Analysis of the model for a 6.60 m flood The reference year is 1999 with a flood height of 662 cm at the Mopti gauge. The course of the 1999 flood deviates quite significantly from the 660 cm reference flood. Indeed, the maximum flood height of 617 cm at Macina, which determines the flooding of the upstream Delta and along the Diaka, stands 39 cm lower than the reference situation which should be close to 656 cm. This deficit is compensated in Mopti by the stronger flood of the Bani River in Sofara (+ 35 cm compared to the reference situation). However, the large time lag between the date of the maximum height in Sofara and that in Mopti (32 days against 8 days on average), as well as the difference in volume contributions between Niger and Bani, result in a lower flood than expected in Akka, at the exit of the Debo lake. Landsat images are available for October 18 (198_050), October 27 (197_050 and 051) and November 28: (197_049, 050, 051). The shape file (Synthèse_660_1999) regroups these different situations : The areas in blue on the map are the flooded areas common to the model and the Landsat images. The pink and red areas (Inon_ less) are two different cases of mismatch: 1- In pink : (value 3 of the item carto) areas are shown vegetation mosaics associating "non floodable" vegetation ( Togge ) or vegetation potentially flooded by run-off first, with floodable vegetation that were poorly flooded in 1999. The former case includes the following vegetation mosaics: AG / TA, AG / TS, ZB / TA, ZB / TS, P / TA, P / TS, VH / TA, VH / TC, VH / TS, ESP / TA, VSP / TA, PAM / TA, PAM / TD, PAM / TS), while the latter case includes the vegetation mosaics PAM, PAN, PAS, PAM / PAN, PAM / PAS, ZB / PAN, ZB / PAS, PAM / P, PAM / VH, PAM / VOR, P / PAN, P / PAS. Together, these mosaics, whose flood status is regarded as "ambiguous", extend over 194,160 ha, of which 154,331 ha were not flooded in 1999, while only 39,829 ha were flooded. 2- In red : (value 2 of the carto item) are shown floodable vegetation formations or mosaics that are flooded according to the model but not on Landsat, amounting to 143,614 ha.These non flooded areas distributed by flood depth level as indicated in the following table: The majority of these non-flooded areas on the 1999 Landsat images are concentrated around Diafarabé, in the Moura plains, in the southern Pondori basin and in the high plains of Sébéra. They result, for the first two cases, from the water height deficit (-39 cm) noted at Ke Macina and, for the other two cases, possibly from a poor flood provision by the Bani despite the higher than expected flood height recorded in Sofara. However, the last statement should be carefully considered: the Pondori basin can be flooded very early (in August) and have only reduced flooded areas in the extreme south by November. In the absence of available images in August, it is difficult to decide. 660_1999.gdb.rar contains : NIV_660 Synthese_660_1999 : Item inon_543: Value 1: flooded area common to Veg7_21 and Landsat Value 2: area flooded by the model but not flooded on Landsat Item carto: Value 1: flooded area common to Veg7_21 and Landsat Value 2: area flooded by the model, not flooded on Landsat and composed of floodable vegetation by the river flood. Value 3: not flooded on Landsat and composed of vegetation whose flood status is ambiguous (list above)
Flood depth (m) Flooded area (ha) [-0.10 à -0.30] 82 090 ]-0.30 à -0.60] 47 258 ]-0.60 à -1.50] 13 625 ]-1.50 à -2.80] 721
Gauge  sites   Flood height   in 1999  (cm)   D ates   Reference flood  660 (cm)   Height gap   (cm)   Mopti   662    26 /10   660    + 2    Ké Macina   617    01/10   656    -   39    Sofara   609    24/09   574    +35    Akka   511    23/11   549    - 38
On the raw matrix model (NIV_660), the flooded areas extend over 1,697,636 ha . The area of floodable vegetation units on VEG7 (without Farimaké) measures 1,737,541 ha, with the smoothing done with the 500x500 filter causing an area loss of 2.3%. The reintroduction of these “lost” flooded areas does not pose any difficulties: it comes down to using VEG7 directly by removing level 11 (non flooded) (shape VEG7_21).
Overlaying Landsat images in VEG7_1999 (shape file VEG7_ 1999) gives a fairly different representation of the flood, with a total flooded and/or heavily vegetated area * of 1,439,517 ha. * excluding areas heavily vegetated by woody plants such as the Togge in the south of the Delta, for example. The entire flooded / vegetated surface of VEG7_1999 (Landsat) is within the limits of VEG7_21 so that the area common to Landsat and VEG7_21 extends over 1,439, 517 ha . However, such a method is only applicable in this particular case. For the other flood levels analyzed, the flood heights do not correspond to an application limit of the model, and we will see that the flooded (and / or heavily vegetated) surfaces on Landsat never completely fall within the limits calculated by the model for the maximum flood height at Mopti. In the model, there are areas flooded in addition, while some flooded areas are missed by the model. We used the spatial operator “intersect” between the area calculated by the model and the "flooded" surface recorded on Landsat and corrected the errors due to the matrix smoothing. The area flooded on Landsat in 1999 was within the area flooded calculated with the model. The ratio between the area flooded on Landsat VEG7_1999 and on VEG7_21, equals 82.8%, a relatively poor score explained in part by the extent of vegetation associations with "ambiguous" flood status, and in part by the characteristics of the 1999 flood, i.e. the strong imbalance between the Niger and the Bani floods. The difference between the model run for a 660 cm height (VEG7_21) and the year 1999 (VEG7_SF_1999) as assessed by the spatial operator is recorded in the shape file (Inon_ less_660_1999) and extends over ​​ 298,024 ha , calculated as flooded by the model and not flooded on Landsat images.
660_1999.gdb.rar
+
Download