texto en 
Inglés (pdf)
Artículo en XML
Referencias del artículo
Traducción automática
Enviar artículo por email
Citado por SciELO 
Similares en
SciELO 
Permalink
Introduction
Mexico is considered one of the main emitters of carbon dioxide (CO2) into the atmosphere (Climate Transparency, 2022), despite having approximately 64 million hectares of temperate forests and tropical rainforests, which represent about 32 % of its territory (Comisión Nacional Forestal [CONAFOR], 2022). According to the Food and Agriculture Organization of the United Nations (FAO, 2022), deforestation and land degradation are among the leading global sources of greenhouse gas (GHG) emissions, contributing between 10 % and 20 % of the total-equivalent to 4 to 6 Gt of CO₂ per year. In Mexico, approximately 30 % of national GHG emissions are attributed to the agriculture, livestock, and forestry sectors (Zamora-Morales et al., 2018). These activities are the second-largest source of emissions after fossil fuel combustion (Sapkota et al., 2022).
Carbon credits are a global management tool aimed at reducing CO2 emissions, while providing companies with a mechanism to offset their emissions through the implementation of projects focused on the conservation of forests, soils, and water (Cruz-Aviña et al., 2022; Ranero & Covaleda, 2018). A carbon credit is a commercial certificate that represents the prevention or removal of one metric ton of carbon dioxide from the atmosphere. These certificates are used as financial instruments in carbon markets. In 2020, the sale of carbon credits generated revenues of 53 billion (MXN); however, the COVID-19 pandemic led to an increase in GHG emissions, causing imbalances in the market (Estrada-Chavira, 2022). The effectiveness of carbon credits depends on reliable measurement methodologies that require detailed examination and documentation processes. These procedures are essential to ensure data authenticity and to demonstrate the positive impact of projects on climate change mitigation (Estrada-Chavira, 2022). For this reason, accurate quantification of biomass and carbon has emerged as a critical research topic for the development of carbon credit projects.
Biomass estimation is essential to determine carbon storage potential (Rincón-Ruíz et al., 2014; Sáenz et al., 2021). To estimate it, allometric equations have been developed based on easily measurable tree attributes such as diameter at breast height and total height (Eker et al., 2017). These equations are derived from the application of destructive sampling and regression techniques (Fonseca et al., 2013; Montes de Oca-Cano et al., 2011; Quiñonez-Barraza et al., 2019). In this context, sampling is a key component in forest inventories, as it allows the estimation of population attributes without assessing each individual tree. Sample representativeness is crucial for the validity of results, as inadequate sampling design introduces biases and lead to misinterpretation (Barrios et al., 2024).
The Forest Management Inventory (FMI) is important for decision-making in the development of management plans and the quantification of volume stocks. Given its relevance in the formulation of management strategies, it is necessary for the inventory to be conducted with high levels of precision while optimizing time and costs. In this context, the inventory design is based on establishing temporary plots, with the quantity estimated to reach the required precision level for volume stock measurements by minimizing the variance among sampling sites (Corral-Rivas et al., 2015).
Regarding forest carbon projects, comparing the sampling schemes of the Forest Protocol for Mexico (PFM) and the Forest Management Inventory (FMI) is essential for evaluating the precision and reliability of carbon content estimates. Analyzing their differences allows for the identification of potential biases, improves data quality, and ensures that the applied methodology estimates carbon stocks with greater accuracy.
The PFM of the Climate Action Reserve (CAR) was specifically developed to estimate the carbon sequestration, storage, and emission capacity of the country’s forests (González-Cásares et al., 2019; Patiño et al., 2018; Pompa-García & Sigala-Rodríguez, 2017) and it plays a crucial role in the sustainable management of forest resources, because it is considered a additional component to the production-oriented forest management program. The protocol's design involves the establishment of permanent plots, with their number proportional to the activity areas of the carbon project. Furthermore, the PFM defines standardized methods for measuring and monitoring carbon stocks and sequestration, ensuring that GHG removals generated by the projects are quantified in a consistent, transparent, and accurate manner.
The objective of this study was to compare the estimates of stored carbon in the forests of three ejidos in the state of Durango, Mexico, using two sampling protocols: the Forest Protocol for Mexico (PFM) developed by the Climate Action Reserve (CAR), and the Forest Management Inventory (FMI). The study discusses the potential use of these two data sources in the development of carbon credit projects and compares aboveground carbon estimates obtained through both protocols. The initial hypothesis suggested that, although statistically significant differences would arise between the carbon estimates produced by the two sampling protocols, both could be effectively integrated for use in forest management planning and carbon credit project development, offering a technically sound and cost-efficient approach.
Materials and Methods
Study area location
The study was conducted in the ejidos Chavarría Viejo, La Cueva y Anexos, and El Brillante, located in the municipality of Pueblo Nuevo, Durango (Figure 1). These forests are managed under the Silvicultural Development Method (MDS) and the Mexican Method for the Management of Irregular Forests (MMOBI) (Lira-Tuero et al., 2019).
Study area
The predominant climate is temperate with summer rainfall (CW) (García, 2004), with an average annual temperature ranging from 12 to 18 ºC and an average annual precipitation of 1 300 mm. The rainy season occurs from June to September, with the first frosts typically occurring in October and the last occurring in June (Instituto Nacional de Estadística y Geografía [INEGI], 2010). Vegetation is characterized by mixed conifer and broadleaf forests. The main pine species are Pinus cooperi Blanco, P. durangensis Martínez, P. engelmannii Carr., P. teocote Schiede ex Schltdl., P. herrerae Martínez, P. leiophylla Schl. & Cham., and P. strobiformis Engelm. The most representative oak species are Quercus sideroxyla Bonpl. and Q. rugosa Née. In addition, other tree and shrub species from the genera Juniperus, Arbutus, and Alnus are also present (González-Elizondo et al., 2012).
Data collected using the Forest Protocol for Mexico v3.0 (PFM)
In each ejido, sampling plots were established (1 356 in total) in accordance with the PFM of the Climate Action Reserve (CAR). This sampling scheme uses two plot sizes: one of 400 m2 (radius = 11.28 m) for measuring large trees (diameter at breast height [DBH] ≥ 30 cm; height [h] ≥ 3 m), and another of 100 m2 (radius = 5.64 m) for small trees (5 ≤ DBH < 30 cm). To ensure comparability between sampling protocols and considering that the minimum recordable diameter in the Forest Management Inventory is 7.5 cm, only trees with DBH ≥ 7.5 cm were included in the analysis.
At each sampling plot, general data were recorded, including date, plot number, latitude, longitude, slope, and aspect, using a GPS device. Detailed tree-level information was also collected, including species, DBH, and h, measured with a diameter tape and a Suunto® analog clinometer. Additionally, tree vigor was assessed and categorized as ‘high or moderately vigorous’, and radial growth (mm) over the past five years was measured. Data for the PFM plots were collected in 2022.
Data from the Forest Management Inventory (FMI)
The FMIs were conducted in 2015, and the data were obtained from the Temperate Forest Planning System platform (SiPlaFor; CONAFOR, 2023). This inventory uses fixed-size circular plots (1 000 m2) randomly established across all areas of the ejidos with timber forest harvesting activities.
Data were collected from 6 615 sampling plots distributed across the three ejidos. In each plot, general information such as plot number, coordinates, slope, and aspect were recorded using GPS and a Suunto® analog clinometer. Additionally, dendrometric data for commercial trees (DBH ≥ 7.5 cm) were collected, including species, dominance, DBH, total height, and crown height (hc, m). The latter variables were measured using calipers and a clinometer. Radial growth (mm) was also recorded by extracting increment cores from one dominant, one intermediate, and one suppressed tree.
Biomass and carbon estimation
Aboveground biomass was estimated for both sampling schemes using the CALCBOSK equations, a tool that integrates equations compiled by the Climate Action Reserve for use in PFM v3.0. Carbon content by species was determined based on carbon proportions reported by Vargas-Larreta et al. (2017). These estimates were calculated at the tree, plot, and hectare levels in order to compare the results from both sampling schemes. Additionally, the biomass of the main species in the region (Pinus, Quercus, Juniperus, and Arbutus) was compared using both the CALCBOSK equations and those developed by Vargas-Larreta et al. (2017). The latter were specifically generated to estimate biomass for the most important forest species in temperate forests located in the northwestern region of Mexico.
Statistical Analysis
First, the Shapiro-Wilk test (P ≥ 0.05) was applied to assess the normality of carbon estimates obtained from both sampling schemes. Since the normality assumption was not met (P < 0.05), the non-parametric Kruskal-Wallis test was used to analyze the existence of significant differences between the sampling schemes and biomass equations. The precision of the carbon estimates was determined using the coefficient of variation (CV) and the sampling error.
To strengthen the statistical robustness of the comparison between sampling schemes, the bootstrap resampling method was employed to estimate the mean, standard deviation, and confidence intervals of the carbon estimates. This approach involves generating multiple subsets of data through random sampling with replacement, allowing the same plot to be selected more than once within a single subset. The bootstrap means, which approximate a normal distribution, were used to calculate the sample confidence intervals. Following recommendations from authors such as Lee-Ing et al. (2012) for evaluating variability and uncertainty in estimates, each plot was resampled 1 000 times. All statistical analyses were conducted using R® 4.3.1 (R Core Team, 2023).
Results
Table 1 shows the number of plots used per ejido for carbon assessment, as well as the sampling intensity and estimated carbon content (Mg·ha-1) obtained with both sampling schemes. In all three ejidos, the number of plots and sampling intensity in the FMI were higher than those in the PFM. Regarding carbon estimates, the PFM yielded an average of 53.56 Mg·ha-1 for the three ejidos, whereas the FMI produced 43.80 Mg·ha-1, representing a difference of 9.76 Mg·ha-1 between the two sampling schemes.
Table 1 Carbon estimates per ejido using the sampling schemes of the Forest Protocol for Mexico (PFM) from the Climate Action Reserve (CAR) and the Forest Management Inventory (FMI).
| Ejido | Area(ha) | Plots | Sampling intensity (%) | Carbon (Mg·ha-1) | |||
|---|---|---|---|---|---|---|---|
| PFM-CAR | FMI | PFM-CAR | FMI | PFM-CAR | FMI | ||
| Chavarría Viejo | 5 772.93 | 413 | 2 865 | 0.29 | 4.96 | 57.05 | 49.75 |
| La Cueva y Anexos | 2 056.31 | 349 | 1 080 | 0.68 | 5.25 | 52.50 | 42.13 |
| El Brillante | 7 220.67 | 594 | 2 670 | 0.33 | 3.70 | 51.13 | 39.53 |
| Promedio | 5 016.64 | 452 | 2 205 | 0.43 | 4.64 | 53.56 | 43.80 |
Table 2 presents the precision indicators for both sampling schemes in estimating carbon for the three ejidos. On average, the PFM showed a coefficient of variation of 60 %, while the FMI recorded 49 %. Similarly, the average sampling error was 2.99 % for the PFM and 0.93 % for the FMI, reflecting lower variability and, consequently, higher precision in the latter.
Table 2 Precision of carbon estimates based on the coefficient of variation (CV) and sampling error in the Forest Protocol for Mexico (PFM) from the Climate Action Reserve (CAR) and the Forest Management Inventory (FMI) sampling schemes.
| Ejido | Sampling | Average (Mg·ha-1) | Variance (Mg·ha-1) | Standard deviation (Mg·ha-1) | CV | Sampling error (%) |
|---|---|---|---|---|---|---|
| Chavarría Viejo | PFM-CAR | 57.05 | 1 031.92 | 32.12 | 0.56 | 3.10 |
| FMI | 49.75 | 506.93 | 22.52 | 0.45 | 0.82 | |
| La Cueva y Anexos | PFM-CAR | 52.50 | 880.06 | 29.67 | 0.57 | 3.11 |
| FMI | 42.13 | 343.85 | 18.54 | 0.44 | 1.11 | |
| El Brillante | PFM-CAR | 51.13 | 1171.39 | 34.23 | 0.67 | 2.75 |
| FMI | 39.53 | 519.09 | 22.78 | 0.58 | 0.86 | |
| Average | PFM-CAR | 53.56 | 1 027.79 | 32.01 | 0.60 | 2.99 |
| FMI | 43.80 | 456.62 | 21.28 | 0.49 | 0.93 |
On the other hand, Table 3 shows the results of the Kruskal-Wallis test applied to the carbon estimates; the analysis indicates statistically significant differences (P < 0.001) between the two sampling schemes for the three ejidos.
Table 3 Kruskal-Wallis test for carbon estimates using the sampling schemes of the Forest Protocol for Mexico (PFM) from the Climate Action Reserve (CAR) and the Forest Management Inventory (FMI) by ejido.
| Comparison | Ejido | P-value < 0.001 |
|---|---|---|
| PFM CAR-FMI | Chavarría Viejo | 1.42E-05 |
| PFM CAR-FMI | La Cueva y Anexos | 1.14E-10 |
| PFM CAR-FMI | El Brillante | 2.91E-15 |
According to Table 4, the average carbon estimates obtained through bootstrap resampling analysis were 53.54 Mg·ha-1 for the PFM scheme and 43.81 Mg·ha-1 for the FMI for the three ejidos. The standard deviation was 1.49 Mg·ha-1 for the PFM and 0.49 Mg·ha-1 for the FMI, indicating lower variability in the latter’s estimates. The 95 % confidence intervals were 49.58-57.47 Mg·ha-1 for the PFM and 42.30-45.02 Mg·ha-1 for the FMI.
Table 4 Carbon estimates using the bootstrap resampling technique in the sampling schemes of the Forest Protocol for Mexico (PFM) of the Climate Action Reserve (CAR) and the Forest Management Inventory (FMI).
| Ejido | Sampling | Bootstrap average (Mg·ha-1) | Standard deviation (Mg·ha-1) | Confidence intervals (Mg·ha-1) |
|---|---|---|---|---|
| Chavarría Viejo | PFM-CAR | 57.00 | 1.56 | 52.94-61.11 |
| FMI | 49.75 | 0.43 | 48.71-50.82 | |
| La Cueva y Anexos | PFM-CAR | 52.52 | 1.58 | 48.34-56.60 |
| FMI | 42.15 | 0.58 | 40.67-43.62 | |
| El Brillante | PFM-CAR | 51.11 | 1.34 | 47.47-54.71 |
| FMI | 39.54 | 0.46 | 38.42-40.62 | |
| Average | PFM-CAR | 53.54 | 1.49 | 49.58-57.47 |
| FMI | 43.81 | 0.49 | 42.30-45.02 |
Figure 2 shows the comparison of tree-level biomass estimates using equations from CALCBOSK and those developed by Vargas-Larreta et al. (2017) for the main Pinus species. In general, the equations by Vargas-Larreta et al. (2017) yielded significantly higher estimates (P < 0.01), except for Pinus leiophylla, P. teocote, and P. strobiformis. On the other hand, Figure 3 presents the tree-level aboveground biomass estimates for species of Quercus, Juniperus, and Arbutus obtained using both sets of equations. In contrast to the observations made for Pinus species, in this case, the CALCBOSK equations produced significantly higher biomass values (P < 0.01), except for Quercus durifolia Seemen.
Figure 2 Mean values (bars) and standard error (whiskers) of the biomass estimated for the main species of the genus Pinus, using the equations from CALCBOSK and Vargas-Larreta et al. (2017). 1 = Pinus cooperi, 2 = Pinus durangensis, 3 = Pinus engelmannii, 4 = Pinus herrerae, 5 = Pinus leiophylla, 6 = Pinus teocote, 7 = Pinus strobiformis, 8 = Pinus lumholtzii, 9 = Pinus douglasiana, 10 = Pinus oocarpa.
Figure 3 Mean values (bars) and standard error (whiskers) of the estimated biomass for the main species of Quercus, Juniperus, and Arbutus, using the equations from CALCBOSK and Vargas-Larreta et al. (2017). 1 = Quercus sideroxyla, 2 = Quercus rugosa, 3 = Quercus durifolia, 4 = Quercus crassifolia, 5 = Juniperus deppeana, 6 = Arbutus bicolor.
Discussion
Carbon storage in forest ecosystems is determined by various factors, including management practices, stand age, and the structure and composition of the vegetation (Galván-Moreno et al., 2024; Medrano Meraz et al., 2017). In this study, the highest carbon stocks were recorded in Pinus species, reflecting the dominant composition of the forest by this genus. This result is consistent with the findings of Silva-Flores et al. (2014), who emphasized that Pinus species represent the main carbon reservoirs in the temperate forests of the Sierra Madre Occidental.
The carbon values obtained using the PFM v3.0 sampling scheme are consistent with those reported by Pimienta et al. (2007), who estimated 51.12 Mg·ha-1 in the ejido La Victoria, Pueblo Nuevo, Durango. Similarly, they agree with the results of Vargas-Larreta et al. (2017), who reported values ranging from 5.12 to 232.94 Mg·ha-1, with an average of 63.80 Mg·ha⁻¹ for the temperate forests of Durango. Likewise, Graciano-Ávila et al. (2019) estimated a total biomass of 130.28 Mg·ha-1 and an aboveground carbon content of 65.14 Mg·ha-1 for tree species in the same region. Unlike this study, where carbon concentration by species was determined through laboratory analysis (Vargas-Larreta et al., 2017), the aforementioned authors applied a fixed conversion factor of 0.50 to estimate the carbon content in Pinus, Quercus, Juniperus, and Arbutus species. Additionally, the results are comparable to those reported by López-Serrano et al. (2019, 2021) for forests in the same region, which further validates the findings of this study
Avery and Burkhart (2015), as well as Alberdi et al. (2016), emphasize that an efficient forest inventory, both statistically and operationally, must include a sufficient number of sampling units to achieve the required level of precision. According to the CAR’s PFM v3.0, the expected sampling error for a forest carbon project should be ±5 % of the mean at a 90 % confidence level, depending on the number of strata defined in the project’s activity area. In this study, the PFM sampling scheme showed a sampling error of 2.99 %, confirming that it meets the established precision level.
CONAFOR (2015), through SiPlaFor, establishes that the sampling error must not exceed 10 % at a 95 % confidence level to obtain acceptable results and comply with current forestry regulations, specifically NOM-152-SEMARNAT-2023 (Secretaría de Medio Ambiente y Recursos Naturales [SEMARNAT], 2023). To meet this criterion, a sampling intensity greater than 3 % is recommended. In this regard, the FMI scheme achieved a sampling intensity of 4.64 % (Table 1) and an average relative sampling error of 0.93 % (Table 2), which is below the limit established by NOM-152 (SEMARNAT, 2023) for the estimation of volume stocks in forest management programs. These results ensure compliance with the required regulatory standards.
The results agree with those reported by Corral-Rivas et al. (2016), who argue that it is possible to reduce sampling intensity in management-oriented inventories in Mexico, particularly on properties larger than 2 000 ha, without compromising data quality. According to these authors, a reduction of up to 50 % in sampling intensity could be achieved while maintaining adequate levels of precision in volume estimates. This reduction would not only help optimize the time and costs involved in conducting inventories but also maintain the accuracy of volume estimates essential for sustainable forest management.
The significant difference observed in biomass and carbon estimates between the two sampling schemes can be largely attributed to the difference in sampling intensity (0.43 % in the PFM and 4.64 % in the FMI), as greater sampling effort helps reduce variability and uncertainty in the estimates (McRoberts et al., 2013). However, additional factors such as differences in plot size and the timing of the inventories may also explain the observed discrepancies in the estimates (Navarro Cerrillo et al., 2024; Roque et al., 2023).
The variability of carbon estimates in relation to sampling intensity was evaluated using the bootstrap resampling technique. This tool is widely recognized for its utility in statistical inference, because it provides a data-driven approach to estimate the uncertainty associated with a population distribution (Quiñonez-Barraza, 2024). According to Egbert and Plonsky (2021), the bootstrap method enables the quantification of the precision of sample estimates and is particularly effective when dealing with small sample sizes or with data that follow unknown or non-normal distributions. Additionally, it has been proposed as a valid method for assessing data homogeneity, validating statistical results, and supporting randomized decision-making processes in forestry contexts.
In this study, biomass estimates for species of the genus Pinus were significantly higher when using the equations developed by Vargas-Larreta et al. (2017), compared to those generated by the CALCBOSK tool, except for P. leiophylla, P. teocote, and P. strobiformis. In contrast, biomass estimates for the genera Quercus, Juniperus, and Arbutus were higher when using the CALCBOSK equations, except for Q. durifolia. These differences reflect variations arising from the use of different sampling schemes and support the observations of authors such as Rahman et al. (2021), who emphasize the importance of selecting appropriate allometric models to minimize uncertainty in aboveground biomass estimates. They also caution that applying equations outside the data range for which they were developed can introduce significant errors into the results.
Conclusions
The sampling schemes showed significant differences in carbon estimates for the three ejidos, with the Forest Protocol for Mexico (PFM) v3.0 of the Climate Action Reserve (CAR) producing the highest values. However, this scheme also exhibited greater coefficients of variation and sampling errors compared to those obtained with the forest management inventory. The differences between the equations developed by Vargas-Larreta et al. (2017) and those implemented in the CALCBOSK tool highlight the need for further research aimed at evaluating biases and precision in baseline estimation for forest carbon projects, especially those implemented under the CAR’s PFM. Combining both sampling schemes could offer an efficient alternative for optimizing inventory costs by leveraging their strengths and facilitating their application in timber management programs and forest carbon projects.
