productivity declines and farmers abandon land and may migrate back into
the pioneer fringe; land is concentrated in the hands of fewer property own-
ers. The zone of decay is subsequently revitalized as originally small plots are
amalgamated for livestock production or higher-technology agriculture is
introduced. The last two stages of Henkel’s model are equivalent to Richards’s
closing frontier, which he characterizes in terms of low or zero availability of
forest land for further colonization and further improvements of infrastruc-
ture and farmer integration with markets. These simple models contribute to
the explanation of the spatial zonation of agricultural frontiers at any time and
the changes that are observed over time at any given place.
Stages in the development of agricultural frontiers are accompanied by
changes in the proportion of the landscape at some stage of the shifting culti-
vation cycle. Smith et al. (1999b) studied these changes in farming communi-
ties near the city of Pucallpa in the Peruvian Amazon, using data derived from
interviews with farmers in settlements characterized as belonging to the
emerging market economy (EM) and closing frontier (CF) stages of Richards’s
framework. Interpreting their results as points in the temporal development of
a single frontier, they conclude that excluding cattle ranches, approximately 50
percent of the original primary forest cover of an area is lost between the pio-
neer and EM stages. Half the cleared land in EM is in fallow at a given time,
and this proportion increases to about 60 percent in CF as residual primary
forest cover plummets. Despite the marked decline of residual forest cover
between the two stages, the net decline of overall forest cover as the frontier
evolves is low because primary vegetation is replaced by secondary (including
fallow). Forest cover in CF is still 40 percent, composed of 23 percent second-
ary and 17 percent primary. However, the rise of cattle ranching in CF means
that the loss of original forest cover and the net overall reduction in forest
cover including secondary vegetation are both greater when the whole land-
scape is considered. A final comment on Smith et al.’s analysis is that whether
net change in total forest cover during the development of swidden landscapes
is limited depends, of course, on the definition of forestthat is used. If the def-
inition includes young fallow vegetation, then net change is indeed limited. In
the context of this chapter, however, it is vital to remember that fallow vege-
tation and secondary forest are very different, biodiversity-wise, from primary
forest.
Agricultural frontiers are sometimes dynamic, as we have seen, but in some
cases shifting cultivation is a stable land use in both traditional and modern
contexts. It probably began in 3500 BCin the Maya Zone of Mesoamerica and
is still practiced there today. This time period naturally involves a switch from
traditional to modern forms of the practice, and the extent of land used and
the size of the dependent population undoubtedly have fluctuated over time
for many reasons, including the widely cited hypothesis that land degradation
was a factor in the decline of Maya civilization (Lambert and Arnason 1989;
158 III. The Biodiversity of Agroforestry Systems