French Guiana through the clouds: first complete satellite coverage

J.-P. Rudant

LGGST, Université Pierre et Marie Curie, ParisAssociate Researcher at ORSTOM Centre, Cayenne

French Guiana, a French overseas 'Département', covers an area of 90 000 km². It lies on the northern (Caribbean) coast of South America between latitudes 2°N and 6°N and between longitudes 52°W and 54°W. Some 90% of the country is covered by dense tropical forests. Apart from a few airfields, navigable rivers are the only means of access to the interior. Most economic development is concentrated along the Quaternary coastal plain, representing less than 2000 km2. The European spaceport at Kourou (also known as the Guiana Space Centre, or CSG) is located on this plain. It is from here that the ERS-1 satellite was launched in July 1991. The southern part of French Guiana is probably the least well mapped of any French territory. Up-to-date large-scale maps are only available for a tiny portion of the country. These were prepared by the BRGM (Bureau de Recherches Géologiques et Minières) for mineral prospecting.

The first complete coverage of the country by satellite imagery was acquired in the nine months from April to December 1992. The coverage required 18 images gathered by ESA's ERS-1 satellite. A mosaic (Fig. 1) was assembled by the imagery laboratory of the ENST (Ecole Nationale Supérieure des Télécommunications) in collaboration with LGGST (Laboratoire de Géologie, Géomorphologie Structurale et Télédétection) of the Université Pierre & Marie Curie, Paris. Before the images could be mosaicked, the raw data had to be processed both geometrically (translation and rotation) and radiometrically. Radiometric processing was required to compensate, particularly over forests, for effects due to variations in the imaging instrument angle of incidence within the instrument swath (Rosaz, Maltre et al.) The finished mosaic represents probably the first complete satellite coverage of a country in the humid tropics. For the sake of convenience, only a small-scale version of the mosaic is presented here. Larger scale versions reveal a great deal more detail The small-scale reproduction gives an excellent overview of French Guiana, clearly revealing drainage patterns and regional scale geomorphology. Because the initial images were not georeferenced, slight geometric distortion remains, in addition to terrain effects. Although these distortion effects have little impact on thematic interpretation, they do limit the mosaic's cartographic accuracy, hence its use as a map proper.

Mosaic of French Guiana
Figure 1: Mosaic of 18 ERS-1 Synthetic Aperture Radar images of French Guiana, acquired between April and December 1992: this is the first complete satellite image coverage of this country Iying in the humid tropics. Radiometric homogenisation has been performed on the inland forest, thus leading to discontinuities between frames on the ocean. Natural frontiers are clearly visible: Maroni river West border with Surinam, Oyapock river East border with Brazil and Mount Tumuc-Humac border with Brazil. At the ocean-continent boundary, the black spots correspond to smooth mudbanks emerged at low tide, and the light spots indicate eroded areas with dead trees on the ground. One can see at short distance from the coast, the Iles du Salut, and on land, the town of Kourou and the Guiana Space Centre from which ERS-1 was launched in July 1991. (Copyright: Télécom Paris-UPMC).

Below we present a synthesis of the results of thematic interpretation studies focusing on coastal dynamics, hydrology, topography, geology and general mapping.

Coastal dynamics

From the Oyapock river in the East to the Maroni to the West, the French Guiana coastline is 320 km in length. The coast is low, swampy and fringed by extensive mangrove forests. The entire coast is strongly affected by, the load dispersal system associated with the Amazon river, which has its mouth some 500 km to the East. Each year, the Amazon pours around one billion tonnes of sediment into the ocean. The suspended load reaching the French Guiana coast is estimated at around 10% of this total. This results in significant changes in the coastline. While silt accumulates in some areas (Fig. 2), it is eroded in others (Figs. 3 & 4). Overall, the series of five major mudbanks currently established along the French Guiana coast migrates westward at an average rate of around 1 km/year. Analysis of the first ERS-1 images (Rudant et al) revealed that Cband radar images are suitable for monitoring coastal dynamics and for quantifying certain associated phenomena, particularly the overall east-west migration of mudbanks, the area of sedimentation zones (soft smooth mud) visible at low tide, and the speed at which the coast is advancing or retreating in sedimentation and erosion zones respectively. Sedimentation zones can be spatially discriminated in ERS-1 imagery by the different types of vegetation they carry (Baltzer et al). This gives clues to the physicochemical properties of the different types of mud and silt occurring in these zones. Analysis of imagery acquired by airborne imaging radars as part of the SAREX campaign in 1992 confirmed the usefulness of this type of imagery for studying coastal environments in tropical regions (Prost et al, Mougin & a/.). The images acquired during this aerial campaign offer higher resolution than ERS-1 radar images, making its easier to identify morphological units and to map the various types of swamp vegetation.

Inlet Silting Up
Figure 2: Crique-Fouillée, an inlet South-West of Cayenne, in the process of silting up. A young mangrove progresses on a mudbank. This picture, taken facing the Sun, shows tidal erosion channels. Smooth, soft, emerged silt zones appear black: we can notice five of them on the mosaic of Figure 1.

Coastal Erosion
Figure 3: Coastal erosion West of Cayenne. Mangrove Iying on the ground (Avicennia trunks, about 20 m) produce strong backscattering, showing up as bright spots on the ERS-1 images and on Figure 1.

Hydrology

Before the ERS-1 campaign, no usable high-resolution remote sensing imagery had been acquired of French Guiana during the wet season. The ERS-1 images acquired in May and June 1992 (i.e. during the extremely cloudy wet season) represent an important contribution to the study of both the smaller drainage basins along the coast and the larger ones further inland. The images clearly reveal the general morphology of the larger inland basins and the location of rock outcrops and associated cataracts that severely limit the traffic on the country's major rivers (Fig. 5). For the smaller coastal basins, ERS-1 images improve our knowledge of the coastal morphology, the annual extent of open water and the types of vegetation cover in swamp zones (Fig. 6). This information has enabled one team (Lointier & al) to put forward a surface water circulation scheme.

Topography and forest facies

Inland of the coastal plain, the moderate slopes (less than 20°) and the dense and highly uniform forest cover make the backscattering coefficient virtually independent of the local angle of incidence, thus making it considerably easier to directly interpret the geomorphology. With the exception of a few areas where the forest facies are different for natural reasons or as a result of human activities (e.g. islands of savannah in the midst of dense forest, patches of clearings, scrub and undergrowth, and areas from which materials have been extracted), we were able to check test sites and show that the backscattering corresponding to a given image pixel could be considered as proportional to the corresponding ground area. This means that radiometric variations can be transformed into slope effects. The integration of these effects can then be used to determine terrain elevations to a satisfactory accuracy using the basic principles of the radarclinometer method. Complementary data were derived from the differences in angles of incidence for a given ground patch imaged as part of overlapping swaths. This enabled us to use radargrammetric techniques for ridges with favourable orientations (Tonon & al).

Geology

Because the high sensitivity of image radiometry to slope effects is unaffected by the presence of vegetation, radar images render geomorphological signatures particularly well. It is thus relatively simple, for instance, to map basalt dikes associated with the South Atlantic opening. Certain shear zones can also be followed across the PreCambrian basement. This is striking since in the field these only appear as occasional outcrops where they cross rivers (Deroin et al) (Fig. 7).

Coastal Erosion
Figure 4: Coastal erosion at the mouth of Sinnamary river. This dead mangrove area is bright on the image due to its high roughness. Analysis of sediment cores (taken here between O and 2 m deep) enables the understanding of spatial and temporal chemical evolution of mudbanks as well as associated vegetabon.

Rock Outcrop
Figure 5: Rock outcrop and associated cataract limiting navigation on Oyapock river at 'Saut Maripa', on the border between French Guiana and Brazil. The biggest cataracts are shown on the mosaic, despite degraded resolution.

General mapping

The excellent orbital stability of the ERS-1 spacecraft ensures the regular acquisition of images of any given area under closely similar conditions as regards the imaging geometry. Repeat coverage imagery can thus be used to reduce the random signal component due to speckle without increasing the pixel size. This filtering method significantly improves the effective resolution of the imagery for all areas where variations in the imaged ground cover due to the cover itself are minimal. Application of this method to three ERS-1 radar images of the Kourou region yielded iconomaps at 1:100 000 (Pénicand et al ). If more scenes are used in this way, it should be possible to further increase mapping capabilities, hence final map scales. In a time-dependent landscape, information extracted from a multidate colour composite image may be richer, but also more difficult to interpret. In the case, for instance, of images acquired from the coastal line at different tides, one can see bathymetric limits corresponding to the water levels existing at the moment of the satellite pass. This results from the physical properties of recent mud deposits the smooth aspect of which makes them appear black on radar images when taken individually (Fig.8).

Conclusion

For countries in the humid tropics, ERS-1 radar imagery opens a wide and varied range of applications. The work presented here will be followed by new studies addressing the ecology and the development of French Guiana. During 1994, we hope to follow the filling of the Petit Saut dam, then to monitor aquatic vegetation dynamics and specifically the anticipated development of surface vegetation on this artificial lake. We also plan to extend the geographical area under investigation to the neighbouring coastal regions of Amapá and Pará states, Brazil. This will enable us to turn the experience acquired in working with imagery of French Guiana to maximum account in a regional context in collaboration with our Brazilian colleagues. Another study focusing on a tropical area is in progress in collaboration with the University of Dakar, Senegal and ORSTOM (Institut (francais) de Recherche Scientifique pour le Développement en Coopération). This concerns the West African coast between the Senegal and Saloum rivers. The aim here is to contribute to the implementation of a coastal development plan tailored to the local environmental conditions. The work presented here was undertaken as part of an ERS-1 pilot experiment entitled 'Coastal and Fluvial Environment of French Guiana' joint project of BRGM-ORSTOM-UPMC with the participation of the following organisations: ADAT, Armée de Terre, CESR, CNES, CRPE, DGA/CEGN, ENST, JRC/IRSA, MRT, French Ministry of Environment and University of Orsay.

References

References quoted here refer to the Proceedings of:

Baltzer F, Plaziat J C, Prost M T, Rudant J P & Dechambre M: Use of SAREX, ERS and Erasme data for a comparison between geochemical measurements on interstitial waters and vegetation distribution in equatorial littoral swamp sediments: an example from Marais Leblond, French Guyana. SAREX 92 Workshop, ESA WPP-76.

Deroin J P, Braux C, Cautru J P, Rudant J P. & Toux L: ERS-1 SAR: an aid for geological mapping and mineral exploration in rain forest areas. ESA SP-361, pp 897-902.

Lointier M, Rudant J P, Sabatier D, Prost M T, Nezry E, Degrandi G, Conway J & Sieber A J: Contribution of ERS-1 SAR data to hydrologic approach in tropical area: example in French Guyana. ESA-SP 361, pp 12671 272.

Mougin E, Lopes A, Hery P, Marty G, Le Toan T, Fromard F & Rudant J P: Multifrequency and multipolarisation observations on mangrove forests of French Guyana during SAREX-92 experiment. Final Results Workshop, ESA WPP-76.

Pénicand C, Champseix N, Ledez C, Lentz C & Rudant J P: Utilisation de l'imagerie satellitaire radar pour la cartographie de base en milieu tropical, 5th Eur. Conf. EGIS, Paris 29 Mar 1 April 94.

Prost M T, Baltzer F, Rudant J P & Dechambre M: Using SAREX and ERASME imagery for coastal studies in French Guyana: example of the Kaw swamp. SAREX 92 Workshop, ESA WPP-76.

Rosaz J M, Maitre H & Rudant J P: Mosaicking ERS-1 images: difficulties, solutions and results on the French Guyana. ESA SP-361, pp. 1221-1226.

Rudant J P, Cautru J P, Lointier M, Charron C, Dechambre M, Deffontaines B, Deroin J P, Prost M T, Raymond D & Vanderhaeghe O: First results of ERS-1 SAR data analysis, pilot project PPF12: Coastal and fluvial environment in French Guiana. ESA SP-359, pp. 835-845.

Rudant J P, Lointier M, Prost M T, Cautru J P, Deroin J P, Sabatier D, Girault D, Tonon M, Nezry E, Degrandi G & Sieber A J: ERS-1 SAR data analysis of the French Guyana coastal plain: some thematic aspects. ESA SP-361, pp. 953-960.

Tonon M, Rudant J P, Sabatier D, Nezry E, Degrandi G & Sieber A: Relief estimation in tropical context (French Guyana) with SAR-ERS-1 imagery. ESA SP-361, pp. 1261-1268.

Wet Season Flooding
Figure 6: Swamp during wet season flooding, Northwest of Kourou, images of the same period allow estimation of the extension of free water, which wasn't yet possible with optical remote sensing.

Examining Outcrop
Figure 7: A geologist examines an outcrop on the bank of the Approuague river. Rivers are privileged observation sites, especially when crossing geological features (shown on ERS-1 images).

Composite of Kourou
Figure 8: Extract from a multidate colour composite of Kourou and surroundings (17 April in red, 22 May in green and 18 December in blue). From East to West (32 km) one can see the western part of the Kourou spaceport and a coastal area with extensive silt accumulation. The different outlines of the mudbank correspond to different tides (0.65, 2.25 and 2.65 m). Multitemporal information allows precise data on spatial extension and geometry of littoral mudbanks as well as on the importance of erosion on the littoral (very bright areas west of the mudbank). On the coastal plain, dark blue spots are swamp zones, flooded in rain season (low radiometry in April and May), and emerged in dry season (high radiometry in December), red spots on the littoral correspond to tidal effects and, in land, to changes in vegetation cover. Areas covered with savannah are strongly backscattered in all seasons and appear in light grey. (Copyright: DGA/CEGN - UPMC).

Photographs of the sites taken by the author.


About | Search | Feedback

Right Up Home ESA EOQ Nr. 44.
Published June 1994.
Developed by ESA-ESRIN ID/D.