Remote Sensing: Seeing the Forest and the Trees

These online materials were specially prepared by INPA's GISLAB to support the real-time interaction you will see during LFRF program 3, on April 21, 1998. Many thanks to researcher Miroslav Honzak and GISLAB head Antonio Nobre for their exceptional cooperation in and support for LFRF. Both researchers enthusiastically believe in the application of high technology to help us better understand the natural world. They enjoy applying cutting-edge computer science to the complex issue of how to comprehend the biodiversity of tropical rainforests in order to find ways to better protect them.

Tropical rainforests are famous for their abundance of exotic animals, birds and insects, but in fact there's an equally impressive diversity of plant and tree species. The thick, apparently uniform green carpet seen from high above hides the greatest complexity of organisms to be found in any ecosystem on Earth. In just two acres of Amazonian rainforest there are about four times more species of tree than can be found on the whole of the North American continent!

Because the forest is so vast and the canopy so high, the identification of tree species using classic botanical methods is laborious and time-consuming. If all tropical botanists active in the world today set out to describe the Amazon rainforest, it would still probably take several generations before humans would know the distribution of biodiversity in the Amazon as well as we know temperate and boreal forests. However, Earth's rainforests are disappearing at a speed of one football field per second, which leaves no time for a slow study.

Recognizing the need for a breakthrough in the identification and mapping of tropical tree species, the Geo Info Systems Laboratory (GISLAB) of the National Institute for Amazonian Research, INPA Brazil, is promoting the marriage of traditional botanical knowledge with modern and powerful computerized pattern recognition techniques. The idea is to "teach" the computer to use digital aerial imagery to find and automatically recognize the crowns of trees botanically; in other words, to use remote sensing to map the distribution of tree species.

This task is not easy. In fact, it involves cutting-edge computer science and complex algorithms, as well as taxonomy and biology. First, the experience of a human botanist must be transferred to a database that will form an essential part of a so-called "knowledge-based system." The core of the initial working system will be an artificial neural network (a computer program) which will perform automated tree species classification using fine spatial resolution data (aerial photographs or, in the near future, fine spatial resolution satellite images). In particular, spectral information, texture information, morphometric information (the shape of leaves or crown), and tree phenology information inferred from the imagery will be used, together with the knowledge-based system, to map the tropical tree species diversity in Amazonia.

Conclusions: to be able to map the biodiversity of tropical forests effectively, before is too late, new technologies and advanced mathematical tools must be employed. The resulting biodiversity maps will then advance our knowledge of how rainforests work as living systems, and will thereby hopefully help establish a strategy for their preservation. This will benefit future generations of rainforest creatures who will live there, and also we humans, who depend in many ways upon the rainforests of our planet.

Links below explain some of the concepts referenced above, and provide additional words and images to bring the ideas to life.

Remote Sensing: Seeing the Forest and the Trees

  Spatial Resolution
  Spectral Information
  Morphometric Information
  Tree Phenology
  Jungle Tower

These pages were developed by Miroslav Honzák ( ), INPA - GISLAB, Manaus, Brazil.