The objective of the workshop was to assess the current status of application of spatial information to the management of water resources in arid and semi-arid developing countries. The presentations included in this publication do not pretend to exhaustively cover all issues relating the use of spatial information systems to the management of water resources in developing countries. However, the richness and the variety of papers prove the interest in hydrological applications related to analysis of spatial information that exists in these countries.
The presentations and the discussions of this workshop were centred mostly upon the utilization of spatial data obtained from satellite platforms (low resolution NOAA, high resolution SPOT and LANDSAT), aerial photography and digital elevation models (DEM). Geographic information systems (GIS) were also covered in many presentations, either for specific applications or for the analysis of data obtained from remote sensing. GIS are used for the management of spatial information, but also as information and decision-making tools through simple (data combination, data aggregation, intersection of overlays) or complex operations (hydrological modelling).
A classification by objective effectively summarizes the two main themes of the workshop:
� objectives related to global or local knowledge of the earth surface for an improved management of water resources, which may also be described as an attempt to describe in order to manage better, particularly in the short term.
� objectives related to modelling, comprehension and analysis of hydrologic events, which consists in describing for understanding and forecasting: quantifying, classifying and validating.
| Themes | Number of presentations |
| Management of water resources | |
| � regional approach | 8 |
| � local approach | 6 |
| Hydrological modelling | |
| � hydrological characteristics of soil | 8 |
| � water balance: evapotranspiration | 4 |
| � hydrological models of surface runoff | 7 |
| Total for the workshop | 33 |
These two main themes are discussed below. The current limitations and possible improve-ments underlined in the different papers and during the discussion are also presented. The table shows the distribution of presentations as a function of these two themes.
A first group of presentations covered the application of remote sensing and GIS on a regional or local scale and provided various examples of the application of these techniques in an operational framework. The examples illustrate a wide range of possibilities: global knowledge (on the African continent scale); assistance to preparation and implementation of master plans; regular monitoring of large study areas or; on the opposite end, information on limited areas for which these techniques appear to be both a reliable and replicable option.
The analysis of the different presentations shows a certain number of objectives assigned to remote sensing and to GIS, and these can be classified as follows:
Cartographic (qualitative and quantitative characterization):
� boundaries between zones (spatial segmentation tools),
� links between zones,
� linear (drainage, fractures) and surface objects, characteristic features that are stable (relief, land use) or variable (seasonal vegetation, humidity, temperature, evapotrans-piration.) over time.
Updated information and monitoring:
� readily available information: maps related to natural hazards,
� change of viewpoint for the description of space: from local to regional information,
� objective and complete information: uniform observations on large surfaces, particularly for areas lacking detailed base maps,
� monitoring of change and evolving phenomena.
Regionalization:
� calculation of regional or global balances,
� extrapolation of local results.
Planning and management:
� inventories, support to the formulation of master plans,
� archive (memory of events) and creation of records,
� management of large units, conflict management.
The description of large land units is related to three levels of information (Lointier):
� accurate mapping of boundaries and segmentation of landscape in homogenous zones. This is the classic use of remotely sensed images for the determination of objects of hydrological significance;
� qualitative mapping of the exchange of superficial and underground water between zones. Objects are characterized by their proximity, their relative position and their thematic links:
� mapping of environmental response to specific events, therefore a monitoring approach.
One paper (Aouni) described the analog interpretation of LANDSAT images followed by the analysis of SPOT images as a tool for the inventory of surface water in Tunisia: waters surfaces, dams, hydrologic network, boundaries of watersheds, wetlands, agricultural resources. In humid tropics, to compensate for the continuous presence of cloud cover, the use of radar images appears to be more appropriate, notwithstanding the rather poor information included in a single image. This solution can be integrated with temporal methods.
The importance of satellite images can be measured by comparing the contribution of the images and of all other material available. In the case of tropical zones for which low-quality maps exist, satellite data, even of low resolution, quickly attain a much higher level of interest than other existing data. For areas that have unclear or variable boundaries (such as tropical wetlands), it is difficult to carry out proper classifications and field checks become essential (Lointier).
The utilization of GIS for the description of regional information in order to obtain improved knowledge and better management (including conflicts in the case of limited resources) was also the subject of several presentations (Bousquet et al., Faur�s, Killmayer). The following were presented:
� the use of data obtained from remote sensing or of data from the intersection of multiple overlays within a GIS;
� the relation between hydrological and territorial units (administrative or political boundaries of the land), and the hydraulic exchanges between territorial units
� the integration in a GIS of data from different sources, including remote sensing data. A practical example is given for Senegal (Killmayer). This describes the updating of maps of rice paddies with the objective of large scale monitoring of cropping and management. The approach is simple, operational and proposes a methodology that is now widely used in remote sensing: classification and integration into a GIS for the annual monitoring of crops. What is shown, can be considered to be a first step in the management of water resources, through the elaboration on an annual basis of a monitoring and information tool for rice crops. In the specific case, the cost of remote sensing represents only 10% of the global cost of the monitoring system, which makes it easy to integrate it into the project.
Several local level application were included in the presentations:
The detection of water surfaces
The detection of water surfaces on an optical satellite image (SPOT, Landsat, etc.) is one of the easiest applications of remote sensing: the very weak radiometry of open water surfaces, particularly in the near infrared, gives them a high level of separability if compared to the other themes of the image. One presentation (Ousmane) gives an example of this type of detection in areas with poor cartographic information. There are two main ideas:
- remote sensing can be a powerful tool, on condition that it is used in conjunction with other local topographic and hydrological data: in the example provided, the combination of remote sensing information (SPOT), climatic data (ground station) and a good knowledge of the local topography provides a simple and operational method to calculate the volume of water surfaces;
- remote sensing can be used as a database, making it possible to reconstruct a series of events. This is an original objective which will become increasingly important with the growing availability of chronological images (for the past) and the possibility for the user to select the viewing angle even at short notice (for the future).
The detection of fractures and their hydrological interest
Two presentations described the identification of linear features on the images as an indicator of geological fractures, therefore potential areas for the utilization of groundwater. Two aspects were covered:
- a cartographic aspect: mapping of linear features with LANDSAT images for geological (fractures) and hydrogeological purposes (for the selection of drilling sites). Aerial photography was used routinely in fault areas, but the use of digital satellite imagery is relatively new. Digital analysis of the images makes it possible to use powerful algorithms for searching linear features and reduces, at least partially, the subjective nature of photo-interpretation;
- an analytical aspect: a study of frequencies and of the connection between drilling sites and fractures through the analysis of the productivity of the sites in relation to their distance from the fractures (through GIS) and the direction of the fractures themselves. The detection of the most productive tectonic events facilitates the choice of areas for new drilling sites. This point is particularly interesting, as it shows how these analyses with a GIS, can provide a practical utilization of the results. Water resources can be supplied to people by greatly increasing the chances of success in the search of high yield drilling sites.
The delimitation of watersheds
The use of digital elevation models (DEM) for the hydrological definition of watershed is becoming increasingly important due to the relative ease of obtaining DEM at reasonable cost and with a degree of accuracy that corresponds to that of the maps currently available. Most of the hydrological applications of these DEM are connected to the use algorithms, particularly for the definition of hydrological networks, the boundaries of watersheds, the measure of slopes, etc. These algorithmic application are at the same time very valuable and rather easy to use, a fact that has caused a multiplication of software available on the market. One example is provided in the presentation of Bergaoui and Camus that describes the application of the DEMIURGE programme created by ORSTOM for the definition of erosion potential and hydrological modelling in Tunisia. The DEM is used to define slopes, the hydrological network and the Beven indices. In the case of small and flat catchments, however, most presentations agree that it is difficult to correctly derive the boundaries of the watershed and the hydrological network from the DEM.
The importance of adapting the platform to the objectives
One presentation (Puech and Carette), shows the results obtained from a variety of platforms for local studies in Africa, illustrated by three examples: amateur aerial photography, use of standard aerial photography and SPOT satellite images. It is important to have a good correlation between the objective of the application, the scale and the platform: always avoid privileging the tool over the theme, avoid carrying out excessive digital analysis on the image if it doesn't contain the required information. The cost factor should also be integrated in the choice of the platform to be used. Aerial photography and satellite images are usually a source of complementary information to be used with caution.
Remote sensing and the management of rural space
Another presentation (Patrick) proposes the integration of environmental information obtained from two complementary sources: remote sensing and the indigenous knowledge of local populations. Remote sensing provides a view at a given moment, while local populations integrate the knowledge of local conditions over a number of years. The presentation suggests integrating indigenous knowledge into an expert system. This method, particularly original in its attempt to combine global information with the local knowledge, needs further testing before its feasibility can be evaluated.
In this section, it is possible to identify two main themes: a descriptive and mapping theme, followed by a modelling one. The objectives of the presentations in this section may be described as follows:
Description and mapping
� determination of the hydrological characteristics of soils: quantitative mapping of parameters that vary over time (soil types, vegetation, crops, surface conditions and runoff index);
� determination of the elements of water balance: quantitative mapping of parameters that vary over time (humidity, evapotranspiration, temperature);
� objective division of space into hydrological units.
Hydrological modelling and interpretation
� collection of spatially distributed information to be used as a support for modelling;
� analysis of the connections between environment and its functions: search for descriptors of environmental mechanisms,
� detailed analysis of hydrological behaviour: renewed interest for the internal hydrological functions of watershed.
Hydrological characteristics of soils
A number of presentations described the activities carried out in the framework of an FAO project for the estimation of discharge in Sahelian regions, and particularly the tests carried out on the utilization of remote sensing for the improved estimation of discharge from small watersheds. The surface feature mapping in Sahel (Lamachere and Puech) was also described. These activities start from the assumption that, thanks to a number of peculiarities, the Sahelian region is favourable for the development of remote sensing tools used for assessing the origin of discharges: absence of detailed base maps, limited relief, well defined seasons, and dominance of mainly surface runoff over interflow. Several attempts have been made in this ares to link remote sensing and hydrology since 1985. The method described during the workshop was based upon the following elements:
� elementary hydrological knowledge based on the standard surface conditions of the Sahel, obtained from a synthesis of measures with a rainfall simulators (1 m2 test areas);
� surface feature mapping obtained from satellite imagery (LANDSAT TM) and ground data;
� hydrological validation obtained with field observations on a number of small Sahelian watersheds.
Through the results it is possible to obtain a reasonable map of the potential runoff, that is to say the basic runoff to be expected from each pixel. This map is an important tool for the comparison of watersheds and the eventual hierarchization of basic runoffs in terms of total flow. However, the map of potential runoff is not sufficient to quantify the runoff at the outlet of the watershed. The downstream transfer of runoff and the modelling of flow within water courses still pose a number of problems that are far from being solved. To use the map of runoff potential at the watershed scale, it is currently necessary to use a calibration function. The parameters of this function have been estimated and can be used for calculating the 10-year floods on small watersheds of West Africa.
One presentation covers the use of the rainfall-runoff model elaborated by the Soils Conservation Service of the United States Department of Agriculture (USDA/SCS) through the runoff curve number (CN) to define the conditions of runoff (Colombo and Sarfatti). This is the most traditional method and the simplest utilization of remote sensing for the estimation of runoff conditions. The method, developed in the US, is tested in this case in Eritrea for the estimation of peak discharges and of annual runoff. It includes two phases: a phase of image splitting based on remote sensing and on the assignment of a global parameter to the hydrodynamic behaviour (production coefficient) which is then applied prorata to land cover. The extrapolation of calculation methods poses the problem of choosing the CN coefficient, considering that these coefficients are used in a context that is different from the one for which they were initially developed.
Another presentation also covers the issue of the hydrological response of landscape units that are considered to be homogeneous from the hydrological point of view (Vin�). The landscape segmentation phase refers to "hydro-landscapes", and insists on the hypotheses of models and of the choice of segmentation. The space is divided into landscape units that are grouped according to a limited number of land cover categories. These categories are defined from the image. The choice of categories is based upon the hypothesis of each one having a specific hydrological response function contribution. Putting in relation the global hydrological response of several small watersheds (on a yearly, quarterly and monthly basis) with the different categories of land use, the unit response for each category is obtained through a numeric deconvolution scheme, based upon a disaggregation technique. This original analytical method appears well adapted to the valorization of remote sensing to provide a global value of hydrological response per landscape unit.
All these approaches put the accent on the uses of remote sensing as a tool for the segmentation of space and emphasize the uses of this segmentation of space as the initial phase for the modelling which is made possible by remote sensing, GIS and DEM.
Mapping of parameters that vary over time (water balance)
Remote sensing is also presented as a tool for mapping parameters that vary over time (temperature, humidity) and that are difficult to obtain from the standard network of point measurements. This aspect is particularly interesting as it poses the problem of changing the approach from point measurements to base surface information.
Two written presentations (Hurtado Santi et al., Chebouni et al.) start from thermal infrared maps and demonstrate the use of theoretical models in the exchange balance to approach the actual evapotranspiration (ET) by using the differential between radiative and aerodynamic temperatures. The applications were carried out in semi-arid regions of Arizona, the Sahel and central Spain. Remote sensing is herein described as a tool for managing spatial information, in this case the objective being the production of ET maps at a regional level.
Finally, one presentation (Gineste) covers the contribution of ERS radar data to the estimation of soil humidity in a small watershed in France. The results are insufficient to produce soil humidity maps as the calibration of the images is still a major problem. One the other hand, they can provide a relative vision of the dynamics and of the spatial distribution of humidity on the watershed. The main interest of this approach resides, therefore, in a visualization of temporal and spatial variations of surface moisture defined by hydrological models. These results can be used to validate the models themselves.
Objective segmentation of space - reference to the concept of hydrologically homogenous areas
Streamflow modelling or simply understanding of the hydrological environment systematically refers to the utilization of spatial data as a tool for segmenting land into homogenous areas. Even if often this tool is not defined as such, it is nevertheless present in most of the presentations.
Remote sensing and GIS appear to be the main tool for segmenting space. Practically all the presentations that refer to land use include this segmentation phase. Some only present the segmentation phase and the cartographic results without proceeding to the modelling phase, which is the main link to the environmental mechanisms that are being studied.
Other presentations propose a reflection on the choice and the meaning of homogenous areas: these areas, homogeneous on the images (according to visual aspects), should also correspond to similar functionality according to the theme which is being studied. It is possible to verify (Vin�), that visual homogeneity does not guarantee homogenous behaviour.
Segmented data can then be included in a modelling process that is usually the final objective: segmentation of space is usually only the initial step of modelling applications.
In a spatial approach to streamflow modelling, there is an evident separation into two phases: first the segmentation of space and then the modelling itself. Apart from the implementation of these two key phases, a number of presentations have dwelled on their implementation, including research and methodological aspects. Segmentation of space and distributed modelling have in fact a number of serious problems of validity and interpretation, particularly as they manipulate multiple plains, an operation that is often simple from a computing standpoint, but very uncertain as far as the meaning and the coherence of results are concerned.
Modelling outside GIS
These are calculations carried out outside of a GIS framework, but whose base data are spatial. Some of the results have been described above. The most effective applications are the simplest and the most typical: hydrological balance or SCS (Soil Conservation Service) models in which a runoff coefficient is applied to each land use.
Two presentations (Nonguierma and Dautrebande, Colombo and Scarfatti) presented the applications of simple SCS models. Remote sensing is used for the segmentation into homogeneous areas: an evaluation of CN by areas is proposed, based on the average soil occupation, with the quantification of the runoff calculated at the rate of each homogenous area.
Another presentation (El Idriss and Persoons), uses a simple and effective conceptual framework (linearity, permanence during a storm of production and transfer functions), to simulate the hydrogrammes of the maximum flood events for the calculation of structures. As it uses classes of pixels in a isochrone position in relation to the watershed outlet, the structure of the grid hydrological model (GHM), enables a quick calculation in comparison to classical distributed models. Its application consists in the creation of scenarios to test the influence of the evolution of the physical characteristics of the watershed on the change in flood regime.
The reference to TOPMODEL modelling on the basis of spatial knowledge of relief (Gineste) should also be noted. This is a simplified conceptual model of water tables that proposes a precipitation-discharge transformation based on the notion of variable contribution areas (during the storm), for calculating flood events. The areas can be defined using their saturation index, that is a function of slope and of the surface drained upstream of the considered point. Starting from a digital elevation model (DEM), it is possible to map this index, then calculate its distribution that enters directly into the modelling. The increasingly easy access to DEM encourages the use of this type of models and valorizes the inclusion of relief in hydrological modelling. As a consequence of the hypotheses on the elementary streamflow process, its application is a priori restricted to humid and temperate areas.
Modelling with GIS
Some examples of modelling based on spatial data managed through a GIS are shown both in simple terms (regional hydrological balance constructed with a continental scale GIS, Bousquet et al.) or more complex ones (modelling including production and transfer functions, Faur�s). One presentation (Perez et al.) illustrates the test of a event distributed modelling tool, the ANSWERS model, that runs in a GRASS environment (a GIS software in image mode) to define runoff and erosion in an agricultural catchment. The model requires the initial knowledge of several parameters, several of which are obtained from field information or from remote sensing, while others are obtained by calibration. The application is made simpler by the direct use of information overlays (relief, land cover) in the modelling.
In the combination of overlays obtained from a GIS there is a problem in the interpretation of map results obtained, particularly on the boundaries between homogenous areas. This problem was raised by a number of participants: at the junction of zones, the variations may appear artificially sudden, while in reality, they are usually gradual.
Even though a number of issues related to precision, validation and interpretation of cartographic treatment of data with GIS remain to be solved, the results obtained are usually more precise, more exhaustive and easier to use that the simple compilation of non systematic point measurements. Regardless of the remaining difficulties, the use of GIS is an interesting option for the extension, extrapolation and interpolation of observation for crossing information overlays of differing nature.
In addition to the results themselves, some presentations provided a basis for reflection on more fundamental questions and they concentrate upon the connections between hydrological models and spatial information. The analysis of discrepancies between the models and the observations in Sahelian region emphasizes the limited precision of the latter. Spatial tools make it easier to take into consideration local variations. In particular, remote sensing in its widest sense can be used to explain the differences between watersheds through the representation of soil occupancy or of drainage. It makes it possible to study watersheds in great detail. However, the majority of hydrological models based on remote sensing assume the linearity and the invariability of the process, and this is far from evident.
Remote sensing is presented as a tool for validating hypotheses. Radar images, for example, can be useful for validating or invalidating the hypotheses on the dynamics of moisture in the watershed. Internal behaviour of watershed can be explained by two spatial elements: morphology (which explains the interest of DEM utilization in modelling) and land cover (from where the interest in remote sensing). Several models using on or the other of these tools have been developed at different scales.
The connection between spatial descriptors and hydrological indicators is also an important subject. It shows the difficulties in associating the visual objects available on the image and the objects required for hydrology, particularly because each connection is related to scale: the description of hydrological processes varies with scale. Therefore, the choice of hydrological objects (homogeneous areas) cannot be limited to one aspect. Their size depends on the objectives of the study and should be based on a compromise between the consideration of the physical phenomena underlying hydrological processes and the field data available for the elaboration of models.
In the Sahelian context, each scale needs a specific approach. Mapping runoff potential is possible at the pixel scale in areas where surface runoff prevails (this is the case for the Sahel) and enables a good comparison between watersheds. However, due to the non linearity of hydrological processes and to the relative difficulty of representing them at the different scales, the discharge at the outlet of the watershed can only be obtained by the simple addition of elementary runoff obtained on each pixel. It is necessary, therefore, to work on hydrological objects of larger scale for which runoff measurements do not exist.
Data obtained from remote sensing and DEM now produce an almost continuous flow of spatial descriptions, which makes it possible to analyse the internal variability of watersheds and to utilize distributed models to simulate its hydrological functions. However, this shows the difficulties in applying these models: the current knowledge of internal processes of watersheds is insufficient to make it possible to use data from remote sensing and DEM in a useful manner. In fact, hydrology, which was developed before the introduction of technologies allowing for the collection and processing of spatial information, is poorly adapted to the efficient use of the full potential or remote sensing and GIS. The empirical methods development for the estimation of peak (flood) discharge and annual contributions, barely use spatially distributed information, and generally only in a statistical form. If the best possible use is to be made of georeferenced information, it is important to undertake the development of new hydrological methods.
At the same time, there is the problem of choosing a model: a physical model based on spatial data, difficult to calibrate and therefore basically unstable, or a conceptual parametric model which is easier to calibrate but in which it is harder to explain the physical meaning of the parameters.
All the results presented above have been positive and encouraging. However, the use of spatial techniques also includes a series of bottlenecks, of failures even. Very important lessons can be drawn from the analysis of these failures. They help to reflect and are very useful for the identification of the various constraints. The different presentations and discussions have raised a number of problems that could explain these difficulties:
Technical problems:
� vision limited to the ground surface
Problems related to inappropriate use of tools:
� attempts to develop universal tools that is in contrast with the impossibility of correctly modelling natural phenomena in their geographical variability;
� underestimation of the role of validation and field measurements;
� use of one technique instead of another without verifying complementarity.
Problems related to methodological limitations:
� fast development of computers, associated with a slower development of methodologies and thematic reflections;
� spatial approach that is based upon new ways of thinking that are not adapted to the traditional calculation methods;
� problems of scale and change of scale not properly managed; high sensitivity of indicators to spatial and temporal scales, improper combination of information at multiple scales;
� bad selection of the so-called useful indicators.
The different presentations of the workshop cover on the one side operational results and on the other some more general reflections on the methods for treating spatial information for the study and management of water resources. Among the operational results, we can essentially find simple cartographic applications, with remote sensing and DEM adding new, pertinent and qualitatively measurable cartographic information. The enormous possibilities in terms of knowledge of the environment and of surface quantification of specific parameters are easily understood. A lot of hopes are also vested in the use of GIS for the integrated management of spatial data. On the other hand, the associated models are still at the research stage.
The applicability of these approaches is in fact highly variable and can be judged through a number of criteria such as:
� region of application (the value increases in little or poorly known environments);
� previous thematic knowledge for validation (importance of field measurements and of observation on experimental watersheds);
� the required precision, determined by the objective of the application;
� the ease of application (simple vs. complex, operational vs. research);
� combined use of remote sensing, GIS and of classic information sources.
The difficulties of using these new tools and these new data appear less and less connected to the techniques themselves. Through the spectacular evolution of these tools, of numeric data and of available software, cartography and numerical analysis have experienced a remarkable development. Digital maps produce documents of proven quality and with such improvements to cause a growing unbalance between techniques and conceptualization. The main problem is not the production of maps, but their analysis, their use and their interpretation for thematic objectives.
The use of remote sensing and of GIS for the study of water resources makes it possible to integrate elements of the landscape into calculation algorithms in an increasingly detailed fashion. Their efficient use cannot be obtained other than with a schematization of reality. Nowadays, the main problems reside in this schematization or modelization. It is now essential to orient research on the utilization of space through the notion of objects and the relationship between objects.
As far as remote sensing per se is concerned, the improvement in utilization should not only be expected from an improvement in the technical characteristics (for example better resolution), but also from a better use of the available information. Often the potential resolution becomes too fine in relation to the objective and it might be useful to accept some degradation of the image in order to obtain the desired results. The search for the optimal resolution, in relation to the objective of the study should be a more fundamental and systematic preoccupation and become a prerequisite of each application.
Many applications start from the idea that is possible to delimit areas that have homogeneous hydrological behaviours. The difficulties reside not only in tracing the boundaries of these areas but also in their definition. In fact, it is not easy to choose the criteria to be used for the definition of these zones, knowing that homogeneity does not correspond to reality. So, it is possible to match each type of image with a segmentation of space based upon homogeneous units defined by visual characteristics, but the segmentation of space should also have a functional meaning related to the studied processes.
On the other hand, the ease of image processing software and of associated applications presents a certain danger. It is now possible to classify images in an anarchic fashion, without field checks, without control or relation to the thematic objective. It is possible to rapidly intersect, combine without validation or control a large number of overlays with different scales and meanings. There is a real danger of being carried away by the ease of digital maps and not to advance in the field of their rational utilization.
Among the most operational applications of data obtained from remote sensing, integration with GIS appears to be expanding the fastest, be it for the intersection of overlays, as a tool for the spatial management of multiple data or for the support to hydraulic management of a given area. These applications will continue to develop in the coming years as GIS make it possible to elaborate synthetic cartographic documents, that are the basis of discussion between stakeholders and therefore essential tools, for example, for the resolution of conflicts on the use of resources. Now, it is known that with the increased uses of water, the number of regions in which this resource becomes rare also increases. Several communications expressed these preoccupations and provided examples of applications of GIS to the management of water resources in sudano-sahelian catchments (Traor�).
The interest of data from remote sensing is also strongly demonstrated by the mapping of the variable parameters of hydrological models, where only point observations were previously available (moisture, evapotranspiration, temperature). Thus, for certain parameters, there is an evolution in the acquisition of data. Other examples of runoff modelling through spatial models provide encouraging results, notwithstanding all the incertitude that is still associated to these. The specific interest of the new data has been shown particularly by insisting on the association between remote sensing imagery representative of surface conditions and of land cover and the DEM, representing the topographic knowledge of the watersheds.
Finally, the images now offer a surface memory of exceptional events (floods for example). With the increase of the image archives, these can be considered as chronicles of landscape evolution or of hydrological events on the same level as records of discharge. Applications connected to the use of this data will certainly experience important developments in the future.
