Eliezer

My gut feeling says the SRTM data would be good for very large rivers with wide flood plains. My experience on a floodline estimation in a small river is that SRTM is not detailed or dense enough to pick up the detailed river geometry required for a hydraulic HEC-RAS flow model. However the surveyor who did the drone LIDAR survey of the river reach, upon request extended the reach length using SRTM, and he found that elevations in the flat floodplain areas corisponded well with the overlapping LIDAR data on these floodplains. (not in the small main river channel though)

Based on this experience, in Namibia, i think it could work.

Regards

Gert

If there is nothing else available, then the SRTM data can be used, but with much caution. The major issues with creating a 2D flood model from the SRTM include:

- Poor vertical accuracy. Vertical accuracy in this dataset is something like +/- 10m. You could be lucky and find that in your region the levels are a close match to survey, or unlucky and find outliers of several meters or more.  Accuracy is also dependent on terrain slope, and also terrain aspect. These are all legacies of the methodology used to capture the survey (two radars on the space shuttle, and triangulation).

- Ability to integrate your SRTM levels with other, known benchmarks in the region. You often end up with the situation where surveyed bridge and road levels bear no resemblance to their underlying terrain. Same goes for flood height gauges. And here, the flatter the region the worse the issue is, because large changes in flood volume cause only minor changes in flood elevation. This makes calibration very difficult.

- Hydrologic connectivity. There are three versions of the SRTM data. Base (DEM), Smooth (DEM-S), and Hydrologically enforced (DEM-H). Only DEM-H has been processed to remove artificial sinks and barriers, to ensure that water will always flow downhill.

In summary, it depends on how you intend to use the information from your flood model. It will probably be ok for high level analysis, such as identification of the general floodplain area, or approximate inundation for a given flow. It will definitely not be very good for detailed flood risk mapping and the like.

I would suggest to use both SRTM as well as ASTER DEM’s and compare both results. From my experience, the two productions require readjustment using a couple of GPS field measurements. However, in my case ASTER presented a better performance.

Good luck

Eyad

Dear Eliézer,

You already have excellent responses to your question. I don't know the area or the length of the sections of the Amazon you are working with (even a big basin has its small and narrow reaches upstream :).. I think it all depends on the question you are trying to address and the accuracy you need to answer that question. Modelling only the main channel will not represent a flooding situation (furthermore, low-quality DEM is particularly evident in the representation of the main channel). You need to consider the computational capacity you have if an higher resolution DEM (based on LiDAR for example) is to be used in wide and/or long reaches). Also consider that you should use a terrain model and not a surface model (HEC-RAS and other programs allow you to add obstructions and/or inefficient flow areas later).

Bom trabalho!