The Tanguro Savannization Experiment

Experimental fire at Tanguro Ranch

"How frequent do fires need to be before a forest is replaced by Savanna?" This is the main question driving our current research on savannization. To replicate the forest fires that take place in the Amazon, we have established a large-scale experiment on Fazenda Tanguro, a private Soy farm in Mato Grosso state, where areas of forest are already slated for destruction to expand soy fields. Before these forests are removed, we will burn 3 square kilometers (approx. 750 acres) of them repeatedly over the next 5 years, and intensively study the impacts of these fires on the plant and animal diversity, microclimate, and general structure of the forest. We expect to learn a great deal about how the forest changes with each successive fire, and hope to ultimately learn whether the forest will be able to survive such treatment, or whether it will be gradually replaced by fire-adapted species from the adjacent Cerrado ecosystem, which is a savanna composed of tall grasses, shrubs, and small trees. We are also coordinating closely with researchers from Stanford University who are collecting and analyzing satellite imagery of the forest pre- and post- burn to link our ground measurements with these images. In addition, we schedule our fires to coincide with overpasses of satellites with thermal imaging capabilities, to provide ground-based temperature measurements and location of the fire at eacah moment, in an effort to add to the knowledge base of space-based fire detection and analysis.

Hot Pixels, in red, show where fire activity was greatest in 2004.

Located on the southeastern fringes of the Amazon forest, in a region that has suffered some of the most frequent and extensive burning in recent years, the Tanguro ranch is ideal for such an experiment. The forest here is very typical of transitional forests all along the southern and eastern Amazonian frontier, where season drought is especially severe and fires are more common. Our experiment should be very representative of processes that are occurring along the most active frontier of the Amazon basin.

Aerial view of the first half of the 2004 burn, with the field sample grid overlaid.

We have established 3 experimental parcels, each with an area of 150 hectares (370 acres). Two-thirds of each parcel will be burned and the remaining area of each will be used as an experimental control. Inventories of all large trees (>40cm DBH) in each parcel are conducted, and all trees are inventoried in key transects and subparcels. Litter traps are laid out over the grid and sampled monthly to determine pre- and post-burn litterfall. Leaf Area Index (LAI) is measured monthly over 650 sample points in each of the 3 parcels, to learn how the fire affects the integrity of the forest canopy. Soil moisture, air temperature, and relative humidity are sampled at several locations in each parcel, to learn how the forest micro-climate and soil moisture content change as a consequence of the burning. Sections of each parcel will be burned repeatedly under different intervals over the next 5 years to simulate repetitive accidental fires.

A researcher examines new sprouts from the roots of a tree killed by the 2004 fire.

About half of this tree's crown was killed by heat from the fire beneath it.

We burned the first square-kilometer parcel in August of 2004, and have had a chance to see some of the impacts the fire has on the forest. Perhaps most striking was the selective nature of tree mortality – certain species appear to be far more resistant to fire than most, and other species may be killed by the fire, but they quickly send out new shoots from their still-living roots. As we burn subparcels repeatedly, we hope to learn whether fire favors certain species in this transitional forest. Some trees were not killed by the fire at their base, but even though the flames never got high enough to touch their leaves, some portions of their canopies were damaged by the intense heat.

Dramatic reductions in leaf area over certain sections of the parcel that were particularly hard hit by the fire have occurred; soil moisture is markedly down in these areas as well, reinforcing theories that trees redistribute water from deep in the soil to closer to the soil surface.

We have documented good correlations between temperatures measured on the ground during the fire, and thermal imagery taken by satellite. We have learned that some coarse-resolution satellite imagery typically used for detecting fire was unable to see our ground fire, even though it was advancing on many active fronts. Apparently higher resolution imagery is needed to detect fires beneath the forest canopy.

More fires are scheduled for August and September of 2005, the end of the dry season in this zone. We will burn two more parcels, each 1 square kilometer in size, and we will also burn half of what was burned last year for a second time. Monthly fieldmeasurements will continue for the duration of the experiment.

©Woods Hole Research Center, 2007