Assuming both datasets are accurate you can subtract or add one from the other. In Esri use Raster Calculator (spatial analyst extension required); You can also use GDAL_Calc on command line or QGIS raster calculator. If your flood data is just the depth of flooding then you can add the two to get the flood level, if it's the bottom of the channel you want to subtract.
In the output raster the values will be near 0 for most of the data and be considerably larger where water sits. Observe the general background value where there is definitely no water and use that as a threshold figure to extract the underwater sections.
By subtracting the DEM (surface) from the model flood data the data is normalized to just the depth which will allow you to get a good depth cross section for any location along the watercourse. If you already have depths as a source then I'd use that to show cross sections, after all it's more important to know how deep/far the water overflows than the correct elevation at the overflow.
You will need to be aware that the data will probably not be perfect and you have to plan for a bit of background 'noise' - that may affect the actual boundary by a little bit but should be nearly insignificant overall.