Conclusion The tests of the flooding model in isolines of equal depth call for a series of remarks: • The first comes from the difficulty to get a good series of Landsat images showing the successive stages of the flood and its extent over the entire Delta. If we are lucky to have had images since 1984 – the lowest flood in the century –, the series of images is, however, discontinuous and of uneven quality, which prevents multiple tests, and the repeatability of these images is not sufficient to accurately determine the maximum flood extension for each year. It would therefore be desirable to have images with high temporal resolution like MODIS in order to follow the progression of the flood across the Delta. • The second remark is liked to the very nature of the floods: we tried to relate each height of our flood-tests to a "normal" model balancing the contributions of Niger and Bani rivers with the water height at the gauges of Mopti and Akka. However, all the floods used during these tests were found to be specific, sometimes with an imbalance between the contributions from the Bani and Niger rivers, sometimes because of an abnormal delay in the dates of these flood contributions, and sometimes for both reasons. However, we have learned from the model: it works satisfactorily in the high to medium flood years, with a confidence ratio that is always above 83% and values which are more or less strongly impacted by the flow imbalances between Niger and Bani. The confidence ratio decreases when the maximum flood at the Mopti gauge decreases from 6.00m to 5.08 m (75%) and again to 4.40 m (63%).

* depending on whether we take into account only the totally flooded surfaces (84.4%) or all of the surfaces that are totally or partially flooded (93.3%). If we assume that partially flooded surfaces are on average flooded at 50 %, the confidence ratio stands at 88.9%. Beyond the confidence that we may have in the model, the tests reveal a certain number of common features: weak water supply in the south of the Delta, for which evidence in the Pondori basin appears when the flood is 6.60 m or lower at the Mopti gauge, and is confirmed for all the test values. This shows a marked imbalance between the southern part of the Delta (the right bank of Niger up to Mopti and the Bani) and the left bank of Niger, which is much better supplied by flood water, in particular along the Diaka, with its very peculiar topographic profile. This "geo-morphological" logic, which is linked to inequalities in the subsidence of the Delta basin, is confirmed by the tests. In addition, at a finer scale, there is a local logic linked to the propagation of the flood: a deeply flooded vegetation association unit, such as VB or B, surrounded by less deep or unflooded associations, will be poorly flooded or not flooded at all, revealing threshold effects common at the edges of the Delta floodplain or in the Peru of Diallube. On the contrary, units of vegetation associations linked to shallow flooding when they are lying between large deeply-flooded basins will be more likely to be flooded and appear highly vegetated on Landsat images. However imperfect it may be, the model works satisfactorily for flood maxima at the Mopti gauge between 6.00 m and 6.60 m, which represent 73% of the flood years between 1922 and 2014 (60 years out of 82 years fully surveyed). The model begins to be less efficient for floods between 6.00 m and 5.10 m at the Mopti scale (22% of flood years), and to markedly deviate from observed floods for flood levels below 5.10 m (4 years out of 82 years, i.e. 5% of surveys). Three questions arise, however: 1 - Do "normal" floods with a balanced contribution in volume and in the propagation times of the Niger and Bani rivers exist or are they a calculation artefact? 2 - The model works in a "flat" way: a bit like a bathtub that is filled at a certain level. Is it possible to transform it into a digital elevation model (D.E.M.) including the actual slopes and altitudes of the Delta? We will try to answer this question further down. 3 – The floods of the Niger and Bani Rivers historically range at the Mopti gauge between 4.40 m in 1984, and 7.39 m in 1924 (with 15 records out of 82 exceeding 7.00 m). The model cannot predict flooding beyond 6.60 m. The difference between the strongest flood (7.39 m) and the reference flood (6.60 m) is 80 cm, which is considerable. Between 6.30 m and 6.60 m, the potentially floodable areas increase by 8% for every increase of 10 cm in the height of the flood. Compared to the highest flood recorded, the increase in the flooded area would be of the order of two thirds of the flooded area at 6.60 m, i.e. an area which would spread from 1,700,000 ha to almost 3,000,000 ha. Towards the eastern edge of the Delta, the topography very strongly limits the possibilities of extension of the flooded area. Indeed, at the edge of the flood plain, there is a glacis with a slope of 10 ‰, which would cause a very slight translation of the limit of the flood, of the order of 80 m for 80 cm of flood height. The situation is radically different on the western edge of the Delta, where the floodplain adjoins the "dead Delta", a former floodplain of the ​​Niger river before it moved eastwards. A digital elevation model of decimetric resolution would be required to assess the area and pathways that the flood might take at its maximum expansion.