Mitigation and adaptation are the main strategies to address climate change. Mitigation aims at limiting the degree of the phenomenon by affecting emissions and removals of greenhouse gases from the atmosphere (IPCC 2001). Adaptation, on the other hand, seeks to hamper the side effects of a changing climate by adjusting social and ecological systems to current or expected climate variability (IPCC 2001). It can be said that mitigation addresses the causes of climate change while adaptation focuses on the consequences (Locatelli et al. 2011). However, as some degree of climate change is unavoidable (IPCC 2014), both strategies are important elements of an integral strategy to tackle the phenomenon. Moreover, mitigation and adaptation are interrelated activities: the cost and benefits of mitigation affect the cost and benefits of adaptation and vice versa (Kane and Shogren 2000).
This interrelation is particularly evident in the land use sector, an area in which practically any activity or policy has a significant effect on the goals of both strategies (Locatelli et al. 2015). Yet, in practice, mitigation and adaptation are treated as two different policy instruments (Duguma, Minang, and van Noordwijk 2014), which results in isolated and often uncoordinated efforts to address climate change. One of the main reasons behind this dichotomy is the complexity derived from managing ecologic-economic systems (e.g. landscapes) with multiple goals in mind. In other words, as the interactions between mitigation and adaptation are not well understood, tools to assess the integrated outcomes are lacking (IPCC 2014).
Previous research efforts mainly relied in a conceptual approach to analyze the interrelation of climate strategies in the sector mentioned (Locatelli et al., 2011). For instance, in the context of tropical deforestation and forest degradation - the topic analyzed here -, activities with potential to deliver co-benefits (mutual mitigation and adaptation gains) were considered as enhanced policy outcomes or synergies, for short. However, it has also been shown that significant trade-offs between climate strategies might arise when mitigation and adaptation goals are considered (Pramova et al. 2012). For example, when the implementation of a mitigation project restraints the access to forest resources to communities that rely on them (Locatelli et al., 2008).
It is important to highlight that from an economic point of view, the presence of co-benefits does not automatically translate into an enhanced policy outcome. If it is considered that majority of environmental services that contribute to mitigation and adaptation goals are treated as externalities (Barbier, Burgess, and Grainger 2010), economic theory suggests that - unless intervention takes place - outcomes are likely to be sub-optimal (Mas-Colell, Whinston, and Green 1995). Likewise, the presence of a trade-off is not necessarily detrimental for social welfare. As long as the marginal benefits generated from an activity are greater than its costs, performing this activity results in welfare gains (Barbier, Burgess, and Grainger 2010). However, once again, the presence of externalities distorts the private gains and cost from the social ones, pushing the equilibrium away from the optimum. Thus it is evident that in order to address managerial aspects of the integrated implementation of climate strategies, the conceptual approach previously followed must be extended to account for the effects described.
The provision of relevant ecosystem services has been used as a framework to better understand the interactions between mitigation and adaptation (Locatelli et al., 2008). Ecosystem services (ES) are the benefits that societies obtain from functioning ecosystems (Costanza et al. 2017) and they are classified as provisioning, regulating, supporting and cultural (MEA 2003). For mitigation and adaptation purposes, the first two categories are the most relevant ones. Provisioning services are the goods obtained from ecosystems (e.g. agriculture products, wood, charcoal) (MEA 2003). Regulating services are the benefits derived from the regulation of ecological processes (MEA 2003) and they range in spatial scale from global (e.g. carbon storage, biodiversity conservation) to regional (e.g. soil retention, waterflows regulation, micro-climate regulation) (Locatelli et al., 2008). In this regard, mitigation is related to the global dimension of regulating services while adaptation to provisioning and regional regulating services.
It has been showed that contrary to isolated efforts (e.g. only mitigation), the joint implementation of mitigation and adaptation measures restores optimality when co-benefits prevail (Lopez 2016). A situation that takes places, for example, when carbon storage (e.g. the amount of carbon contained in biomass pools) positively correlates with other ES relevant for adaptation. However, it has also been found that carbon storage might have a low correlation with other ES (Locatelli, Imbach, & Wunder, 2014), which dilutes welfare gains of the integrated approach. Furthermore, as it was mentioned before, the possibility of triggering important trade-offs exists when both strategies are considered. Hence, there is a need to explore implications of the integrated approach in more general situations. It is, when co-benefits but also trade-offs between different ES are explicitly considered as side effects of land reallocation process. The main goal of this paper is to contribute to fill-in this knowledge gap.
One useful framework to analyze the implications of explicitly considering co-benefits and trade-offs between climate strategies is ecosystem services elasticity, which measures how human welfare responds to changes in key environmental stocks (Daw et al. 2016). In this analysis, an adapted version of the concept mentioned was implemented by means of a static optimization model. The model was then used to analyze environmental and welfare implications of different policy configurations in the context of tropical deforestation using exploratory modeling and analysis (EMA) (Kwakkel and Pruyt 2013), a methodology designed to deal with complex and uncertain issues. The mentioned approach helped to identify policy configurations with synergic properties, which were seen as interventions able to simultaneously improve mitigation and adaptation indicators with respect to the unregulated situation.
In order to accomplish the goal of the paper, the rest of the manuscript is structured as follows: in the next section the description of the methods is presented. A special emphasis is made on how the concept of ecosystem service elasticity was translated from a conceptual map to an economic optimization problem. Likewise, details about the implementation of EMA are offered. In the third section, the performance of different policy configurations is presented and analyzed with the aim of identifying best policy configurations. The paper finalizes with some concluding remarks. An appendix deals with technical details regarding equilibrium determination (A.1), the specification of regulating measures (A.2) and a brief explanation of forest transition model and the classification of countries according to it (A.3).
Based on the classification of ES proposed by Millennium Ecosystem Assessment (MEA 2003) and the concept of ES elasticity (Daw et al. 2016), a conceptual framework that explicitly shows the assumed causal chain between land allocation and welfare for a hypothetical ecologic-economic system was developed. This framework was translated into a 2 × 2 production model (Mas-Colell, Whinston, and Green 1995), which was augmented with the presence of environmental externalities. Following the EMA approach, different policy configurations were simulated under deep uncertainty conditions. The economic model was implemented in
The following subsections show in detail the implementation of each of the steps mentioned above.
At an aggregated level, the concept of ES suggests a positive relationship between their provision level and human welfare. However, at a lower scale, and especially when one considers their interactions, this relation becomes complex and often indirect (Daw et al. 2016). In the context of tropical deforestation - the main issue analyzed here - there is a trade-off between provisioning services from agriculture and forestry. Both productive activities compete for land, thus - holding everything else constant - the output in one cannot be increased without reducing the output in the other. In fact, agriculture expansion is recognized as the main driver of deforestation at the global level (Hosonuma et al. 2012).
The relation between regulating and provisioning services is even more difficult to observe. However, there seems to be a co-benefit between forestland and agriculture. Empirical evidence supports that forest in the proximity of croplands have a positive effect in yield (Reed et al., 2017), which suggests a positive relation between regional regulating services and agriculture output. In addition, there are interactions among regulating services at different scales. For instance, carbon storage - an important ES for mitigation - exhibits a mixed effect with respect to different regional regulating services. Soil related services (e.g. soil protection or fertility) tend to increase with higher carbon stock (Locatelli et al., 2014). However, water runoff declines with a denser forest (Duncker et al., 2012; Farley, Jobbágy, & Jackson, 2005; Locatelli et al., 2014). Putting it in another way, carbon storage may be related to both co-benefits and trade-offs.
Taking into consideration the co-benefits and trade-offs related to carbon storage, the impact of deforestation and forest degradation Forest degradation must be interpreted as a reduction of biomass with respect to a natural forest without involving a land use change.
The conceptual framework developed for this analysis is displayed in Figure 1. It shows the assumed causal chain from land allocation to social welfare. The figure demonstrates that land can be kept as a natural forest (
Land allocation determines the output of environmental goods (
The reallocation of land, however, has a negative impact on carbon stock. Two main consequences follow from this: firstly, hydrological services (
Based on the conceptual framework presented in the previous subsection, a 2 × 2 production model augmented with an externality was developed with the purpose of deriving equilibrium allocation of resources and provision level of the ES under consideration. The model derives the equilibrium allocation by solving the optimization problem of the representative firms in the agriculture and forestry subsectors. As it was indicated before, the provision level of ES is determined by land allocation and it has an impact on agriculture output. In what follows, the mathematical representation of these processes is shown. The appendix shows further details regarding equilibrium determination and regulating measures.
To begin with, the representative firm in each subsector (agriculture and forestry) is assumed to be competitive and its goal is to maximize profits subject to a technological constraint. In addition, the profit function considers the policy followed, which takes the form of a Pigouvian tax/subsidy on land. In mathematical terms, the problem of the firm is:
A small open economy setting is followed, which implies that output prices are exogenously determined. Thus, the overall value of the production of the given economy is:
In this analysis the GDP is used as the social welfare measure and changes in this variable are interpreted as an adaptation indicator as a proxy to regional welfare. In addition, the last term in equation (2) represents a compensation scheme dependent on policy configuration. In the unregulated setting, this term is zero as no land is left as a natural forest.
At this point, it is important to clarify that the use of GDP as a measure of welfare is subject of debate and alternative measures have been proposed to better assess this dimension of sustainable development (Giannetti et al. 2015; Kubiszewski et al. 2013). However, as the model presented here is formulated, the GDP formulation simply captures the behavior of output value, leaving aside any further complexity that in reality might take place. This implies that a weak sustainability approach is followed, or in other words, a non-declining consumption is considered the indicator of sustainability (Ang and Van Passel 2012; Davies 2018).
Land allocation directly affects carbon stock through deforestation and degradation. These impacts affect
The previous equation can be interpreted as follows: if no production process is performed, carbon stock is at its maximum capacity (the first term on the right hand side of equation (3)). However, if land is reallocated to agriculture,
Changes in the carbon stock affect the provision level hydrological services and the state of the environment. More specifically, it is assumed that the provision of
This formulation implies that when carbon stock is reduced, either by deforestation or degradation, the provision of hydrological services increases (e.g. due to lower evapotranspiration).
The state of the environment is specified by a normalized CES function (Barro and Sala-i-Martin 2004) that depends on
This formulation captures the key idea that ES at aggregated level increase welfare but, at a lower level, their effect on the indicator mentioned is ambiguous and indirect. For example, a decrease in
Figure 2 illustrates the effect of land use change on the state of the environment at aggregated level, which results from the combination of equations (4), (5) and different sets of parameters. It is precisely this relation, the one treated under deep uncertainty conditions, as it will be detailed later. The dotted line shows the unregulated situation: it is inaccurately perceived that land use change has no effect on the state of the environment. However, the reduction of the carbon stock through land reallocation has an initial ambiguous effect on the state of the environment. For instance, a combination of parameters might immediately deteriorate it, as the black line shows. Other combination of parameters might exhibit an initial smoother response, as indicated by the blue curve. There is even the possibility of initially observing a relatively neutral response in the state of the environment, as the red line illustrates. Excessive land use change, nevertheless, leads to a deterioration of the environment and lower productivity in all cases. Situations like the one exemplified in by the gray curve are considered unrealistic, as excessive land use change would lead to an improvement in the state of the environment.
The EMA approach is a research methodology designed to analyze complex systems under deep uncertainty conditions. These are situations where it is reasonable to incompletely enumerate multiple possibilities without being able to rank how likely these are judged to be. In the context of EMA, a single model run is considered an experiment and it reveals how the system would behave if various guesses were correct. By conducting multiple experiments, it is possible to explore the implications of those guesses (Kwakkel and Pruyt 2013).
In this exercise, the model shown in the previous section was simulated using different policy configurations considering deep uncertainty in parameters specification. Specific details about model quantification and simulation settings are shown in the next subsections.
In order to derive a numerical solution of the model presented in the previous section a Cobb-Douglas production function exhibiting constant returns to scale was used for both subsectors. The selected parameters, presented in Table 1, show that the assumption of relative factor intensities was followed. In particular, it is assumed that forestry sector is relatively intensive in land while agriculture sector is relatively intensive in labor. Additionally, to reflect that deforestation is driven by agriculture expansion, the price of agriculture output was set 35 percent higher than forestry output. Parameters selected replicate average land allocation in tropical countries at the late transition phase, which is around 21 percent for degraded forest, natural forest is nearly zero in the phase mentioned (Hosonuma et al. 2012).
Parameters of the model.
α | 0.3 | Exponent agriculture land in production function | Jointly determined to replicate late transition phase as reported in Hosonuma |
β | 0.7 | Exponent agriculture land in production function | |
pa | 1.3 | Output price agriculture | |
pf | 1 | Output price forestry | |
γ | 0.25 | Exponent of state the state of the environment in externality function | Value selected to replicate forest cover along the different phases of forest transition as reported in Hosonuma |
Cmax | 1 | Maximum carbon stock (Normalized) | The values could not be derived from the literature directly. Nevertheless they are consistent with what can be found there |
1 | Land endowment (Normalized) | ||
1 | Labor endowment (Normalized) | ||
k | 0.25 | Inflection point of Hill function | Assumption consistent with mathematical formulation |
n | 1 | Steepness Hill functions | Assumption consistent with mathematical formulation |
Random Parameters | |||
s | N(0.5, 0.083) | Carbon share | Assumption consistent with mathematical formulation* |
α | U(0.1 – 0.85) | Elasticity of substitution | Assumption consistent with mathematical formulation |
d | U(0.1 – 0.5) | Degradation | Specht |
Source: author’s own elaboration based on indicated sources.
Implausible shapes for equation 2.5 are seen if parameters are outside the specified ranges.
The externality associated to the state of the environment - the function
Resource endowments and maximum carbon storage were normalized with the purpose of simplifying the interpretation of results. In this way, allocation of resources is measured in relative terms and changes in carbon stock are reported as the fraction remaining with respect to its maximum level. The provision of hydrological services (see equation (4)) was assumed to respond smoothly to initial changes in carbon stock. Notice that parameters described so far were fixed during simulations.
Parameters of equation (5) were randomly specified with the purpose of capturing the context specific nature of adaptation. In other words, it is considered that the same system structure applies in all regions (see Figure 1). However, parameters defining the relationships – in particular those capturing the externality associated from forest ecosystems to agriculture (see doted lines in Figure 1) -, differ from region to region. It must be considered that the functional form of the equation mentioned along with complementarity of regulating services restrain the plausible range of values that the relevant parameters might take. The carbon share parameter, for instance, is naturally defined for range of values between zero and one. It is additionally assumed that balanced profiles, where both ES are important to define the state of the environment (e.g. the parameter
A similar reasoning allowed to specify the range of values for the elasticity of substitution. More precisely, given equation (4) and the assumed variation of the parameter
The variation in the degradation parameter was defined according to values reported in the literature. The main drivers of degradation at the world level are timber extraction and wood fuel collection (Hosonuma et al. 2012). Based on reports for Brazil, fuel wood collection represents extraction level between 19% and 35% of the standing biomass (Specht et al. 2015). Likewise, depending on its intensity, timber extraction reduces standing biomass between 20% and 48% with respect to natural forest (Gerwing 2002). The model developed here does not distinguish between the different sources of degradation. Hence, to capture the full range of values reported, the variation of this parameter was specified following a uniform distribution ranging from 0.1 to 0.5.
Five different policy experiments were performed. The purpose of these experiments was to identify policy approaches consistent with the definition of synergy. A general overview of the configuration of these scenarios is shown in Table 2.
Policy configurations.
Unregulated | 0 | No | - | - |
Reserve | 7 | No | - | - |
Compensation | 7 | Yes | - | - |
Mitigation | 7 | No | Tax on C | - |
Adaptation | 7 | No | - | Subcidy on H |
Policy integration | 7 | No | Tax on C | Subcidy on H |
Source: Author’s own elaboration.
Notes:
Percentage of total land endowment.
When compensation mechanism is active (yes), the amount of land kept as natural forest (
A tax implies that
Variables calculated for the numerical solution of the model and their use.
Πj | Profit | Not reported but known to be zero |
Yj | Production | Calculate GDP |
Lj | Land use | Model calibration |
Tj | Labor use | Not reported |
λj | Shadow price | Not reported |
Factors Markets | ||
w | Wage rate | Not reported |
r | Opportunity cost of land | Calculate GDP (depending on policy configuration) |
Ecologic sector | ||
C | Carbon stock | Mitigation indicator |
H | Hydrological services | Indirectly contained in GDP calculation |
E | State of the environment | Calculate GDP |
Source: author’s own elaboration.
Average forest cover in 2010 per phase of forest transition process.
Pre transition | 13 | 69.9 |
Early transition | 39 | 46.6 |
Late transition | 33 | 21.3 |
Post transition | 15 | 27.3 |
Source: author’s own elaboration based on World Bank data.
In the
The
To perform EMA analysis, one thousand random numbers for each parameter specified in Table 1 were generated. These numbers were used to determine the equilibrium according to policy configuration and parameterization of the model. Each experiment was classified as a synergy or a form of trade off. In other words, carbon storage (mitigation indicator) and resulting GDP (adaptation indicator) in the unregulated economy were used as a benchmark to determine the impact of the specific policy design: a synergy implies that both criteria are improved with respect to the unregulated economy. On the contrary, a tradeoff indicates that the policy improves one criterion by worsening the other. Two categories of trade-offs are possible: mitigation trade-off (MT) and adaptation trade-off (AT). The former indicates that a gain in carbon implies a reduction in GDP while the latter indicates that a gain in GDP requires further carbon losses. Lastly, a fourth category was defined as disjunction, which indicates that policy intervention deteriorates both indicators with respect to the unregulated situation.
For each policy scenario considered, one thousand (reproducible) simulations were performed according to the specifications described in the previous section. Unsuccessful (equilibrium was not found) and unrealistic (
Figure 3 displays the classification of outcomes of each policy scenario analyzed. As it can be seen from the figure, the scenarios based on current policy approaches (reserve and compensation) perform poorly compared to the scenarios implementing alternative policy schemes, based on the internalization of externalities (adaptation, mitigation and policy integration). For example, a disjunction was observed in 22% of the times in the case of reserve and compensation scenarios. Likewise, a mitigation trade off was observed 74% of the times in the case of reserve and 16% of the times in the case of compensation scenario. And a synergy was observed 4 and 62% of the times respectively. These last numbers show that the main effect of a compensation scheme is to switch some of the instances from a mitigation trade off to a synergy, which reflects the relevance of the mechanism mentioned in the design of current environmental policy in the land use sector.
A drastically different behavior was observed when alternative policies were implemented. As the figure shows, only a small fraction of instances corresponded to disjunction (2%) or mitigation trade off (16%) when adaptation policy was implemented. Moreover, synergies were observed in all cases under mitigation and policy integration scenarios.
It is worth noticing that synergies associated to each policy have varying quantitative properties, which were not possible to distinguish in the qualitative analysis presented previously. In other words, it was not possible to elucidate whether improvements in mitigation and adaptation indicators were substantially higher in a specific policy intervention. Figure 4 shows a notched box-plot of synergic outcomes in environmental (C in left panel) and welfare terms (GDP in right panel), no overlapping notches indicate an statistically different median in different groups (Wickham 2016). As the figure reveals, the overall performance of current policy approaches is weak compared to alternative ones in both indicators. The median equilibrium carbon stock is considerably lower in the reserve and compensation scenario (0.13 and 0.15 respectively) than in the alternative policies (A:0.28, M:0.38, PI:0.42). It follows that forestland conservation tended to be higher in the alternative policies.
A similar behavior is observed in GDP with the difference that this indicator shows a higher variability in reserve and compensation scenarios, as it is shown by the width of the boxes and the outliers. This reflects that only under very specific conditions current policy approaches are comparable to alternative ones in adaptation terms.
The results obtained in reserve and compensation scenarios can be anticipated using the Rybczynski theorem (Mas-Colell, Whinston, and Green 1995). Specifically, the net effect of decreasing the endowment of land is an increase in the amount of this factor allocated to agriculture, the relatively less land intensive sector. The magnitude of this effect is fully explained by technologies and output prices. As the parameters that defined those relations remained fixed, the resulting allocation of land was the same for all experiments in these configurations. It follows that synergies are determined by extraction level in forestry sector (degradation parameter) and compensation level (see equation 2, in particular the third element in the right hand side). In the case in which these effects overcome GDP loss due to land endowment reduction and extraction in forestry sector allows a net carbon gain, a synergy is realized. However, as land allocation remains fixed, only relatively weak synergies were expected under these schemes.
The mechanism described so far also takes place in the other three configurations. However, it is complemented by the internalization of the relevant externality. Depending on policy configuration, the opportunity cost of land is modified to account for the effect that the economic subsector has on a particular environmental service. As a consequence, the allocation of factors is determined not only by the parameters mentioned before but also by the impact that land reallocation has on the provision of ES. At this point it is convenient to recall that ES determining the state of the environment (see equation (5)) were treated under deep uncertainty conditions. Thus, variability was involved and only general patterns could have been identified.
Figure 5 summarizes the results obtained for the alternative policy scenarios. The scatterplot plot in this figure shows two distinctive patters: mitigation and PI tend to group on a cloud located in the top right corner while adaptation exhibited a positively sloped shape. As the carbon density plot reveals (top panel), the main characteristics of the clouds formed by M and PI is that, in general, these configurations conserve a higher fraction of carbon compared to adaptation. In addition, these configurations delivered a higher GDP (see the right panel). The figure explains why synergies are stronger in M and PI scenarios. However, PI exhibited an even better performance, as carbon storage tended to be higher under this configuration without sacrificing GDP.
In this paper the joint implementation of mitigation and adaptation measures was analyzed assuming co-benefits and trade-offs among different provisioning and regulating ecosystem services. Concretely, a hypothetical ecologic-economic system in which deforestation and degradation of forest directly impacted output level of relevant economic activities was constructed. It was additionally considered that land use change indirectly affected output through the provision of competing regulating ecosystem services. Due to current knowledge gaps, this last relation was treated under deep uncertainty.
Different policy scenarios grouped in two main categories, current mechanisms and internalization of externalities, were tested. The simulated outcome suggest that the latter approach overcame the former in economic and environmental terms. Putting it differently, a higher carbon stock and GDP were associated to the second group of policy mechanisms. Moreover, policy integration - a mechanism combining mitigation and adaptation measures in the land use sector as a whole-, exhibited the best performance among the interventions in the second group.
The mentioned results indicate that synergies may be more common and accessible than previously considered, however, current dichotomy approach (e.g. adaptation and mitigation) and mechanism (e.g. compensation) may not be the best tools to trigger them. Furthermore, the importance of conceiving the land use sector as an entity rather than as a set of isolated components (e.g. agriculture and forestry) was highlighted.