Curiaú is a peri-urban Quilombola territory, approximately 3,130 hectares and around 505 households, where livelihoods rely on small-scale agriculture, extractivism, wage labor, and public transfers [Figure 1]. Quilombola communities are Afro-Brazilian traditional communities with collective land rights^[46], historically associated with social marginalization and heightened exposure to socio-environmental risks^[5]. Its mosaic of native forest, floodplains, and smallholder land-use lies in the surroundings of Macapá and is increasingly affected by urban expansion and climatic variability. Recent flooding events in the Amazon, Vale do Ribeira, and southern Brazil have underscored the frontline position of Quilombola communities in the country’s climate crisis: entire territories have been isolated, crops and basic infrastructure damaged, and long-standing vulnerabilities linked to limited public investment and historical marginalization exposed^[47,48].

Following the standard methodology^[42], the study design operationalizes a sequence of linked analytical steps applied to the Curiaú case [Figure 2].

First, a Livelihood Vulnerability Index (LVI) module^[43] is applied to assess Curiaú’s livelihood vulnerability, derive household- and community-level indices, and determine whether Curiaú meets the vulnerability-based eligibility threshold defined in the methodology (Section “Livelihood vulnerability indexes”). Second, the resulting vulnerability profile is interpreted diagnostically to identify the main socio-climatic needs or constraints affecting livelihoods and the dimensions in which targeted support would be most relevant (Section “Livelihood vulnerability indexes”). Third, a dynamic, locality-specific baseline for deforestation in smallholder-dominated mosaics is constructed using the MapBiomas 10-metre land-cover product^[49] and a reference region composed of Quilombola territories and agrarian-reform settlements in Amapá (Section “Remote-sensing and dynamic baseline data”). Fourth, this adjusted baseline is combined with forest carbon-stock factors and conservative deductions to estimate the annual volume of potentially creditable emission reductions (Section “Estimation of potential credits and climate-resilience analysis”). Finally, an exploratory analysis evaluates how the associated carbon finance could curb deforestation and enhance ecosystem resilience in Curiaú in a sustainable way (Section “Estimation of potential credits and climate-resilience analysis”).

In February 2024, a non-probability snowball sampling approach was used to recruit 123 of the 505 households in the Curiaú community. Initial interviewees were identified in geographically defined neighborhood clusters, including areas near the main access road and more hard-to-reach or hidden remote sectors, and were then asked to refer further households. These clusters were defined solely on a geographic basis and serve as entry points for recruitment rather than as socio-economic strata. This approach was chosen for field feasibility and to ensure coverage of both peri-urban and more distant parts of the community. However, because the sampling was non-probabilistic, some degree of selection bias is likely, and the resulting sample is interpreted as characterizing the community, rather than providing precise estimates for specific sub-groups. If treated as a simple random sample, this sample size would correspond to a 90% confidence interval with a margin of error of 7.7%, but these metrics should therefore be interpreted as indicative rather than as formal design-based.

Data were collected through face-to-face interviews at the households’ premises using a structured questionnaire that had been pre-tested and validated and was designed to capture the full spectrum of livelihood vulnerability [Table 1]. A complete list of survey items and summary statistics is provided in Supplementary Table 1. Responses were recorded digitally using an open data kit platform on mobile devices.

The collected data were normalized to a 0-1 scale to generate two distinct community-level indices. First, the Smallholder Farmer Livelihood Vulnerability Index (LVI-SF) was calculated by aggregating indicators across the Sustainable Livelihoods Framework capitals, providing a synthetic measure of overall livelihood vulnerability on a continuous scale from 0 (least vulnerable) to 1 (most vulnerable). Second, the IPCC Livelihood Vulnerability Index (LVI-SF-IPCC) was derived by reorganizing the same set of indicators within the IPCC’s conceptual framework of “exposure-sensitivity-adaptive capacity”^[5], following the LVI-IPCC formulation proposed by Hahn et al.^[45], and synthesized in empirical studies as well^[50]. This formulation applies equal weighting through simple arithmetic means across subcomponents and introduces the inversion of some original items that belong to adaptive capacities only at the final aggregation stage. As a result, the LVI-SF-IPCC yields values ranging from -1, indicating the lowest level of vulnerability, to +1, representing the highest level of vulnerability. Eligibility for carbon finance is assessed using the operational vulnerability screening thresholds, with at least moderate thresholds (LVI-SF > 0.30; LVI-SF-IPCC > -0.30) used to identify contexts in which socio-economic constraints and climate-related exposure and sensitivity are likely to outweigh adaptive capacity^[42,43]. Additionally, both indexes could further inform the prioritization of contexts in which socio-economic vulnerability and climate-related risks justify targeted mitigation, adaptation and livelihood-protection interventions.

The dynamic deforestation baseline was conducted in accordance with the standard methodological requirements, including the definition of the reference region, class reclassification rules, accuracy assessment and adjusted area estimation. Land-cover and land-use change analyses use the MapBiomas 10-metre land-cover product for Brazil (Collection 2, beta), derived from Sentinel-2 MSI Level 2A imagery at 10-metre resolution and processed according to the Algorithm Theoretical Basis Document for the MapBiomas 10-metre project. QGIS (version 3.42) was used for spatial preprocessing, such as coordinate harmonization, polygon preparation, area calculations, raster extraction and class transitions. Project and reference boundaries were obtained from federal cadastral datasets. The Curiaú Quilombola territory was sourced from the Cadastro Nacional de Territórios Quilombolas (CNTQ) maintained by the Instituto Nacional de Colonização e Reforma Agrária (INCRA), and the boundaries of other Quilombola communities and agrarian-reform settlements were obtained from the Sistema de Informações de Projetos de Reforma Agrária (SIPRA) and related INCRA public datasets.

Natural forest formations and anthropic land-use classes were defined following the MapBiomas legend and the reclassification rules^[42] [Figure 3]. Natural forest area was mapped for the year 2022, which represents the baseline observation year adopted in this study, consistent with the subsequent assessment of forest-to-anthropic transitions in 2022-2023 and with the reference period captured by the livelihood vulnerability survey. The project area was defined as the largest contiguous block of natural forest within the officially demarcated Curiaú territory; other land uses and smaller forest fragments inside the legal boundary, including additional forest patches and agricultural areas, were not included in the project polygon but remain part of the recognized Quilombola territory. All spatial data were harmonized in a projected coordinate reference system suitable for area estimation.

The reference region was defined by selecting Quilombola territories and agrarian-reform settlements whose land-use systems, tenure and governance arrangements are comparable to Curiaú and that meet the methodology’s minimum robustness criteria: at least ten units and a total forest area in 2022 of at least fifty times Curiaú’s 2022 forest area. To exclude extreme outliers in terms of scale, only units whose natural forest area in 2022 lay between the 5th and 95th percentiles of the distribution across all candidates were retained, thereby removing very small and very large units while maintaining a coherent sample. The resulting reference region is used to characterize deforestation dynamics in mosaics dominated by smallholder and traditional communities.

For each unit in the reference region, the forest area in 2022 and the area of pixels that transitioned from natural forest in 2022 to anthropic classes in 2023 were computed. For this study, only a single annual transition, the 2022-2023 forest-to-anthropic change, was considered as an annual deforestation rate. For each reference unit, an annual deforestation proportion was calculated as the ratio between 2022-2023 deforestation area and 2022 forest area. These proportions were summarized across the reference region, separately for Quilombola territories and settlements, and an overall reference deforestation rate for smallholder-dominated mosaics was derived as an area-weighted mean, using each unit’s 2022 natural-forest area as the weighting factor.

To ensure the integrity of the dynamic baseline, an independent accuracy assessment was conducted through 10-metre land-cover maps. A probability-based stratified random sample of reference points was drawn across the project area and the reference region, with strata defined by mapped class (natural forest and anthropic use) and spatial domain (Curiaú and the external reference units). The sample size was chosen to meet the minimum requirements^[42] for each stratum and to allow estimation of user’s accuracy with reasonably narrow confidence intervals. Each sample point was visually interpreted using multi-temporal Sentinel-2 MSI Level-2A imagery at 10-metre resolution from the 2022 baseline year, to verify consistency with the annual land-use and land-cover classification used in the study. Class-specific confusion matrices were constructed to estimate user’s accuracy for map validation, and were used to support a conservative accuracy-based adjustment following Olofsson et al.^[29] and IPCC guidance^[30], together with the overall accuracy of the map. Unbiased adjusted area estimates were then obtained by applying the overall map accuracy level to the nominal 2022-2023 deforestation areas in both Curiaú and the reference region. The numerical results of the accuracy assessment and the magnitude of the adjustments are reported in the section "RESULTS".

These procedures yield a dynamic “micro-jurisdictional” baseline. Activity data are derived from externally produced land-cover maps, corrected using probability-based accuracy assessment and design-based area estimators, and summarized for a locality-matched reference region rather than generated through counterfactual modelling. For contextual purposes, the resulting baseline structure was qualitatively compared with other consolidated REDD+ frameworks currently applied in the voluntary carbon market.

Potential avoided deforestation is quantified as the difference between the dynamic baseline deforestation within Curiaú and a project scenario in which deforestation is avoided. For the year applied, avoided deforestation area is converted into avoided CO₂ equivalent (CO₂e) emissions using carbon-stock factors from the latest Brazilian Forest Reference Emission Level (FREL)^[51] for the forest type corresponding to Curiaú in the official Instituto Brasileiro de Geografia e Estatística (IBGE, the Brazilian national statistical and geographic agency), vegetation map. Leakage deductions and other conservative adjustments are then applied, using parameter values that fall within the range typically adopted by major REDD+ standards in the voluntary carbon market, in order to obtain conservative estimates of net emission reductions that could, in principle, be issued as carbon credits - which are labelled as Social Carbon Units (SCUs)^[52] under the applied standard.

Complementarily, Curiaú’s potential carbon credit volumes are linked to an illustrative climate-resilience analysis focusing on flood-related damage to agricultural plots and on the gradual adoption of regenerative practices on existing cropland and pasture. The analysis is explicitly stylized and uses only externally sourced parameters. First, a set of plausible carbon-price scenarios for high-integrity forest and REDD+ credits is defined based on recent market reports documenting transaction prices for credits with social and environmental co-benefits. Under each scenario, annual revenues that could flow to the community are approximated and compared with per-hectare cost benchmarks for re-establishing flood-affected cassava-based cropping systems, establishing diversified agroforestry and rehabilitating degraded pastures in Amazonian contexts.

Second, an exploratory regenerative-agriculture component assesses the potential additional contribution of soil organic carbon (SOC) sequestration in Curiaú’s agricultural mosaic, using indicative sequestration factors and cost ranges drawn from studies on conservation agriculture, agroforestry and improved pasture management. This second step provides only order-of-magnitude estimates of additional removals and associated revenues and is treated as a complementary, hypothesis-generating exploration of how regenerative practices could form an additional mitigation and finance pillar alongside avoided deforestation, rather than as a basis for project-level crediting.

The LVIs indicate that Curiaú exhibits an overall moderate level of livelihood vulnerability over the survey period. Because LVI-based metrics are primarily designed to compare vulnerability across communities or social units within a given study context, rather than to establish universally applicable absolute cut-offs, their interpretation should be treated as contextual^[45,50]. In line with this, empirical studies tend to apply interpretive ranges that vary according to the regional setting, the variables included, and the scale of analysis^[45,50]. Accordingly, the thresholds adopted here are interpreted as operational eligibility screening values under the applied methodology^[43], rather than universally applicable benchmarks. Against this interpretive background, Curiaú is classified as meeting the methodology’s eligibility screening thresholds [Table 2].

Within the applied methodological framework, livelihood vulnerability is not treated solely as a descriptive metric, but as a criterion to identify contexts where carbon finance can generate additional social and land-use decision effects^[42,43]. In this sense, the use of vulnerability-based eligibility thresholds is consistent with the broader logic of carbon finance approaches, which prioritize contexts where financial incentives are more likely to influence land-use decisions and enhance adaptive capacity, thus reducing reversals and enhancing permanency.

The LVI-SF equals 0.39, which falls within the vulnerability bands commonly interpreted in empirical applications as "vulnerable" (0.21 to 0.40)^[50], while the LVI-SF-IPCC equals -0.04, which falls within the "moderate vulnerability" range in IPCC-consistent applications (-0.19 to 0.20)^[50]. Livelihood strategies, financial stability, social networks, natural disasters and climate variability, and socio-demographic characteristics exhibit the highest values, ranging from 0.34 to 0.69 [Table 1; Supplementary Table 1]. These results indicate moderate to severe vulnerability across multiple livelihood dimensions, with particular concentration in structural and socio-economic domains. In contrast, basic household needs, such as food, housing and drinking water display a markedly low score (0.01), reflecting low vulnerability in this domain and suggesting that the community has access to essential goods and services sufficient to support developmental processes.

Within the LVI-SF-IPCC framework, the results reveal a moderate yet structurally significant exposure to climate-related stressors, with the aggregated “Exposure” component reaching 0.34 - the highest among the IPCC vulnerability components. This exposure profile is consistent with the documented flood-proneness of Amazonian floodplain and delta systems in the lower Amazon region, where seasonal and extreme flooding constitute recurrent hazards for agricultural livelihoods and settlement infrastructure^[53-55]. Perceived impacts of climate variability are already evident within agricultural production systems, particularly through constraints on water availability, which show a comparable magnitude (0.24), indicating that exposure is not merely theoretical but already translates into observable livelihood pressures.

Sensitivity-related indicators further reinforce this interpretation. Elevated scores associated with agricultural production and market constraints indicate that hydroclimatic shocks are effectively transmitted into livelihood stress through coupled biophysical and economic pathways. In particular, production and yield-related indicators average 0.66, reflecting exposure to crop and livestock losses and reliance on external inputs, while rainfall dependence and episodic water scarcity contribute an additional sensitivity of 0.24. Market access constraints, although less dominant, remain present, with a market-related sensitivity score of 0.16, indicating that climate-induced production shocks can propagate into income and access pressures.

At the same time, agricultural diversity indicators point to a structural source of sensitivity, with the agricultural diversity component reaching 0.59 (higher value, lower diversity), indicating limited diversification of cropping and livestock systems. This constrains households’ ability to absorb and redistribute production and market shocks under recurrent flooding. This pattern is consistent with empirical studies from Amazonian floodplain and smallholder systems showing that where present, diversified cropping portfolios, high on-farm agrobiodiversity and flexible land-use strategies can partially buffer production and market impacts, even under high flood exposure^[53,56,57].

The most binding structural constraint emerges in the adaptive-capacity domain. High pre-inversion scores in adaptive-capacity-related indicators - particularly those capturing limited access to agricultural training, weak exposure to alternative production systems and restricted learning and knowledge-sharing networks - translate into low effective adaptive capacity. This pattern suggests that, while some degree of livelihood diversification is present, the community’s capacity to anticipate, absorb and transform in response to recurrent climate stressors remains constrained by institutional and informational barriers.

It is important to note that the LVI applied here represents a time-bound diagnostic, based on recent household conditions reported in the survey, and therefore does not imply persistence of vulnerability over time. While livelihood conditions may evolve in response to climatic shocks, market dynamics, or policy changes, the role of the LVI within the applied framework is to establish initial eligibility and contextualize additionality at project entry^[43,52]. The temporal dynamics of land-use outcomes are instead captured through the annually updated ex post baseline, which operates independently of the LVI and reflects observed deforestation behavior in the reference region^[43]. At the same time, socio-economic constraints and climate-related impacts on livelihoods, such as reduced productive capacity or delayed recovery of agricultural systems, often persist beyond a single year, reinforcing the relevance of vulnerability as an entry condition^[36,38,39]. Under the applied standard, project cycles operate over defined timeframes (e.g., 10 years), allowing vulnerability to function as an entry condition while enabling its reassessment in subsequent cycles as socio-economic conditions evolve.

The independent validation of the 2022 MapBiomas 10-m land-cover map using 200 visually interpreted points yielded an overall accuracy of 94%, with user’s accuracy of 98% for natural forest and 90% for anthropic classes [Table 3]. These values exceed the class-level user’s accuracy threshold set by the applied methodology for 10-m products without published regional validation (above 90%) and are within or above the range of forest and overall accuracies reported for widely used global 10-m land-cover datasets such as ESA WorldCover and Esri Land Cover^[58,59]. Binomial 95% confidence intervals for class-level user’s accuracy were ±2.7% points for natural forest and ±5.9% points for anthropic classes, confirming the adequacy of the sample size. Although uncertainty was higher for anthropic classes, confidence intervals remained sufficiently narrow to support the application of the map within the adopted baseline-adjustment framework.

Across the 47 reference units (Quilombola territories and agrarian settlements), encompassing 531,272 ha of natural forest in 2022, the weighted mean annual deforestation rate between 2022 and 2023 was 0.74%. Classification error was accounted for using the probability-based area-adjustment procedure defined in the avoided deforestation framework^[42], based on the overall map accuracy and the class-specific confusion matrix structure. Following this adjustment, the baseline deforestation rate applied to Curiaú was conservatively rounded to 0.70%.

In 2022, Curiaú’s forest area was estimated at 915 hectares through remote sensing. Based on this dynamic baseline, an expected deforestation of approximately 6.4 hectares was projected for 2023 in the absence of targeted interventions. The spatial pattern of 2022-2023 forest loss in a subset of the reference smallholder and traditional territories is shown in Figure 4.

The official deforestation data for the region of comparison come from PRODES (Satellite Monitoring Program for Deforestation in the Brazilian Legal Amazon). PRODES provides annual clear-cut deforestation estimates for the entire Legal Amazon since 1988, based on interpretation of Landsat, CBERS and Sentinel satellite imagery. For 2022-2023, the published deforestation was approximately 9,000 km², equivalent to approximately 0.18% of the biome’s total area^[60]. This macro-scale percentage is not directly comparable to the proportional 0.74% reference-region rate used in this study, as PRODES measures absolute annual clear-cutting across the entire Legal Amazon rather than proportional forest loss within mosaics of smallholder and traditional territories. For contextual comparison at the jurisdictional scale relevant to the study area, PRODES reports that, in the same monitoring year, the state of Amapá accounted for only 15.7 km² of clear-cut deforestation, one of the lowest totals in the region. This places Curiaú and structurally similar reference territories within a low-deforestation state-level context.

Complementing this, MapBiomas data^[61] indicate that the state of Amapá experienced approximately 11,324 ha of vegetation suppression in 2023 (combining primary and secondary vegetation suppression) against an estimated 11.66 million ha of natural forest in 2022. This corresponds to an annual rate of approximately 0.10%, which is consistent with PRODES in identifying Amapá as a low-pressure jurisdiction within the Brazilian Amazon.

While these macro-scale percentages provide important contextual information, they include a large proportion of intact forest with much lower deforestation rates compared with smallholder-dependent and currently fragmented forest. The higher proportional rates observed reflect the small-denominator effect typical of smallholder mosaics. For Curiaú community, the baseline shows a modest deforestation risk, consistent with a peri-urban Quilombola landscape where forest conservation coexists with increasing external pressures. Such results align with recent evidence that carbon losses derived from the edge effect induced by fragmentation affect 97% of global forest^[62]. By deriving the baseline solely from recent transitions from natural forest to anthropic use in structurally similar territories and adjusting for classification error using design-based area estimation, this dynamic baseline strengthens best practices for uncertainty-aware deforestation baselines in smallholder landscapes^[15,29] and promote the social equality as well.

We further contrast our approach with consolidated jurisdictional frameworks. In consolidated REDD+ frameworks such as Verra’s VM0048^[63] methodology and its VMD0055^[64] module, project baselines for avoiding unplanned deforestation are anchored in observed land-cover change but are constructed at much larger jurisdictional scale as dynamic baselines that are periodically updated. Jurisdictional activity data on unplanned deforestation are generated by the owner institution or contracted data service providers using high-resolution satellite imagery, forest-cover benchmark maps and sample-based area estimation over a recent historical reference period; The dynamic baseline implemented for Curiaú under our framework follows the same broad principle of grounding baselines in observed deforestation but does so at a micro-jurisdictional scale and with a fully ex post temporal framing tailored to smallholder mosaics. Instead of receiving a multi-year allocation of jurisdictional activity data, Curiaú’s baseline is calculated directly from last-year forest-to-anthropic transitions observed in a reference region of 47 other, socioeconomically comparable smallholder territories. All underlying deforestation areas are first corrected using a probability-based accuracy assessment of the 10 m land-cover maps, so that uncertainty is incorporated through design-based area estimation rather than post hoc discounting of baseline activity data. Relative to different REDD+ frameworks such as VM0048 and VMD0055, this approach appears to offer a comparatively simple, equitable and transparent baseline that can be updated ex post year by year as new land-cover data become available, while still aligning with emerging integrity expectations by anchoring baselines dynamically in externally observed behavior beyond the project boundary and limiting the scope for project-developer discretion in baseline setting.

This temporal design implies that each credited year is benchmarked against observed deforestation dynamics within the same period, rather than against an aggregated historical average. While multi-year reference periods can potentially reduce sensitivity to atypical conditions in individual years, they also smooth temporal variability and introduce assumptions regarding the selection of historical windows. By contrast, the annual ex post construction adopted here captures year-to-year fluctuations and provides real carbon emission reduction directly.

Collectively, these design features reinforce the integrity of the carbon emission reduction. The combination of a dynamic ex post baseline anchored in observed deforestation, locality-matched reference regions that constrain discretionary baseline setting, and probability-based accuracy assessment with design-based area estimation strengthens baseline credibility and data robustness in response to concerns that methodological choices can inflate avoided-deforestation claims^[26-28].

Within the applied framework^[42], livelihood vulnerability enters the additionality logic as a screening condition that complements conventional additionality assessment procedures. As in established AFOLU methodologies, additionality is demonstrated through the identification of plausible alternative land-use scenarios with the application of structured additionality tests (e.g., barrier or investment analyses)^[42,63]. The inclusion of a vulnerability-based filter therefore introduces an additional layer that constrains eligibility to contexts in which socio-economic limitations, restricted adaptive capacity and non-trivial deforestation pressure coexist, increasing the likelihood that climate-finance revenues are materially relevant to land-use decisions. In this sense, the framework differs from actor-neutral REDD+ approaches, including VM0048^[63], not in the structure of additionality demonstration itself, but in explicitly conditioning where it is applied. This is particularly relevant in light of broader implementation and equity constraints observed in natural climate solutions and smallholder contexts^[35-37]. Framed in this way, vulnerability-based screening may reduce the likelihood of crediting areas where forest retention would plausibly persist in the absence of carbon revenues, while also helping to counterbalance structural barriers associated with transaction-cost-intensive systems that tend to favor larger or better-capacitated landholders^[38,39,41]. At the same time, by identifying binding livelihood constraints, it provides a basis for linking revenues to interventions that address the underlying drivers of land-use change, thereby strengthening the conditions under which additionality becomes effective in practice^[65].

The forest-type specific emission factor in alluvial dense ombrophilous forest from Brazil’s national Forest Reference Emission Level (FREL)^[51] is adopted. Total live biomass carbon stocks (above- and below-ground) is 108.25 tons of carbon per hectare. Applying the dynamic baseline deforestation rate of 0.70% for 2023 yields 2,540 tCO₂e of baseline emissions [Table 4]. Following a conservative 20% deduction, consistent with buffer-pool requirements in standards such as Gold Standard^[33], the American Carbon Registry^[66] and the UK Woodland Carbon Code^[67], 80% of avoided emissions is treated as creditable. This deduction reflects standard approaches to account for non-permanence and related risks, ensuring that credited volumes represent net climate benefits after conservative adjustment, in line with the integrity-oriented design discussed in Section “Accuracy assessment and dynamic deforestation baseline”.

This results in an annual credit of approximately 2,032 tCO₂e. This volume is much smaller relative to large project- and jurisdiction-scale REDD+ initiatives that issue millions of credits, such as national and jurisdictional programs in Guyana and Gabon under ART-TREES standard^[68], but it is consistent with and essential to Curiaú’s forest-dependent smallholders.

Generally, the baseline estimates (Section “Accuracy assessment and dynamic deforestation baseline”), avoided emissions, and resulting credit potential (Section “Potential emission reductions and carbon credit generation”) illustrate how the proposed framework operates in a smallholder forest landscape such as Curiaú. Similar outcomes are expected in settings with comparable livelihood structures and levels of vulnerability (Section “Livelihood vulnerability”), together with similar ecological conditions - where equivalent or higher forest carbon stocks and deforestation pressure are present - as observed in other traditional and Indigenous territories across the Pan-Amazon^[19]. Under these conditions, the dynamic micro-jurisdictional baseline can be applied consistently to estimate avoided emissions and potential credits. By contrast, the climate-resilience finance implications (Section “Climate-resilience finance”) depend more directly on locally derived vulnerability profiles and livelihood conditions, as reflected in the LVI results.

The present application has several limitations that should be made explicit. First, as discussed in Section “Livelihood vulnerability”, the LVI thresholds used here are operational screening values under the applied methodology rather than universally valid vulnerability cut-offs^[43,45,50]. Relatedly, the LVI functions as a time-bound entry diagnostic, based on recent household conditions, and therefore does not capture the full temporal evolution of vulnerability over the project duration. Although the applied standard allows reassessment across project cycles, this design implies a trade-off between operational feasibility and temporal representativeness in the use of vulnerability as part of additionality. Second, the livelihood diagnosis is based on a non-probability household sample and should therefore be interpreted as characterizing the community rather than providing statistically representative estimates for all sub-groups. Third, the empirical application is restricted to a single Quilombola territory, which limits direct extrapolation of the observed LVI profile, baseline behavior and finance implications to other smallholder or traditional-community settings with different socio-ecological configurations.

Additional limitations arise from the baseline and finance components. As discussed in Section “Accuracy assessment and dynamic deforestation baseline”, the annual ex post baseline is more responsive to observed deforestation dynamics than multi-year averaging, but this same responsiveness may generate interannual variation in baseline levels, credit volumes and expected revenues, which can complicate long-term financial planning in community-based projects^[28]. Baseline robustness also remains conditional on the stability and classification performance of the adopted land-use product, even when conservative accuracy adjustments are applied^[29]. In turn, the climate-resilience analysis in Section “Climate-resilience finance” is intentionally illustrative: it does not fully internalize certification, transaction, monitoring and organizational costs, and therefore likely overstates the share of gross revenues that could be effectively appropriated for livelihood recovery and adaptation. The same applies to the regenerative-agriculture component, which is framed as exploratory and order-of-magnitude only, rather than as a project-crediting estimate. Taken together, these limitations do not invalidate the proposed framework, but indicate that its strongest contribution at present lies in structuring a credible and socially targeted entry point for smallholder forest carbon initiatives, rather than in offering a fully generalizable or financially complete implementation model.