|Home | Archives | About | Login | Submissions | Subscribe | Contact | Search|
Copyright © 2011 by the author(s). Published here under license by The Resilience Alliance.
Go to the pdf version of this article The following is the established format for referencing this article:
Bell, A. Reid. 2011. Environmental licensing and land aggregation: an agent-based approach to understanding ranching and land use in rural Rondônia. Ecology and Society 16(1): 31. [online] URL: http://www.ecologyandsociety.org/vol16/iss1/art31/
Research Environmental Licensing and Land Aggregation: An Agent-Based Approach to Understanding Ranching and Land Use in Rural Rondônia
1School of Natural Resources and Environment, University of Michigan
Agricultural development and climate change will be two of the major stressors on the Amazon natural-human system in the decades to come. Environmental licensing for rural properties is being implemented in several states in the Brazilian Amazon with the goal of restoring forests in agricultural landscapes and mediating the impacts of these stressors. This study presents an agent-based model of ranching and land exchange, informs it with empirical results from social research in the Ji-Paraná River Basin, Rondônia, Brazil, and investigates the social, economic, and environmental outcomes that can be expected as a result of environmental licensing in the context of climate change. Model results informed by these data suggest that although an environmental licensing scheme with monitoring and enforcement may increase the level of forested land in ranching landscapes, it may do so at the expense of the small producer. To the extent that effective monitoring and enforcement exist, a focus on larger holdings will help to mediate this negative social impact. These results suggest that a middle ground can be found in cases where current environmental goals conflict with legacies of past colonization and resource-use regimes.
Key words: agent-based modeling; Amazon; coupled natural-human system; environmental licensing; farm structural change; land-use change
Two major stressors on the Amazon natural-human system are an advancing frontier of agricultural development and global climate change. The agricultural frontier, driven into the Amazon by the aggressive colonization policy in the 1970s, waves of migration of poor landless peasants, and growing domestic markets for beef and international markets for soy (Simon and Garagorry 2005), threatens the Amazon system by clearing trees, destroying habitat, polluting water, and displacing indigenous peoples. In regions along the frontier, the presence of roads and land speculation are commonly cited as the major proximate drivers of land-use change (Faminow 1997, Caviglia-Harris 2004, Soares et al. 2004). Where access is created, small and large farms alike claim new land far from current markets in the expectation that further frontier expansion will drive up the land’s value, though this ‘dragging effect’ (Fearnside 2007) has been demonstrated to be most strong when moderate levels of local infrastructure already exist (Pfaff 1999, Pfaff et al. 2007). Behind the advancing frontier, where most land parcels have been claimed or allocated, the conversion of forest into agricultural use or disuse is in the hands of the property owner. Environmental licensing for rural properties is emerging in several Brazilian Amazon states as a means of regulating land use on active agricultural properties (Lima et al. 2005, ambientebrasil 2010). However, the ability or willingness of a rural producer to maintain forested lands on his or her property may depend strongly on cost structure and the ability to turn a profit from the remaining productive land, which in turn is a function of farm size (Ellis 1993). This is a constraint for any attempts to regulate land use in the region because for smaller properties, stringent environmental regulations may mean either an inability to comply, or an inability to remain in production.
The Amazon Region will be affected in the coming decades by climate change. Simulation results from the most recent International Panel on Climate Change report in 2007 suggest it will be warmer, and probably drier, and there is an expected rise in the frequency of extreme weather events, i.e., longer droughts and stronger storms (Magrin et al. 2007). One impact, on large and small farms alike, will be to make agricultural activity, such as raising cattle on pasture, more expensive because vegetation growth is negatively affected. By simultaneously restricting the area of, and reducing the productivity by, active agricultural land, the joint stressors of environmental licensing and climate change have the potential to pressure production in the rural Amazon.
This study develops an agent-based model of a ranching landscape to investigate the potential social, economic, and environmental outcomes of a new environmental licensing scheme being implemented in the state of Rondônia, under the additional stressor of a change in climate. Agent-based analysis of changes in farm structure is a relatively new field of research (Zimmermann et al. 2009), and the model in this study incorporates features of particular relevance to the Amazonian context, i.e., land sale by struggling farmers and climate variability, that have not appeared in other agent-based approaches to farm change. The coupled model of ranching and climate asks whether the joint pressures of licensing and a changed cost structure due to climate change will act to force producers on small properties off of their land, and whether this social impact can be mediated while still achieving landscape-scale land-use goals.
This study finds that environmental goals can be harmonized with social and economic goals in the ranching landscape, but that this will require particular care in implementation, with monitoring programs that emphasize larger properties. The current work will be of value both to the nascent literature on agent-based approaches to analyzing rural policy, and to the broader discussion within natural resource management of how, in a socially just manner, to match today’s goals for environmental and ecological services with the legacies of colonization and resource-exploitation regimes of the past.
Ranching and environmental licensing
This study focuses on ranching, the dominant agricultural land use in Rondônia, with 5,000,000 ha of pasture compared with only 500,000 ha of cropland in 2006 (IBGE 2006). The rates of land-use change across properties of different sizes in Amazônia tend to be different, with smaller plots needing to deforest proportionally more of their lots than larger plots to meet needs (Aldrich et al. 2006, D'Antona et al. 2006). There is a broad distribution of property size in Ronônia, with nearly 30 properties greater than 2000 ha in size declared in the 1996 Census, along with more than 15,000 properties smaller than 100 ha and hundreds in between (IBGE 1996). There is also a slow process of land aggregation in the Amazon, with many smaller ranchers selling land to meet financial obligations (D'Antona et al. 2006). If climate change affects the profitability of ranching activity, it is reasonable to expect that there will be some impact on the extent of land sale among ranchers. To the extent that ranchers operating at different scales of production deforest at different rates and maintain their land in different ways (Ewers and Laurance 2006), it is reasonable to expect that changes in land distribution will affect environmental outcomes beyond the direct impacts brought about by an increase in storms and droughts.
To confront the environmental problems brought about by land-use change, the State Secretariat for Environmental Development (SEDAM-RO) is following other states in the Amazon region in implementing a program of environmental licensing for rural properties, the Licenciamento Ambiental em Propriedade Rural (LAPRO; SEDAM-RO 2008). At present, to receive any form of rural credit from Brazilian banks, property owners must obtain an environmental license, or for some smaller property owners, simply declare that their properties are in accordance with law. Eventually, SEDAM-RO plans to close off access to markets for those properties not licensed. To obtain a license, rural property owners must generate a management plan for recuperation of forests over a 30-year period in areas of permanent preservation (APP), including riparian buffer strips along all watercourses and forests on all steep hill slopes, and legal reserve (LR).
The requirements for LR are a point of tension for SEDAM-RO; under LAPRO, properties with less than 50% of land in LR prior to 1998 must recuperate up to 50% within the 30-year period. In contrast, owners wishing to clear new land on property that was forested as of 2005 must maintain 80% of the land as LR, a move that clearly favors those who have already committed infractions. Further, to many, the requirements of LAPRO feel like a complete reversal by the state; although on paper the Federal Forest Code has long required rural property owners to maintain 50% of their land as LR, in practice, colonization policies that brought many farmers to the region in the 1970s and 1980s rewarded those who added value to their land by clearing it (Hecht and Cockburn 1989, Fearnside 2001). For large cattle ranchers not currently possessing 50% LR, licensing will mean a big drop in income; for many smaller family properties, licensing that requires proportionally the same from them as from large properties may mean their properties will become unviable as the sole sources of income to maintain the household. Although a number of activities implementing agroforestry systems (SAF) are permitted within APP and LR, including rubber, açai palm, and coffee, there is no guarantee that many of these small farmers have the resources or skills to switch to these activities, or that markets will support them.
A number of proposals have arisen recently to try to minimize the way in which LAPRO will affect the small farmer. One proposal being put forward by several organizations in the state proposes a modification to the Forest Code such that:
This paper tests the following hypotheses, regarding the impacts of climate change and environmental licensing on the rancher-water coupled natural-human system: (H1) decreases in precipitation will drive increased rates of land aggregation; (H2) environmental licensing will lead to better environmental outcomes on properties currently lacking significant forest cover; (H3) reduced access to markets through environmental licensing will drive increased rates of land aggregation; and (H4) reduced licensing requirements for small properties will lead to lower rates of land aggregation.
Agent-based modeling in agriculture: filling a current gap
The last decade has seen the development of several agent-based approaches to looking at farm production and change (Berger 2001, Parker et al. 2003), but they are still relatively uncommon (Zimmermann et al. 2009). The model in the literature that most closely relates to the current work, looking at the ties between policy and farm structural change, is the AgriPoliS model of Happe et al. (2008), a sophisticated agent-based approach to rural economics problems that allows farmer agents to make technological and structural change to their farms by purchasing equipment and renting additional plots, and to make land-use choices in response to shifts in policy, prices, and costs (Happe et al. 2006). Developed to look at European, with particular attention to German, agriculture, AgriPoliS has been applied to several policy-relevant issues in common with the current study, i.e., the effect of a switch in policy regime on farm structure (Happe et al. 2008), and the factors that may cause farmers to leave the agricultural landscape (Happe et al. 2009).
However, AgriPoliS lacks the capacity to model several of the features that characterize the Amazonian frontier and postfrontier agricultural landscapes. First, although land rental does occur, land purchase and aggregation under successful farmers is much more common than in the European context for which AgriPoliS was developed. Second, climate variability, one of the focal stressors in the current study, is an important decision making factor for local farmers, and strongly shapes the productive capacity of pastures for cattle. Most ranchers in the sample reported using pasture conditions rather than market prices as the primary decision factor in stocking pasture. A model of the Rondônian postfrontier ranch landscape must incorporate the local practice of selling off land parcels to cover financial needs, as well as the link between climate, pasture productivity, and rancher decision making. The model developed for this study fills this particular gap.
This study employs an agent-based model of a ranching landscape, informed by and validated through survey data collected from February to April 2009. The survey was applied to a sample of 241 small to medium cattle producers (up to 320 hectares in size) from three municipalities, i.e., Ji-Paraná, Cacoal, and Machadinho do Oeste, in the Ji-Paraná River Basin in Rondônia, Brazil (Fig. 1). Farmers were interviewed as they visited local offices of the state agency for rural extension services, EMATER-RO, and the sample was poststratified by size. Rondônia boasts the most intensive agricultural production of the Amazonian states with 37% of its land committed to pasture and cropland (IBGE 2006). Within Rondônia, the Ji-Paraná Basin is the most developed, for example, most of the length of the BR-364 highway in Rondônia passes through the basin, and is an ideal site to investigate cattle ranching.
The survey research yielded important baseline data with which to inform and calibrate the agent-based model. Specifically, data on the use and ownership of tractors, as an indicator of mechanization, on the rate of recuperation of pasture and the annual maintenance costs incurred, on the annual costs to supplement cattle diets during drought, and on the kinds of information used to decide how many cattle to stock in pastures were obtained. These data, and the role they played in informing model development, i.e., directly translated to a model parameter or interpreted in concert with other literature or anecdotal data, are given in the full model description (Appendix 1) and the section on calibration and validation (Appendix 2). Where available, other data to parameterize the model were obtained or interpreted from literature sources; where source data were unavailable, reasonable parameter values are assumed. These parameters, and explanations of their source and interpretation, are summarized in Appendix 3.
Full source code for the following model, implemented in MATLAB, is available as Appendix 4 to this study. The following is a summary of model logic; the complete description of model mechanisms and state equations, along with data on land use and mechanization from the sample by which the model is informed, can be found in Appendix 1.
Rancher agents raise cattle on an n x m grid of land representing a rural Amazon watershed. Each agent begins with an allocation of grid cells, with land in each cell allocated entirely to pasture, the source of grass for cattle growth. Cattle consume grass to meet their dietary needs when grass growth is sufficient to support them; when grass growth is insufficient, such as during a drought, ranchers must purchase supplements to meet cattle needs. This simple mechanism captures the idea that rather than selling cattle off as a first response, ranchers may pay nominal costs to maintain cattle health during drought periods, a practice observed in our sample through the purchase of supplements, commitment of land to the growing of sugar cane leaves for cattle, or by the renting out of under-utilized neighboring pastures.
At each time period, rancher agents choose to modify a portion of their land, i.e., clearing forest for pasture and restoring degraded pasture to pasture or pasture to forest, to stock their land with cattle, and to purchase or sell land from their neighbors. Land-use change decisions are made based on the present value of land under the particular use with a discount rate d, and conversion is limited by both the financial and time resources of the agent. Ranchers who fall into financial deficit sell cattle and land to attempt to remain solvent. Parcels of land put up for sale are auctioned to the highest bidder among neighbors of the property from which the parcel is being taken. The cattle stocking rate is a function of the grass growth rate.
After an initial start up period of 10 years, an environmental licensing program is implemented for the remaining 30 years of the simulation (Fig. 2). Under the license, ranchers must achieve a set level of reforestation each year to maintain their licenses and enjoy the premium market price that is given only to license holders. A random selection of agents is monitored at each time step, and those ranchers that are far off from meeting their licensing obligations may lose their licenses. The selection of agents is made by a uniform random selection of grid cells, so that larger properties are more likely to be fined. Agents are informed of the monitoring of other ranchers by communicating with other ranchers in the landscape, which in turn informs their expected incomes when calculating the present value of each land use. The strength of communication among agents is thus a determinant of how well ranchers can predict the expected costs of clearing forest. All ranchers share a network link with all other ranchers, they are a ‘clique’ in the network sense; the strength of each link, i.e., the likelihood that a rancher will communicate with another particular rancher in a time period, is normally distributed.
Daily precipitation is drawn from exponential distributions of mean λi, with a different λi for each month i of the year. Climate change is treated as an equal, fractional decrease in all λi and thus, in overall annual precipitation. The direct impact of climate change is to increase supplement costs for cattle diets during drought periods.
In the experiments discussed below, ranchers are granted an initial allocation of land based on the distribution of properties observed in the field sample from the Ji-Paraná Basin. During an initial start up period of 10 years, no charges are levied or land sales permitted while ranchers stock their land and clear away forest to make room for more pasture. At the 10-year mark, land sales are permitted, monitoring and enforcement for licenses begins, and the model is run for an additional 30-year period. Ranchers must continually reforest their property, at a rate that allows them to meet the established goals by the end of the 30-year period, to keep their licenses. The outcome for each experiment at the end of this period is measured by the forested fraction of the landscape, the average profit per hectare of property per year, and the distribution of land among all ranchers originally present on the land.
This study presents results from a set of experiments across 12 scenarios: four sets of assumptions about the structure and value of rancher networks and communication for each of three policy scenarios.
For each of the 12 scenarios, a set of n = 10 replicate model runs with different seeds was performed across the values of the independent variable pairs ΔEI and ΔPrec (in scenarios 1 and 2) and ΔEI and pmon (in scenario 3) to generate a response surface (Table 1). The dimension DEI represents the fractional change in expected income from the sale of cattle when not in possession of a license, and is a signal of how strictly market access for those without licenses is controlled. The price ranchers without licenses obtain for cattle is simply (1-DEI) times the market price. The dimension DPrec represents the change in overall precipitation relative to the base case; in month i, precipitation is drawn from an exponential distribution with mean (1-DPrec)li. The dimension pmon represents the likelihood of a particular grid cell being targeted for a site visit, and is a signal of effort invested into monitoring.
The total number of runs for each response surface is 11 x 11 x 10 = 1210 runs.
The first two of the three scenarios investigate the interaction of the ΔEI and ΔPrec stressors across two different implementations of environmental licensing, a basic approach, and an approach with tiered requirements for farms of different sizes. The goals of the licensing are treated in terms of a target fraction of forest cover at the end of the 30-year licensing period, ftarg,final, so that the two approaches are:
ftarg,final = 0.3 for properties less than or equal to 120 hectares but greater than 60 hectares, and
ftarg,final = 0.5 for properties larger than 120 ha
The third scenario returns to the basic, nontiered licensing of scenario 1 and investigates trading off effort in monitoring and enforcement while holding climate constant. In both of scenarios 1 and 2, pmon (the probability of a grid cell being selected and the corresponding property monitored) was set to 0.075, meaning that a property composed of 10 cells, for example, had a 7.5% chance of being monitored in a given year. In scenario 3, ΔEI and pmon are varied to investigate the way in which these two parts of the monitoring and enforcement process (site monitoring and the control of market access) may substitute for one another within the basic, nontiered case.
In each policy scenario outcomes are evaluated across four scenarios of rancher networks, based on two dimensions of communication strength and size effect (Table 2). These dimensions arise from two major assumptions to be made about how ranchers interact in the system: first, that they interact to exchange information, e.g., about costs, practices, or having their site monitored, etc., and second, that they interact to share labor and equipment, thus cutting their costs and allowing them to act like larger, more mechanized farms. Over 75% of the sample reported membership in local rural syndicates and producers associations, and many reported that this membership gave them access to equipment and discounts they would not otherwise have. However, the survey did not otherwise shed much light on the extent to which these assumptions of information, labor, and equipment sharing might be true in the region.
Communication strength refers to the mean strength of connections among ranchers in the system, and thus the probability that a given pair of ranchers will share information, such as about land values or levied fines. Size effect refers to the extent to which ranchers who share a strong connection also share resources, e.g., labor, tractors, etc., and is thus a measure of how well smaller farms are able to act, from a cost and land-use perspective, like larger farms. The mathematical details of these scenarios are presented in the model description in Appendix 1.
Lacking precise knowledge of how well networked ranching communities may be, or how costs may vary across scale, an exploration of these alternative scenarios sheds insight into the ways that communication and economies of scale may affect the trajectories of land aggregation and environmental quality throughout the simulations.
The complete set of experiments, i.e., four rancher network scenarios within each of three policy scenarios, and 1210 experimental runs to generate the surfaces in each experiment, results in a total number of 3 x 4 x 1210 = 14520 experimental runs.
Figures 3 through 6 present the three outcomes of (1) fractional forest cover, (2) average profit per year per hectare of land, and (3) property GINI for the rancher scenario of low communication and low size effect; the complete set of results across all rancher scenarios is included as Appendix 5.
The state of land aggregation in the basin is expressed as a ‘GINI’ coefficient in the model results. The formula for the property size distribution GINI is:
where Ai is the size of a ranch and n is the total number of ranches. This coefficient ranges from close to 0, implying a more even distribution of land among ranchers, to 1, signifying most or all land concentrated under a single or small number of ranchers.
Joint pressures of environmental licensing and climate
In the nontiered environmental licensing scenario, there is some support for hypothesis H2, that licensing can in fact bring about better forest outcomes on denuded properties (forested fraction initially rises as DEI increases from 0 in all scenarios and for all tested levels of DPrec; Fig. 3). However, this comes at the expense of revenue, i.e., average per hectare profits strictly decrease as the market price available to nonlicensed properties drops. Above some threshold value of DEI, profits on average drop below 0, and ranchers do not have the resources to commit to forest restoration or even their own ranching. Forested fraction peaks and then decreases as the stricter environmental licensing makes ranching unviable.
Moving along the dimension of decreasing precipitation, the peak forested fraction that is achieved drops, suggesting a lack of resources to commit to forest restoration, which in turn is reflected by the smooth decrease in average profits per hectare as DPrec drops further.
Overall, higher forested fractions are achieved when network connectivity is high, i.e., when ranchers are better able to assess the risk of having a license stripped and their perceived opportunity cost of losing the license is much higher (Fig. D.1). Put simply, for a policy to be effective, those it is meant to govern must be well informed. When size effect is high, meaning that strong network connections allow smaller ranches to behave much like larger ranches, the peak and drop in forested fraction is much less pronounced, i.e., small and large ranches alike are more able to turn a profit because their costs are lower, reflected in higher average profits per hectare.
In all network scenarios except for the high communication/high size effect scenario, both DEI and DPrec act as drivers of land aggregation, measured as an increase in the property GINI, which provides some support for hypotheses H1 and H3 (see also Appendix 5). However, as was also the case for forested fraction, these relationships have a single peak, beyond which they decrease. The explanation is that although all ranches are impacted by stricter licensing or by drier weather, smaller ranches have less of a financial buffer once their basic needs and costs are met, and will be the first to need to sell cattle or land to make up for an expensive year. Larger ranches will be in a position to buy up this land and, initially at least, increases in DEI and DPrec lead to higher property GINI values. However, as these stressors increase further, the profit margin for even larger properties disappears, leaving them unwilling or unable to purchase neighboring plots, and the property GINI peaks or drops off. If an incremental increase in either stressor makes smaller ranches more willing to sell faster than it makes larger ranches less willing to buy, it leads to a net increase in property GINI. In general, across the four rancher scenarios, the peak in property GINI is diminished when the two stressors are acting jointly, along the back edges of each of the surfaces. When size effect is high and all ranches have similar cost structure, any effect on land aggregation from DEI is minimal, which is to be expected because the smaller ranches share similar cost structure and mechanization to larger ranches in this scenario.
Tiered environmental licensing
The tiered environmental licensing option leads to several distinct outcomes relative to the untiered case (Fig. 4). The shapes of the curves remain similar, however, and these distinctions may be better viewed by looking at the difference in environmental, economic, and social outcomes between the two experiments (Fig. 5).
First, the forested fraction achieved is lower across all conditions relative to the untiered case, as would be expected. The effort to improve equity across ranch scale requires that smaller ranchers be held to a looser environmental standard, and the overall area of restored forest is reduced. The difference in average profits in the tiered case rises with DEI, reflecting the relative ease that the lower environmental standard gives to the ranching landscape; this difference is less significant under conditions of lower precipitation, suggesting that the additional climate stress helps to equalize any differences between the two approaches.
The major result is that under a small range of conditions, the tiered approach to licensing does result in lower rates of land aggregation (as DEI initially increases from 0). This effect is most pronounced in the scenario where communication is high and size effect is low. In this scenario, ranchers are very well informed of the risk of being monitored and the costs those in their network are incurring, but gain little else through their network connections (Figure A5.3 in Appendix 5). Conversely, the effect is least pronounced in the case where both communication and size effect are high, i.e., where smaller ranches are able to act much like larger ranches and thus are less disproportionately affected by environmental licensing (Figure A5.3 in Appendix 5).
There is at best mixed support for hypothesis H4 however, since as conditions worsen (DEI and DPrec increase further) the tiered case appears to lead to higher levels of land aggregation than the nontiered case (the initial dips in Figure 5 in property GINI as DEI increased from 0 now rise). The implication is that, rather than eliminating the problem of land aggregation, the tiered approach simply shifts the domain in which DEI and DPrec act as drivers of land aggregation farther out. That is, under moderate climate or policy stress, the tiered approach can ameliorate some of the pressure on small properties, but if the stressors intensify, the same issue may return. This is even clearer when looking at the relative standard deviations across replications (Figures A5.5, A5.6 in Appendix 5). As ranches begin to fail, the variance in profitability across the landscape increases, i.e., some farms are doing well while others are failing, and then falls off as conditions worsen, i.e., all farms are failing. The sharp ridge on the surface for profitability in Figures A5.5 and A5.6 in Appendix 5 marks the threshold, as a function of both climate and licensing stressors, where ranches begin to fail. Comparing the two figures, these ridges move farther out from the origin (DEI = DPrec = 0) in the tiered case, suggesting a boundary shift or distortion rather than a problem solution.
Another approach to achieving a socially equitable outcome
The tiered licensing proposal is unpopular among those who do not stand to benefit. Hence, it is worthwhile to look for other means of achieving more equitable results under licensing. Rather than creating explicit tiers that may or may not map well onto functional groups of ranchers it should be possible to design a monitoring and enforcement scheme that implicitly lessens the burden that licensing places on smaller ranchers.
The monitoring and enforcement process in this model has two parts: (1) monitoring of land use on individual properties and allocation or stripping of licenses, and (2) verification of licenses at the point of sale of cattle, such as at a slaughterhouse. In the model, properties are selected for monitoring based on size; a fraction of the cells in the grid is selected randomly, and the properties to which they belong are selected. In this way, larger properties are at a higher risk of being caught. This is a reasonable representation, because any real agency with limited, and perhaps minimal, resources would likely choose to target a smaller number of relatively large targets over a large number of smaller targets. However, all ranchers in the simulation forfeit the same proportion of their revenue DEI when they lose their licenses. If DEI is interpreted as a measure of the difficulty of unloading cattle, this too is reasonable, since all truckloads of cattle present themselves to slaughterhouses in much the same way, regardless of how large the property from which they come.
The agency tasked with monitoring and enforcement must choose how to divide effort between the two parts of the process described above to maximize some objective function. The monitoring of individual properties would likely involve the use of real-time satellite imagery of the property in question, as well as a site visit and consultation with the property owner regarding his or her management plan. The verification of licenses at the point of sale would require the stationing of an agent at a slaughterhouse or the provision of incentives to the slaughterhouse to require licenses as a part of the sale. Considering these processes for monitoring and enforcement together, the implication is that if the objective is to make equitable the burden placed by licensing on ranches, in the modeled system, more effort should be allocated to site monitoring, and less to point-of-sale verification of licenses.
This implication plays out in the experimental results. Figure 6 shows a set of outcome surfaces generated by varying both DEI and pmon. Forested fraction increases along both DEI and pmon dimensions until it peaks, so that curves of equal forested fraction, ‘isoforest’ curves, can be drawn that show how monitoring effort (pmon) can substitute for control over market access (DEI) to give equivalent forest outcomes. Per-hectare profit strictly decreases along both DEI and pmon, so that similar isoprofit curves can be drawn. In general, the isoforest and isoprofit curves map closely onto each other, which is to be expected; similar areas in forest should indicate similar areas in pasture, and thus similar levels of revenue generated on average across the landscape. However, the same relationship does not hold for the property GINI.
With the notable exception of the high communication, high size effect case, the property GINI is generally higher when DEI is high and pmon is low, and downward sloping as DEI decreases and pmon increases (Fig. D.4). Thus, moving along the isoprofit and isoforest curves associated with the peak forested fraction from higher DEI toward higher pmon, the property GINI decreases, implying lower rates of land aggregation and a social outcome that is more favorable for smaller ranches (Fig. 7). This effect is most pronounced in the low communication, low size effect case; where ranchers on small properties are the least informed of the risks they face and where their costs are considerably higher than those for larger properties, they have the most to gain by shifts in policy that place more burden on larger properties.
Because site monitoring is likely to be the more resource intensive component of the monitoring and enforcement process, it is important that this social equity benefit be emphasized. Figure 7 shows that site monitoring and point-of-sale enforcement can substitute for one another to preserve forest cover and average profit levels, but that site monitoring will not simultaneously preserve the capacity for smaller ranches to produce.
A key process in making insights derived from modeling results useful to real-world situations is to step back through the set of simplifying assumptions upon which the model is built and understand how these insights change as the assumptions are relaxed. In this model, a number of simplifying assumptions were made regarding land market structure, the monitoring and enforcement of fines, and land-use decisions, and it is important now to discuss how the more complex, real-world versions of these processes might modify the study results.
First, only parcels offered by ranches in deficit entered the land auction in the model. This is certainly a major component of land that gets sold in real ranching landscapes, but is incomplete; successful ranchers may also be aggressively attempting to buy up local properties in an effort to grow. This mechanism is excluded here to avoid introducing further assumptions about how ranchers choose to invest their money; the impact of excluding this mechanism is likely to be a more conservative estimate of land aggregation, since only some of the means through which successful ranchers can buy up neighboring land are included.
Another important simplification in the model is that there are no wholly unexpected costs borne by the ranchers. In reality, the failure of equipment as well as illnesses and injuries among family members are unpredictable shocks and can drive the need to sell off cattle or land in a pinch. It is reasonable to assume that the risk of injury or illness is uniform across the population, if not higher among poorer ranchers, and that richer ranchers will be better prepared to weather these shocks. Again, this simplifying assumption likely leads to a more conservative estimate of the rate of land aggregation.
The mechanism through which changes in climate influence production in this model is simple: a decline in precipitation results in decreased grass growth, which in turn increases the cost to the rancher to supplement cattle diets during dry periods. Although such a relationship has a basis in the literature (Svoray et al. 2008, Hirota et al. 2010), it is certainly not the only way climate might affect the growth of grass or other crops. In reality, shifts in the mean levels as well as the temporal distribution of precipitation and temperature may have positive or negative effects on grass growth depending on whether they pull conditions toward or away from what is optimal for the plant. Thus, it is worth interpreting the climate effects more loosely, as in, to the extent that changes in temperature or precipitation inhibit grass growth, they may act as drivers of land aggregation in the Rondonian ranching landscape. Integrating more sophisticated relationships between vegetation growth and climate into models focused on social processes, like this one, is an important direction for future work.
It is worth noting as well that, particularly on smaller farms in Rondônia, land use is diversified away from exclusive cattle production and into crop production, such as coffee or beans, for consumption and markets. Diversification is commonly cited in the development literature as a coping strategy to climate variability. However, in this particular system where cattle production and markets are so well developed, it is more difficult to argue that production of coffee, for example, alongside cattle helps reduce climate-related vulnerability. It is more likely that when climate impacts are significant enough to affect cattle health, other crops would be more significantly affected, so that impacts on small farmers with production more diversified than the cattle producers in this model would be more severe.
The other major assumptions that steer model results relate to monitoring and enforcement. It is assumed in this model that all ranchers will have the same difficulty marketing their product without a license, and thus the same DEI is applied to all ranchers. In reality, it would not be unreasonable that larger, more powerful ranches would be better positioned to circumvent rules and obtain good prices than might smaller ranches; this is another effect that might tip outcomes in favor of larger ranches.
In sum, this first set of assumptions in this model provide what is likely a conservative assessment of the role that environmental licensing in Rondônia could play, in concert with expected climate stresses, in driving rates of land aggregation, given some nontrivial capacity for monitoring and enforcement of the licensing scheme. Relaxing these assumptions, one could expect more severe impacts on smaller property holders in the real system than demonstrated in the model, and further aggregation of land holdings. In terms of the study hypotheses, this means a stronger case for the findings for H1, H3, and H4 (Table 3).
The last major assumption of the model is that effective monitoring and enforcement occurs at all. This is a key assumption because strong evidence exists to suggest that little enforcement of policy does take place. IBAMA, the federal environmental protection agency, recently estimated that they collected only a small fraction (less than 5%) of the fines that they levied (Hall 2008). In the sample, only a small fraction (less than 20%) of properties reported even having their properties visited by members of a public agency for the purposes of observing environmental quality. The process of visiting sites also requires an accurate and current land registry; although registration for LAPRO will help facilitate this, a complete registry should not be taken as given. It is clear that under such conditions the real impact that environmental licensing may have is trivial to evaluate, i.e., little will happen; reports from licensing schemes elsewhere in Amazônia do not yet suggest much success elsewhere in the region (Lima et al. 2005). This reality means that the study findings with respect to H2 must be interpreted with caution (Table 3). The model developed in this study is not the appropriate tool to investigate why such monitoring and enforcement does not occur, nor how it might be encouraged. The value of this study is in highlighting the benefits that can arise from effective implementation of environmental licensing, and in examining how the social impacts of licensing can be managed, under the assumption of some real capacity for effective implementation.
This study contributes to what is still a small body of agent-based models tied to empirical data (Berger and Schreinemachers 2006), and an even smaller body that examines structural change in farms (Zimmermann et al. 2009), by tying rancher decision making in with climate and representing the set of conditions particular to the Amazonian frontier and postfrontier. The Rondônia case is just one of many where land-use practices established through colonization or resource exploitation are in conflict with present-day goals for environmental preservation, but are depended upon to preserve livelihoods. Beyond POLONOROESTE, through which much of Rondônia’s settlement was funded, the World Bank funded projects in the 1980s in Indonesia, Asia, and the Congo that included as goals the transmigration of peoples and the liquidation of forest assets as a means to economic development (Fearnside 1997, Ekoko 2000, Rachman et al. 2009). As long as legacies of these projects remain, they will continue to present conflict among environmental, economic, and social goals. The results presented in this study should offer some hope that these dissonant goals may be harmonized, and that tools like agent-based models allow explicit study of the tensions among them.
Responses to this article are invited. If accepted for publication, your response will be hyperlinked to the article. To submit a response, follow this link. To read responses already accepted, follow this link.
ACKNOWLEDGMENTSThis study was funded by grants from the Graham Environmental Sustainability Institute and the Rackham Graduate School, both at the University of Michigan. The questionnaire developed for this study drew from material previously developed by the ACT Institute at Indiana University. Initial support in project planning and development was provided by Tatiane Checchia and Paulo Schröder at the Amazônian Protection System (SIPAM). Logistic support in the administration of questionnaires was provided by the agency for technical assistance and rural extension for Rondônia, EMATER-RO. Finally, the author is grateful for assistance in project coordination from Dr. Margarida Marchetto of the Federal University of Rondônia (UNIR) and for the hard work of the research team – Thiago Emanuel Figueiredo Possmoser Nascimento, Ingrid Alexandrina Martins Veronez, Emerson Andrade de Souza, Mônica Chagas Cerquiera, João Paulo Papaleo Costa Moreira, Carina Stre Holanda, and Leonardo Rosa Andrade – from the Department of Environmental Engineering at UNIR Ji-Paraná.
Aldrich, S., R. Walker, E. Arima, M. Caldas, J. Browder, and S. Perz. 2006. Land-cover and land-use change in the Brazilian Amazon: smallholders, ranchers, and frontier stratification. Economic Geography 82:265.
ambientebrasil. 2010. Licenciamento Ambiental em Propriedade Rural na Amazônia. [online] URL: http://www.ambientebrasil.com.br/composer.php3?base=./gestao/index.html&conteudo=./gestao/programas/licenciamento_amazonia.html.
Berger, T. 2001. Agent-based spatial models applied to agriculture: a simulation tool for technology diffusion, resource use changes and policy analysis. Agricultural Economics 25:245-260.
Berger, T., and P. Schreinemachers. 2006. Creating agents and landscapes for multiagent systems from random samples. Ecology and Society 11(2): 19. [online] URL: http://www.ecologyandsociety.org/vol11/iss2/art19/.
Caviglia-Harris, J. L. 2004. Household production and forest clearing: the role of farming in the development of the Amazon. Environment and Development Economics 9:181-202.
D'Antona, A. O., L. K. VanWey, and C. M. Hayashi. 2006. Property size and land cover change in the Brazilian amazon. Population and Environment 27:373-396.
de Jesus, A. 2009. Altera e acresce dispositivos à Lei n°4.771, de 15 de setembro de 1965, Código Florestal Brasileiro. PL 5020/2009 Brasília, Brazil. [online] URL: http://www.camara.gov.br/sileg/Prop_Detalhe.asp?id=429594.
Ekoko, F. 2000. Balancing politics, economics and conservation: the case of the Cameroon Forestry Law reform. Development and Change 31:131-154.
Ellis, F. 1993. Peasant economics: farm households and agrarian development. Cambridge University Press, Cambridge, UK.
Ewers, R. M., and W. F. Laurance. 2006. Scale-dependent patterns of deforestation in the Brazilian Amazon. Environmental Conservation 33:203-211.
Faminow, M. D. 1997. Spatial economics of local demand for cattle products in Amazon development. Agriculture Ecosystems & Environment 62:1-11.
Fearnside, P. M. 1997. Transmigration in Indonesia: lessons from its environmental and social impacts. Environmental Management 21:553-570.
Fearnside, P. M. 2001. Land-tenure issues as factors in environmental destruction in Brazilian Amazonia: the case of Southern Para. World Development 29:1361-1372.
Fearnside, P. M. 2007. Brazil's Cuiabá-Santarém (BR-163) Highway: the environmental cost of paving a soybean corridor through the Amazon. Environmental Management 39:601-614.
Hall, A. 2008. Better RED than dead: paying the people for environmental services in Amazonia. Philosophical Transactions of the Royal Society B-Biological Sciences 363:1925-1932.
Happe, K., A. Balmann, K. Kellermann, and C. Sahrbacher. 2008. Does structure matter? The impact of switching the agricultural policy regime on farm structures. Journal of Economic Behavior and Organization 67(2):431-444.
Happe, K., K. Kellermann, and A. Balmann. 2006. Agent-based analysis of agricultural policies: an illustration of the agricultural policy simulator AgriPoliS, its adaptation, and behavior. Ecology and Society 11(1): 49. [online] URL: http://www.ecologyandsociety.org/vol11/iss1/art49/.
Happe, K., H. Schnicke, C. Sahrbacher, and K. Kellermann. 2009. Will they stay or will they go? Simulating the dynamics of single-holder farms in a dualistic farm structure in Slovakia. Canadian Journal of Agricultural Economics-Revue canadienne d' agroeconomie 57:497-511.
Hecht, S., and A. Cockburn. 1989. The fate of the forest: developers, destroyers, and defenders of the Amazon. Verso, London, UK.
Hirota, M., C. Nobre, M. D. Oyama, and M. M. C. Bustamante. 2010. The climatic sensitivity of the forest, savanna and forest-savanna transition in tropical South America. New Phytologist 187:707-719.
Instituto Brasileiro de Geografia e Estatística (IBGE). 1996. Censo Agropecuário de 1995-1996. Instituto Brasileiro de Geografia e Estatística, Brasília, Brazil. [online] URL: http://www.ibge.gov.br/home/estatistica/economia/agropecuaria/censoagro/1995_1996/default.shtm.
Instituto Brasileiro de Geografia e Estatística (IBGE). 2006. Censo Agropecuário de 2005-2006. Instituto Brasileiro de Geografia e Estatística, Brasília, Brazil. [online] URL: http://www.ibge.gov.br/home/estatistica/economia/agropecuaria/censoagro/2006/default.shtm.
Lima, A., A. Rolla, M. Wathely, C. C. Augusto, and R. A. Alves. 2005. Mato Grosso, Amazônia (i)Legal: Desmatamentos de florestas em propriedades rurais integradas ao Sistema de Licenciamento Ambiental Rural entre 2001 e 2004. Instituto Socioambiental, Brasília, Brazil.
Magrin, G., C. Gay García, D. Cruz Choque, J. C. Giménez, A. R. Moreno, G. J. Nagy, C. Nobre, and A. Villamizar. 2007. Latin America. Chapter 13 in M. L. Parry, O. F. Canziani, J. P. Palutikof, P. J. van der Linden and C. E. Hanson, editors. Climate change 2007: impacts, adaptation and vulnerability. Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK.
Parker, D. C., S. M. Manson, M. A. Janssen, M. J. Hoffmann, and P. Deadman. 2003. Multi-agent systems for the simulation of land-use and land-cover change: a review. Annals of the Association of American Geographers 93:314-337.
Pfaff, A. S. P. 1999. What drives deforestation in the Brazilian Amazon? Evidence from satellite and socioeconomic data. Journal of Environmental Economics and Management 37:26-43.
Pfaff, A., J. Robalino, R. Walker, S. Aldrich, M. Caldas, E. Reis, S. Perz, C. Bohrer, E. Arima, W. Laurance, and K. Kirby. 2007. Roads and deforestation in the Brazilian Amazon. Journal of Regional Science 47:109-123.
Rachman, N. F., L. A. Savitri, and M. Shohibuddin. 2009. Questioning pathways out of poverty: Indonesia as an illustrative case for the World Bank's transforming countries. Journal of Peasant Studies 36:621-627.
Secretaria de Estado do Desenvolvimento Ambiental (SEDAM-RO). 2008. Manual Operacional Para a Licença Ambiental em Propriedade Rural no Estado de Rondônia. Secretaria de Estado do Desenvolvimento Ambiental, Porto Velho, Rondônia, Brazil.
Simon, M. F., and F. L. Garagorry. 2005. The expansion of agriculture in the Brazilian Amazon. Environmental Conservation 32:203-212.
Soares, B., A. Alencar, D. Nepstad, G. Cerqueira, M. D. V. Diaz, S. Rivero, L. Solorzano, and E. Voll. 2004. Simulating the response of land-cover changes to road paving and governance along a major Amazon highway: the Santarem-Cuiaba corridor. Global Change Biology 10:745-764.
Svoray, T., R. Shafran-Nathan, Z. Henkin, and A. Perevolotsky. 2008. Spatially and temporally explicit modeling of conditions for primary production of annuals in dry environments. Ecological Modelling 218:339-353.
Zimmermann, A., T. Heckelei, and I. P. Dominguez. 2009. Modelling farm structural change for integrated ex-ante assessment: review of methods and determinants. Environmental Science & Policy 12:601-618.
|Home | Archives | About | Login | Submissions | Subscribe | Contact | Search|