Land used for extensive grazing in the Neotropics has increased continuously over the past decades and most of this increase has been at the expense of forests. Ranching-induced deforestation is one of the main causes of loss of some unique plant and animal species in the tropical rainforests of Central America and South America as well as carbon release in the atmosphere. Although the driving forces changed, this process is known to be still ongoing, while more and more evidence is becoming available on the often unsustainable use this represents. With time productivity declines and severe degradation problems occur.
From this standpoint, it is urgent that alternatives to extensive livestock production in Latin America are found. To be able to address this need, a first prerequisite is to know the location of the problem areas: Statistics from different sources demonstrate the evolution of livestock production and deforestation at the nation level, but their relation is known to be highly location dependant. To help decision-making address the problem it is therefore essential to dispose of a spatially specific information on the relation between different land uses, as well as its evolution.
The LEAD programme therefore seeks to identify “grazing induced degradation hot spots”, i.e. areas where unsustainable exploitation leads to degradation of ecological, environmental and productive quality. The current mapping effort focuses on deforestation hot spots, which are considered more critical and urgent to deal with than pasture degradation. Livestock related deforestation hot spots are defined as large forested areas that will be replaced by pasture in the near future. On the basis of data, literature and expert knowledge, the objective of the spatial modelling approach is to explain spatial correlation between various processes, and predict where they will take place in the near future. The focus is here on spatial extent and location (at regional scale), which is a clearly contrasting and complementary approach with respect to the large body of literature analysing causes of deforestation (at local scale).
The spatial modelling framework
Of all the factors that have been reported to influence the location of the deforestation process (and thereby the other land uses!), we need to know where they have a significant influence and what their relative weight is. We then can spatially distribute quantitative change projections. This information is obtained through a statistical systematic modelling approach. We selected the Conversion of Land Use and its Effects (CLUE) modelling framework, prepared by Veldkamp and Verburg at the Wageningen University.
The main result is a map of the Neotropics depicting predicted land use changes in the future (2010). It highlights where forest is most likely to be replaced by pasture or crop land.
Distinct patterns of deforestation are found within and between countries, pointing to different types of land use change processes. Looking at the region as a whole, lowland deforestation frontiers clearly account for the bulk of deforestation. Forest in these frontiers is mainly converted to pasture, except for a few cropland expansion areas.
National and local policies are important determinants to the deforestation process and its location. Although not explicitly included in the location factors, they are taken into account in the analysis as the land use change demand at country level integrates the effects of policies on the aggregated land use changes, and the regressions are calculated at a sub national level, assessing how current policies are translated into current land use patterns. If policies do not change in the near future, we assume that the land use change should follow the same “rules” as it has done in the near past. As a decision support exercise, sensitivity analysis and scenarios assessment would also allow to test future trends if policies are changed.
These results are summarized in the table below regarding the quantitative summary per country of the share of pasture and cropland estimated to replace forest. The percentage refers to the total deforested area.
Pasture expansion into forest
Cropland expansion into forest
%of total deforested area
%of total deforested area
All study area in Central America
All study area in South America
Results show that:
in all countries over 50% of the deforested area is occupied by pasture
in Central America this is over 60%, although near insignificant in Costa Rica
in South America large differences are found between countries
many other changes are predicted, although of lesser magnitude
in general a net replacement of cropland by pasture is found, leading to crops “moving" into the forest
an inverse situation is found in Ecuador, Guyana and Venezuela
The case of protected areas
The protection status of protected areas is mostly unclear. Over 80 different protected area denominations were reported throughout Latin America and some authors report that the protected status of many areas is not respected due to their large size and the little means for control (Jaramillo and Kelly, 1997). Instead of prohibiting the model to allocate any deforestation in protected areas, we therefore included protected areas as a location factor in the stepwise regression modelling: if protected areas are strongly dominated by forest this suggests that the area’s status is more or less effective and they will remain relatively protected during change allocation procedures (except Panama, where, as recommended by local experts, no deforestation was authorised in the Panama Canal watershed). Three classes of protected areas were used in the regression models. Overlaying the change allocation results and protected areas showed that in general this approach leads to respecting the protected status of forest. Some deforestation though was allocated in protected areas, which is worth noting. In Central America, significant pasture expansion into forest is foreseen in the Maya Biosphere reserve in Guatemala’s northern Petén region, mainly in the Laguna del Tigre national park. In South America, a few parks would be severely threatened; the Formaciones de Tepuyes natural monument in eastern Venezuelan Amazon, the Colombian national park Sierra de la Macarena and the Cuyabeno reserve in north-eastern Ecuador. This has been confirmed by national experts for the last two cases. Although this represents a limited portion of total deforestation, it may have a considerable ecological significance. The Macarena national park for example is the only remaining significant corridor between the Andes and the Amazon lowlands. Small spots, that could be the beginning of worse, are also noted at the high end of the Carrasco Ichilo national park on the Andes slopes between the Bolivian highlands and the lowlands towards Santa Cruz. In all cases, the majority of the deforested area would be occupied by pasture.
Without stating that livestock production causes deforestation, one can surely say it is associated with this process, being the main land use replacing forest after clearing. Indeed, the results produced estimate that the expansion of pasture into forest is greater than that of crop land, and above all they show where each of these processes are expected to dominate. The results allow distinguishing different spatial patterns of change. The presented hotspot maps indicate the main contiguous areas at risk. In addition large areas with a far more diffuse land use change pattern, but that may deserve attention too, were delineated. While hotspots may serve as political and scientific focal points of attention, the model results also help to increase the awareness that a large part of agricultural expansion into forest may well take place outside hotspots.
Land use change modelling, especially if done in a spatially-explicit, integrated and multi-scale manner, is an important technique for the projection of alternative pathways into the future. This work, and the regional insights provide, sufficiently detailed to constitute a decision support tool for decision makers. The spatial information as such may serve to target policies and focus activities (i.e. conservation, payment for eco-services, incentives for sylvopastoralism, adapt land tenure regulations). To this end it will be particularly interesting to cross the results with other policy information as biological corridors, protected areas, conservation programmes, etc. The model and its input themselves also constitute an interesting decision support tool, allowing to evaluate the impact of different scenarios of change. This role can be achieved through testing aggregated land use demand scenarios or through modifying location factors (e.g. new human settlement, new transport infrastructure).
From a research perspective, the results provide a framework to the local case-studies on the land use change realised in the recent past, filling the gap between such studies and national quantitative estimates of change. We also hope our work will support the identification of further detailed studies, by giving insights as to the factors to be considered (the location factors of our study) and the identification of dynamic frontiers (hotspots) and diffuse deforestation areas.
Raw model outputs, hotspot maps as well as baseline land use input are available in GIS format through the Geonetwork at the following URL: http://www.fao.org/geonetwork
An approach including expert knowledge
Modeling change for the entire Neotropics represents an unprecedented scale of application of the model. Model assumptions, parameters and results therefore needed to be evaluated. The same account for model inputs, as it is far from straightforward to obtain consistent, reliable and spatially detailed information at this scale. A panel of national experts was established for this purpose.
over thirty experts contributed by filling out a questionnary addressing above isuues
a workshop gathered international experts to: discuss the modelling scenarios in terms of input data, estimated land use demand changes and elasticity of land use conversion, discuss preliminary results of the model, discuss the implications of regression results in terms of causes of deforestation, including qualitative “driving forces” that could have been missed
This study has been submitted for publication with Agriculture, Ecosystems and Environment under the title “Predicting land use changes in the Neotropics: the geography of pasture expansion into forest.”