Shifting Cultivation: A New Old Paradigm for Managing Tropical Forests

Shifting cultivation, or swidden farming, is often held to be the principle driving force for deforestation in tropical Asia (Myers 1993). National governments in Southeast Asia, notably in Indonesia, the Philippines, Thailand, and Vietnam, have been inclined to blame shifting cultivators, usually members of ethnic minorities, for rapid loss of forests (Dove 1984, Do Van Sam 1994, Le Trong Cuc 1996, Rambo 1996). In Vietnam, the official view of shifting cultivation has been particularly negative, reflecting a combination of the ethnocentric assumptions of the numerically dominant Kinh (lowland Vietnamese) about the cultural superiority of wet rice farming and the Marxist view that swiddening represents a primitive stage in the cultural evolutionary sequence (Jamieson 1991, Rambo 1995). Resource managers in these countries invariably see shifting cultivation as a single, simple system of farming in which the forest or scrub is slashed and burned to make swiddens. These fields are cultivated for only one or two seasons before soil fertility is exhausted or weed growth overwhelms the crops (Padoch et al. 1998). The field is then abandoned and the farmers move on to clear a new field elsewhere in the forest. From this perspective, swidden farmers are “forest eaters” whose unending search for new forests to clear is a major cause of deforestation. However, to view swiddens as just temporary fields surrounded by abandoned land under wild growth is wrong. More than four decades ago, Harold Conklin (1957) pointed out that “shifting cultivation may refer to any one of an undetermined number of agricultural systems” (p. 1). Spencer (1966) described 18 distinct types of shifting agriculture within Southeast Asia alone. Brookfield and Padoch (1994) argued that swidden agriculture is not one system but many hundreds or thousands of systems. This article seeks to describe the agroecosystems of a hamlet in northern Vietnam in which traditional shifting cultivation has not resulted in extensive deforestation, but it has altered the character of the vegetative cover. Through analyses of this agroecosystem, we seek to develop a better understanding of the swidden agricultural systems found in the region and the effects of these systems on land-cover change in the area over the last 40 years. Our findings have implications for understanding the role of shifting cultivation at more macro levels, including its contribution to global climate change. Shifting cultivation, the subject of this article, differs from the common practice of using fire simply to clear forests for permanent cultivation, pasture, or further development. The latter practice was employed on a massive scale during the European settlement of the frontiers of the eastern United States (Rambo 1990). This slash-and-burn cultivation, as practiced by the millions of lowland Vietnamese resettled in upland areas since the 1960s, turned large areas into virtual lunar landscapes (Le Trong Cuc et al. 1990). Of course, some traditional forms of shifting cultivation can also cause long-term environmental degradation. “Pioneer” shifting cultivation, for example, is a system in which new areas of forest are cleared for fields, allowed to remain under cultivation too long, and then abandoned in a degraded condition (Kunstadter et al. 1978). In northern Vietnam, Hmong farmers have converted large areas in the highlands to grasslands as a result of pioneer swiddening. Nevertheless, in focusing on destructive forms of slash- and-burn cultivation, national governments and resource managers have ignored the nature of more common and more sustainable forms of swidden cultivation. Alcorn (1990) calls swidden farming “managed deforestation,” a system built around patchy, pulsed removal of trees but not of the forest. She suggests that indigenous farmers work to manage deforestation in sequential agroforestry systems that integrate secondary successional vegetation—everything from grass and bushes to young open-canopy tree communities to mature closed-canopy tree communities. Brookfield and Padoch (1994) suggest that the concept of abandoned fallows is being displaced by research that shows that the plants found in any stage of secondary successional vegetation are in large measure the result of conscious planning. Even where the forest that succeeds farming is not closely managed, it is used. (The “fallow” is the period when land is left to recuperate and vegetation allowed to regenerate. It is not “abandoned”land, which the swidden farmer does not intend to use again. Rather, it is part of the land-use system in which the farmer will return to this plot again. Meanwhile, while it lies fallow, the farmer may use it to collect numerous food, timber, and nontimber forest products. Popular conception, however, is to view this land as “abandoned fallow.”) Failure to understand the swidden agricultural system, and its associated secondary vegetation, has led scientists to overestimate the amount of “deforestation” that has occurred in Southeast Asia. Potter et al. (1994) suggest that as much as 26% of all land in Southeast Asia falls into the “other” category, which includes scrub, brush, pasture, waste, and other land-use categories that often represent not deforestation but forest fallow or secondary regeneration. Likewise, Kummer and Turner (1994) suggest that approximately 33% of the land cover in the Philippines falls into the “other” category, and this category grew by more than 20,000 ha between 1948 and 1987. In Thailand, the field plan of the Tropics Program of the GEWEX Asian Monsoon Experiment (GAME-Tropics) is based on a land-use breakdown in which fully 49% of the nonforested land in northern Thailand is “unclassified” (GAME-Tropics 1996). Because swidden agricultural systems are so little understood, many governments have implemented (mostly unsuccessful) large-scale resettlement programs that are intended to convert swidden cultivators into farmers of permanent agricultural fields. Moreover, failure to understand the role played by secondary successional vegetation in swidden systems has meant that resource managers have not correctly identified the impacts, both positive and negative, of swidden agriculture on species diversity, watershed hydrology, and carbon sequestration (Skole et al. 1998), all of which have important implications for biodiversity conservation and global warming. Shifting cultivation has been practiced for centuries, if not millennia, in the northwestern highlands of Vietnam. Yet, contrary to the popular conception that shifting cultivation always causes deforestation, the area under forest cover in many parts of this region has not changed significantly, despite rapid population growth, over the past 50 years. Indeed, a recent study of the Da River watershed (Nguyen Duy Khiem and Van Der Poel 1993) found no correlation between the occurrence of shifting cultivation and the extent of deforestation. In those districts having the greatest extent of swiddening (12–36% of total area), the percentage of land under forest cover ranged from 6% to 48%, whereas in those districts having almost no shifting cultivation (less than 6% of total area), forest cover ranged from 8% to 48%. Tat, one of 10 hamlets in Tan Minh village (Da Bac district, Hoa Binh province) is representative of many swiddening villages in Southeast Asia. The hamlet is positioned alongside the Muong River, a tributary of the Da River, at an altitude of approximately 360 m above sea level. The Muong valley ranges from a few hundred meters to approximately 2 km wide and is surrounded by peaks reaching from 800 m to over 1100 m. The valley walls are extremely steep, with slopes often exceeding 60% crisscrossed by many small streams flowing into the Muong River. The valley is shaped like an amphitheater, with buttress-forming ridges extending up to the surrounding peaks. Our understanding of the hamlet's ecosystem is based on an integrated spatial database that incorporated topographic maps, aerial photographs, satellite images, and a digital elevation model with information on elevation, slope, and aspect. Information on land-use practices collected through interviews with farmers and other key informants was georeferenced to the spatial database. This database served as a framework for analyzing changes in land cover and forest patterns through time and as a tool for analyzing the information and insights collected in semistructured informal interviews. The spatial database was developed on the basis of aerial photographs (nominally 1:40,000) taken in 1952 by the French military, as well as a 1995 Landsat Thematic Mapper image. The aerial photographs were taken as part of a comprehensive mapping of Vietnam.1 The photographs were manually interpreted and classified into three land-cover categories—secondary regeneration, swidden, and paddy. The secondary regeneration or successional vegetation category was further subdivided into three classes: closed-canopy forest (closed-canopy forest mapped from the 1952 aerial photographs may or may not have been primary forest, but for our purposes we are classifying these forests as secondary vegetation); open-canopy forest; and grass, bamboo, and scrub. These photographs were registered on a map base, and the land-cover categories were digitized and entered into a geographical information systems (GIS) database (Arc/Info software on a Sun Sparc workstation). The Landsat image was classified into the same land-cover categories. Ground-truth points were collected in the field using differential GPS (global positioning system). The Landsat image was registered to the same map base as the 1952 photographs. Spectral signatures of the different land-cover types were derived from a subset of the ground-truth points that were then overlain on the imagery. A supervised classification of the image was done using a maximum-likelihood rule (ERDAS Imagine software). An accuracy assessment of the 1995 satellite-derived land-cover classification showed 95% of the 155 checkpoints correctly classified after accounting for GPS surveying and image registration error. The socioeconomic database was developed through interviews with many Tay residents of Tat hamlet and provincial, district, village, and hamlet government officials. Researchers documented changes in national and regional policies influencing land use (e.g., tenure, taxation, credit, import and export regulations) as well as changes in infrastructure (roads and markets). We interviewed residents of the village to learn more about the socioeconomic factors contributing to their decision to create or maintain forest fragments in their area. Key informants were used to assess, among other factors, local peoples' perception of the forest. Researchers conducted semistructured informal interviews with villagers to identify the socioeconomic and institutional factors influencing use and management decisions regarding forestland and forest vegetation. The people of Ban (hamlet) Tat are mostly members of the Tay ethnic minority.2 They speak a language belonging to the T'ai family. Local oral history maintains that migrants from Son La settled the hamlet a little over 100 years ago. Mobility appears to have been relatively high, with several waves of immigration and emigration. A large group of people migrated from Ban Tat to Nghia Lo, just across the hamlet's northern boundary in Phu Tho province, and kinship ties are still maintained today between these two hamlets. According to one elderly informant, only seven households, or approximately 50 people, lived in the village in 1954. Today the hamlet has grown to 69 households with a population of 389 people. This represents a population growth rate of roughly 4.9% annually—probably the result of a natural growth rate of 3–3.5% and in-migration. During this period, population density grew from approximately 10 people to 75 people per km2, which is approximately twice the average population density for the Da River watershed. For as far back as any can the Tay of Tat hamlet have been (Rambo that households manage permanent wet rice fields in the valley and shifting swidden fields on the and wild of the forest. as practiced by the Tay is an of the total agricultural system, not an of an more swiddening that is in the of being by more is swiddening a recent to rapid population growth that has the of the wet rice fields and people to their farming the as the Tay have practiced both wet rice farming and swidden agriculture as an integrated system of for and is a in Ban Tat at the of the fields are in the with cleared swidden fields in the surrounded by secondary vegetation. systems are found among the of and northern Thailand 1998), the of in and the of the in the Philippines (Dove In the of Ban Tat, elderly informants that their that employed both systems when to the valley at 100 years ago. that the area was by closed-canopy forest and was no of land on which to make fields in the valley The area of fields was much than it is and forestland was and for the It have been for households to have only cultivated fields or only cleared but are to have done are in a more Indeed, the resource system of the Tay is for its of a wide of A Tay a The of Ban Tat is a of cultivated and fields with forest areas by the 2 shows the of swidden fields, secondary vegetation, and tree cover that the Ban Tat does not to be any to this fields can be found on the from the of the to the A village that is for land and regarding forest use the This land to villagers from the two large areas Ban Tat where swiddening is and In villagers their swidden fields between the two areas several one of the areas to fallow and regenerate. In however, land is both are used at the same the fields more than 2 ha per the villagers manage their patterns of between cultivated crops and fallow in the hamlet to the forests to their fields under the of the village The use of a however, no long-term or use over that and after one of cultivation it is not to return to the same relatively to the when the soil is the common of swidden cultivation is years of rice and 2 years of by years of The of the fallow appears to be by the for and households can to maintain a The fallow period has been in recent reflecting the of land for swiddening by population density and by government to the and tree and are also important of the agricultural are used to and plants for More as and have been to of recent are from the in areas used also for swiddens. are grown for local and for as and and in fields, and swiddens and secondary forest also and between of the (Rambo and Le Trong Cuc the spatial of land use is settlement and wet rice fields at the of the on the above the tree and swiddens up the slope, rice swiddens on the and secondary forest on the which are by the can be found across the as well as the and tree The three types of primary forests primary or secondary secondary and watershed a national the government households to secondary forest to their and scrub not classified in one of these categories can be cleared for swiddens. The for forest is 800 in per however, the of being are relatively and the high, so the system is not an to forest in the more areas where government is forests are on the part of the some are found on the part of the above the with swidden fields on the and some are found on the of the with swidden fields on both forests are also found in the These are under the of the hamlet and swidden is not In some particularly on slopes above their farmers the forest to maintain for The and of the United deforestation as of forest land to and includes all land where the forest cover has been and the land converted to as permanent cultivation, shifting cultivation, and of on the other to a in the extent and of the forest cover to factors as and forest that deforestation as both a change in land cover loss of forest and a change in land use converted to other permanent on the other is simply a change in the of forest cover. have that of tropical deforestation for several including surrounding the of forests that have been a of forests is the rate of deforestation is in the of the rate of change in areas et al. Turner et al. et al. In other if the forest the be forest not deforestation. changes in land cover and in Ban Tat between 1952 and 1995 are in and in During this period, the area by secondary regeneration or successional vegetation from to of the The area by and open-canopy forests from to and to of the while the area by grass, bamboo, and scrub from to of the Our that of the of Tat hamlet under secondary more than a of shifting that traditional swiddening does not permanent but only temporary use of we not describe the secondary regeneration that after shifting cultivation as deforestation. We found only deforestation in land used for from to for the study area over the years between 1952 and the amount of land under secondary regeneration relatively over the period, and open-canopy forests have degraded to of Ban Tat as many swidden agricultural landscapes in Southeast Asia, is not so much deforestation as a change from a forest cover and to a cover of secondary vegetation. Our work suggests that the other major land-cover change from the long-term practice or of shifting cultivation is a in the of forest changes have also occurred in the spatial of that land cover. 1952 and the number of secondary regeneration fragments grew from 18 to and the from ha to 2 ha 1). et al. found in northern The of forest is well also in the on (e.g., et al. 1991, and et al. and 1996). have only to study the effects of on biodiversity in on species of forest and on forest in Southeast Asia have that the of species in forest fragments is often with that in more mature forests al. 1998). in (Padoch and et al. 1998), showed that of both plants and is in swidden areas than in the primary forest, some have of the of natural wild (e.g., are more by forest The in more (e.g., wild at Ban Tat is more the result of than it is a of The effects of forest on the and of in across a forest to effects extending the of a forest and 1995). the spatial of from forest can changes in and and et al. 1993). and cultural factors important in both the and of forest fragments and 1996). the between factors and the and of forest fragments are not understood, to that the of a when many have traditional agricultural with from or migrated to areas in search of in land use and land cover in Ban Tat are at with the that shifting cultivation, under of population in extensive deforestation. of the is to we deforestation as to forest but two other factors are if not and is the of the swiddening system employed by the which have not Because can be as population density fields with swiddens in a system some of the to clear more forest. in Tat hamlet approximately of their from their fields, these fields are much in area than the swiddens. The also to the Tay system, is the of secondary and the to maintain a of land cover across the of the of forest This indigenous land-use system both the of and the percentage of the by secondary vegetation. Our work to too much has been on the swiddening has on land cover from forest to cover of secondary and too little on the of swidden agriculture as the land-use system in this In other if not just forest agriculture result in land cover by trees (e.g., or or in one of (e.g., and upland tree that the hamlet may be to maintain a percentage of tree cover. In the or however, as by the number of species found on the et al. 1998). be more than those under secondary vegetation of the traditional swidden system et al. sequestration be approximately the same under an system as it is under but it be under an system et al. 1998). The of from secondary vegetation to a however, be than under the swidden the a global well a move from swidden to permanent agriculture over the few Tat hamlet has been in being to the changes of the last several decades with in the total area by secondary vegetation 1). and to that the of tropical forest has been a change from a relatively forest to a and cover of secondary vegetation. The land cover may be degraded in of However, this secondary vegetation, a of the swidden land-use system, well be the species and and land cover In swiddening is the land-use system for the of the local suggests that tropical biodiversity conservation is a in which forest and secondary growth forests are being for their in the conservation of These a new in the management of tropical biodiversity that conservation to al. 1998). Failure to see the as well as the of secondary vegetation and the swidden agricultural system has led to government policies for swidden of which have been A more as well as be to in research on of the biodiversity associated with swidden fallows while their and Failure to understand secondary successional vegetation has also meant that resource managers have often to the both positive and negative, of swidden agriculture on watershed hydrology, and carbon sequestration (Skole et al. 1998). of global change have been based on an of forest to degraded or 1996). Failure to for the effects of may that effects of land-cover change are not being cultivation is an built around the temporary removal of trees but not of the forest. we the new we well to the of this for tropical forest This is based on research conducted in Tat hamlet since by the for and of the Vietnam National and the for our field research has been by to from the of the the and the National and by a to and the from the and of on land-cover change in Tat hamlet was by a to from the cover and in Ban Tat in 1952 and The swidden system in the fields are in the with swidden fields in the surrounded by secondary vegetation. The of Ban Tat in the wet the of swidden fields, secondary vegetation, and tree cover. cover and in Ban Tat in 1952 and in The Ban Tat land-cover classification map is in