Spatio-temporal characterisation of the sub-equatorial forest canopy based on multiscale approach from field and space observations
(2002)
Files
deWasseige.pdf
Restricted Access - Adobe PDF
- 9.18 MB
Details
- Authors
- Supervisors
- Defourny, Pierre
- Abstract
- Satellite remote sensing is a major source og information for the tropical rain forest monitoring. While applications such as forest extend mapping are operational many basic questions are not documented yet. Why the vegetation index values of the rain forest are found lower than those pbtained for secondary forests? Why seasonal reflectance variations have been regularly observed over evergreen forests? The poor understanding of the remotely sensed signal with regards to the ecosystem complexity limits the developpement of advanced applications wich could assess forest type degradation or transformation. The Tropical Forest Experiment jointly carried out by the TREES project and a consortium of universities was a pioneer exercise towards a multi-scale study of the tropical forest from field to satellite measurements. This unique field experiment has been completed in the forest of Ngotto (Central African Republic) for a 8 months period. This research specifically focused on the spacio-temporal variability of tropical forest foliage based on SPOT-XS and NOAA-AVHRR times series acquired synchronously with field campaigns at a compatible scale. First, Leaf Area Index (LAI) measurements made on transects, were analysed using geostatistical tools in order to design the field sampling protocol for the seasonability study. Monthly LAI measurements acquired on four permanent 4 km² sites highlighted the foliage seasonality of such a tropical forest often considered as evergreen. A foliage seasonality of 0.5 LAI was compared to the temporal profiles of both satellite series. The NDVI temporal profiles were built from 6 SPOT-XS and 12 NOAA-AVHRR images selected out of hundreds of observations. This empirical comparison provided unexpected results further investigated through a sensitivity analysis built on a virtual 3-D tropical forest. The radiative transfer simulations indicated that a temporal variation of 15% on the SPOT near infrared reflectance could be split into 3 contributions : the surface change of 0.5 LAI (3%), the atmosphere (4%) and the geometry og observation (7-8%). To improve the relationship between LAI and satellite signal two methods were proposed to reduce the geometric factor G provided better results and has been applied to smooth out the temporal profile. At the opposite, these dominant directional effects are also taken as a source of information for the characterisation of the forest structure. Three ground-identified forest types could be remotely discriminated on the basis of the angular signature of their texture, driven mainly by the gaps size and distribution of the dominant trees. Towards a sustainable tropical forest management, a monitoring system of selective logging activities has been finally evaluated based on the logging trails detection as a function of to the spatial resolution, the temporal resolution and the geometry of acquisition of the satellite images.
- Affiliations
Citations
de Wasseige, C. (2002). Spatio-temporal characterisation of the sub-equatorial forest canopy based on multiscale approach from field and space observations. https://hdl.handle.net/2078.5/124232