This paper presents a part of a reservoir engineering study initiated in 1978 to evaluate secondary recovery projects by water-injection in so producing areas of the Algerian HASSI MESSAOUD giant producing areas of the Algerian HASSI MESSAOUD giant oil field. Paper mainly deals with the first part of the study : an in-depth description of the sandstone reservoir anatomy. Based on a concept of individualized sedimentary units, the description characterizes and classifies these units. It uses a systematic treatment of such as numerical filtering. The resulting picture of the reservoir, is an organization of good sand bodies distributed among a poor sandstone medium, into which are dispersed discontinuous shale breaks. Good sand bodies and shale breaks were characterized by their densities on the vertical, and by the frequency of their occurrence with depth. Permeability of the good sand bodies was found to be related to their density. When numerous, good sand bodies are individually more permeable than when they are scarce. The inferences of such a description on flow parameters, such as the effective horizontal and parameters, such as the effective horizontal and vertical permeability, and the effective critical gas saturation were extensively studied. Results were a better understanding of reservoir behaviour and a more rigorous simulation of dynamical phenomena. Thanks to this method, history matching was easily completed, thus saving time for further studies. Introduction The selection of data for reservoir simulation is generally a difficult task, specially in heterogeneous formations. Modern reservoir simulators are proved to be reliable on mathematical and physical proved to be reliable on mathematical and physical basis. Therefore the quality of the results depends on the quality of the data. Not only on the quality of each data by itself, but also on the imaginative concept which presides at their gathering and assembling into a model. Many unsuccessful attempts of predicting reservoir behaviour with numerical simulators predicting reservoir behaviour with numerical simulators were due to the lack of geological perspective, the necessity of which has been emphasize in the recent literature. When data assembling comes to a "cells filling", history matching is often difficult, tedious, and the resulting picture of the reservoir poorly reliable. poorly reliable. This paper tries to summarize studies which were performed on a producing area of HASSI MESSA field in Algeria. Main objective was the appraisal of the various possible drive mechanisms. Such a objective could only be achieved by the use of mathematical simulations. It was therefore decided, prior to the use of simulators to devote a large prior to the use of simulators to devote a large part of the effort to the reservoir description. part of the effort to the reservoir description. Results obtained from the simulations proved that such an effort was worthwhile. HISTORY SUMMARY Located in the Northern part of the Algerian Sahara desert (Figure 1), HASSI MESSAOUD field was discovered in 1956. Producing formation is a Cambrian sandstone encountered at 3 100 meters subsea. With oil in place over 2 000 × 10(6) m3, HASSI MESSAOUD is among the giant oil fields of the world. The oil is light - 45 degrees API. Original pressure is 473 bars. Production started in 1959, and after a few years, it appears that the whole field could be divided into relatively independant producing areas (Figure 1). Geologically and dynamically these areas are not completely separated from their neighbours, but the fluid exchanges are in most cases highly restricted by low permeabilities on the borders generally due to faults, in addition to similar rythms of depletion. Until 1964 the production mechanism was only depletion above bubble-point. At that date a miscible gas injection started, firstly limited to a portion of area 13, but which now extends to several areas (2, 6, 9, 14, 15, 19, 20, 25). For other areas (4, 7, 11, 12, 16, 17) a water-injection program was scheduled. One of the main problems of water-injection was to evaluate the best starting date.