Conclusions & Recommendations

Through the above description of ERS-SAR image interpretation, a number of advantages and limitations to the use of radar data for mapping, monitoring and management of natural resources have become evident. The most important are:

• Cartographic maps and land cover maps can be considered one of the basic needs in African countries, to ensure their sustainable regional and national development. The scales of 1:500,000 for evaluations at national level and 1:100,000 for regional planning can be considered as most adequate, containing a relatively small number of classes. This reduced number of land cover classes should be considered at least for those countries that extend geographically over more than one climatic zone.

  ERS-SAR data fulfil these basic mapping requirements. Scales of 1:100,000 to 1:500,000 can be achieved without any major problem while taking the reduced number of classes into account. In order to better identify the different land cover classes, the multitemporal approach has been shown to be the best solution. However, it is not necessary to acquire a large number of images, but only 2 to 3. The images analysed above were taken in July and August, which seems to be a season when a large number of features can be identified. Images taken in other seasons should equally be investigated.

  The processing of the data has been reduced to mandatory step reflection around an axis and the (not mandatory) transfer from 16 to 8 bits. This latter step is specifically recommended as the size of the original data set (120 megabytes) makes it difficult to handle when using less sophisticated hardware. To reduce speckle, an averaging filter, available in almost all image processing software packages currently on the market, was used. The size of the moving box should be chosen based on the image size. For a full scene, a 10 x 10 box is useful for overview purposes. For the interpretation of the images, a computer-aided visual approach was preferred (through digitalization on the screen). Currently available methods for automatic classification were tested for certain images but did not produce any significant improvements.

• Within the application area of Forestry , the capabilities of ERS to detect mangrove forests have to be specifically emphasized. The highly valuable results obtained through a visual inspection of the images allow us to state that mangrove forests can be identified with almost one hundred percent accuracy. Thus, monotemporal data sets seem to be a sufficient support. Confusion in radiometric terms only appears with towns that have a similar grey level (very bright). The identification of tropical forests is more difficult to achieve. Four cases are possible:
  • The region of interest is almost completely covered by forest - the data will show up in an uniform grey. When there are some mountains or hills they can be seen in a quasi-three- dimensional way. Beyond the identification of these forests, no further distinction between different types is possible. It has to be assumed that the mountain ranges are also covered by forests.

  • There are clearings within the forests; in the majority of cases, these clearings can be identified when they are large enough and not covered by bushes. Cuts through shifting cultivation are very difficult to identify as trees normally remain in the fields.

  • Forested areas remain in a region used for agriculture. The identification of the forests largely depends on the crop types cultivated. The higher the crops (bananas, etc.), the lower the chance of distinguishing the forests from the crops (similar reflectance).

  • Forests only remain in some remote locations which are normally mountain ranges. Due to the reflection of the radar signal in hilly terrain, the detection of forests is very difficult. The use of DTM's has been shown to give some good results. One should, however, analyse the benefits as the generation of DTM's is generally very expensive and forest interpretation results are not automatically improved.

• The mapping and monitoring of hydrological phenomena can be considered one of the most adequate application areas for ERS-SAR data. Monitoring of lake levels and extensions as well as flood monitoring have given very good results. Visual interpretation of the images is recommended and sophisticated image-processing equipment is not mandatory.

• Tropical agriculture is very difficult to monitor as crop signatures tend to be confused with the surrounding natural vegetation. The monitoring of rice with ERS-SAR is the only but very important exception as certain stages (flooded fields) give a unique reflection. Swampy areas might be classified as rice. Also here, no complex image interpretation equipment is necessary for the processing of data. Automatic classifications are possible, but algorithms are still under development .

• The use of radar data for urban applications should be assessed very carefully. Due to the resolution of the images (about 20 m), their capabilities for the detection of inner-city roads and distinction of different housing areas is of less interest. An adequate application was, however, found in terms of mapping remote villages and housing areas, which clearly appear as white spots in the radar data.

• The detection of geological features to be used for general purposes as well as the later exploitation of minerals can be considered very worthwhile. Images can be used for overview mapping accompanied by field surveys. Scales around 1:500,000 are considered the most valuable.

In addition to the above points related to selected applications, a number of advantages of a more general nature should not be forgotten. First and foremost, one should consider the ERS-SAR's ability to penetrate cloud cover. With ESA's current and planned missions (ERS-1, ERS-2, Envisat ) this implies the long term availability of data acquired with each satellite pass (every 15 days in general). Unfortunately, the availability of ERS-SAR data over Central African countries is more problematical as no receiving station (apart from the mobile station of DLR) can cover the whole region. The area between the South African station and the one at Maspalomas will remain uncovered (in terms of ERS acquisitions) until a receiving station has been installed in the area or a satellite is capable of recording data on-board (Radarsat and Envisat will have this capacity).

Until then, users in Central Africa will have to rely on data which were acquired in 1994 and 1995. For a large number of the applications described above, this is however not a limitation (at least to start with). Among these applications, regional cartography could easily use the already acquired data sets for a first up-dating of maps. This also applies to forestry and mangrove issues where ERS-SAR data could demonstrate its usefulness. Existing radar images can therefore be used extensively to prepare for the future operational availability of SAR imagery over Central Africa.

However, one of the major necessary prerequisites is further support to co-operative projects between European and African institutions in order to better introduce the new technique into national plans and adapt it to their particular needs. A number of pilot projects of national, regional or even global interest could be carried out for this purpose:

• A west African mangrove mapping project: ERS's capabilities for mangrove forest detection and mapping were impressively demonstrated in chapter "Selected Interpretation Results". Knowledge of the current mangrove status (location, pressure, etc.) on a regional scale would allow the countries to manage their development better. Precise data could be introduced into global databases on climate change and bio-diversity.

• A cartography demonstration project: FAO is currently carrying out a project called Africover dealing with the description of the available cartographic bases in Africa. The ultimate goal is to create a general map for regional and national scale applications. It should be re-emphasized here that the availability or otherwise of topographic maps over a particular country or region strongly influences its development.

• A Lake Chad Observatory: This is aimed at long-term study of the lake's level. Projects are currently underway, but radar data have not yet been introduced due to their non-availability in the past. In addition, by exploiting the specific capabilities of ERS imagery in relation to hydrological applications, flooding or in more general terms projects dealing with the operational observation of rivers and lakes, could be investigated. Strong interest on the part of scientific groups dealing with the global hydrological cycle can be expected, in addition to that from national and regional institutes.