories for estimating Tier 3 emission factors (EFs) and potential carbon benefits. However, few primary studies have been conducted in Vanuatu (e.g. [22]) that report on forest carbon stocks, or changes in forest carbon stocks, resultant from LULCC interventions. The current study helps address knowledge gaps through:

1. Assessing carbon stocks and structural characteristics of tropical forests at project sites;

2. Analysing the drivers and rates of deforestation and forest degradation, and factors impacting forest health, to develop appropriate responses which reduce pressures (e.g. avoided deforestation and/or forest degradation via Forest Protection) and improve forest health (e.g. Forest Restoration, Improved Forest Management); and

3. Estimating potential project-derived carbon benefits from implementing responses using a carbon gain-loss approach [26].

Outcomes are intended to help the Government of Vanuatu and other large ocean states (LOS), alternatively known as small island developing states (SIDS)—particularly those of the South Pacific—with LULCC decision making, strengthening the platform from which other PES projects could be developed and implemented.

Box: Definitions for the various interventions reported in this study:

• Forest Restoration: The process of assisting the recovery of forest from a degraded state [45]

• Forest Protection: Prevention of deforestation and/or forest degradation [45]

• Improved Forest Management: Introduction of management practices that increase carbon stock of forest lands managed for wood production (PM001) and forest uses such as agroforestry where coffee, fruit and nut producing trees—which have a direct community utility—will be planted in a number of project sites/locations. This is considered a more appropriate reflection of the scope of interventions (as opposed to Forest Restoration) given project areas will primarily be managed to improve the yield and production of community resources

• Carbon benefits: Carbon benefits are synonymous with the term ‘emissions reductions and removals’ and used in this study to estimate potential increases in carbon from the implementation of interventions. An emissions ‘reduction’ equates to a reduction in atmospheric GHG emissions (atmosphere → land). It is typically used for avoided deforestation (Forest Protection) interventions where projected emissions that would have occurred in the baseline (or business as usual) scenario are reduced because of project activities. An emissions ‘removal’ is synonymous with carbon sequestration and represent the removal of GHGs from the atmosphere and applied here for Forest Restoration and Improved Forest Management interventions. This terminology is consistent with the approach adopted by the IPCC [27]

• Deforestation: The permanent conversion of forest land to non-forest land (e.g. where clear fell logging is followed by conversion to pasture or settlements, Food and Agriculture Organization [20])

• Forest degradation: The reduction in forest biomass, canopy cover, ecological integrity, or carbon stocks without changing the land-use category (e.g. forests are logged but subsequently regrow, or are degraded by invasive vine overgrowth. Food and Agriculture Organization [20])

Methods

Study area

The study focused on three sites in Vanuatu; Whitegrass (Site 1) and Middlebush (Site 2) in Central Tanna (Tanna Island), and South River (Site 3) on Erromango Island (Fig. 1). Whitegrass encompasses ~300 ha of agroforestry, grasslands, and primary and secondary tropical rainforest within a mixed landuse mosaic, while Middlebush is focused on the riparian margins of Imanaka Creek which have been degraded through the loss of canopy cover and vine overgrowth. South River is situated on an alluvial coastal plain and surrounded by steeply dissected, forested hills which display a series of raised riverine terraces.

Fig. 1.

Life and Land Project locations of South River (Site 3: 18°56′20.66"S, 169°10′30.04"E), Erromango, Whitegrass (Site 1: 19°26′24.37"S, 169°16′56.23"E) and Middlebush (Site 2: 19°27′52.51"S, 169°18′25.22"E), Tanna, Republic of Vanuatu

Project sites exhibit a tropical maritime climate characterised by high humidity, variable rainfall, and mean high temperatures that range from 19 °C in July–August to 24 °C in February on Tanna, and 21 °C in August and 31 °C in February on Erromango (Climate-Chart.com, n.d.). The dry season extends from May to October and the wet season, when cyclones can occur, from November to April [64].

Tanna is volcanic in origin, with sections of its coastline characterised by raised coral reef terraces formed through tectonic uplift [37]. On Tanna (Sites 1 and 2), the soils are mature, fertile, saturated andosols derived from volcanic ash deposited during the Pleistocene Manaaki Whenua – Landcare Research Manaaki Whenua – Landcare Research [29] and are frequently supplemented by fine ash deposits from Mt Yasur located in the southernmost part of the island. Erromango (Site 3) is also volcanic in origin and formed by an ancient underwater volcano, raised 100–300 m above sea level (a.s.l) by tectonic uplift [36].

Carbon stock and ecosystem health assessments

Stratification of project sites and forests

The geopolitical boundaries of project sites are delineated by traditional community ownership, management, and land use rights. These structures were used to define and demarcate project boundaries and the spatial scope of interventions. Assessment sites and forest strata were delineated according to the major vegetation types and canopy cover present as well as current and intended land uses using a random sampling design. Thirty-eight systematic flora plots were conducted across project sites, with 911 individual trees measured in total. We used a large number of relatively small plots across agroforestry, secondary, and old-growth tropical forests—rather than fewer, bigger plots—to better capture the small-scale heterogeneity typical of the shifting cultivation mosaic landscape in the study region [66]. Ground surveys conducted across the majority of the Forest Protection area were used to identify and spatially delineate recent forest clearance and degradation, including fine-scale and sub-canopy disturbances that may not be detectable using satellite imagery, and to document the associated drivers of land-use and land-cover change (LULCC) following the procedures outlined by Plan Vivo [44, 45]. Disturbances encountered were classified by type (e.g. garden creation or logging), with their extent and spatial distribution recorded using GPS-derived boundaries and/or point locations.

Field assessments of forest structural characteristics and biomass

Field assessments employed a modified version of plot-based methodologies advocated by Beets et al. [4] and Hurst et al. [25] to suit site characteristics (see Supplementary Materials 1 for a plot layout schematic). Square plots of 10 m × 10 m were established and all trees ≥1.4 m in height and with a DBH ≥5 cm recorded. Lianas over 2.5 cm in diameter were measured at 130 cm from the root—soil surface [52]. Larger plots were used at Middlebush (50 m × 40 m) to accurately characterise the sites riparian zone (20 m either side of the centre of the stream), and 30 m × 30 m plots were used at the Cattle Grazing/Coconut Plantation Zone (Whitegrass) given sparse tree coverage and density. All trees were measured for the riparian margins of South River and GIS analysis used to assign area (1.6 ha). Seedling and sapling (≤ 1.4 m in height, ≤5 cm DBH) abundance were assessed using three 1 m × 1 m, randomly selected, nested sub-plots. A scoping survey conducted in September 2023 identified a number of large, mature trees at Whitegrass that had been felled for timber along an approximately 350 m long and 5 m wide track. Tree species and diameter at cut height were recorded to estimate aboveground biomass loss.

Coarse woody debris (CWD) was estimated using a modified version of the planar line intercept technique [8, 63] that differed by measuring both the diameter and estimated length of large debris [4]. Two perpendicular, linear transects were established through the centre of each plot and any downed woody material (fallen and detached twigs, branches, or tree stems recorded. Fragments were classified into three categories based on diameter at intercept point,medium (> 2.5—7.5 cm), large solid (> 7.5 cm) or large rotten (> 7.5 cm). The same assessor was used on both transects to avoid double counting any overlapping pieces. Fine and small pieces of debris, along with leaf litter, were considered too numerous to assess.

Allometric equations for calculating biomass and carbon stocks

Given the absence of specific allometric equations of relevance for tree species of Vanuatu, the pan-tropical allometric equation developed by Chave et al. [14] was used to estimate above ground biomass (AGB_est):

1

where E is a measure of environmental stress and calculated based on temperature seasonality, climatic water deficit, and precipitation seasonality [14]. The value for E (0.1775431) was calculated in R with the ‘computeE’ function in the BIOMASS package, using the coordinates for Whitegrass, Tanna [48]. P is the density of wood (g/cm³), and D is diameter at breast height. Wood densities for individual tree species were sourced from the Global Wood Density Database [68]. While the equation is widely used, it does have inherent limitations and can produce large discrepancies in carbon stock estimates when compared to locally derived, species-specific allometric equations and wood density values [15, 39]. Below ground biomass was estimated by using the shoot: root ratio (1:0.24) developed by Cairns et al. [10] for tropical forests whilst noting limitations given ratios can vary as a function of stand age, edaphic characteristics, tree size, and site productivity [10, 33]. Coconut palm (Cocos nucifera) biomass was estimated through species-specific equations developed by Mugasha & Zahabu [34] that reference tree height (H) for above (Eq. 2) and below ground biomass (Eq. 3).

2

3

The biomass of lianas was calculated using Eq. 4 (Schnitzer et al. 2006).

4

Large Ficus obliqua and Ficus glandifera individuals exhibited extensive hollow stems and internal cavities, making application of standard allometric equations likely to substantially overestimate above-ground biomass. As no published allometric correction factors or destructive biomass studies were identified for hollow F. obliqua and F. glandifera spp., a conservative correction factor of 0.25 was applied to account for the substantial loss of internal woody volume observed in the field. This approach was adopted to minimise overestimation of biomass and carbon stocks and is acknowledged as a methodological assumption, noting these species comprised 4 records out of a total of 911 trees measured.

The volume (m³ ha⁻¹) of large woody debris was calculated using Huber’s formula based on diameter and length [3], while medium size classes were determined using quadratic mean diameter equations as described in Brown [8] and van Wagner [63]. Specific gravity (mass) of all CWD size classes followed Tue et al. [60] to convert debris volume to mass per unit area. Carbon content was calculated by multiplying both above and below ground biomass by a factor of 0.464 following IPCC [27] conventions. All biomass carbon pools were scaled to Mg C ha⁻¹.

Metrics for assessing ecosystem health

We used the revised Index of Structural Complexity (rCI) developed by Cameron et al. [13] as the main proxy for comparing the relative ‘ecological quality’, or current state, of assessment sites. When combined with in-situ surveys of the extent and scope of pressures (e.g. invasive, smothering vines), the rCI helps provide a numerically based target to gauge intervention progress. It includes the indicators of tallest tree height within a site (H), basal area (BA), tree density (D), and species diversity (N originally proposed by Holdridge [23] while also adding living biomass (B, Mg C ha⁻¹), coarse woody debris biomass (CWD, Mg C ha⁻¹), canopy cover (CC, estimated as the % of canopy obscured by foliage), and the number of structural classes present (SC). Each indicator is transformed using the natural logarithm (In) and divided by 10^{^2} to ensure results are manageable (Eq. 5).

5

Hill’s numbers (q = 0, q = 1, q = 2), the Shannon-Weiner Species Diversity Index (H’), and the Pielou Evenness Index (J'') were also calculated and used as proxies for forest quality. The effective Shannon diversity (q = 1) is calculated by the relative abundance of each species recorded, and effective Simpson diversity (q = 2) is the effective number of dominant species in a community. H’ was calculated from the number of tree species recorded in plots per site, the proportion of species relative to the total number of individuals, and sums the proportion multiplied by the natural log of the proportion for each species. J’ is a measure of how evenly species are distributed in a community and calculated using H’ and the total number of species in a sample.

Statistical analysis

Site means were calculated for each variable before applying a Shapiro–Wilk test which identified a non-normal data distribution across variables (e.g. biomass, seedling/sapling density, tree density, DBH, CWD, and basal area). Data was logarithmically transformed to meet normality assumptions, and a one-way ANOVA was used to compare variables between sites. A post—hoc Tukey honestly significant difference test was applied to identify sources of significant differences using site-based pairwise comparisons. Tree density remained non-normal after logarithmic transformation, and Kruskal–Wallis and post-hoc Dunn tests were used instead to identify site differences. The significance threshold (α) was set at 0.05 for all tests, and data reported as the mean ± the standard error.

Development of emissions factors and activity data

Landcover change assessment at Whitegrass (Site 2) to determine drivers and rates of forest degradation and loss

Normalized Difference Vegetation Index (NDVI) analysis was used to corroborate estimates of the rate and extent of deforestation and forest degradation at Whitegrass, which were primarily determined through in situ ground assessments (Sect. "Stratification of project sites and forests") and community consultation (Sect. "Community consultation"). Areas of Interest (AOIs) across sites in Tanna were defined, with AOI-1 encompassing the Forest Protection Zone at Whitegrass and AOIs 2–4 selected as reference regions. NDVI was calculated for each AOI from 2020 to 2025 using the Sentinel-2 bands B8 (NIR) and B4 (Red):

6

Images with >20% cloud cover were excluded, and a cloud mask was applied to the remaining imagery. The NDVI difference (ΔNDVI) was calculated using 2019 as the baseline year due to incomplete data availability for earlier years. Vegetation loss was identified using a threshold calculated as μ ± n·σ, where σ is the standard deviation of the ΔNDVI image and n is a visually selected scaling factor [31]. Values of n between 1 and 2 have been shown to reliably detect natural and anthropogenic forest change across different forest ecosystems [31]. Although a Kolmogorov—Smirnov test indicated that ΔNDVI values were non-normally distributed, visual assessment identified 1σ as the most appropriate threshold for detecting vegetation loss, corresponding to a ΔNDVI value of 0.048. Forest loss was subsequently estimated for each AOI as the area (ha) with ΔNDVI values exceeding 0.048.

Estimation of potential carbon benefits from implementing project interventions

Emissions factors (EF) and activity data (AD), developed in accordance with IPCC [27] guidance, were used to estimate potential carbon benefits from implementing Forest Protection, Improved Forest Management, and Forest Restoration interventions using the gain–loss approach [26]. The gain—loss approach quantifies carbon stock changes as gains from sequestration (e.g. tree growth) and losses through emissions (e.g. deforestation or forest degradation) for each activity type [69]. EFs were calculated from differences between current carbon stocks measured in situ at each site and projected carbon stocks derived from empirically observed differences among the forest types and conditions assessed. For Improved Forest Management and Forest Restoration interventions we use a linear timeframe for stand regrowth of 66 years for secondary forests to obtain the equivalent biomass of mature reference forests based on a meta-analysis of secondary tropical forest regrowth by Poorter et al. [46]. For the Forest Protection intervention at Whitegrass (Site 1), EF are calculated based on measured carbon stocks within the mature forest and logged tract multiplied by AD, or the estimated rate (ha⁻¹ year⁻¹) of deforestation and forest degradation (Sects. "Stratification of project sites and forests" and "Landcover change assessment at Whitegrass (Site 2) to determine drivers and rates of forest degradation and loss").

Community consultation

A series of structured Driver-Pressure-State-Impact-Response workshops facilitated by the authorship team and Nasituan were held between 2023–2025 involving dozens of community members from Whitegrass (representatives from Lonamilo, Lowlipang), Middlebush (representatives from Louahao, Lounahuru, Iwel, and Lamnatu), and South River (representatives from South River, Port Rosak, and Antioch). Mapping and discussion-based activities identified links between social and environmental changes and explored desired future outcomes. Appropriate responses were then developed as Forest Protection, Forest Restoration, and Improved Forest Management interventions. The Life & Land Project’s approved Project Design Document, publicly available on the Plan Vivo website (https://www.planvivo.org/projects/life-and-land-vanuatu), provides a full description of the consultative process. The workshops also helped to centralise the role of traditional governance in the project through revisiting historical land management practices and motivations, with the harmony between protecting biodiversity and forest health and using natural resources at the forefront Table 1.

Table 1.

Structural characteristics and carbon stocks of sites assessed in this study

Location	Whitegrass (Site 1)						Middlebush (Site 2)		South river (Site 3)
Site	1. (a) Mature tropical rainforest (a)		1. (b) Logged tropical rainforest trees (b)		2. Coffee grown under native shade trees (c)	3. Cattle grazing/coconut plantation area (d)	4. Reforestation of Imanaka Creek riparian margins (e)		5. Agroforestry zone (f)		6. Forest enhancement zone (g)		7. Erromango Kauri Reserve (h)		8. Reforestation of South River riparian margins (i)
Metadata/general forest descriptors
No. plots	18		N/A		N/A	3	4		2		6		5		N/A
Species dominance per site/overall*	Macaranga sp.		Dysoxylum bijugum		Coffee	Coconut	Ficus sp., Coconut, Acalypha sp.		Hibiscus tiliaceus (17.6%), Semecarpus vitiensis (14.7%)		Macaranga dioica (37.6%), Hibiscus tiliaceus (17.6%)		Hopea sp. (12%), Calophyllum neoebudicum (9.8%)		Naval (25.5%), coconut (23.4%)
Species richness	51		7		N/A	1	24		18		24		32		13
No. vegetational structural classes**	5		N/A		3	2	2		4		3		5		1
Max height (m)	27		N/A		N/A	23.9	24		35		40		45		24
Canopy cover avg. (%)	71		N/A		75	10	11		55		84		70		5
Average DBH (cm)	20.2 ± 4.4 (b)		80.5 ± 9.1 (g, h)		3.7 ± 0.2 (Coffee), 28.7 ± 2.9 (Shade trees)	36.6 ± 1.5 (g, h)	29.0 ± 17.0 (g)		22.4 ± 6.5		14.5 ± 4.1 (i)		17.2 ± 4		45.3 ± 3.2
Basal area (m2 ha−1)	97.5 ± 3.0 (d, g, i)		94.8 ± 0.8		35.7 ± 0.1	11.4 ± 0.1 (f, h)	37.5 ± 1.8		141.2 ± 4.7 (g, i)		28.0 ± 0.8		60.8 ± 2.6 (i)		6.5 ± 0.0
Tree density (trees ha−1)	1,833.3 ± 143.8 (d, e, i)		143 ± 0.1		933 ± 91.7 (d, e, i)	111 ± 6.4 (h)	115.0 ± 34.3 (g, h)		1,700 ± 400		1,278 ± 296.5		1,840 ± 399.5 (i)		32.9 ± 0.0
Seedlings/sapling density per plot (stems per ha−1)	249,630 ± 71,855		N/A		N/A	N/A	N/A		135,000 ± 20,209		114,444 ± 51,001		152,667 ± 36,255		N/A
Biomass (above and below ground biomass, coarse woody debris), Mg C ha−1
Live and dead trees	431.7 ± 17.2 (d, g)		714.0 ± 7.0 (d)		131.2 ± 0.2	51.2 ± 0.1 (f)	107.3 ± 10.7		679.9 ± 26.5 (g)		82.1 ± 3.3		251.8 ± 12.6		34.1 ± 0.1
CWD	59.0 ± 15.4		N/A		N/A	N/A	N/A		49.2 ± 0.1		42.1 ± 18.8		30.6 ± 10.6		N/A
Total biomass	490.7 ± 32.6 (d, e, g)		714.0 ± 7.0 (d)		131.2 ± 0.2	51.2 ± 0.1 (f)	107.3 ± 10.7 (f)		729.0 ± 26.6 (i)		128.2 ± 22.7		282.4 ± 23.2		34.1 ± 0.1
Structural metrics and indices
Revised Index of Structural Complexity (rCI)	816.8		N/A		N/A	0.2	14.4		620		225.7		527		0.3
Hill diversity (q = 1, q = 2)	26.3	16.7	4.9	3.8			16.2	12.3	14.9	11.6	11.2	5.8	7.7	5.7	23.4	18
Shannon-Weiner Index of diversity (H'')	3.38		1.6		N/A	N/A	2.78		2.66		1.8		3.14		2.04
Pielou Evenness Index (J'')	0.86		0.82		N/A	N/A	0.88		0.92		0.57		0.91		0.79

^*Most abundant tree spp. recorded within plots

^**Out of: Emergent, canopy, subcanopy, understory, and groundcover

Significant differences from ANOVA post-hoc Tukey HSD pairwise results for structural metrics identified in superscript

Results

Carbon stocks and structural composition of forests

Statistically significant, high variability in forest carbon stocks and structural composition was observed between sites (Fig. 2, Table 2), while 129 distinct tree species were recorded (Supplementary Materials 2). ANOVA results, for example, showed significant differences in total carbon stocks between sites (F = 6.48, DF = 8, p = 5.21e−5) reflecting differences in landuse, landcover and environmental pressures. The mature, healthy, predominantly primary tropical rainforest of Whitegrass had an average carbon stock of 490.7 ± 32.6 Mg C ha⁻¹, 12% of which was contained in CWD, while the large trees extracted for logging had an estimated biomass (above and below ground, exclusive of CWD) of 714 ± 7.0 Mg C ha⁻¹. Carbon stocks were significantly higher than the Cattle Grazing/Coconut Plantation Zone (p = 0.001), Imanaka Creek of Middlebush (p = 0.027), and the Forest Enhancement Zone of South River (p = 0.018).

Fig. 2.

Site photos. (a) Mature, healthy primary tropical rainforest Whitegrass; (b) Damage from logging at Whitegrass; (c) Recent forest clearance for garden development at Whitegrass; (d) Quary development on Mount Tagen at Whitegrass; (e) Invasive liana overgrowth at Imanaka Creek, Middlebush; (f) Fallow gardens overrun with Big leaf rope in the Forest Enhancement Zone of South River; (g) Degraded forest dominated by Macaranga spp. and overrun with Big leaf rope at the Forest Enhancement Zone, South River; (h) Riparian margins of South River

Table 2.

Initial forest area, area of deforestation and forest degradation, and annual rate of loss

Scope	Forest area at project start	Area converted to non-forest/degraded (2020–2025)	Average annual % loss
AOI-1	748.29	43.14	1.15%
AOI-2	997.16	63.36	1.27%
AOI-3	973.38	78.15	1.61%
AOI-4	984.12	28.84	0.59%
Avg. AOI 2–4	984.89	56.78	1.15%

AOI-1 encompasses 1,055 ha and inclusive of the project and leakage area at Whitegrass

At the shade coffee site, almost all biomass was contained within mature trees (130.0 ± 0.5 Mg C ha⁻¹), with a 1.49:1 ratio of understory coffee to shade trees and stem densities of 558.3 ± 54.8 and 375.0 ± 36.9 stems ha⁻¹ respectively. The degraded, invasive vine smothered forests of Imanaka Creek (Middlebush) and the Forest Enhancement Zone (South River) had similar carbon stocks of 107.3 ± 10.7 Mg C ha⁻¹ and 128.2 ± 22.7 Mg C ha⁻¹ respectively, while also displaying comparatively low basal area (37.5 ± 1.8 m² ha⁻¹ and 26.6 ± 0.8 m² ha⁻¹). Notably, almost a third of total biomass of the Forest Enhancement Zone was necromass (42.1 ± 18.8 Mg C ha⁻¹). Primary forest in Erromango also displayed comparatively high biomass carbon stocks, with 282.4 ± 23.4 and 729.0 ± 26.6 Mg C ha⁻¹ recorded for the Kauri Reserve and Agroforestry Zone respectively. In contrast, the low density (111 ± 6.4 trees ha⁻¹) of coconut trees within the Cattle Grazing/Coconut Plantation Zone resulted in a corresponding low biomass (51.2 ± 0.1 Mg C ha⁻¹), noting that almost a quarter is composed of fallen trees. This is similar to the overall biomass recorded for the grassland-dominated riparian margins of South River (34.1 ± 0.1 Mg C ha⁻¹, 32.9 ± 0.0 trees ha⁻¹).

Structural complexity and diversity indices followed a similar yet more nuanced pattern than biomass results, with basal area, tree height, canopy cover, and the number of structural classes in the mature primary tropical forest of Whitegrass (in addition to CWD and live and dead tree biomass) all contributing to higher-end rCI results (rCI = 816.8). The diversity of species recorded (51) resulted in the highest effective Shannon diversity (q = 1, 26.3) and H’ scores (3.38), but not J’(0.86) or effective Simpson diversity (q = 2, 16.7). At the other end of the rCI ecological quality spectrum were the Cattle Grazing/Coconut Plantation Zone (Whitegrass) and riparian margins of South River (0.2 and 0.3 respectively), with results driven by low scores across all indicators within these low density, mixed agroforestry-grassland landscapes. The rCI of the Forest Enhancement Zone of South River (226.9) was less than a third of the Erromango Kauri Reserve and Agroforestry Zone (527 and 620 respectively), with the effective Shannon diversity (q = 1, 11.2), effective Simpson diversity (q = 2, 5.8), H’ (1.8) and J’ (0.57) scores similarly lower. The vine smothered margins of Imanaka Creek recorded an even lower rCI score of 14.4 than the Forest Enhancement Zone, despite displaying greater species diversity (q = 1, 16.2; q = 2, 12.3; and H’ = 2.78) and evenness (J’ = 0.88) Fig. 3.

Fig. 3.

Carbon stocks of sites assessed in this study. Note that values above zero indicate carbon stored in above ground biomass and coarse woody debris, and values below zero indicate carbon that is stored in belowground biomass

Rates and drivers of deforestation and forest degradation, and factors impacting forest health

Whitegrass (site 1)

Results from NDVI analysis revealed annual rates of forest loss at Whitegrass and the immediate surrounding area (AOI-1) of 1.15%, or 8.6 ha per year⁻¹, which was the same average rate detected for reference areas (AOI-2, 3, and 4. Table 2 and Fig. 4).

Fig. 4.

Top: Areas of Interest (AOIs) used in the land use change analysis. AOI1 includes both the Whitegrass project and the surrounding area to determine ‘leakage’, while AOI2-4 represent regions with similar deforestation and forest degradation pressures. Bottom: Change in NDVI from 2019 to 2025 for each AOI, using a threshold of μ ± 1·σ

Ground truthing showed similar patterns, with logging directly observed along a ~350 m long, ~5 m wide track (1,750 m²) enabling truck access within core primary forest (Fig. 4. a). Twenty-five large trees had been felled, with significant damage to adjacent forest evident from fallen trees and extending ~10 m from the stump. The diameter at cut height ranged from 25 cm-168.5 cm, with the majority of trees (18) >50 cm (average diameter 80.5 ± 9.1 cm). Additionally, a further 150 m section of the logging track had a similar observed density of logged trees, although records of diameter at cut height and species were not taken. The total extent of damage from logging was estimated to be ~1 ha.

Field observations also revealed several patches of forest that had recently been disturbed to create gardens, with the undergrowth cleared and a number of sub-canopy trees (≥ 20 cm DBH) felled within discrete areas of approximately 50 m × 50 m (0.25 ha). Consultation with the Whitegrass community (see Sect. "Community consultation") revealed that the 29 families with land titles each clear approximately 3—4 gardens per year, with an average size of roughly 20 m × 20 m (400 m²). When scaled to a yearly rate and extent of clearance (3 gardens per family, 29 families, 400 m² per garden), this equates to an estimated annual clearance of 34,800 m² (3.48 ha⁻¹ year⁻¹). For the Life & Land Project, the estimated rate of forest degradation through garden development is conservatively scaled back to 1.5 ha⁻¹ year⁻¹ (0.61% year⁻¹).

Quarry operations and associated infrastructure were another key driver of forest loss and degradation at Whitegrass (Fig. 4d). The China Civil Engineering Construction Corporation (CCECC) was granted a permit from the Vanuatu Government (Public Works Department and Department of Geology and Mines) to extract material from both Mount Tangen (within the project area) and Mount Toualeu (outside the project area). The roading network on north and central Tanna is largely unsealed, with constant replenishment of base material required to maintain use and functionality given high rates of rainfall and erosion. This means that base material will continue to be sourced from quarries, including that on Mount Tangen. The current quarry footprint on Mount Tangen is 1.21 ha, and the total permitted extraction footprint is 18,780m² (1.878 ha). The quarry footprint is projected to expand southward, clearing a further 0.66 ha and leading to the loss of topsoil, alluvial stones, native trees, grasslands, and wild cane resources. The ancillary impacts of quarrying (e.g. continued road works, infrastructure development) are likely to manifest through edge effects, wind throw, and on-going, incremental forest loss, fragmentation, and degradation.

On-going, incremental deforestation, fragmentation and degradation from logging and garden development (1.02% year⁻¹, 3.05 ha⁻¹ year⁻¹) and quarry operations (0.83 ha⁻¹ year⁻¹) is likely to result in the loss of an estimated 3.55 ha⁻¹ year⁻¹ of primary and secondary forest at Whitegrass in the absence of any intervention.

Field surveys identified multiple additional factors impacting forest health and condition at sites aside from forest degradation through logging and garden development and deforestation from quarrying (Table 3, Fig. 4). Smothering and overgrowth of forests by invasive vines was the primary pressure at Imanaka Creek (Middlebush) and the Forest Enhancement Zone (South River), while livestock grazing prevents the establishment of seedlings to replace lost trees at the Cattle Grazing/Coconut Plantation Zone (Whitegrass) and riparian margins of South River.

Table 3.

Site characteristics, forest health, and key impediments to restoration in the absence of project interventions (‘business as usual’ scenario)

Assessment site	Project intervention type	Site description and key impacts
Whitegrass, Tanna (Site 1)
1. (a) Mature tropical rainforest; 1(b) Logged tropical rainforest trees	Forest Protection (avoided deforestation/forest degradation from logging, gardens, and quarry development)	The old growth, mature forest at Whitegrass (Fig. 2.a) covers ~300 ha (excluding fallow gardens, cleared areas, housing, and infrastructure) and is currently subjected to ad-hoc logging (Fig. 2. b) and on-going, intermittent clearance for garden development (Fig. 2.c). Additionally, the northern flank of Mount Tangen and associated secondary forest has been removed through quarry operations (Fig. 2.d) which is likely to lead to on-going forest loss and degradation through exacerbating edge effects and windthrow
2. Fallow gardens	Improved Forest Management (fallow gardens → shade coffee), and Forest Restoration (fallow gardens → native forest)	Existing fallow gardens are currently largely overgrown with smothering vegetation, including invasive vines such as big leaf rope (Merremia peltata) and blue morning glory (Ipomoea indica) which hinders productivity and prevents forest development through the growth and recruitment of seedlings
3. Cattle grazing/coconut plantation area	Improved Forest Management	There are ~24 ha of open grasslands used for cattle grazing at Whitegrass interspersed with sporadic coconut and fruit trees (e.g. lemon) at low densities (Fig. 2.e). An estimated ¼ of mature coconut trees have been lost through tropical cyclone wind-throw, while persistent grazing pressure prevents the growth and establishment of new coconut and fruit trees
Middlebush, Tanna (Site 2)
4. Reforestation of Imanaka Creek riparian margins	Forest Restoration	The riparian margins of Imanaka Creek at Middlebush (Fig. 2.f) are currently almost entirely smothered in vines and creepers, with blue morning glory (Ipomoea indica) predominant. The stream itself exhibits intermittent surface flows, and while feed by a number of springs, the actual stream course is now largely undefined given liana overgrowth which prevents the scouring of stream bed material necessary to develop and maintain channels and enable permanent flow. There are sporadic occurrences of emergent trees, mostly coconut (Cocos nucifera), Acalypha spp. and Ficus spp., along parts of the stream course, although relative tree density, basal area and hence biomass remain comparatively low. Macaranga spp. are the most prevalent tree species (14.1% of all trees recorded)
South River, Erromango (Site 3)
5. Agroforestry zone (planting of fruit and nut producing trees)	Improved Forest Management	Across the river from the community of South River, on the true left bank, is an alluvial plain overlying karst bedrock. The forest in this area is an important habitat for coconut crab (Birgus latro). There are a number of large, mature remnant native trees within this area interspersed by an understory dominated by Hibiscus tiliaceus (17.6%)
6. Forest enhancement zone (degraded forest and fallow gardens)	Forest Restoration	The area designated for forest enhancement is located on the true left bank of South River along a former river terrace that forms a plateau ~200 m asl. Approximately half of this site (5 ha) is composed of open forest dominated by Macaranga dioica (47%) which is subject to significant liana loading and vine/creeper overgrowth from species such as mile a minute (Mikania micrantha), big leaf (or American) rope (Merremia peltata), and Blue morning glory (Ipomoea indica. Figure 2.g). Intermittent, fallow gardens degraded by on-going pig damage and vine/creeper overgrowth occupy the other half of the forest (5 ha. Figure 2.h). This forest enhancement area used to be referred to by the community as ‘dark bush’, a term used to describe mature, open, tall forest with high canopy coverage that is easy to walk through. Damage from successive tropical cyclones and the ad-hoc removal of mature canopy tree species through logging have led to the proliferation of invasive vines as canopy gaps emerge
7. Erromango Kauri Reserve	N/A (reference forest for South River)	The Erromango Kauri Reserve is located on the western coast of Erromango near the village of Happy Land. The reserve comprises the upper catchments of the Lampunari river and of the Pongkil river, with the highest peak of Mount Vet Pop at 802 m a.s.l. The reserve was initially proposed back in the early 1970’s as means to protect a 3,257 ha remnant block of primary forest dominated by emergent Pacific Kauri (Agathis macrophylla) with associations of Calophyllum neoebudicum, Hernandia moerenhoutiana, Weinmannia denhamii, Syzygium spp. and Bischofia javanica from logging by the French company Societe Agathis [58]. The company ceased operations in 1974 without logging the area, and today it remains as one of the few remaining stands of unlogged primary forest on Erromango. In this study, it is used as a reference forest to assess potential forest structural composition and biomass carbon stocks resultant from restoration activities within the Forest Enhancement Zone at South River
8. Reforestation of South River riparian margins	Forest Restoration	The riparian margins of the true right bank of South River are currently in pasture to enable grazing by cattle (Fig. 2.i). There are a number of trees directly adjacent to the riverbank, although they are not at sufficient density to consolidate stream banks or prevent erosion. Coconut trees have also been planted sporadically across this area, with a number of mature trees lost through windthrow impacts due to successive tropical cyclones. Similarly to Whitegrass, cattle grazing prevents the establishment and growth of seedlings and saplings

Estimated carbon benefits using a gain–loss approach

Overall carbon benefits from the Life & Land Project are estimated at 6,688 Mg CO₂e year⁻¹ (Supplementary Materials 4). The majority of projected benefits (86.37%) are emissions reductions derived from Forest Protection at Whitegrass, with Forest Restoration and Improved Forest Management providing 10% and 3.67% of overall benefits through emissions removals respectively (Table 4).

Table 4.

Carbon benefit estimates for project interventions

Location	Whitegrass (Site 1)					Middlebush (Site 2)	South river (Site 3)
Site	1. (a) Mature tropical rainforest (logging and gardens)	1. (b) Mature and secondary tropical rainforest (quarry)	2. (a) Fallow gardens converted to shade coffee	2. (b) Fallow gardens converted to native forest	3. Cattle grazing/coconut plantation area	4. Reforestation of Imanaka Creek riparian margins	5. Agroforestry zone	6. (a) Forest enhancement zone (fallow gardens)	6. (b) Forest enhancement zone (degrade forest)	8. Reforestation of South River riparian margins
Intervention	Protection		Improved Management	Restoration (gardens → forest)	Improved Management	Restoration	Improved Management	Restoration
Extent (ha)	299.7	25	5.1	4.8	24.2	8.5	2.8	5	5	1.6
Rate of loss per annum (ha)	3.05	0.83	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
Biomass emissions removals, Mg C ha−1	N/A	N/A	182.5	482.9	51.2	383.3	377.7	498.1	377.7	471.7
Biomass emissions reductions, Mg C ha−1	419.1	293.7	N/A	N/A	N/A	N/A	N/A	N/A	N/A	N/A
NPP, Mg C ha−1 year−1	N/A	N/A	2.8	7.3	0.78	5.8	5.7	7.5	5.7	7.1
Soil C emissions removals/reductions*, Mg C ha−1 year−1	N/A	63.2	1.04	1.04	0.22	1.04	0.5	0.5	0.5	0.4
(a) Total C gain/avoided loss, Mg C ha−1 year−1	419.1	356.9	3.8	8.4	1.0	6.8	6.2	8	6.2	7.6
(b) Total C gain/avoided loss, Mg CO2e ha−1 year−1**	4,714	1,092	14	30.7	3.7	25.1	22.7	29.3	22.7	27.9
(c) Mg CO2e ha−1 year−1 total***	4,714	1,092	71	166.7	89	213	64	147	113	45

*Soil organic carbon (SOC) burial rates (Mg C ha−1 year−1) are sourced from Table 5,[17] where SOC is divided by time after land use change as follows:

• Grassland to secondary forest: 0.4 Mg C ha−1 year−1;

• Cropland to secondary forest: 1.04 Mg C ha−1 year−1;

• Primary forest to cropland: 0.7 Mg C ha−1 year−1; and

• Primary forest to secondary forest: 0.5 Mg C ha−1 year−1.

For ‘Primary forest’ to ‘Cropland’ and ‘Secondary Forest’, we assume the reverse applies whereby ‘Cropland’ and ‘Secondary’ to ‘Primary Forest’ enhances soil carbon burial by 0.7 and 0.5 Mg C ha⁻¹ year⁻¹ respectively

^**(b) Conversion of Mg C to Mg CO₂e = 12/44 (or *3.670)

^***(c) Improved Management and Restoration = Extent (ha) * Total C gain. (c) Protection = annual rate of loss (ha) * Total C gain

^Conservatively assumes soil carbon burial is half the rate of ‘grassland to secondary forests’

NPP = net primary productivity (annual rate of biomass gain), C = Carbon, N/A =not applicable

Discussion

Forest carbon stocks, structural composition, and comparison to tropical forests of the Indo-Pacific

The wide range in carbon stocks, rCI values, and diversity indices observed between sites is directly attributable to the extent of disturbance regimes with the healthy, intact, and diverse mature tropical rainforests of Whitegrass and the Erromango Kauri Reserve (South River) juxtaposed against the vine smothered margins of Imanaka Creek (Middlebush) and the Forest Enhancement Zone (South River). The ability of forests to repair and regrow post-disturbance (e.g. from logging, garden development, or windthrow damage) can be severely limited by the presence and prevalence of vines that prevent new growth [51], while the productivity and growth of remaining trees can be restrained by lianas which quickly proliferate as light wells open up [30]. At the Forest Enhancement Zone in particular, impacts are compounded by high volumes of coarse woody debris (largely downed Macaranga spp.) while liana overgrowth (mostly Merremia peltata) almost completely covers the canopy, preventing light from reaching the ground and limiting access for birds and bats.

The average biomass of the mature primary tropical forest at Whitegrass is high when compared to reported values in primary tropical forests of the Indo-Pacific (range 170.2—356.4 Mg C ha⁻¹, Table 5). Median carbon stocks (293.7 ± 7.24 Mg C ha⁻¹) are more similar to those reported for tropical Asian rainforests (226.3 ± 50.7—237 ± 59.6 Mg C ha-1, [55] and [24] respectively) than they are for Australian and Oceania tropical forests (149.9 ± 2.2—175.1 ± 1.2, [24]). This is largely due to the skewering effect produced by two plots (out of 18 in total) located on an elevated river terrace where large, high biomass trees were recorded (e.g. Bischofia javanica and Dysoxylum bijugum with DBHs of 118.1 cm and 106.6 cm respectively), leading to high carbon stocks (1,565.2—1,655.6 Mg C ha⁻¹). Similarly, the estimated biomass of felled trees along the Whitegrass logging track is over two thirds higher than upper ranges reported in comparable literature (e.g. [55] and [24]). However, results are considered an accurate reflection of overall carbon stocks given the high spatial heterogeneity that often occurs within primary tropical rainforests. Large trees commonly cluster in intermittent, highly localised patches (such as that observed along the logging track and riverine terraces) with an uneven distribution typically associated with hydraulic and edaphic properties conducive to high productivity. This is evident in geomorphic settings such as river terraces, stream valleys, or older alluvium deposits due to enhanced nutrient availability and water access [32].

Table 5.

Comparison of biomass carbon stocks (Mg C ha⁻¹, mean ± SE) between primary tropical rainforest sites assessed in this study and selected published studies of tropical forest ecosystems from the Indo-Pacific region

Location and country	Forest type	Dominant species	Canopy height (m)	AGB	BGB	Total biomass	Data source
Whitegrass, Vanuatu	Primary tropical forest with intermittent pockets of secondary, regenerating forest	Macaranga spp. (11%); Dendrocnide latifolia (8.5%); Myristica inutilis (7%)	35	407.2 ± 29.2	83.5 ± 3.3	490.7 ± 32.6	This study
Erromango Kauri Reserve, Erromango, Vanuatu	Primary tropical forest	Hopea spp. (12%), Calophyllum neoebudicum (9.8%)	45	233.7 ± 20.7	48.7 ± 2.4	282.4 ± 23.2	This study
Vanuatu	Dense tropical rainforest	N/R	N/R	56.2 ± 29.1	18.1*	74.3 ± 29.1	Ministry of Forestry, Vanuatu (2022)
Vanuatu	Open tropical rainforest	N/R	N/R	40.0 ± 25.8	12.9*	52.9 ± 25.8	Ministry of Forestry, Vanuatu (2022)
Malua Forest Reserve, Sabah, Malaysia	Regenerating tropical lowland dipterocarp rainforest (selectively logged)	Dipterocarpaceae spp.; Macaranga spp.	N/R	105.1 ± 6.4	16.5 ± 0.5	121.6 ± 6.9	Saner et al. [50]
Danum Valley Conservation Area, Sabah (Borneo), Malaysia	Tropical lowland primary rainforest primary	Dipterocarpaceae spp.	N/R	207 ± 71	49.7**	256.7 ± 71	Asner et al. [1]
Wanang Conservation Area, Papua New Guinea	Tropical lowland primary rainforest	Pometia pinnata (Sapindaceae), Mastixiodendron pachyclados (Rubiaceae), Celtis latifolia (Cannabaceae)	N/R	137.3 ± 14.0	32.9**	170.2	Peck et al. [43]
Robson Creek, Queensland, Australia	Tropical primary rainforest	Lauraceae spp., Rutaceae spp., Proteaceae spp., and Elaeocarpaceae spp.	N/R	194.2	46.6**	240.8	Bradford et al. [7]
North-east Queensland, Australia	Tropical primary rainforest	N/R	N/R	287.4	69**	356.4	Murphy et al. [35]
Pasoh Forest Reserve, Peninsular Malaysia	Tropical lowland primary rainforest	Dipterocarpaceae spp.	N/R	248.7	44.5	293.2	Niiyama et al. [40]
Rewa Delta, Fiji	Deltaic mangroves	B. gymnorrhiza; Rhizophora spp.	10–25	234.5 ± 29.2	56.0 ± 6.5	290.6 ± 35.7	Cameron et al. [12]
Ba delta, Fiji	Deltaic mangroves	R. selala; R. stylosa	2–20	58.8 ± 6.7	5.3 ± 1.3	64.2 ± 8	Cameron et al. [12], weighted average of scrub and riverine mangroves within the delta
South-eastern coast, Viti Levu, Fiji	Riverine and coastal fringing mangroves	B. gymnorrhiza	15–25	190.7 ± 42.7	47 ± 8.8	237.7 ± 51.4	Cameron et al. [12]
Regional estimates for primary tropical rainforest ecosystems
Australia	Tropical rainforest	N/R	N/R	120.9 ± 2.2	29.0	149.9 ± 2.2	Hunka et al. [24]
Asia	Tropical rainforest	N/R	N/R	191.7 ± 59.6	46**	237 ± 59.6	Hunka et al. [24]
Asia	Tropical rainforest	N/R	N/R	182.5 ± 50.7	43.8	226.3 ± 50.7	Slik et al. [55]
Oceania	Tropical rainforest	N/R	N/R	141.2 ± 1.2	33.9	175.1 ± 1.2	Hunka et al. [24]

^*BGB calculated using the root to shoot ratio (1:0.323) reported by MoF. No standard error (±) was reported for BGB

^**BGB calculated using the root to shoot ratio (1:0.24) developed by Cairns et al. [10]

AGB is inclusive of DDWD where reported

N/R = not reported

Developing appropriate responses to reduce deforestation and forest degradation and improve forest health

The findings demonstrate that Forest Protection, Forest Restoration, and Improved Forest Management interventions developed under the Life & Land Project are directly aligned with the observed drivers of forest loss and degradation and the structural condition of forests at each site, with multiple ecological benefits that can be realised through community implementation.

At Whitegrass, estimated rates of deforestation associated with logging and quarry impacts (0.4% and 0.6%, respectively) exceeded the average annual forest loss reported by [21] for West Tanna, Middlebush, and North Tanna combined (0.084%,0.21 ha⁻¹ year⁻¹). Both estimates, however, are likely to underestimate the contribution of forest degradation. Garden creation and selective clearance frequently occurred at small, spatially heterogeneous scales and involved sub-canopy vegetation removal while retaining canopy cover, making these disturbances difficult to detect using remote sensing [67]. Rapid vegetation regrowth, canopy closure of over logged areas, and liana smothering may further constrain NDVI-based detection where spectral reflectance becomes indistinguishable from intact forest. For example, the logging track at Whitegrass (Site 1) – which, by definition, constitutes a change in landuse via deforestation—could not be detected through drone reconnaissance ~1 ½ years post-logging due to canopy closure and rapid regrowth. Our findings, therefore, highlight the critical importance of integrating ground-based assessments and community knowledge when quantifying fine-scale forest degradation and its drivers for projects involving small-scale disturbance.

Logging and quarry development also generated impacts beyond direct tree removal. Tree felling and access-track construction damaged adjacent vegetation, while quarry-associated forest fragmentation and road development created edge conditions that facilitate structural degradation and invasive species establishment. Edge effects can reduce forest biomass and carbon stocks substantially [42], while roads provide high-light, low-competition corridors that favour liana proliferation [47]. This was directly observed at Whitegrass, where Ipomoea indica had established 20—50 m into mature forest adjacent to recently widened and constructed quarry roads. The establishment of Forest Protection areas restricting logging and new garden creation therefore directly addresses the principal anthropogenic drivers identified in this study. Complementary timber and fuelwood planting may further reduce extractive pressure on mature forest, while vegetated buffers adjacent to quarry areas could reduce edge and wind exposure.

The structural condition of fallow gardens and degraded forests also provides a clear rationale for active Forest Restoration and Improved Forest Management. At Whitegrass, fallow gardens were characterised by dense liana thickets that inhibited tree establishment, while degraded riparian forest at Middlebush and open forest at South River were dominated by Macaranga spp., taxa commonly associated with early successional and disturbed tropical forests [16]. Similarly, the understory of degraded agroforestry forest at South River was dominated by Hibiscus tiliaceus, a species associated with disturbed coastal and riverine forests [57]. These findings indicate arrested or early successional forest states in which targeted liana clearance and enrichment planting of native canopy and agroforestry species may accelerate structural recovery and carbon accumulation. Liana removal may also increase the growth and productivity of established trees, while canopy development and increasing shade can suppress light-demanding lianas and reduce the need for continued intervention [19].

Finally, the absence or limited establishment of seedlings in cattle-grazed areas at Whitegrass and along South River riparian margins indicates that grazing pressure is constraining natural regeneration. Excluding cattle and protecting or planting seedlings therefore directly addresses an observed barrier to forest recovery. Collectively, these results demonstrate the value of linking site-specific assessments of forest structure, carbon stocks, and degradation drivers with locally appropriate management responses. Rather than applying uniform restoration or protection measures, the interventions developed for each site target empirically identified pressures and forest conditions, providing a basis for reducing continued carbon losses and enhancing forest recovery.

Estimated carbon benefits from project implementation

In mixed-intervention forest carbon projects, such as the Life & Land Project, Forest Protection (or avoided deforestation/forest degradation) interventions often produce the highest carbon benefits per unit area compared to other interventions (c.f. Forest Restoration, Improved Forest Management), yet require the least amount of sustained, active community work and involvement to implement [2]. From a pure cost–benefit carbon generation perspective, this illustrates cost effectiveness and efficiencies of scale and also highlights why it remains vitally important to protect remnant, carbon dense ecosystems such as forests in perpetuity. However, as Forest Restoration and/or Improved Forest Management interventions require more active engagement, they offer more opportunities for paid labour (e.g. vine clearance, seedling collection, nursery rearing, out-planting, and site maintenance). Job opportunities in developing countries are cited as one of the most important societal outcomes from developmental assistance and PES projects [6]. While the Pacific Australia Labour Mobility (PALM) scheme in Australia and the Recognised Seasonal Employer (RSE) scheme in New Zealand, for example, offer opportunities for seasonal employment, it also means that people—particularly younger men—often aren’t present to undertake physically demanding maintenance on family plots such as garden clearance and maintenance or the removal of downed woody debris post-severe windthrow events. The lack of available labour is one of the main reasons identified by communities for the persistence of fallow gardens, or why agroforestry systems can remain overrun by invasive lianas. The Life & Land Project, in contrast, provides an opportunity for people to remain with their communities and undertake paid work which restores the health of agroforestry and forest ecosystems.

A key limitation of PES projects, however, is that once initial project funding ceases, so too does the ability to pay community members to continue core activities (e.g. forest maintenance) which may reverse environmental gains by undoing the ‘permanence’ of carbon sequestered through degradation of previously restored sites. Part of the rationale for monetising carbon benefits under PES projects is the ability to reinvest a portion of the revenue generated to support on-going project management and the continuation of activities once short-term funding ceases. Developing a long-term, sustainable revenue stream is vitally important to insure against the risk of reversal by providing secure, on-going employment for community members while potentially enabling the temporal, spatial, and sectoral expansion of projects. However, the uncertainty around voluntary carbon markets and fluctuations in demand must be acknowledged. Whilst credit demand outstripped supply for a number of years pre-Covid 19, the glut of supply now outstrips demand as the voluntary carbon market contracts [18]. This is offset to some degree by an increasing preference for ‘high integrity’ carbon credits, or credits that are genuine, verifiable, and contribute to meaningful climate action and co-benefits for biodiversity or communities, with demand remaining strong and continuing to outstrip supply (Ecosystem [18]). With regards to Forest Protection (i.e. avoided deforestation and forest degradation) initiatives in particular, ensuring that carbon credits meet the threshold of ‘high integrity’ is paramount given recent concerns regarding additionality and substantial over-crediting and the need for ex post certification to demonstrate legitimate, causal reductions in deforestation and forest degradation [56].

For SIDS/LOS such as Vanuatu, the scale at which carbon credits can be generated from protection and restoration projects can have significant revenue implications where the volume of credits available as a function of area (hectares) may not align with the volume that corporate organisations seek to offset. For example, while projected carbon benefits under the Life & Land Project per unit area (18.4 Mg CO₂e ha⁻¹ year⁻¹) are higher than average values for afforestation/restoration (AR) and avoided deforestation (AD) projects registered with the Verified Carbon Standard (9.5–10.4 Mg CO₂e ha⁻¹ year⁻¹ respectively, [11]), the volume and scale (6,688 Mg CO₂e year⁻¹ across 362.7 hais an order of magnitude lower than similar projects (average annual issuance for forestry and land use projects ~245,000, Ecosystem Marketplace 2024; average AR project size 5,632 ha, [11],median AD project size 62,361 ha, [54]) or the average annual purchase volume of carbon credits per corporate buyer (26,700 Mg CO₂e year⁻¹. [62]). It is critically important to note, however, that scale and volume are also fundamentally relative to the location and geographical scope from which carbon benefits are generated in the first place. While these factors may not meet corporate buyer expectations, this does not detract from the importance of developing and implementing PES initiatives in SIDS/LOS such as the Life & Land Project which have real world, tangible benefits for communities.

Conclusions

This study improves carbon stock estimations in tropical rainforest and mixed-use agroforestry systems, an important outcome given the lack of published LULCC studies from Vanuatu and the paucity of terrestrial forests studies from the South Pacific. Implementing project interventions under the Life & Land Project can reduce the impacts of deforestation and forest degradation from logging, shifting cultivation, and quarry development (Forest Protection), reducing carbon emissions. Additionally, sustained Forest Restoration and Improved Forest Management initiatives can remove carbon from the atmosphere and reduce the extent and prevalence of pressures such as invasive liana overgrowth which remains one of the most pressing environmental issues Vanuatu faces. Successfully implementing these interventions will also likely provide co-benefits beyond carbon such as biodiversity conservation, provisioning ecosystem services, sustainable rural development and the diversification of livelihoods. Importantly, it also provides a template and example for encouraging the uptake of further PES projects in the region.

Author contributions

CC conceived, designed, led research, analysed data, wrote and edited the manuscript; OB conducted primary analysis of landuse and landcover (NDVI) results and wrote subsequent sections; OB and AF wrote and edited the manuscript; JL reviewed and edited the manuscript; AF and JL conceived and developed the project from which this study is based.

Funding

This project was funded by the New Zealand Government, Ministry of Foreign Affairs and Trade.

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The Authors declare that there were no competing interests associated with the research, authorship, or publication of this study.