Research project OR/02/001 (Research action OR)
Objectives
Sustainable management of forest resources requires quantitative baseline data including spatial attributes and stand parameter estimates. In addition to terrestrial techniques, a number of remote sensing based methods have proved to be valuable. For many land use / land
cover applications at the local and regional level the spatial resolution of the traditional high resolution satellite data, like Landsat and SPOT images, has proven to be inadequate. This is particularly true for highly fragmented forest landscapes. Scattered forest parcels fulfil
diverse environmental, social, cultural and economic roles and various benefits could be obtained thought their wider integration in all land-use systems (rural and urban). In this context, the upcoming deployment of Pleiades and COSMO-SkyMed instruments constituting
ORFEO, presents an unique opportunity to develop innovative methodological techniques exploiting the synergy of optical and radar very high resolution (VHR) data.
Based on simulated data, this project aims at developing new image analysis techniques to prepare and promote the applicability of the ORFEO imagery for mapping and inventory of small forest patches. Two main research objectives can be formulated:
1. Mapping: design of an object-oriented classification routine to map forest patches
2. Inventory: assessment of the added value of radar data in the estimation of four forest
biometric variables: stand density, crown size, canopy closure and species composition
The project determines whether newly developed techniques, which eventually might lead to operational procedures, may replace established forest mapping and inventory processes. Consequently, decisive answers are formulated on how the new ORFEO sensor data may
contribute to the use of remote sensing technology in forest-related local and regional decision making. It is expected that reliable estimation of relevant forest parameters will aid governments in implementing their policy of sustainable forest management and conservation
of biological diversity.
Methodology
This project consists of two research modules in compliance with the two above stated
research objectives.
1. Module 1: SYNOPRA-FM# forest patch mapping
Design of an object-oriented segmentation/classification routine including:
- image segmentation
- object feature calculation (e.g. texture, shape, spectral and wavelet measures)
- feature selection (with genetic algorithms)
- object classification (with classifiers like neural networks and/or support vector machines)
Research questions:
- What are the important feature types?
- Are the relevant features derived from optical or radar VHR data
- To what degree does radar backscatter provide complementary information?
2. Module 2: SYNOPRA-FI# forest patch inventory Design of object-oriented segmentation/modelling routines to estimate stand density, crown size, canopy closure and species composition. For this purpose, both novel techniques will be designed and operational
tools previously developed at the Laboratory* will be fine-tuned and adapted.
Research questions:
- Is the joint use of optical and radar data appropriate? Or, are the stand variables more
accurately estimated based on one of both data types?
- What feature types can be distinguished and which ones are relevant?
- What additional information does radar backscatter provide?
Expected results
Expected project outcome
- Semi-automated method to map forest patches
- Semi-automated methods to derive estimates of stand density, crown size, canopy closure and species composition
- Scientifically founded statement whether VHR radar data provide an added value to VHR
optical imagery in forest patch mapping and inventory expected products
Expected project deliverables
- In-house developed software: application of developed methodologies and services - Scientific papers in international peer-reviewed journals
- Tools that assist in mapping, managing and monitoring of small forest patches
Study area
Extended Meix study area, South-eastern Belgium; Baux-de-Provence
Data
VHR optical and radar data (e.g. Pelican, Ramses, Cosmo, TerraSAR-X)
