Topological spatial data, linked to the thematic data relating to vegetation types (flora, ecology, production), are managed in specific layers, all of which are derived from a primary layer called VEG4. This layer, which describes the boundaries and the spatial distribution of vegetation types, is made up of 14 535 polygons, which are areas corresponding to the various vegetation types (with the exception of 5 polygons corresponding to open water - the minor beds in Niger, Bani, part of the Diaka and the lakes - and 731 polygons corresponding to built-up areas). Each unit is made up of a polygon, bearing an acronym designating its vegetation type which may be a plant combination or a mosaic of plant combinations. VEG4 , the essential information layer in the GIS, derives directly from the CIPEA/ODEM rangeland map realized by Pierre Hiernaux and Mark Haywood, who also played a key role in the realization of this layer of GIS information by performing the digitization of the map (30 sheets 40x25 cm to 1: 100 000) on an A3 format digitizing table, using the drawing software programme (Autocad). These maps were captured within a 5x5 km squares grid, rigorously using the method developed by Mark Haywood for the realization of the original maps constructed on a UTM grid (spindle 30) at a 5 km step, each digital file containing six 5x5km squares, usually arranged in a 3x2 format. Mark Haywood was also careful to note in a text file, for each polygon with a recognition mark, the unit acronym and the mark coordinates in x and y. This file, which is independent from the 163 DXF graphics files that represent the inland Delta rangeland map (approximately 978 squares of 5x5 km), contains 31,083 lines (600 pages of word processing). This data base entry in a standard exchange format (DXF format) cost Mark Haywood three years of work and greatly facilitated our task. However, the data from these 163 files had to be translated into a single ARC/INFO layer and a number of methodological and technical problems had to be solved. For this we transcoded the 163 files into 163 ARC/INFO layers, then merged them into a single layer. There were, however, several problems to be solved: a unit corresponding to a vegetation type is an object, therefore it must be represented by a polygon. However, the presence of the construction grid could lead to the division of one area into two, three, four, or more adjacent polygons that had then to be merged back into a single polygon. On the other hand, Autocad does not allow the transfer of unit attributes, in particular, the acronyms identifying the vegetation types. To do this, a VEGPOINT layer was created from the text file including the x, y coordinates and the acronym of the labeled point inside each polygon. We therefore had a layer containing all the arcs i.e. the faces of the future polygons, and all the points destined to become the labels of the polygons representing the vegetation units. Yet it was imperative to have one single label per polygon, with the right acronym identifying the vegetation type of the polygon (in total, 31,083 points and as many elementary polygons). The position and the acronym of 31,083 points then had to be checked, a process which took several months, and the adjacent polygons bearing the same acronym were aggregated to get the final layer (Veg1). From VEG1 to VEG4: simplifying thematic information VEG1 contains 14,605 ​​polygons, each identified by an identification number (ID) and an acronym referring to its vegetation type, i.e. a total of 209 different acronyms, and a number indicating the occurrences of the acronym. A number of these abbreviations are "script" errors made when drawing maps in Addis Ababa. For example, the acronyms “TA/Z” or “TD/B” indicate mosaics that do not exist. In some cases, a particular unit is given two acronyms representing the same vegetation type, one belonging to the inland Delta nomenclature and the other to the nomenclature used in the «dead Delta», for which another vegetation map was drawn (and has remained unpublished) ... On the other hand, some of the mosaics are ecologically "unlikely" and are only represented in very few occasions : 35 mosaics are represented only once, 19 only twice, and so on. In general, we considered that all mosaics represented in 10 units or less should be replaced by the ecologically closest mosaic. This transformation also extended to mosaic considered unlikely although present in more than 10 units (VSP/P, AG/TT, BP/B). Finally, the TB/TC mosaic (made of two dry woody savannas on “ togge ”, very close to one another), indicated in 95 units, was simplified as TB. The TB/TC mosaic was attributed to “TOPO” units, which represent the built-up area. This convention makes it possible to suppress “TOPO” from the vegetation map (by replacing it with a strongly anthropized woody savanna TB/TC without losing the polygons which can at any time regain their status as built-up surfaces. Table 1 :

Ultimately, we retained 121 vegetation types (including RIVER which represents open water and TB/TC that represents built-up areas). The acronym modification leading to the 121 vegetation types is a large simplification (from 209 to 121 vegetation types). However, the resulting loss in spatial information is very limited, with a reduction in number of polygons from 14,605 to 14,535. This final layer is called VEG4 and becomes the main layer of the GIS. It has two functions: 1) to document the spatial distribution of the vegetation types; 2) to produce derived layers, FLORE1 for flora, CRUE 3 for ecological data, SOIL for soils data and PRODUCT for rangeland fodder production.

F i gure n° 1 : The process leading to the organization of ARC/INFO covers to create VEG4

GIS organization (1) SPATIAL DATA CREATION