Abstracts
Résumé
Le Projet, « suivi de la végétation en zone tropicale » de l'Institut des Applications de la Télédétection au Centre Commun de Recherche des Communautés Européennes d'lspra, a mis au point et alimente en permanence une banque de données satellitaires NOAA/AVHRR. On présente ici la méthodologie développée pour une utilisation de cette banque de données, à des fins hydrologiques, et les résultats d'une première application dans le cadre d'une étude menée en commun, par le Projet et le Département des Eaux Continentales de I'ORSTOM.
Cette étude réalisée sur la Falème, affluent du fleuve Sénégal, le Haut Niger et le Konkoure donne des résultats encourageants quant à la possibilité d'utiliser des descripteurs d'états de la surface, tirés de la banque de données NOAA/AVHRR, comme données d'entrée pour la modélisation hydrologique, au pas de temps mensuel, des grands bassins de l'Ouest Africain.
Il serait peut-être ainsi possible de suivre les modifications des régimes hydrologiques de ces grands fleuves liées à des modifications éventuelles, naturelles ou anthropiques, de leur environnement.
Mots-clés:
- Grands bassins versants,
- télédétection,
- régime hydrologique,
- Afrique de l'Ouest,
- modélisation,
- couvert végétal
Abstract
The project « Monitoring Tropical Vegetation » (MTV) of the Institute for Remote Sensing Applications-Joint Research Center of the European Communities, Ispra Establishment, has created and continually updated a NOAA:AVHRR Imagery data bank. We describe here the proposed methodology to use this data bank for hydrological purposes and the results of an initial application, within a joint research, performed by the MTV Project and the Department for Continental Waters of ORSTOM.
The knowledge and the monitoring of the hydrological regimes of large watersheds, especially in the inter-tropical zone, are obviously major topics for hydrology. Interest shown by the scientific community of hydrologists catches up with the goals that are environment protection and « reasonable » sustainable socioeconomic development of these inter-tropical environments, for which a good understanding of the hydrological cycle and water balance is necessary.
A lot of studies, which have already been carried out in order to reach these issues, very often include rainfall runoff relationship modeling, about which it is well known that there is a strong influence of surface conditions. But if many research program dealt with that particular point, few things in fact have been done on small scale basis, for large river basins. Taking into account the kind of information needed for the characterization of surface conditions at the required time and space perception levels (i.e. seasonal and interannual dynamics, large geographical expanses) and due to the fact that it was possible to utilize classical tools and techniques for data collection, satellite-based remote Sensing techniques were obviously the only ones which could contribute to start in answering the question.
The aim of this study is to demonstrate that a real possibility exists to obtain, by an appropriate use of the NOAA imagery, relevant informations related to the time and space dynamics of the surface conditions, (more precisely about eh evolution of the vegetation cover status), which can be used as quantitative inputs for hydrological models for the West African large watersheds.
This study has been conducted through three phases.
First, five groups of basins have been selected along a north to south transect which intersects pile all the major ecological systems of this part of the African continent, from the soudano-sahelian zone to the guiean one. Using scientific and technical criteria, but also taking into account the fact that some of this basins are of particular interest for development projects conducted by the EC in West Africa, 3 watersheds have been selected. The Falkeme river, a tributary of the Senegal river, the Upper Niger and the Konkoure. The main characteristics of these basins are given in the text.
Then, we have defined the methodological approach with 3 major steps :
- creation of date banks for runoff and rainfall, satellite time series and cartographic information (basin boundaries),
- extraction, from the satellite imagery, of the indexes we selected to describe the surface conditions,
- joint analysis of hydrological and remotely sensed data for 2 years with important hydrological contrasts.
The hydrological data bank contains the monthly mean discharges and rainfall for 5 selected basins, from year 1980/1981 to year 1988/1989 (hydrological year).
The creation of the satellite data bank started with the preparation of a NOAA-AVHRR HRPT time series including the years 1987, 1988 and a part of 1989. Each of the image we choose (no clouds or as less as possible and with orbital characteristics allowing a 3° W flight over the Equator in order to mitigate the image distortion) has been processed using a processing chain (installed al MTV project).
Radiometric corrections have been made to avoid calibration differences between the AVHRR sensors of NOAA 7, 9 and 11.
The computation of the derived channels is done through the transformation of the original data provided by the AVHRR sensor into values which are easier to interpret in terms of surface conditions.
Three radiometric indexes have been kept :
- the vegetation index (NDVI), computed by combination of records in the red and mean infrared bands, which is a good indicator of the vegetal chlorophilian caver on the land surface (Townsed and Justice, 1986),
- the surface temperature (TS), computed by the « split window method, using the brightness temperatures in the AVHRR 4 and 5 channels,
- the signal, in numerical count, recorded by AVHRR 3 channel. This mean infra-red channel (3.55-3.93 µm) is of great interest. First it records the radiation emitted from the surface and, being very sensitive to the high temperature emissions, is very useful for bush fires detection. Secondly, it records a reflected composent which gives information on clearings within vegetated surfaces. By combining the 2 signals, it is possible to detect the bush fires which are known to affect, each year, large extents of vegetation cover, and to made the differentiation between different vegetation covers, particularly in forest areas or forest-savannah transition zone.
Resulting images have been geometrically corrected and resampling of radiometric signals done by cubic convolution method. The final product is a 5 bands image (TS, NDVI, albedo, atmospherical water contents and channel 3 inverted) of 2400 x 1600 pixels (pixel size is 1 km), with geometrical rectification and centered 7° W, 10° N. It must be borne in mind that the radiometric signals are not corrected for the atmospheric effects. That explains why we are rather more interested in the evolution in time of the indexes values than in their absolute values. From these 5 channels images, sub-images of 500 x 500 km are extracted to cover areas of interest for the study of the selected river basins.
The graphic data base contains the digitized basins boundaries. These files have been geometrically corrected using reference points, both on the 1/1 000 000 map and on the NOAA:AVHRR HRPT images, in order to be superimposable to these images. After that, the basins are located on the images and using the CHIPS software statistical descriptors of watersheds surface characteristics are derived.
In a third step, we have applied our methodology to the Faleme river basin. The main hydrological data for the 4 hydrological stations within the basin are given in the text. 1987/1988 an 1988/1989 years has been compared among to each others and to the long term average values. One can remark that, as far as rainfall is concerned, the selected years are very close to the long term average annual rainfall, while runoff during 1987/1988 is clearly lower than the long terra average discharge, and that is true for all the basins. Runoff increase for 1988/1989, compared for the year before, is more than 100 %, while annual rainfall are quite equivalent. That can he explained by a difference in the time distribution of rainfall during these two hydrological years. In fact it is observed that runoff for 1988/1989 starts and ceases earlier than for 1987/1988. So the high values of rainfall at the very beginning of the rainy season 1988/1989 have been propitious to runoff, whereas a more homogeneous rainfall repartition during the rainy season, the year before, has given a higher rate of infiltration and probably an easiest use of water by vegetation.
These differences in terms of total annual rainfall and in time distribution along the year, should have bar, an effect on vegetation dynamic for the two years, and subsequently on the radiometric responses of the surface.
In fact, analysis of dynamics of the signal in the NDVI and the medium infrared band shows that :
- a minimum is reached for the NDVI value during February for 1987/1988, December for 1988/1989,
- a clear shifting between the curves related of the different basins. At the same date, the mean value is always higher for the upstream parts of the basin, with a denser vegetation cover, than for lower part of the basin integrates downstream flat areas,
- a very low mean value for the NDVI as early as December 1988, lower than 0.10, which is to be compared to the threshold value 0.05 given as the normal value corresponding to a very sparse herbaceous cover (1 % coverage),
- much steeper slopes immediately at the end of the rainy season and beginning of the dry season 1988/1989 than for 1987/1988. This indicates a much faster drying up process for the herbaceous stratum.
When looking at the dynamics of the signal in the mean infra-red band, one can see that for the Faleme basin at Kidira during the year 1987 the shape of the distribution histogram changes as we are sinking deeper into the dry season. Starting from a bimodal form it reaches finally an unimodal one. This phenomenon indicates a gradual drying up of the entire basin with as a result a homogenization of the radiometric values.
A date la date comparison of the 1987 and 1988 histograms shows that, as early as November 1988, the distribution is unimodal with a mean value higher than in 1987. The fact that homogeneity kept on during the warn-up period indicates that the main part of the vegetation cover, and particularly the herbaceous stratum, was already dry in November 1988.
All these observations tend to show that it is possible to perceive, on one hand space-time distribution of rainfall and on the other hand the relative importance of grass in the vegetation.
Finally this study has been extended to the Upper Niger and the Konkoure basins. All the above observations related to the Faleme river basin are corroborated by the temporal evolution of the standard deviation values of NDVI.
In conclusion, we can say that this study bears encouraging results as regard as to the possible use of descriptive indicators of the surface conditions derived from the NORA AVHRR data bank, as inputs for the hydrological models, on a monthly basis, of large West African watersheds.
Thus, it may be possible to evaluate the changes in these large rivers regimes, in relation with changes, either natural or due to human activities affecting their environment.
Keywords:
- Large watersheds,
- remote sensing,
- hydrological regime,
- NORA-AVHRR,
- West Africa,
- modeling,
- vegetation canopy
Download the article in PDF to read it.
Download