Estimation de la biomasse forestière et caractérisation de la structure verticale des peuplements de conifères par radar VHF et radar sondeurs aéroportés

Abstract : Interest about monitoring and predicting the evolution of forest resource has considerably increased these last years, notably to supervise and predict the impact of the growing anthropogenic pressures on forest ecosystems. These efforts require a better characterisation of the forest covers. Microwave remote sensing ("radar" remote sensing) appears as a powerful tool for these problems, considering the sensitivity of microwaves to the biophysical properties of the forests. However, retrieving variables of interest linked to forest covers, using current radar spatial sensors, is not sufficiently reliable. This thesis aims at carry on with previous works on radar remote sensing of forests, focusing on the potential of originals airborne sensors. This is, using an approach starting from the experimental analysis of the data, supported by theoretical modelling of microwave interaction with forest media, towards the development of semi-empirical retrieval algorithms. The two test-sites are composed of monospecific plantations of Maritime pine and Austrian pine. A great amount of forests measurements are available, and integrated, for one site, in a GIS. The first sensor under study is a helicopter borne, non-imaging, ranging scatterometer, HUTSCAT, which gives backscattering profiles inside tree canopy with a vertical resolution of 68 cm at two wavelengths (X-band : 3 cm and C-band : 5,6 cm). The second sensor is an imaging airborne SAR operating at VHF frequencies (20-90 MHz, wavelengths of 3 to 15 m). For the interpretation of the data, two models were used : a radiative transfer model (RT) simulating only the propagation of the energy, and a coherent model, describing the propagation of the electric field (both amplitude and phase of the microwave). These models require a precise description of the geometry of the trees, difficult to obtain through ground measurements, even more necessary when the task is to model high resolution remote sensing data. This precise description is given by the architectural plant model AMAP (Atelier de Modélisation Architectural de Plantes du CIRAD). Results of the experimental analysis of HUTSCAT show that the height of the trees can be estimated with an absolute accuracy of about 1 m. From these estimations, others variables of interest can also be retrieved : stem volume, site index. CARABAS radar backscattering coefficient is found to be closely related to forest variables in a large range of stem volume values : up to 900 m3/ha. The signal is strongly correlated to the trunks geometric characteristics (diameter, height, volume) and the relations of the backscatter versus stem volume appear to be very close between both sites. A strong dependence to the topography (slope) is also observed. For the derivation of biomass retrieval algorithms, it would be necessary to develop a semi- empirical correction. Interpretation of HUTSCAT vertical backscatter profiles requires a model that can accommodate the vertical variability of the canopy. The RT model, coupled with AMAP, is appropriate because it describes the vegetation as a superposition of horizontal and infinite layers. Results of simulation demonstrate that the infinite layers assumption is not valid at shallow incidence angles and a correction is proposed taking into account the shape of the tree crowns. From there, comparisons of simulated and experimental profiles show a good agreement. The model show that the needles are by far the main scatterers. These simulations prove, at centimetric wavelengths and for low incidence angles, that all the part of the canopy contribute to the signal measured by the sensor. Based on these conclusions, a retrieval algorithm of the foliage distribution within tree canopy is developed and validated using ground measurements. Theoretical study of CARABAS data is conduced using the coherent model in order to take into account coherent interactions between scattering mechanisms such as double bounce scatterings between the ground and the vegetation. For both species, result show that the trunk is the main scatterer and the branches contribute lowly to the total signal. The decrease of the backscatter as a function of the topography is explained by the loss of the preponderant scattering mechanism (bistatic trunk-ground reflection) as the slope increases. We conclude that for species presenting a trunk larger compared to all other vegetation elements (in terms of volume), as for the coniferous, the measured backscatter will be strongly correlated to the stem volume and little variation will be observed between species. These observations demonstrate that using metric wavelengths allow to diminish the impact of the tree species and the forest complexity. All the presented results show the interest and the complementarity of airborne sensors working at high resolution and at both edges of the radar spectra (X-band and VHF). Even if these sensors configurations cannot be applied to satellites in a near future due to technical difficulties, these data will be of great interest for forest managers to monitor and predict the evolutions of forest resource at local to regional scales.
Liste complète des métadonnées

Cited literature [2 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00000036
Contributor : Ecole Agroparistech <>
Submitted on : Tuesday, March 19, 2002 - 8:00:00 AM
Last modification on : Friday, April 12, 2019 - 4:22:30 PM
Document(s) archivé(s) le : Friday, September 10, 2010 - 3:11:19 PM

Identifiers

  • HAL Id : pastel-00000036, version 1

Citation

Jean-Michel Martinez. Estimation de la biomasse forestière et caractérisation de la structure verticale des peuplements de conifères par radar VHF et radar sondeurs aéroportés. Sciences du Vivant [q-bio]. ENGREF (AgroParisTech), 2000. Français. ⟨pastel-00000036⟩

Share

Metrics

Record views

1504

Files downloads

5094