Terra-Care

Terra-Care
TerraCare focuses on how Infoterra uses remote-sensed geoinformation to study how and why the Earth’s natural elements contained in each biome change.

Introduction
There are two complementary methods for monitoring and mapping the Earth’s changing natural features: field survey and analysis of remotely sensed imagery that is the spectral and spatial information contained in a satellite image.

In a semi-automated classification approach, land cover maps can be derived directly from remotely sensed data by a variety of analytical procedures, including statistical methods (supervised and unsupervised classification algorithms) and photo interpretation. These maps are categorical, dividing land into categories of land use and/or land cover (thematic mapping; land classification). Techniques now allow the mapping of land use and land cover as continuous variables or as fractional cover of the land by different land cover and land use categories, such as tree canopy, herbaceous vegetation, and bare soil (continuous fields mapping).

Land cover and land use datasets may be compared between time periods using geographic information systems (GIS) to map and measure change detection at local, regional, biome and global scales.

There is not a single approach to generate land cover and land use datasets and Infoterra RS experts emphasize on the spectral signature of the satellite image and the biophysical characterization that infers from it. The approach used is endorsed by the scientific community.

The Method
Infoterra’s method to map the Earth’s natural features and to carry out a standardised, high quality Land Cover project follows the steps below:

• Collection of field or reference data (information from library, published maps, older LC maps, existing generic LC), • Selection of suitable satellite images from the time baseline onwards (for mapping changes, several years are necessary), • Development of an agreed classification nomenclature through consultation with project partner (definition of the different classes,…), • Pre-processing of satellite images (ortho-rectification,…), • Biophysical processing, i.e. processing the image to produce an absolute, physical description of the image in terms of pixel nature (vegetation, soil, ..) and attached physical characteristics. This is achieved by the so-called biophysical processing implemented in the Overland software. The output information layers are further named “biophysical layers”, • On some cases an operation of image segmentation is necessary to produce “image objects” having manageable size and homogeneous features, • Classification using processes and algorithms under the operator’s control, i.e. the qualification of the “physical state” of each image object. The physical characterization obtained at pixel level from the biophysical layers must be supplemented by other features such as object texture or heterogeneity. A list of the accessible physical states, with attached criteria, is established and all image objects are classified according to this physical state nomenclature.

Photo-interpretation are performed at the different stages (digitization by an operator of land-cover elements, on the basis of a visual inspection of the image).

Quality control to check the overall quality level of the land use, land cover map.