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%).                                              
Flood at Mopti  gauge  (m) Confidence ratio (%) 6.6 83.6 6.21 {93.3 - 84.4}* 5.97 85.6 5.08 74.7 4.4 63.3
*    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.