Converting forests to coffee cultivation (Coffea arabica) significantly reduces organic carbon sequestration payments in the Peruvian Amazon

1 Introduction

The Peruvian Amazon has experienced large losses of forest cover due to changes in land use, which contribute to the increase of CO₂ in the atmosphere. This is due to the frequent change of use from forests to crops, mainly corn, generating critical degradation in forests (Mendoza-Lopez et al., 2024). After conversion from primary forests to crops, the stratification rate of SOC, total nitrogen, and available potassium decreases significantly by 58.98%, 59.49%, and 59.66%, respectively. Likewise, erosion is reduced by 10.76% in soils between 0 cm and 20 cm deep, showing an alarming loss of nutrients with negative impacts on soil quality. These findings emphasize the value of maintaining natural vegetation (primary forests) in these areas to retain nutrients and protect the soil from runoff and erosion (Solórzano-Acosta et al., 2025); likewise, establishing crops in association with shade trees generates a sustainable environmental impact.

Plant cover with tree species has a positive impact on soil organic carbon sequestrations in the soil, influenced by altitude and temperature along the altitudinal gradient, due to slower decomposition activities. Furthermore, only in secondary forests were there significant differences in SOC between the different districts of the Peruvian Amazon (Vallejos-Torres et al., 2023). To fully understand carbon sequestrations and their variability in these endangered ecosystems, we suggest follow-up studies on carbon sequestrations and emissions in different land uses in the Peruvian Amazon rainforest, which can lead to appropriate policy recommendations (Vallejos-Torres et al., 2024a). In addition, primary forests have the highest SOC content and density, followed by intervened and deforested forests, observing an opposite pattern in terms of soil erosion. These patterns are especially marked in the first 40 cm of soil depth (Vallejos-Torres et al., 2024b).

These results, found by researchers in the Peruvian Amazon, highlight the need to protect and conserve forest trees established in primary and secondary forests and agroforestry. It is important to note that a diameter limit that emphasizes the protection of carbon sequestrations would ideally protect trees from a DAP limit of ≥41 cm. Given the great carbon storage potential of different forest species in the Peruvian Amazon, the loss of these carbon sequestrations could have serious climatic consequences, especially if these large trees are cut down (Lozano et al., 2026). Primary, secondary, and agroforestry forests in the Peruvian Amazon are known to store large amounts of carbon in both soil and plant biomass.

Quantifying the economic values of total carbon sequestrations and annual carbon sinks of agroforestry ecosystems is necessary for carbon trading and forest carbon projects. The value of carbon sinks (VCOs) is often used to assess the value of a forest's ecological services. Therefore, the estimation of VCO in forest vegetation has received considerable attention (Zhang et al., 2018). The issue of financing the necessary climate measures, including those of the agroforestry sector, to drastically reduce greenhouse gas (GHG) emissions and global warming, is important for both national governments and international organizations (Zhou et al., 2024). The economic values of this environmental asset are remarkably high, even when considering a low market price for carbon emissions trading (Ascioti et al., 2017). Currently, we do not know if carbon payments are viable to protect Amazonian forests from logging, even though the Amazon is home to the largest reserves of untapped timber and an expanding timber sector (Bousfield et al., 2022).

In this context, the present research work aimed to analyze how carbon payments can improve agroforestry profitability and sustainability in the Peruvian Amazon, establishing the specific objectives of (i) analyzing agroforestry diversity in the Peruvian Amazon, (ii) analyzing carbon sequestration in agroforestry systems of the Peruvian Amazon, and (iii) analyzing carbon costs in agroforestry systems of the Peruvian Amazon.

2 Materials and methods

2.1 Study sites and sampling

The study was carried out in four agroforestry systems in the San Martín Region, Peru: (i) Coffee as a monoculture (CAF-MON) with 05 ha, located in the district of Jepelacio, province of Moyobamba, whose coordinates were 6°10′48″ S and 76°52′48″ W at 1,066 m above sea level, (ii) Coffee with Inga (CAF-GUA) with 05 ha, located in the district of Jepelacio, province of Moyobamba, whose coordinates were 6°10′48″ S and 76°53′35″ W at 1,078 m above sea level, (iii) Coffee with agroforestry systems (CAF-SAF) with 05 ha, located in the district of Jepelacio, province of Moyobamba, whose coordinates were 6°10′12″ S and 76°54′00″ W at 1,070 m above sea level, (iv) “Primary Forest (BOS-PRI)” of 182,000 ha, located in the town of Pardo Miguel Naranjos in the province of Rioja coordinates at 5°9′16.64″ S 77°1′32.81″ W at 1,625 m a.s. (Figure 1). These agroforests are threatened by extensive agricultural practices such as corn plantations and climate change (Vallejos-Torres et al., 2021). The San Martín region covers an area of 51,253.31 km² and is dominated by a subtropical and tropical climate. Throughout the year, temperatures generally range from 24° C to 31° C, with an average annual rainfall of 1,000 mm. To study carbon payments that can profitably improve agroforestry sustainability in the Peruvian Amazon, we selected a total of 16 plots [four plots per agroforestry site (Figure 1)], each plot consisting of 500 m².

Under this sampling design, each evaluated system was represented by a main sampling site, within which four plots of 500 m² were selected. These plots were used as sampling units to characterize the internal variability of each system. Therefore, the study design should be interpreted as an observational comparison between the evaluated ecosystems.

2.2 Estimation of above and below-ground biomass and carbon in forest trees

Allometric models were used to estimate aboveground biomass based on diameter at breast height (DBH), total height, and wood basal density. Wood basal density values for each identified species were obtained from the “Global Wood Density Database” (Zanne et al., 2009). In cases where species data were not available in the database, genus or family density values were used. Aboveground and root biomass were estimated using the regression equations developed by Chave et al. (2014). This generalized equation incorporates variables such as diameter at breast height (DBH), tree height (H), and wood density (ρ) to obtain biomass. Likewise, Cairns et al. (1997) determined this based on the natural logarithm of the aboveground biomass. The biomass values obtained for each tree within the same plot were used to calculate the total biomass of the trees in the plot, and the resulting value was extrapolated to obtain the biomass and carbon stock per hectare. The carbon fraction of 50% is a standard value widely used in biomass and carbon studies (Toochi, 2018); however, several studies have shown that the carbon content in woody tissues of tropical species exhibits natural variability. The IPCC recommends default values between 0.47 and 0.50 for tropical woods (IPCC, 2014). In addition, Martin and Thomas (2011) demonstrated that the carbon content in the wood of tropical species can vary considerably among coexisting species, with an average of 47.4% ± 2.51%. Since this variability has been documented for Amazonian species and could influence carbon estimates, a sensitivity analysis was conducted. Taking these references into account, to estimate the amount of underground and below-ground carbon, the values obtained for underground and below-ground biomass were multiplied by 0.5, assuming that the total amount of underground and below-ground biomass had a carbon content of 50% (Aye et al., 2022; Eshetu and Hailu, 2020).

2.3 Calculation of plant diversity

The identification of the plants was carried out in the field using their names and trees, following the methodology of Chemeda et al. (2022). Species diversity was studied using the Shannon index (Shannon and Weaver, 1963), which is a widely used index to compare diversity between various habitats (Okpiliya, 2012). Simpson index (Simpson, 1949) and the Pielou index (Pielou, 1977) were used.

2.4 Calculating the price of captured carbon

The calculation of the price of carbon captured from the most dominant species was carried out considering the social price of carbon in the social evaluation of investment projects in typologies; the opinion of prices by the CIGDES Organization established in the San Martín region was considered. Finally, the social cost for a ton of carbon in Peru was considered at a rate of 7.1 U$D (Ministry of Economy Finance, 2019).

2.5 Statistical analyses

The statistical analysis was performed in R version 4.5.0 (R Core Team, 2024), using the RStudio development environment. Initially, a descriptive exploration of the data was performed using boxplots and summary statistics, including measures of central tendency, dispersion, and shape. Subsequently, the effect of the ecosystem on wealth, alpha diversity indices, and variables related to carbon storage and their economic valuation was evaluated using generalized linear models implemented with the glmmTMB library. For each response variable, different models were adjusted, selecting the most appropriate distribution family and link function according to the nature of the data and its best statistical performance, evaluated using the Akaike information criterion corrected (AICc) with the MuMIn library. In the selected models, the explanatory capacity was evaluated using the coefficient of determination (R²) with the performance library, and the diagnosis of residues was carried out with the DHARMa library. Finally, comparisons between ecosystems were made from the estimated marginal means, with a significance level of 0.05 and adjustment for multiple comparisons using the FDR method, using the emmeans library.

3 Results and discussion

3.1 Alpha wealth and diversity

Table 1 shows significant differences between the evaluated ecosystems in species richness and alpha diversity indices (Shannon, Simpson, and Pielou), with p < 0.01 in all cases, which shows clear differences in plant diversity between systems. For wealth, a Conway-Maxwell-Poisson (C-M-P) distribution with logarithmic linkage was selected; for diversity indices, an asymmetric normal distribution with logarithmic linkage was selected. The AIC values support the choice of these models, and the high adjustment percentages (R² between 94.00% and 97.40%) confirm their good explanatory capacity. These results indicate that species richness and alpha diversity were significantly associated with the evaluated ecosystem type.

Table 1. Wealth and alpha diversity indices: ANOVA, Selection and adjustment of statistical models.

C-M-P, Conway-Maxwell-Poisson distribution; N. Sk, Asymmetric normal distribution.

Figure 2 revealed significant differences in tree diversity between the evaluated ecosystems. Species richness was highest in the primary forest (B-PRI) and the coffee agroforestry system (C-SAF), intermediate in coffee with Inga (C-GUA), and lowest in coffee monoculture (C-MON). The primary forest showed the greatest diversity across all metrics, indicating a greater number of species, high overall diversity, and a more even distribution of individuals. The C-SAF also maintained high species richness (6.25), although its lower evenness value (0.42) suggests the dominance of a few species. In contrast, C-GUA exhibited lower richness (2.0) but moderate evenness (0.58), whereas C-MON presented minimal diversity, reflecting a simplified community dominated by very few species.

These results demonstrate that vegetation cover and agroforestry management play a key role in maintaining tree diversity and community structure. The superior performance of the primary forest may be associated with its higher elevation (1 625 m a.s.l.) and greater precipitation and humidity, conditions that favor water availability and the coexistence of a larger number of plant species (Madonsela et al., 2021; Fundisi et al., 2022). Similar patterns have been reported in tropical forests, where climate, topography, and anthropogenic disturbances strongly influence tree diversity (Gadallah et al., 2025; Kanda et al., 2026). Likewise, studies in the Peruvian Amazon have documented high levels of tree diversity, with 59 recorded species and consistently high values across the main diversity indices, reflecting both high species diversity and a balanced distribution of individuals among forests (Lozano et al., 2026). Although agroforestry systems supported lower diversity than the primary forest, they still maintained considerable numbers of native tree species, highlighting their potential contribution to biodiversity conservation and sustainable land management (Esquivel et al., 2023). Overall, these findings reinforce the importance of conserving primary forests while promoting agroforestry systems as complementary strategies for maintaining biodiversity in tropical ecosystems.

3.2 Carbon storage

Table 2 shows significant differences between the evaluated ecosystems in the variables related to carbon storage (p < 0.01). The number of plants was not retained in the final models, as it provides redundant information and does not improve their explanatory capacity. For these variables, a Gamma distribution with logarithmic linkage was selected, suitable for continuous, positive, and asymmetric responses. The AIC values support the choice of models, and the very high adjusted R² (99.70% to 99.99%) shows a high explanatory capacity. These results indicate that the differences in stored carbon and its economic value are mainly associated with the type of ecosystem evaluated.

Table 2. Carbon storage: ANOVA, selection, and adjustment of statistical models.

AGC, total air carbon; BGC, total underground biomass carbon.

Figure 3 shows the differences between ecosystems in AGC, BGC, and total CO₂ (A–C). In all cases, the primary forest (B-PRI) presented the highest values, differing significantly from the other systems. The coffee systems exhibited considerably lower values, with specific differences between monoculture coffee (C-MON), coffee with Inga sp (C-GUA), and coffee with SAF (C-SAF), according to the variable evaluated. The figure shows that the B-PRI system presents significantly higher values in the three variables evaluated (AGC ≈ 133.74 t ha⁻¹, BGC ≈ 40.12 t ha⁻¹, and CO₂ ≈ 638.05 t ha⁻¹), differing statistically from the rest (group “a”), showing greater carbon storage than C-MON, C-GUA, and C-SAF systems, whose values are considerably lower. Among the latter, minor but significant differences are observed: C-SAF tends to have relatively higher values in AGC and CO₂. At the same time, C-GUA stands out in BGC, and C-MON generally shows the lowest values. Together, the results show that B-PRI clearly dominates in all variables, while type C systems have a more limited behavior and less pronounced internal variations.

Amazonian forests store large amounts of carbon in aboveground biomass, belowground biomass, and soil, with the highest SOC values reported in primary forests (3.54% and 2.51%) and the lowest in deforested forests (Mendoza-Lopez et al., 2024). This greater storage is associated with dominant species such as Manilkara bidentata and Brosimum alicastrum, which exhibit high aboveground and belowground carbon sequestrations (Lozano et al., 2026). Similarly, primary forests have been reported to store up to 118.99 t ha⁻¹ of carbon, compared with approximately 65 t ha⁻¹ in coffee systems (Solórzano-Acosta et al., 2025), mainly because of their greater tree diameter, height, and species diversity. Ecosystem carbon sequestrations range from 103–132 t ha⁻¹ aboveground and 21–26 t ha⁻¹ belowground (Khan et al., 2024). However, coffee agroforestry systems also represent important carbon sinks, particularly when associated with shade trees such as Inga sp. Biomass carbon sequestrations of 118.2 t ha⁻¹ have been reported in shaded coffee plantations, compared with 132.2 t ha⁻¹ in secondary forests and 76.5 t ha⁻¹ in coffee polycultures (Vallejos-Torres et al., 2024b), while carbon sequestration in coffee-based agroforestry systems ranges from 18.8 to 48.5 t C ha⁻¹ (Niguse et al., 2022; Vallejos-Torres et al., 2023).

The results of Figures 2, 3 suggest that ecosystems with greater species richness and higher values of the Shannon, Simpson, and Pielou indices also have a greater carbon storage capacity, which is consistent with the biodiversity-functioning theory of the ecosystem. This relationship can be explained because greater diversity increases the complementarity in the use of resources (light, water, and nutrients), favors a more complex forest structure, and increases the probability of coexistence of species with high biomass production and high wood density, which translates into a greater accumulation of carbon in the aerial biomass and, subsequently, in the soil. In addition, more diverse forests tend to be more resilient to environmental disturbances, more stably maintaining their function as carbon sinks (Warner et al., 2023; Yao et al., 2023; Crockett et al., 2023).

However, a limitation of the present study is that the primary forest (B-PRI) was located approximately at 1,625 m asl, while the coffee systems (C-MON, C-GUA, and C-SAF) were between 1,060 and 1,080 m asl, representing an altitudinal difference greater than 500 m. (Eckhardt et al. 2026) reported that changing from forest to agroforestry significantly reduces total carbon stocks, regardless of altitude. Meanwhile, studies show that in forests there is a tendency for soil carbon to stabilize or even increase with elevation due to lower decomposition rates and lower temperatures. However, this pattern is interrupted when the conversion of the forest occurs, since the loss of aerial biomass dominates the carbon balance at all altitudinal levels (de la Cruz-Amo et al., 2020).

3 Economic valuation of carbon storage

Table 3 shows significant differences between the evaluated ecosystems in the variables related to carbon storage (p < 0.01). The number of plants was not retained in the final models, as it provides redundant information and does not improve their explanatory capacity. A Gamma distribution with a logarithmic link was selected for all variables, suitable for continuous, positive, and asymmetric responses. The AIC values support the choice of models, and the very high adjusted R² shows a high explanatory capacity. These results indicate that the differences in economic value are mainly associated with the type of ecosystem evaluated.

Table 3. Economic valuation of carbon storage: ANOVA, selection and adjustment of statistical models.

AGC, total air carbon; BGC, total underground biomass carbon.

Figure 4 shows the differences between ecosystems in economic terms of AGC, BGC, and total CO₂ (A–C). In all cases, the primary forest (B-PRI) presented the highest values, differing significantly from the other systems. The coffee systems exhibited considerably lower values, with specific differences between monoculture coffee (C-MON), coffee with Inga sp (C-GUA), and coffee with SAF (C-SAF), according to the variable evaluated. Thus, the figure compares four systems (C-MON, C-GUA, C-SAF, and B-PRI) in terms of environmental costs expressed as AGC, TBS, and CO₂ (USD ha⁻¹), showing clear and consistent differences between them. In all three panels, B-PRI presented much higher estimated values (≈949.52 in AGC, 284.86 in TBS, and 4,530.15 in USD t C ha⁻¹), showing a considerably greater economic valuation of carbon than C-MON, C-GUA, and C-SAF, which maintain much lower values. Among the latter, a trend is observed where C-SAF tends to register the highest values within the coverage group, followed by C-GUA and then C-MON, with the lowest. However, the differences between them are relatively small compared to B-PRI. The letters indicate significant statistical differences, confirming that B-PRI is consistently different and superior in costs, while systems with coverage present minor but still differentiable variations.

The economic value of carbon sequestration in the primary forest was higher than that reported by SENDECO (2020), whereas the values obtained for C-MON, C-GUA, and C-SAF were lower than the reported average of US$438.82 ha⁻¹ in carbon credits. Forests are expected to contribute approximately 10% of the mitigation required to limit global warming to 1.5° C, which would require carbon prices to reach about $281 t⁻¹ CO₂ by 2055 (Austin et al., 2020). By comparison, the average carbon sequestration cost in northwest China was estimated at US$4.77 t⁻¹ CO₂ (Yao et al., 2024), lower than the average value of US$7.1 t⁻¹ CO₂ estimated for Peru in this study. These differences emphasize the important role of forests and agroforestry systems in climate change mitigation and the need to prioritize their conservation to maintain cost-effective carbon sequestration. In the Peruvian Amazon, protecting forests may require investments ranging from US$7.97 to 10.45 t⁻¹ CO₂, supporting the implementation of conservation strategies and increasing the economic value of standing forests (Baumbach et al., 2023).

Advances in carbon cost assessments can provide a stronger scientific basis for forest carbon projects, carbon markets, and climate mitigation policies aimed at achieving carbon neutrality (Niu et al., 2025). However, direct comparisons among studies remain difficult because estimates often differ in the definition of life-cycle costs and the cost categories included (Nunes et al., 2020; Skidmore et al., 2014). In addition, economic factors such as timber prices may influence sequestration costs, as higher wood revenues can offset project operating expenses and reduce the overall cost of forest carbon sequestration (Callaghan et al., 2019; Ge et al., 2023).

4 Sensitivity analysis under alternative carbon pricing scenarios

The sensitivity analysis showed that the economic value of CO₂ increased proportionally as the price of carbon increased in all the evaluated ecosystems. The magnitude of this increase was related to the total CO₂ storage of each ecosystem. Primary forest (B-PRI), even in low-price scenarios, presented a markedly higher gap with respect to coffee systems, remaining widely separated by its higher CO₂ storage values (Figure 5A). Because of this difference, Figure 5B presents only the coffee systems, making it possible to more clearly observe their internal variations. In this panel, the differences were reduced in the lowest price scenarios, especially for 5 and 20 USD t⁻¹ CO₂, where economic values tended to approximate. However, from 50 USD t⁻¹ CO₂, the gap between systems became more evident, with C-SAF standing out with the highest economic values, followed by C-GUA and C-MON. This trend was clearer in the 100 USD t⁻¹ CO₂ scenario, where the greater plant complexity of the agroforestry system was associated with a more visible relative economic advantage.

5 Comparative synthesis of ecosystems

Figure 6 visually integrates the main ecological and economic results obtained in the four systems evaluated. A clear gradient is observed from monoculture coffee to primary forest, in which the increase in the complexity of vegetation is associated with greater values of species richness, alpha diversity, total CO₂ storage and economic valuation of CO₂. The primary forest presented the highest values in all the indicators, while the coffee in monoculture showed the lowest values. Meanwhile, the agroforestry system with coffee occupied an intermediate position, indicating that greater structural complexity within coffee systems may be associated with biodiversity conservation and CO₂ storage, although without reaching the levels observed in the primary forest.

The results also support the principles of REDD+ initiatives, which aim to conserve forests and increase carbon stocks through economic incentives. According to Wunder et al. (2020), the most successful REDD+ projects are those that integrate conservation mechanisms with direct benefits for local producers, favoring both the reduction of emissions and the conservation of biodiversity. In this sense, the observed transition between C-MON and C-SAF represents a trajectory compatible with REDD+ objectives by simultaneously increasing stored carbon and biological diversity. In relation to Payments for Ecosystem Services (PSE), the increase in the economic value of carbon observed in the figure, from 174.06 USD ha⁻¹ to 4,530.15 USD ha⁻¹, shows the economic potential associated with the conservation of tree cover. Shen et al. (2023) highlight that market mechanisms, such as PES and REDD+, generate effective incentives when there are strong institutions, transparent monitoring systems, and an equitable distribution of benefits among the actors involved.

Regarding voluntary carbon markets, the results show that agroforestry systems have a greater potential to generate carbon credits due to their greater CO₂ capture capacity. This evidence coincides with the review by Blanton et al. (2024), who point out that Latin America constitutes one of the main global suppliers of forest carbon credits thanks to the high carbon sequestration potential of its tropical ecosystems. Similarly, Kotsialou et al. (2022) indicate that the certification of carbon credits continues to face challenges related to the additionality, permanence and transparency of projects. Meanwhile, these results support net-zero emission strategies, given that the progressive increase in stored carbon reflects greater climate mitigation potential. Voluntary carbon markets and REDD+ mechanisms represent relevant instruments to achieve these objectives; however, their effectiveness depends on the environmental integrity, governance and active participation of local communities (Morita and Matsumoto, 2023; Kotsialou et al., 2022).

Regarding the uncertainty in carbon valuation, the results show a positive gradient between the structural complexity of vegetation and associated ecosystem services, evidenced by the increase in species richness, Shannon diversity, carbon storage, and the economic valuation of carbon from coffee monoculture to primary forest. This behavior coincides with what was reported by Lugo-Pérez et al. (2023), who showed that the incorporation of shade trees significantly increases the carbon stored in coffee plantations due to the increase in aerial biomass and soil organic carbon, with shade trees being the main explanatory factor for carbon variability in these systems. The marked difference between the primary forest and agricultural systems also coincides with the evidence synthesized by Molina-Alvarado et al. (2025), who point out that the structural simplification of agroforestry systems through the reduction of the canopy or tree diversity significantly decreases the carbon reserves of the soil and biomass. The economic valuation of carbon, increasing from 174.06 USD ha⁻¹ in monoculture to 4,530.15 USD ha⁻¹ in primary forest, shows that the potential economic benefits derived from ecosystem services increase as carbon storage does. Tinoco et al. (2026) highlight that, although agroforestry systems represent an opportunity to access carbon markets, there are still important gaps related to economic valuation, certification, and spatial variability of stored carbon.

6 Multivariate patterns of diversity and total carbon between ecosystems

The main component analysis (Figure 7) showed that the first two axes explain 94.07% of the total variation (PC1 = 86.14% and PC2 = 7.93%), evidencing a robust representation of the multivariate structure. The first component (PC1) defined the main ecological gradient, positively associating variables related to the diversity and functionality of the system, such as species richness (S), Shannon diversity (H′), Pielou equity (J′), the Simpson index (D), and total carbon (CO₂). For its part, the second component (PC2) allowed for the differentiation mainly of the wealth (S), with positive charges, from the total carbon (CO₂) and the equity (J′), with negative charges, suggesting variation in the structure and operation of the communities.

In the biplot, a clear differentiation of the evaluated ecosystems is observed. Primary forest (B-PRI) is mainly associated with high values of CO₂, J′, H′, and D, indicating more complex, equitable systems with greater carbon storage capacity. In contrast, the agroforestry system (C-SAF) is closely related to species richness (S), evidencing greater taxonomic diversity. Coffee in monoculture (C-MON) is located at the opposite end of the gradient, reflecting simplified systems. In contrast, coffee with Inga sp (C-GUA) has an intermediate condition close to the origin, suggesting a transition between systems of low and high ecological complexity.

Recent studies show that small variations in soil respiration (CO₂) can have disproportionate impacts on the carbon balance, which explains their high load on the main components and their ability to discriminate land treatments or uses (Xu et al., 2024). Likewise, recent research applying ACP in agricultural soils and systems such as coffee or agroforestry confirms that the first components usually capture gradients of fertility, organic carbon, and biological activity, clearly separating managed vs. natural systems, in accordance with the differentiation observed between C-MON, C-GUA, C-SAF, and B-PRI (Romero-Chávez et al., 2026). In addition, recent works show that changes in land use and management (for example, tillage, agroforestry, or deforestation) significantly modify CO₂ emissions and carbon sequestrations, generating multivariate patterns similar to those observed in your analysis, where certain treatments are positively or negatively associated with this vector (Kulmany et al., 2022). Along the same lines, recent studies on carbon dynamics indicate that the stability, vulnerability, and persistence of soil carbon—including the CO₂ breathed—depend on interactions between physical, chemical, and biological factors, which are usually grouped into a few main axes in multivariate analyses (Minich et al., 2026). Finally, research on different environmental gradients and land uses shows that the ACP is a robust tool to identify these dominant “drivers,” especially those related to organic carbon, microbial activity, and environmental conditions, reinforcing the interpretation that the PCA1 in your results reflects a key functional gradient of the system (Samaniego et al., 2025).

The economic valuation of carbon estimated in this study supports the incorporation of agroforestry systems in Payment for Ecosystem Services (PES) schemes. At the same time, PES are more effective when they integrate economic incentives, environmental monitoring, and direct productive benefits for farmers, increasing the permanence of sustainable practices and reducing land use changes (Börner et al., 2020). In this context, the carbon quantification carried out in this study constitutes a technical basis to estimate the potential of coffee systems as alternatives for climate mitigation and complementary income generation for rural producers (West et al., 2023). Therefore, the integrated approach applied in this study provides evidence to support mitigation policies, contribute to compliance with Nationally Determined Contributions, and promote coffee production models compatible with international emission reduction objectives (Intergovernmental Panel on Climate Change, 2023; United Nations Environment Programme, 2023).

Although the models presented a high statistical adjustment (adjusted R² > 99%), the carbon estimates and their economic valuation maintain uncertainty associated with ecosystem variability, biomass measurement, and selection of allometric equations. Therefore, the values of AGC, BGC, and CO₂ must be interpreted as limitationsconsidering confidence intervals and sensitivity analysis, since small changes in the estimates of the stored carbon can modify the economic valuation of the ecosystem service (Vorster et al., 2020).

7 Conclusions

The intensification of land use reduces diversity, while maintaining or imitating the agroforestry structure with coffee helps to conserve it. However, it does not reach the levels of the primary forest. This study shows that the complexity of the system increases both the wealth and the balanced structure of the community, with the primary forest being the most diverse and stable system, standing out in all three variables (AGC, BGC, and CO₂), with statistically significant differences. In contrast, the C-MON, C-GUA, and C-SAF systems present much lower values, although with some internal differences where the C-GUA system stands out in BGC. In another aspect, the magnitude of the difference suggests that the B-PRI system has a completely different behavior (much more intensive or efficient, depending on the context of the study). Costs are much higher when conserving forests, followed by the agroforestry system with coffee. Converting forests to crops could generate economic losses in carbon. This study is important since it allows the implementation of forest carbon capture projects and market transactions, contributing to achieving carbon neutrality objectives and promoting sustainable development in the Peruvian Amazon.

This approach allowed a comparative analysis between coffee monocultures, agroforestry systems, and primary forest, evidencing that agroforestry systems constitute an alternative land use that increases carbon storage and its economic value, while favoring the conservation of biodiversity. In addition, the incorporation of biophysical and economic indicators provides a scientific basis to support climate financing mechanisms, such as carbon markets and payments for ecosystem services, as well as the design of incentives aimed at the conservation and sustainable management of coffee landscapes.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Massimiliano Mazzanti, University of Ferrara, Italy

Reviewed by: Halus Satriawan, Universitas Almuslim, Indonesia

Lisset Perez-Marulanda, International Center for Tropical Agriculture (CIAT), Colombia