Apport du canal moyen infrarouge (1,55-1,75 µm) à la cartographie géologique en milieu couvert : la région d'Elven, Bretagne

Middle infrared wavelength (1.55-1.75 µm) contribution to geological mapping below soil cover: the Elven area, Brittany
G. Delpont, J.-F. Desprats, F. Girault
Numéro revue: 
Numéro article: 


En région Couverte peu affleurante et à morphologie peu contrastée, la cartographie géologique nécessite, du fait de la faible densité des affleurements, l'utilisation de données d'observation indirecte de la surface terrestre. A la géophysique et la géochimie, classiquement utilisées quand existent des études antérieures, s'est ajoutée, ces dernières années, la télédétection dont les données sont comparativement peu onéreuses et facilement accessibles. L'essentiel de l'information extraite de ces données est, le plus souvent, relative à la morphologie mais l'étude de la radiométrie des sols nus peut venir compléter utilement cette information. En effet, la relation existant entre les roches mères et les sols surincombants peut être considérée comme assez étroite, hormis dans le cas d'accumulation de formations superficielles exogènes. Dans la région d'Elven, en Bretagne centrale, le canal moyen infra-rouge (1,55-1,75 µm) du capteur Landsat Thematic Mapper (TM) permet de différencier des pointements leucogranitiques et des dépots pliocènes récents de leur environnement schiste-gréseux briovériens et de cartographier leurs limites. Cette différenciation est effectuée grâce à une analyse des variations de la radiométrie des sols nus qui recouvrent les formations géologiques sous-jacentes. Les variations sont corrélées à celles de leur taux d'humidité résiduelle qui s'avère fortement contrôlé par la composition granulométrique des sols. Le rapport argiles/éléments grossiers, qui dépend en grande partie de la composition de la roche mère, peut être considéré, avant la matière organique, comme le facteur primordial déterminant l'humidité, et donc la radiométrie. La précision des limites géologiques obtenues par l'interprétation de telles images est comprise entre 0 et 50 m dans 62 % des cas (124 points test) et entre 0 et 100 m dans 73 % des cas. Le niveau de précision est compatible avec l'ambition de ce type d'étude qui est de fournir un document de travail amont fiable au géologue cartographe. L'intérêt de ce type d'approche sera d'autant plus grand que l'infrastructure géologique du pays étudié sera limitée.


The Brittany region is a peneplaned sub-horizontal plateau that was re-incised by recent regressive erosion. This type of morphology, where rare outcrops combine with intense agricultural landuse, renders geological mapping very difficult, and most knowledge of the geological substratum must be deduced from surface indications. The indirect observation of the Earth's surface, using geochemistry or geophysics, thus represents an important source of alternative information. In this context, remote sensing forms a valuable tool for obtaining preliminary field information, as well as being a guide for certain types of mineral exploration and in some cases even an instrument for final compilation. Morphological analysis generally contributes the most pertinent geological information that is contained in a satellite image. However the data from the Thematic Mapper (TM) sensor carried by Landsat, in particular those obtained in the Middle infra-red (MIR) band, enable a differentiation based on the spectral response of bare soil, which is caused by variation in the soil-moisture content. Apparently a relationship exists between the physico-chemical composition of autochthonous soil and that of the subsurface. The present paper confirms the validity of this approach. Geographic and geological setting, and existing maps. The study area lies to the southwest of Rennes, in central Brittany, and has a mild and humid temperate climate. Agriculture is mostly based on animal husbandry and fodder crops. Geologically (fig. 1), the area lies at the contact between the Central Armorican Domain (DCA) and the Ligerian-Vendean Domain (DLV), bounded by the northern branch of the South Armorican Shearzone [Cassard and Chantraine, 1990; Braux et al., 1991] In the north, it consists of Brioverian schist-sandstone that has been injected by Hercynian leucogranite (DCA), whereas to the south the study area is underlain by Paleozoic schist/sandstone and the Landes de Lanvaux granite of the DLV unit. Locally, outcrops of Pliocene sand and gravel are known. Geological maps of the area mostly are the old 1:80.000-scale edition, which was the basis of evaluating the interest of using satellite images. The relationship between reflectance of soil and its moisture content. The relationship between reflectance intensity and moisture content of a soil has been the subject of much work [Stoner and Baumgardner, 1981; Louhala, 1987; Courreau, 1989; Escadafal, 1989; Bedidi, 1992; Razia Garcia, 1992]. Their general conclusion is that an increase in humidity Will cause a decrease in reflectance for all wavelenghts. However, the interpretation in the MIR band is the most direct, in view of an intensification of water-absorption bands around 1.45 and 1.95 µm. Image used, choice of optimum spectral band and main treatment steps. The TM image 202-27 used was recorded on 11 March 1987 after a rather dry period, thus leading to hydrous contrast that in principle should be good. This favoured the distinction between light, sundried soil and heavy soil that is still humid. The spectral band was chosen after studying histograms of the image which showed that Band 5 had better dynamics than Band 7. Most of the study was thus based on the potential interest of this wavelenght. After geometric correction, the image was restituted as a false-colour photographic document, combining bands 4 (near infrared), 5 (middle infrared) and 2 (green), and which forms the main support of visual analysis of the images. The relevance of Band 5 (MIR) for the differentiation of soil was confirmed by various classifications and by the image of the 5/2 ratio that also turned out to be quite suitable for discrimination. Method of image analysis. The images resulting from the various approaches were visually analysed. Comparison with the available geological maps indicated several points that Will be discussed hereafter. TM image 452 (pl. 1a). This image shows good geographical correlation between plots that are green (in principle dry soil) and blue (in principle moist soil), and, respectively, granite and Brioverian rock as shown on the existing geological map. The image also reveals many detail differences between the outlines of these geological units as seen on image (fig. 2a) and map (fig. 2b). However, it cannot be said that there is a clear correlation between colours, i.e. moisture content and radiometry of soils, and geological formations. Northeast of the Lizio granite, for instance, the green plots correspond to Pliocene sands and gravel deposits, the sun-drying of whose light soil cover should take about as long as a soil over granite. The granite also shows several anomalies, e.g. green bare soil over Paleozoic schist and blue bare soil over granite, which may be due to local variations in vegetation and/or morphology. Classification and ratio-type processing (pl. 1b and 1c). Such processing does not enable the continuous tracing of boundaries between rock types, but confirms the main lines of the interpretation. Classification and ratio show up granite within schist formations, and also indicate anomalies that correspond to "humid" soil within granite (green and blue patches), or "dry" soil (mauve and red patches) within Brioverian or Paleozoic schist. They also reveal the Pliocene formations in the northeast corner of the image. Radiometry and soil composition. Soil analysis of the radiometrically most characteristic plots within the various schist and granite units explains several of the observed phenomena. A total of 24 samples were taken and investigated for their grainsize (5 fractions, including clay), equivalent moisture, organic matter and iron content. After grouping for each sample of the mineral elements that can cause high water retention, i.e. clay and fine silt particles of < 20 µm, the results were analysed. This consisted in comparing the control of moisture factors on effective soil moisture, as well as the latter's influence and that of iron on the resulting radiometry. This is plotted in graphs (fig. 3) that require some comments: - Soil over granite is grouped under low moisture values, whereas soil over schist has high moisture values (fig. 3a), this distribution may be due to the higher clay content of the latter soil type. - Adding organic matter as a moisture factor (fig. 3b) does not fundamentally change soil dispersion, which indicates that the influence of subsurface rock types is dominant. - Comparison of numerical values of the analysed criteria and those of radiometry of the corresponding plots on the image, shows a continuous decrease in dispersion of Band 5 (fig. 3e) and Band 6 (fig. 3f), passing through bands 4 and 2 (figs. 3d and 3c). This shows that the relationship between soil moisture and radiometry can be used at least locally for lithological discrimination, provided this is supported by detailed field calibration. - Finally, the ratio TM5/TM2, compared to moisture factors (fig. 3g and 3h), leads to a very clear grouping of soil over granite for high TM5 and low TM2 values, and to an inverse ratio over schist. Recognition of units mapped by remote sensing. Soil analyses thus confirm that radiometric data extracted from satellite images especially from the MIR Band, correlate quite well with the degree of moisture of bare soils on the date the image was taken, and that such information can be used for distinguishing granite from surrounding schist. Classification and ratio show the presence, through low moisture content, of Pliocene detrital deposits, whose composition would justify a moisture content close to that of soil over granite. Radiometrically "dry" soil is found within the Brioverian without being allochthonous. This might betray the presence of more sandy patches in such soil, or of fractures and veins filled by quartz or igneous rock. However, it should not be forgotten that this type of approach is based on statistics, and that the variability of natural features can create anomalies that are not always easy to correlate with geological phenomena. Field validation. Validation was based on 124 observation points, which were defined in a random manner such as road intersections with image and map boundaries. The reliability of the interpreted satellite data was evaluated by comparing them with the existing geological map. For each observation point, the distances between geological and image-interpretation boundaries, and between geological and map boundaries were measured under the same conditions. These values were split into five classes: 0-50 m, 51-100 m, 101-150 m, 151-200 m and > 201 m. The first of these corresponds to the order of precision of geometrical image correction (1.5 pixels, or 45 m), and the second to that of field mapping in Brittany, where outcrops are rare. Table 1 shows the distribution of these distances and confirms that, in this case, a better accuracy can be expected from the image than from the existing 1:80,000-scale map. Of the geological boundaries seen on the map, 46% can be considered as precise, i.e. < 100 m, against 73% of the boundaries deriving from image interpretation; the strongest differences, i.e. > 200 m, decrease by 60% for image boundaries. The fieldwork provided the following information :- Visual interpretation of the false colour image 452, which shows the greatest variation, provides the best picture of geological reality. - The heterogeneous land-use pattern is a major limiting factor for the mapping of geological formations on the basis of radiometric soil analysis. - Recognition of small or fragmented plots is not easy because of the difficult recognition of local radiometric variations that are unrelated to geology, e.g. the Pliocene sand and gravel deposits. Conclusions. Fine morphological analysis of covered ground from airborne or satellite data generally is the main contribution of remote sensing to geological mapping. Nevertheless, this seems to be usefully complemented by radiometric data in those cases where a relationship can be established between the spectral response of bare soil and underlying geological formations. In our example of the Elven area in central Brittany, the medium-infrared (1.55-1.75 µm) band, or Band 5. of the Landsat V Thematic-Mapper sensor enables the zoning of bare soil according to its moisture content. This zoning, which is largely dependent upon the type of weathering product of the underlying bedrock, provides the mapping geologist with a predictive geological sketch map of the area of interest. After field verification, this can be used for guiding and speeding up detailed mapping, as the degree of accuracy of the geological boundaries thus obtained is compatible with the scale of this type of work. This approach does not explain all radiometric variations, but, as an early investigation method and in combination with morphological interpretation, the method seems to be a valuable source of indirect information that Will help the geologist who works in areas with few outcrops.

Dernière mise à jour le 28.07.2015