Applying a systems approach to assess carbon emission reductions from climate change mitigation in Mexico's forest sector

In consultation with CONAFOR, we identified potential mitigation scenarios to evaluate in the states of Durango (DGO) and Quintana Roo (QROO) due to their sound institutional coordination of forest policy implementation and relevance of their forests for community-based management (Bray et al 2003, Garcia-Lopez 2013, Ellis et al 2015). These states provide contrasting biophysical characteristics, historic land-use changes, and contributions to national timber production (INEGI 2015a, 2015b). DGO (total area 12.3 Mha), containing principally coniferous forests, has had low historic rates of deforestation and produces a large volume of commercial timber. QROO (total area 4.4 Mha), in contrast, has had relatively higher rates of deforestation (twice that of DGO) in mainly tropical forest types with a low volume of timber production (figure 1). See supplementary information available at stacks.iop.org/ERL/13/035003/mmedia for detailed descriptions of study areas.

We quantified the mitigation potential of the selected scenarios in the forest sector as the sum of the changes in net emissions, relative to a business as usual scenario. We used the Carbon Budget Model of the Canadian Forest Sector (CBM-CFS3) and the Carbon Budget Modelling Framework for Harvested Wood Products (CBM-FHWP). Both models are consistent with IPCC Guidelines for national GHG reporting (IPCC 2006). They have been adapted to represent Mexican conditions using data from Mexico. The scientific approach and necessary inputs for the parameterization of these models have been extensively documented in literature (Kurz et al 2009, Stinson et al 2011, Kull et al 2011, Pilli et al 2013, Smyth et al 2014, Kim et al 2016) and the supplemental material. The CBM-CFS3 implements the Gain-Loss method of the IPCC to estimate annual GHG emissions and removals in forest ecosystems. The CBM-FHWP model receives input from the CBM-CFS3 and tracks the fate of carbon in harvested biomass converted to wood products for various categories, uses, and landfills. Finally, we also estimate substitution benefits such as GHG emission reductions obtained from the use of wood products and biomass for energy (Smyth et al 2016).

Previous studies with the CBM-CFS3 in Mexico used a spatially-referenced framework that stratifies the country into 94 units based on the intersection of the 32 federal states with seven ecoregions of the North America Level 1 classification (Olguin et al 2015). The two states used in this study included six of these 94 spatially-referenced units (figure S1). We compiled and harmonized data from national, state and municipal levels on: forest distribution and forest growth (from Mexico´s National Forest and Soil Inventory, CONAFOR 2012), rates of fires, commercial harvests and land-use changes, ecological parameters (e.g. litterfall and decomposition of dead organic matter), climate, and data on harvested wood products and displacement factors (i.e. the change in emissions resulting from substitution of wood for more emissions-intensive products and fossil fuels). We estimated forest GHG emissions and removals in annual times steps for the period 2000–2050.

We constructed a business as usual (BAU) baseline scenario and 4 mitigation scenarios (with 2 sub-scenarios). The BAU baseline estimates the GHG fluxes if forest management and disturbance rates observed in the past continue into the future (2018–2050). We extended the average annual gross rates from the last 10 year period of available activity data for land-use change (LUC) (2000–2010) in ha/year, harvests (2005–2014) in m³ of wood/year and area burned (2007–2016) in ha/year. Net ecosystem CO₂e balances for the two states were generated as the sum of all GHG emissions and removals corresponding to carbon transfers in above- and belowground biomass, dead wood, litter and mineral soil forest carbon pools. We also estimated the net emissions from HWP production and use in the BAU by assuming that neither changes in policies nor changes in forest carbon cycling would occur due to climate change.

We estimated the biophysical mitigation potential (relative to BAU) if, by 2030, the following activities would be fully implemented (table 1): (M1) net zero deforestation rate, resulted from the reduction of gross deforestation rate (conversion from forest land to non-forest land) to equal gross forest recovery rate (conversion from non-forest land to forest land), (M2) M1 plus 10% increased net forest recovery rate, and (M3) increased forest productivity and timber production. For M3, we examined four sub-scenarios resulting from changes in the HWP component: (i) increased harvest using the same commodity proportions as in BAU; (ii) the increased harvest volume directed entirely to long-lived products (LLP); (iii) avoiding GHG emissions from more emissions-intensive materials (substitution benefit) using low displacement factor values; and (iv) avoiding GHG emissions estimated with medium displacement factor values. The last scenario (M4), combines all the activities (M2 and M3, including sub-scenarios). In all cases, we simulated a linear transition from BAU in 2018 to the full implementation of the mitigation actions in 2030.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Net GHG emissions in all scenarios were calculated as the sum of the GHG fluxes in the forest ecosystem, HWPs and substitution benefit components. To assess the mitigation potential of the proposed strategies, we subtracted from each mitigation scenario the net GHG emissions of BAU, and reported both annual and cumulative emission reductions to 2030 and 2050 at the state level. Net emissions before 2018 were identical for all scenarios as estimated in the BAU baseline.

The relatively coarse spatial and temporal resolution in the available land-use/land-cover maps (i.e. four- to nine-year periods, 25 ha minimum mapping unit) could lead to the underestimation of gross deforestation and forest recovery rates (Mascorro et al 2015). Thus, as a sensitivity analysis, gross deforestation rates were doubled (in ha deforested per year) and gross forest recovery rates increased (ha/year) such that the net deforestation rate remained the same in BAU and in mitigation scenarios to assess the possible impact of underestimating the gross conversion rates of forest land to other land uses.

Activity data. The historic and projected deforestation and forest recovery areas for the period 2000–2050 are shown in figure 2(a) for DGO (left) and QROO (right). In the historic period, rates of deforestation in DGO are low but variable, while forest recovery remained at a relatively stable rate. In QROO, the deforestation rate was relatively constant while forest recovery was more variable and both rates were much higher than in DGO.

The municipal level data on area burned are highly variable in both states. DGO has a minimum annual area burned of 615 ha yr⁻¹ and a maximum of 51 755 ha yr⁻¹. The mean of 18 711 ha yr⁻¹ is projected into the future for the baseline, despite the high standard deviation (±17 230 ha yr⁻¹). QROO had corresponding values of 447 ha yr⁻¹ and 79 161 ha yr⁻¹ for the minimum and maximum annual area burned, respectively, and a mean of 18 083 ha yr⁻¹ with a standard deviation of ±22 077 ha yr⁻¹ (figure 2(a)).

Harvest rates in DGO are almost an order of magnitude greater than in QROO (figure 2(a)). DGO produces nearly 1/3 of all harvested wood (mean 436 051 Mg C yr⁻¹) recorded in Mexico's national statistics with the variability in production driven by economic conditions (SEMARNAT 2014).

Emissions. Both states were estimated to be net sinks throughout the analysis period (figure 2(b)): 2000 to 2050. QROO was a sink of −14.1 Tg CO₂e yr⁻¹ compared to −7.96 Tg CO₂e yr⁻¹ for DGO. The contribution of different land categories varied with the strongest sink in forest land remaining forest land (FLFL).

Net GHG emissions in FLFL respond to (1) forest age structure which drives overall uptake rates by forest type, (2) emissions from forest fires and harvests, and (3) changes in forest area. Both states are strong sinks in the historic period (2000−2017), mean −9.7 and −17.7 Tg CO₂e yr⁻¹ in DGO and QROO, respectively. The sink strength in both states decreased over time in the baseline estimates: in DGO from −8.7 Tg CO₂e in 2017 to −6.1 Tg CO₂e in 2050 and QROO from −15.9 Tg CO₂e in 2017 to −8.0 Tg CO₂e in 2050; due to forest ageing and continuous reduction in forest area. The faster growing, relatively younger forests of QROO were a stronger sink in the historic period and at the beginning of the baseline, but they approach the sink strength of the forests in DGO by 2050. The variability seen in the historic period arises from varying incidences of disturbances, with reductions in sink strength corresponding to years with high rates of fires and harvests. This variability is removed in the projections because we use average annual fire and harvest rates in the BAU (figure 2(b)) and mitigation scenarios.

For forest land converted to other land (FLOL) during the historic period, emissions vary with the gross deforestation rates. FLOL emissions throughout the historic and baseline periods in DGO are low (1.51 Tg CO₂e yr⁻¹) compared to QROO (5.91 Tg CO₂e yr⁻¹). Both states show increasing emissions as more lands are deforested. The CBM-CFS3 simulates decay of wood residues over time and thus emissions increase as cumulative FLOL area increases. Forest recovery (OLFL) contributes a weak sink in DGO (−0.381 Tg CO₂e yr⁻¹) and QROO (−1.57 Tg CO₂e yr⁻¹) strengthening slightly over time as recovered forest area is added. Non-forest land (OLOL) emissions are shown here for completeness and represent small emissions on lands deforested more than 20 years ago. This analysis does not include emissions from land use of non-forest lands.

The cumulative mitigation benefits are summarized for forest, HWP and substitution (figure 3). Negative values represent an actual mitigation benefit, while positive numbers represent an increase in emissions with respect to the BAU scenario. In both states scenario M2 (net zero deforestation plus a 10% increase in forest recovery) achieved the greatest cumulative emissions reductions through 2050 of −24.4 Tg CO₂e in DGO and −110.9 Tg CO₂e in QROO. The greatest contribution within this scenario was achieved through a ∼50% reduction in gross deforestation rates relative to BAU, the same annual reduction reached in the net zero deforestation rate (M1 scenario). Thus, both M1 and M2 scenarios gave DGO a cumulative net emissions reduction of 2% in 2030 and 7% and 8%, respectively in 2050 (figure 3(d)). In QROO, the emissions reduction was 6% in M1 and M2 scenarios in 2030, and 23% and 24% respectively in 2050 (figure 3(d)). The more than threefold mitigation potential in 2050 of QROO compared to DGO is because QROO has greater changes in gross deforestation and forest recovery rates (table S2), as well as more rapid carbon accumulation rates and higher forest carbon density (figure S3, table S6).

Increasing forest productivity combined with increasing harvest rates (scenario M3) always yielded a relative increase in the forest emissions component relative to BAU because the increased C uptake is unable to offset the emissions associated with the greater increase in harvest (figure 3(b)). The managed forest area in QROO is relatively small and contributes only 0.6% to Mexico's annual harvest. Thus, even at year 2050, the cumulative carbon loss from increasing harvest rates is quite small in the M3 and M4 scenarios (figure 4(d)). In contrast in DGO, where more than half of the forest area is under silvicultural management, the carbon loss from harvesting 50% more and using the lower displacement factor (0.5 MgC avoided/ MgC wood used), generated 14% more CO₂e emissions by 2050 relative to BAU (figure 4(b)). The cumulative mitigation benefit became positive, with a 4% emissions reduction relative to BAU, only when the increased harvest rate was combined with higher forest recovery rates as in M4 scenario, and all the extra carbon transferred to HWPs goes to sawn wood and we assumed a displacement factor of 2 MgC/MgC (figure 4(b)).

The average annual benefit varied by decade for the different mitigation scenarios (table 2). For both states, scenario M2 provided the most emissions reductions, however the rankings of other scenarios varied over time. The carbon losses in the M3 scenario decreased towards the end of the simulation (table 2). Had we assumed full implementation of this strategy earlier than 2030 (thereby increasing the number of harvest cycles of relatively fast-growing forest) the mitigation potential would have increased. These results show that the assumed increase in productivity does not offset potential C stock reductions resulting from increased harvest rates. While increases in forest productivity through forest management can be achieved, to maintain C stocks, the overall rates of harvests need to be limited if increasing forest C stocks is also a goal.

Maintaining the net deforestation rate but increasing the magnitudes of gross deforestation and forest recovery rates reduced the mitigation benefit in the M1 scenario because there are more emissions in the new BAU (e.g. −98.2 Tg CO₂e in 2050). In contrast, under the M2 scenario, a 10% increase in the forest recovery rate yielded a 33% mitigation benefit relative to the new BAU by 2050 as more overall forest area was recovered (figure 5 (a)).

National policies have different effects on actual emissions reductions depending on local forest characteristics and historic rates of LUC and it is therefore important to consider different policies at the state or even municipal levels when designing mitigation strategies. For example, a mitigation target based on a change of activity data (e.g. net zero deforestation rate) may yield similar types of benefits, but at very different magnitudes depending on where it is implemented. In this study, avoiding deforestation has significant mitigation benefits in QROO and much less so in DGO, because of the different deforestation rates in the baseline case and differences in forest carbon density at maturity (higher in QROO and lower in DGO).

In the case of DGO, if harvest is increased by 50%, it will generate higher CO₂e emissions relative to BAU. However, implementing forest management actions sooner (e.g. transitioning sooner to higher productivity and harvest cycles) can ameliorate the increase in CO₂e emissions in the M3 scenario by 2050 and provide other co-benefits such as increasing revenues from LLP and diversified job options. In addition, timing is also relevant for enhancing the mitigation potential of other strategies, such as increasing benefits from enhanced forest recovery (the earlier the better) or the duration of carbon stored in HWPs.

In the M4 scenario (where all forest strategies are combined), adding more LLP and changing the displacement factor from a low to a medium value in DGO, changed the mitigation potential so that M4 became the best mitigation scenario that included increased harvest for that state (last sub-scenario in figure 4(b)). Although more local data on displacement factors are required, this example shows the role that HWPs can play in achieving forest carbon mitigation targets and highlights the importance of including them in national GHG inventories.

One of the key sources of uncertainty in assessing impacts of future scenarios was the rate of LUC. These data were derived from a change assessment over multi-year periods. However, given the length of the observation period and the high rates of forest growth, some areas may have had undetected forest cover loss and regrowth between observations. For example, Urquiza et al (2007) and Lawrence and Foster (2002) report that basal area of secondary forest stands (∼25 years old) in QROO, already had reached 40% of that of old-growth (>50 years old). If such disturbance/regrowth events occured between observations, then the gross rates of land-use changes would be higher than those assumed in our analyses.

New annual land-use/land-cover maps are being produced at various spatial and temporal resolutions (e.g. Hansen et al 2013, Gebhardt et al 2014/MAD-Mex system) and these could be used in the future to derive more accurate estimates of annual gross rates of LUC. Re-running the analyses for QROO using the higher gross LUC rates but same net deforestation rate (figure 5(a)) increased the estimate of GHG emissions, and reduced the size of the sink in the later years of the BAU and mitigation simulations (figure 5(b)). In the adjusted BAU (dotted black line), the cumulative CO₂e sink is 23% lower (−46.7 TgCO₂e) than the original BAU (solid black line). However, the differences in the cumulative mitigation benefits are quite small (figure 5(c)). While the available activity data (derived from relatively coarse spatial and temporal resolution) might be underestimating gross rates of LUC and therefore net CO₂ emissions, the rank order of the mitigation strategies remains unchanged.

Reducing gross deforestation could provide significant biophysical mitigation potential, particularly in QROO. Mexico´s Emissions Reduction Initiative (FCPF 2016) identifies several activities for QROO including improved silvopastoral/agroforestry systems, sustainable forest management, environmental services payments, and strengthening local regulations and governance. In Mexico and in other countries in Central America, community-based forest management and ecosystem services payments programs are among the most effective policy options for improving forest cover (Min-Venditti et al 2017). Forest management currently has little impact on emissions in QROO. However, if the area under management is increased greatly, the benefit would be twofold; it would help to reduce the overall rate of deforestation and avoid forest degradation through illegal logging practices (the last two not considered in this analysis). It could also increase the number of policy actions and co-benefits (Kapos et al 2012, Skutsch et al 2017), expand the collaboration with other sectors (e.g. HWP for the tourism architecture in the Maya Riviera, Sierra-Huelsz et al 2017) and still maintain the positive mitigation benefits within the state.