Context and objectives
The initial objectives of the first phase of the project (focussing on Costa Rica) were the following:
1. To determine the ability to detect forest perturbation using ERS1-imagery, in relation to the extent, geometry and intensity of the perturbation.
2. To determine the influence of topography and viewing angle upon the monitoring precision, using images of ascending and descending orbits.
3. To study techniques for radar Image enhancement, texture classification and radar specific geometric correction.
4. To define the optimal frequency of data acquisition in relation to land use monitoring.
The initial aims were adjusted as follows: to analyse the capabilities of ERS.SAR.PRI products for
1. tropical forest classification, with emphasis on discriminating mangrove forests and seasonal forest.
2. monitoring of land use in a region where major land uses are oil palm plantations, rice fields and different types of pastures ranging from intensively used to almost fallow.
3. monitoring of urban encroachment, aimed at small villages and Isolated farm constructions in a rural area, and a coastal area under high pressure by the tourist industry.
The study area of the second phase was selected in Congo. Objectives were:
1. To study techniques for radar image enhancement and classification, with focus on forest.
2. To determine the ability of ERS-SAR to detect forest perturbation in relation to the extent, geometry and intensity of the perturbation.
3. Differentiation of vegetation types in relation with soil moisture content using ERS-EAR data.
Expected scientific results
Following conclusions regarding the usefulness of ERS-1.SAR.PRI data for classification and monitoring of land resources in Costa Rico and Congo could be reached:
Only data from areas with slopes less than about 6% can be used directly for analysis.
ERSA.SAR.PRI date provide information on general land use for flat areas. Classes which can be discriminated:
-superclass based on woody vegetation, -superclass based on grassy vegetation, -villages and towns,
-superclass based on annual crops.
No discrimination could be made between different forest types in flat regions.
Large rivers (first and second order) are easily identified in ERS.SAR.PRI.
Villages and isolated large buildings can be identified. Monitoring of the urban encroachment appears thus feasible with ERS.SAR.PRI.
The noise (speckle) in the ERSA.SAR.PM data is substantial despite the 3-look averaging which was applied. The speckle in SAR-imagery can be eliminated (partially) by specific filters, such as the LEE-, FROST- or MAP-filter.
The FROST-filter proved In this study to be the most suited.
Texture (apart from the speckle effect) is present in ERS-II.SAR.PRI products. However, it appears only related to meso- or macro-scale phenomena (e.g. parcel structure and geology respectively).