Co-Principal Investigators

Richard E. Bilsborrow¹ and Stephen J. Walsh^2
1Research Fellow, Carolina Population Center; (919) 966-2157 (tel), (919) 966-6638 (fax), bilsb.cpc@mhs.unc.edu (e-mail)
²Director, Spatial Analysis Lab, Department of Geography; Director, Spatial Analysis Unit, Carolina Population Center; (919) 962-3867 (tel), (919) 962-1537 (fax), swalsh@email.unc.edu (e-mail)

Investigators

Lawrence Band³, Aaron Moody³, and Laura Murphy^4
3Department of Geography; ⁴Department of City and Regional Planning and Carolina Population Center
University of North Carolina, Chapel Hill, NC 27599

I. Introduction and Rationale

The rate and degree of landuse/landcover change (LULCC) associated with deforestation and agricultural extensification in the Ecuadorian Amazon will be indirectly examined over nearly three decades from the early 1970s to the late 1990s through the development of a satellite time-series and GIS coverages. The driving forces that ultimately underlie the process of deforestation in the Ecuadorian Amazon and yield the patterns discerned from the satellite time-series will be assessed relative to social, biophysical, and geographical influences. From 1990 to 1998, however, the driving processes will be examined in detail through longitudinal population surveys of 450 households in 1990 and re-surveyed in 1998, taking as its starting point the presence of small farmers in the frontier. The surveys will provide an opportunity to better understand the landuse/landcover (LULC) practices of migrant farmers -- the effects on agricultural extensification, habitat fragmentation, secondary plant succession, carbon sequestration, and settler threats on sensitive ecological habitats through the deforestation of corridors, riparian zones, and established biological reserves within the study area. Decisions made regarding the use of the land are the outcome of a dynamic process occurring at multiple scales, the household or finca, sector, and region.

Small farm households are considered the primary driving force of conversion of primary tropical forest to other land uses, so it is essential to understand the impacts of their land use patterns on forest succession, standing biomass and leaf area, and regional carbon sequestration in order to develop a dynamic model of deforestation and agricultural extensification. Individual-level factors, farmers with previous agricultural experience or who are more educated are hypothesized to deforest less because they are likely to use more sustainable methods of production, resulting in higher production and income per unit of land area cleared. For household-level factors, we hypothesize that larger families will tend to clear more land since their consumption needs are greater. We also expect a larger number of adult males to be positively related to deforestation since there are more brazos ("hands") to clear the forest and work the land. In fact, an important hypothesis is that the conversion of forest cover to agriculture use, as well as specific uses, relates closely to the family life cycle (Marquette, 1995). Variables measuring household consumption units vs. labor supply units will be included in the model. A larger plot size is positively related to the total area cleared, but inversely to the proportion of the plot cleared. The closer the plot is to a road, the greater the proportion of the plot cleared and the greater the proportion of land in cash crops as well because road access facilitates marketing farm output. Finally, duration of residence of the family on the plot is positively related to both the overall extent of land clearing and the proportion of the cleared area in pasture. The former is due to having more time to clear the forest, and more time for crops to reduce soil fertility and yields (given the almost complete lack of use of fertilizer observed in the region). The areas of land in fallow and secondary forest are also related to duration. Land potential on the settler's plot, as determined through the satellite time-series and GIS derivations of site suitability, may also have positive effects on deforestation: with good soils, crop yields are higher and farmers need to deforest less simply to sustain their families, but rational farmers with high consumption aspirations may in fact deforest more.

The broader context in which households make land use decisions is reflected in a number of community-level or contextual factors as well. Among the potentially most relevant in the Amazon region are the following: (1) characterization as an urban area, or proximity to the nearest large town, since this indicates access to off-farm employment which provides an alternative source of income that may alleviate pressures on settlers to clear more of their plots; (2) population size of local towns, since a larger local market may facilitate market sales of farm produce and/or better prices; (3) distance by road from the local produce market to the provincial capital or other major market town in the Amazon region, for similar reasons; (4) availability of credit, or distance to the nearest source, Banco Nacional de Fomento, is positively related to land use, including the area in pasture since cattle are often accepted as collateral; and (5) proximity to an agricultural extension center may reduce land clearing by enhancing agricultural intensification.

The proposed study site in the northeastern Ecuadorian Amazon region offers a particularly suitable laboratory for examining LULC behavior of the population of semi-commercial spontaneous settlers in the western Amazon basin, where settlement has a different character than that of eastern Amazonia. The rural population in Ecuador’s Amazon region consists largely of migrant settlers with high fertility and mortality, where the "natural" process is not confounded by chronic, violent conflict and land disputes (as in much of Brazil) or by extensive coca growing (as in Peru). Nor are rapid urbanization, large-scale logging operations, or even large ranches found in Ecuador as in the other frontier countries. The nature of frontier settlement in our study area thus lends itself to a "cleaner" understanding of patterns and process of land use and possible agricultural intensification in a context of increasing population pressure on a declining resource base.

The focus on the Ecuadorian Amazon region is also particularly significant for environmental and socio-economic reasons. The western-most region of the Amazon River basin, bordering the Andean mountains and occurring at the headwaters of the Basin, possesses several major centers of endemism (Whitmore and Prance, 1987; Myers, 1994). One site, the Napo center (Prance, 1987) overlaps the proposed study area in Napo and Sucumbios. The region is also one of Myers' (1988) global "hot spots" of biodiversity deserving preservation, yet is continuing to undergo rapid settlement and deforestation. Two of the major national conservation areas in Ecuador are immediately contiguous to the colonization areas of the 1990 survey, Cuyabeno and Yasuni, and are being intervened by illegal spontaneous migrant settlers. This process invasion into biological reserves will be measured over time for the first time in the proposed project through the satellite time-series.

II. Research Questions and Objectives

A. Deforestation, Agricultural Extensification, and Land Management Practices

I. What are the rates of different types of land conversions at the finca, sector and regional levels? Typical land conversion types include forest-to-crop, forest-to-pasture, crop-to-pasture, crop-to-fallow, pasture-to-fallow, fallow-to-forest (abandonment). It is likely that conversion rates vary across scales from the finca to the region. It is also likely that conversion rates change over time and that the spatial and temporal considerations are conversion-type specific.

II. What are the social, biophysical, and geographical factors and processes that drive land-use/land-cover conversion characteristics? Social factors include family size and demographics, educational level, past agricultural experience, family life cycle, duration of residence, economic status; biophysical factors include soil conditions, topography, water supply, surrounding land uses, existing LULC characteristics at time of settlement; and geographical factors include proximity to a road, proximity to a town, proximity to a school, size of local market, access to banks (credit and loans). In this case "land-conversion characteristics" refer to the rates and types of conversion.

III. Are there clear and consistent patterns in the way these factors and processes are manifested in the landscape? That is, based on question II above, are there some fundamental patterns in the variables that drive land conversion that produce consistent patterns of conversion across the region? For example, do roads, towns, and existing agricultural settlements drive land conversion in ways that are consistent enough to be used in predictive models of LULC conversion?

IV. Are there scale dependencies in the links between these factors and observed land conversion rates and patterns at the finca, sector, and regional levels? For example, changes at the finca level and changes at the regional level are likely driven by different primary factors.

B. Land Management and LULC Structure

I. Are there consistent patterns across fincas in household-level land management decisions that govern the structure (i. e. composition and spatial organization) of landuse within fincas? That is, do most fincas exhibit the same patterns of land conversion and land use as driven by the social, biophysical and geographical factors investigated in section A above? Composition refers to the proportions of a finca that are in each LU or LC. Spatial organization refers to the way those LULCs are arranged spatially within the finca. Composition and spatial organization together are referred to here as "structure".

II. Does the lag time between sector settlement and finca development affect the structure of land management within the finca? For example, do fincas that were developed early in the evolution of a settlement undergo different land-conversion processes than fincas that are developed well after a settlement has been established?

III. Does finca structure converge on a steady state over time? If so, how long does this process take? That is, do the rates of change, spatial organization, and composition of a finca settle into a steady state after a certain period after initial settlement?

IV. What are the sensitivities of within-finca land management practices to changes in social, biophysical, and geographical factors and processes? That is, it is expected that the social, biophysical and geographical factors will change over time. How do these changes affect the management and structure of fincas that are already well established?

C. LULCC and Site Suitability for Agricultural Extensification

I. Is there apparent variability in agricultural sustainability throughout the study area given typical finca management practices? Agricultural sustainability can be determined through stability in land-use practices between the two dates of the survey for the sampled fincas.

II. Are there biophysical, social, or geographical factors that explain this variability?

D. Household Decision-Making and Secondary Plant Succession

I. Is land abandonment or long-term fallow part of the typical land management regime in the study area? This region is unique from others in the Amazon basin in that it has more fertile soils. Also, because of the slash-and-mulch approach (as opposed to slash-and-burn) used in the region there is a longer term release of nutrients into the soil. As a result, land abandonment may not be part of the agricultural cycle or it may require a longer time frame.

II. If so, what is the nature of secondary plant succession on these abandoned or fallow lands and how do they differ from undisturbed forest?

E. Deforestation and Landscape Ecology

I. Given the spatial organization of conversion patterns at the finca, sector, and regional scales, what are the implications for effective viable habitat area at these scales? Actual habitat area may include relatively undisturbed areas that are largely dissected by disturbances associated with agricultural development. Therefore, much of the "actual habitat" may not actually be usable by many species of organisms. "Effective habitat" includes only those areas that are large and homogeneous enough to provide viable habitat to the original inhabitants and populants of the area. This question will involve the identification of habitat corridors between effective habitat areas, evaluation of forest fragmentation as an indicator of habitat viability, and proximity measures to determine the relative isolation level for remnant forest patches. At the regional scale we will focus primarily on general spatial organization of the forest/non-forest matrix.

II. What are the rate, extent, and spatial patterns of LULC conversion in the Cuyabeno Reserve and Yasuni National Park? What are the nature of threats imposed as a consequence of deforestation and land fragmentation to lands occurring adjacent to and within biological reserves and riparian zones within the study area, and do such activities affect biodiversity and ecosystem sustainability?

F. LULC and Carbon Sequestration

I. How are the observed land conversions likely to affect the carbon budget in the region? A key question here is whether finca development leads to a net increase, net decrease, or neutral effect with respect to biosphere-sequestered carbon in the region. This will be addressed by linking survey and satellite-based estimates of LULC conversions to literature-based parameters regarding carbon storage and carbon assimilation rates for the observed and modeled LC types. This component will also incorporate life-cycle changes in carbon assimilation rates.

G. Remote Sensing and Population Survey Data for Multi-Scale Modeling

I. Can we develop a remote-sensing based model of LULCC for the entire region by linking ground-based data to a time-series of remotely sensed data? The longitudinal survey data will serve as a reliable source of training and validation data for a remote sensing model of LC conversion. This model can then be used to extend data on land conversion throughout the region where ground data are not available.

II. Can we enhance and improve the above remote sensing model through input of social data and known plot development processes? Given an understanding of the social, biophysical, and geographical drivers of land conversion it may be possible to develop a GIS-based model of conversion that integrates information on new roads, towns, locations and dates of new settlement initiation, increased mechanization, and economic factors such as changes in market prices for certain commodities with remotely sensed data to model and predict future land conversion scenarios.

III. Background and Significance

A. Theoretical Approaches: the Human Dimensions of LULCC

Recently, there has been considerable debate about the nature of the links between human populations and the environment. The neo-Malthusian argument that population growth leads to environmental degradation through extensification of landuse has not been supported by scientific studies (i.e., Bilsborrow, 1994; Marquette and Bilsborrow, 1994). Within the natural sciences, research on Amazon LULCC has elucidated ecological processes but neglected human dimensions (i.e., Uhl and Jordan, 1984). A more comprehensive approach is necessary, in which the impacts of a growing and changing population on the immediate environment are mediated by the standard of living, technology, public policies, and socio-cultural factors (Allen and Barnes, 1985; Bilsborrow and Geores, 1992; Jolly, 1994), together with the biophysical and the geographical domains (Sioli, 1985; Walsh et al., 1997). Conceptualizing the population-environment nexus as involving relationships among people, their institutions, and natural resources is termed a "regional political ecology" approach (Blaikie and Brookfield, 1987; Schmink and Wood, 1987). In this approach resource use decisions of households are linked to broader socio-economic and political forces that affect LULCC. Fortunately, recent research on small farmers and their role in land degradation has begun to link economic, cultural, and environmental factors, emphasizing the economic rationality of small farmers as well as their ability to adapt and persist under stressful conditions (e.g., Netting, 1993; Collins, 1986; Little and Horowitz, 1987). Applied to LULC in Amazonia, this model suggests the need to understand the decision-making processes of individual migrant farmers regarding the use of their plots. This is likely to vary according to household needs, labor supply, location, output prices, and national policies, international market effects, and natural resource endowments. In addition to recognizing more complex linkages between the human population and ecological impacts, spatial patterns must be taken into account by explicitly locating individual actors within their biophysical landscape, thus relating resource use decisions to natural characteristics such as soil quality, terrain, and climate across different natural units or scales and across time.

B. Previous Amazon LULC Research: Empirical Studies

Governments in Amazon-basin countries have long promoted colonization of rainforest frontier regions to relieve population pressures elsewhere, secure borders, increase agricultural production, and defuse political problems resulting from an inegalitarian agrarian structure, landlessness, and poverty. The underlying causes of deforestation or LULCC by migrant small farmers in Amazonia thus must be attributed to the combination of such macro-level factors as national population growth, poverty, high unemployment, land tenure regimes, urban-biased development policies, and the mechanization of agriculture (c.f. Pichón and Bilsborrow, 1992; Adger and Brown, 1995). These are the forces that drive people to migrate to the frontier.

While the fundamental causes of deforestation and LULC conversion in Amazonia vary significantly across and within countries, the clearing of forests for crops and cattle pasture is generally seen as the principal direct agents. This is certainly true in Ecuador (Southgate and Whittaker, 1992; Rudel, 1993). Numerous authors have commented on the social, political, and economic origins of colonization and forest conversion by Amazon settlers, with the bulk of the work on Brazil, i.e. Hecht (1981, 1985, 1992), Schmink and Wood (1984, 1992), Moran (1981, 1983, 1987), Fearnside (1983, 1986), Browder (1989), Eden (1990), Ozorio de Almeida (1992), Schneider (1995). However, most household-level surveys in the Amazon region have relied on fairly small convenience samples of households. Homma (1976) used data from 96 farmers along the Transamazon Highway in Para, Brazil; Ozorio de Almeida (1992) used data from 498 farm households in planned settlements in Pará and Mato Grosso, Brazil; Schmink and Wood (1992) utilized a random sample of some 200 households from a single Brazilian Amazon town; Walker et al. (1993) used data from 68 upland farms, and Walker et al. (1994) used 132 small farm producers, also in Para. In the case of Ecuador, the only other recent survey in the (quite different southeastern) Amazon draws on single visits to 68 households (Rudel and Horowitz, 1993). All of the above samples are small, and most, moreover, are not based on probability sampling, and hence do not allow generalization beyond the immediate households concerned, not even to the small geographic areas involved. Many have also focused on directed settlements and towns, rather than on spontaneous settlers, the vast majority of migrants farmers in the Amazon.

Observations of the process of settlement of these small populations of settlers over time in the Brazilian Amazon settlement, with the rapid conversion of primary forest to crops and pasture, led to a model of land use called the "peasant pioneer cycle" (World Bank, 1992). This cycle begins with the conversion of forest land by a new settler for subsistence crops, who switches to pasture as soil quality declines, and ultimately abandons the degraded land, moving on to repeat the cycle. This simplistic model has resulted in serious misconceptions about the complexity and diversity of LULC patterns, rates of forest clearing across the Amazon, and served to overestimate deforestation and its impact on the regional and global environment. It is particularly important to examine the process in more detail in the western Amazon where so many physical and policy parameters are different. Recent research has emphasized the adaptability of settlers and variations in LULC practices rather than an inevitable degrading cycle leading to further deforestation.

The ecological importance of not only the level of clearing by small farmers, but also its pattern over time and space has been made evidenced by ecologists examining edge effects, fragmentation of forest, and patch dynamics (i.e., Lovejoy et al., 1986; Skole and Tucker, 1993; Walsh et al., 1997). Taking into account the spatial patterns of LULCC and links to carbon fluxes and other biophysical and ecological impacts requires accurate spatial coverages of patterns of LULC at the appropriate scales of land management decision-making and administration leading to agricultural extensification and secondary plant succession. This requires remotely-sensed data confirmed by ground-based observations for LULC characterization and biomass estimation. Spatial analysis approaches to quantify landscape patterns as a means of understanding landscape function are also needed. Wood et al. (1996) link satellite and census data to study Amazon deforestation in Brazil, but census data, since they have to be limited to collecting only minimal information at a point in time, are completely inadequate for investigating the household decision-making processes which underlie land clearing. Moran and colleagues utilize a GIS to combine Landsat imagery with field data to understand the impacts of different land uses on successional vegetation in a caboclo region of Brazil (Brondizo et al., 1994). Moran et al (1994) and Moran and Brondizio (1996) also integrate multitemporal satellite imagery with ground studies to study forest succession in one (Altamira) settlement area along the Transamazonian highway in the eastern Brazilian Amazon. However, this research has not yet utilized household- or community-level socio-economic data. Dale et al. (1993) used satellite data and simulation modeling to assess deforestation and to relate the rates of LULC change to land management in Brazil by incorporating unstructured interviews with farmers. Currently, no studies, however, have been conducted that combine a satellite time-series with household data from a probability sample in Brazil nor the Ecuadorian Amazon except for our baseline survey in 1990; and none have involved a longitudinal data set for the same plots and households. With the 1990 in-hand, we will construct a unique panel survey data set by re-surveying the original 450 households for linkage with LULC, biophysical, and geographical data.

C. The Study Site: Ecuador's Amazon Region

The proposed study will take place in the Amazon region of Ecuador, the headwaters of the Amazon and a place of extremely high biodiversity that differs in crucial ways from most of the Brazilian Amazonia. Agricultural settlement has been almost entirely spontaneous, average soil fertility is higher, and year-round rainfall prevents burning. Ecuador's Amazon region comprises five provinces: from north to south, they are Napo, Sucumbios, Pastaza, Morona Santiago, and Zamora Chinchipe (Figure 1, Appendix A). Ecuador's 1990 census population for the region was 371,000 of which 273,000 are rural. The Amazon region continues to experience a high rate of population growth, over 5% per year in the last intercensal period, 1982-90, double that of the country as a whole. Almost one-half of its population was born outside the region, two-thirds coming from the Sierra, and most since the mid-1970s (INEC, 1991). Government policies have encouraged this migration, as the Amazon region has been perceived as an area with almost infinite space and resources (CLIRSEN, 1986; Larrea, 1987) and thus an "escape valve" to relieve socioeconomic imbalances in other regions.

Access to the northern Amazon was initially made possible by the petroleum boom, which began in the early 1970s and led to road construction to support petroleum exploration and extraction. Petroleum has since provided over one-half of Ecuador's export earnings and federal government revenues. Access and land occupation in the two northern Amazonian provinces have been in part by-products of petroleum production. The view of the Amazon as an escape valve for farmers from the crowded Sierra also has appeal for geopolitical reasons, since it allows high rates of population growth to be accommodated without major social upheavals such as land reforms.

Public forest lands in the Ecuadorian Amazon have been occupied first by settlers who then seek legalization of land claims from IERAC. IERAC certifies these claims, generally to parcels of 40-50 hectares, once settlers present evidence of land clearing for agriculture, according to the 1978 Law for (Amazon) Settlement (INCRAE, 1987). Migrants follow the new roads into the region, which lead to the major towns in the study site -- Lago Agrio, Coca, and Shushufindi, occupying lands along the roads in a landuse pattern known as respaldos, layers of landholdings developed parallel to the main road. The first migrants tended to occupy the first layer, or linea (line) along the road; subsequent settlers occupied lines increasingly farther from the road. This process has continued up to 14 respaldos in the 1990 population survey area, though two to five is far more common. Initially, the settlement policy produced a relatively homogeneous size distribution of land plots. However, with increasing in-migration and declining soil productivity, it is believed that land sales and consolidations have occurred subsequently, though no hard evidence exists.

Land quality and soil type vary widely across the sample: 46 percent of the settlers farm what they reported to be black (good) soil; another 24 percent red soils, which are acidic and of low fertility; 5 percent, lowland alluvial soils with poor drainage; and 25 percent, a combination. Soil type is dominated by ferralsols (FAO/UNESCO, 1971) or oxisols (USDA classification) which predominate in the Amazon river basin. These sedimentary soils are of low natural fertility, but possess excellent physical structure; the main limitation for agricultural is its fragility when subject to cultivation and the rapid decline of nutrients (FAO, 1976). The western fringe of the Amazon Basin, including the proposed site in Ecuador, is dominated by xanthic ferralsols.

A polyculture system has evolved in the study area based on "slash-and-mulch" cultivation, which can be distinguished from the "slash-and-burn" system used elsewhere, in that the felled vegetation is not burned. Due to the lack of a dry season and high annual precipitation (2,800 mm/year) in Ecuador, the slash-and-mulch system is a rational adaptation and differs from that of most other Amazon frontier regions. The polyculture system integrates annual crops such as corn and rice; semiperennials such as plantains, bananas, and yucca; and tree cash crops such as coffee and cacao. Eighty-five percent of the households marketed some crops, mostly coffee. Cattle were raised on two-thirds of the farms, but most owned only a few head. Families also raised chickens and pigs.