Newbie to Itzi, questions

[Sidharth Subramaniam]

Hi all!

Im a really happy to have found Itzi model, thank you to Lauren and team for this great software.

I work for Numer8, and we are striving to provide more and more GIS information to local fishermen in coastal areas, for free. One of my objective is to look flood simulation softwares and try them out. So far Itzi has been the only one tightly coupled with GIS, hence I have begun more exploration.

I am following the template of the tutorial to simulate flooding in Chennai City, India.

Im using ALOS World3D 30m res DEM, and when I reach the step of, r.water.outlet and give coordinates of an end of a river in Chennai that's just about to reach the sea. The area selected as Watershed is quite small compared to the size of the city, is it due to the city being mostly flat?

Is there any other way to simulate a flood across the a large part of the city. Im going to explore rainfall to be given as STRDS file too, so that I can try out how it looks compared to the floods of 2015 in Chennai City.

Thank you for any inputs.

Great work on making Itzi! :)

Sid

[Sidharth Subramaniam]

Hi,

I figured it out, it was a rookie question.

Maybe a not so rookie question would be this - "How to define Friction/Infiltration by using a LULC map?" 

By which I mean, where is the infiltration rate/friction values given? In the maps attribute itself?

Thanks

Sid

[Laurent C.]

Hi Sid,

Using r.water.outlet is only one way of defining the study area. If the area of interest is mostly flat, and you are using a global DSM, then the results of the catchment definition are unlikely to be trustworthy.

For your case, it is likely that the point you defined is not exactly on the flow path. You can check that with the accumulation map of r.watershed.

Please keep in mind that if you are using AW3D30 DSM, the representation of urban area might doubtful.

Best,

Laurent

--
You received this message because you are subscribed to the Google Groups "Itzï" group.
To unsubscribe from this group and stop receiving emails from it, send an email to itzi+uns...@googlegroups.com.
To view this discussion on the web, visit https://groups.google.com/d/msgid/itzi/cc0c7e47-2727-4493-9b59-0c1c1aedb34dn%40googlegroups.com.

[Laurent C.]

Sid,

One common way to estimate infiltration is using soil parameters, more than LULC maps. You can correct the obtained values using LULC maps to take into account anthropogenic activities.

I suggest reviewing the literature on the subject, as I am unaware of any definitive solution.

For mostly urbanized catchment and extreme rainfall, some authors even neglect the infiltration. But this depends on your study area and event.

For friction, there are tables (mostly from Ven Te Chow's Book) that relate land cover to the Manning number. However, the Manning number is usually used for calibration, so you must evaluate if using a spatially variable Manning number is useful to you.

Best,

Laurent

--
You received this message because you are subscribed to the Google Groups "Itzï" group.
To unsubscribe from this group and stop receiving emails from it, send an email to itzi+uns...@googlegroups.com.

To view this discussion on the web, visit https://groups.google.com/d/msgid/itzi/cf028fa5-dc35-42e8-be0f-c0bbeb9add09n%40googlegroups.com.

[Sidharth Subramaniam]

Hi Laurent, Hi all!

Thank you very much for your valuable feedback! Im happy to share some part of the GRASS code below to show you how I did it, and what results I got, which are attached as PNG files.

Just some more details for you -

I defined my watershed by first using "r.watershed" to get streams, then using an add-on "r.stream.basins -l dir=dirs stream=streams basins=basin_main" to get basins in the chosen area.

Then I selected one of the bigger basins, which is automatically numbered by GRASS as "64" using the following code -

GRASS 7.8.2 (chennai2):~ > echo '64=64
> * = NULL' > tmp1

Then selected that basin only.

GRASS 7.8.2 (chennai2):~ > r.reclass input=streams output=sel_streams_1 rules=tmp1

Made that basin into a separate layer.

GRASS 7.8.2 (chennai2):~ > r.stream.basins dir=dirs stream=sel_streams_1 basins=basin1

Memory swap calculation (may take some time)...
Reading raster map <dirs>...
 100%
Calculating basins using streams...
Reading raster map <sel_streams_1>...
 100%
Finding nodes...
 100%
Delineating basins for 1 outlets...
 100%
Writing raster map <basin1>...
 100%

Had a look at it, with d.mon it looked fine

GRASS 7.8.2 (chennai2):~ > d.mon wx1
GRASS 7.8.2 (chennai2):~ > d.rast basin1

Then converted that basin layer into a vector.

GRASS 7.8.2 (chennai2):~ > r.to.vect basin1 output=boundary3 type=area

Then using v.type converted it from area/boundary to line type, just did this to be sure that its a "line" type

GRASS 7.8.2 (chennai2):~ > v.type boundary3 output=boundary_1 to_type=boundary

GRASS 7.8.2 (chennai2):~ > v.type boundary_1 output=boundary_2 to_type=line

Then gave this as input parameter required for Itzi

GRASS 7.8.2 (chennai2):~ > v.to.rast input=boundary_2 type=line output=bctype use=val value=4
Writing raster map...
 100%
v.to.rast complete.
[Raster MASK present]
GRASS 7.8.2 (chennai2):~ > v.to.rast input=boundary_2 type=line output=bcvalue use=val value=0
Writing raster map...
 100%
v.to.rast complete.
[Raster MASK present]
GRASS 7.8.2 (chennai2):~ > r.mapcalc exp='rain=10'
 100%
[Raster MASK present]
GRASS 7.8.2 (chennai2):~ > r.mapcalc exp='n=0.02'

Then went on with config file and ran simulation.

The results are attached below, I attached "_h_max", which is the last frame of simulation which was run for 24 hours with 20 min interval, for maximum height of water after 24 hours. Also attached the legend.

I actually have crowd sourced "flooded streets" of 2015 Chennai city flood event. Surely that may not also be accurate, but its the only good dataset available for validation, since SAR satellite based flood mapping is very coarse and very generalized.

When I overlay the flooded streets with itzi output as geotiff, it looks like a good match, visually I would say it 60-70% match. Surely I want it to be much better!

This is why I wanted your feedback whether this is a good way to proceed? and how to improve it. For Urban areas, yes the DSM will be less accurate due to buildings but I may not have access to LIDAR or some other 3D data for urban buildings. So I may have to proceed with this data for now.

Any input and feedback will be great!

Thanks a lot again for all your valuable feedback I will keep it in mind.

Sid

[Sidharth Subramaniam]

Attaching a screenshot of overlaid flooded streets data.

Thanks

[Laurent C.]

Sid,

I am glad you were able to get the results you wanted. As for the model calibration, it is not specific to Itzï; I recommend following the literature for that.

Regards,

Laurent

To view this discussion on the web, visit https://groups.google.com/d/msgid/itzi/0d740b99-ee54-485a-86ce-c17272dde1a8n%40googlegroups.com.