Urban flood risk mitigation model (area of analysis)

[Diego_Guarin]

Dear Stacie,

Thanks so much, of course ! , all assistance is welcome

Best,

Diego

[Diego_Guarin]

Dear Stacie,

I have run the model again and it worked. I have run it 2 times (Test 1 larger pixel size) and (test 2 smaller pixel size).

However, in the layers of the outputs of test 2, some patches and overlapping or informed areas with no data appear (Apparently there is an error in the LULC and soil group raster layers). Which I wanted to check in the following way:

Test 1: Model with a larger pixel size (for a detailed analysis, the size of the pixel is not accurate to compare it with the urban polygon).

Data:

Land Use / Land Cover layer (Study_area_buffer_2km.tif) pixel size: 300 / -300
Soil hydrologic group layer (soil_clip.tif) 255/ -255

Example of results obtained: layer runoff_retention

There appears to be no missing data or overlapping pixels (if this can occur).

Test 2: Taking into account that when running the model the resolution of the layers to be used is not relevant. I used the previous soil layer (soil_clip.tif) pixel size: 255 / -255, and a LULC layer this time with a pixel size approximated to 10 m (ESA_LULC_10m1_sty_area) pixel size 9. 26/- 9.26, considering that the soil layer is resampled to the most important layer which would be the LULC layer (I hope I’m not wrong).

However, these were some of the outputs. (i.e. runoff_retention):

Some areas appear as if they had not been resampled, so they look like larger patches or as if they are missing data. Nevertheless, when looking at the pixel size of the layer output, it appears to be 9.26/- 9.26. I attach the logfile for test 2 for review.

test_2_InVEST-natcap.invest.urban_flood_risk_mitigation-log-2023-07-27–14_40_11.txt (16.3 KB)

I also attach the 2 layer rasters of test 2 in case there is something not correct.

soil_clip.tif (29.8 KB)
ESA_LULC_10m1_sty_area (1).zip (378.0 KB)

So I would like to know why the layer is displayed that way.

As an alternative option, I also tried to readjust/resample the pixel size of the soil_clip layer from 300 to 10, so that this time there were no empty areas or patches in the outputs.

For this I used tools such as the raster calculator, (conversion:translate convert format) and GRASS r.resampling. I’m not sure if those tools only work for DEM layers or climate data, or maybe I missed incorporating information (I don’t have much experience in resampling pixels).

So I would be very grateful if you could tell me how to solve this concern.

Best regards and thanks so much again.

Diego

[swolny]

Hi Diego -

When the soil layer is resampled, the cell size changes, but the cell values do not (especially if a resampling type of Nearest Neighbor is used, and in this case, the values can’t vary between 1-4 or the model won’t work). For example, in the original raster there may be a single pixel of soil group value 2 that covers a 255m x 255m area, which gets resampled to a bunch of smaller (9.26m) pixels, but all of them still have a value of 2. So in the result, the pixel size will be the same as the LULC raster, but you’ll still see the larger patterns from the soil raster. This is the same in all models where some of the layers are more coarse, like soil and climate.

~ Stacie

[Diego_Guarin]

Hi Stacie,

Thanks for the explanation. However, I would like to know how to interpret the pixels that correspond to water bodies (white pixels),

Because, as you can see in the image below, these areas (white pixels) were not incorporated, how could this be explained?.

It would be correct if those pixels could somehow be combined or integrated with another similar layer without affecting the quality of the results?

Thanks so much again

Best,

Diego

[swolny]

The white pixels correspond to places where your soil raster has NoData cells. Anywhere one of the input layers has NoData, the output will also have NoData.

~ Stacie

[Diego_Guarin]

Hi Stacie,

Thanks so much, I will work on the outputs and let you know how it goes.

Best,

Diego

[Diego_Guarin]

Hi Stacie

I am sending you a summary of the results obtained, which you could review, to know if I am on the right track.

This summary includes a partial interpretation along with some related questions (in yellow).

As always I thank you very much for your time, comments and recommendations, thank you so much

Best,

Diego

[swolny]

Hi @Diego_Guarin -

A few notes from looking through your writeup.

On page 2, it says that Q_mm_urb is “the maximum precipitation (mm) recorded in a storm event that can be retained in the area”. This isn’t correct. Q is the potential runoff (water that can, for example, contribute to a flood during a storm), given the characteristics of the landscape, mainly driven by curve number. So high values of Q mean that the landscape is not retaining water. Page 3 says “the amount of runoff Q_mm_urb not retained by the pixel”, which is correct.

Page 3 asks about interpreting runoff retention per m3. Whether to use it or not depends on your study. It may be adequate to provide “runoff retention”, if you just need to show the places on the landscape that are generally retaining more or less water. But if you need an actual volume of water retained by each pixel from this particular storm event, then you’d use runoff retention per m3.

Page 5:

• The values in m3 are per storm event, not per year.

• Micro-watershed 7 seems to have relatively low retention values, not high.

• Remember that the per-watershed results are an average (or sum) across the whole watershed. The urban area is a much larger portion of watershed 7 than watershed 3. And it’s just a small portion of
  watershed 3, so the non-urban part of the watershed must have good retention abilities, and they are going to dominate the results.

• The model is very heavily driven by the land cover map and curve numbers, which determine the proportion of precipitation that is retained. The volume outputs are additionally heavily influenced by the precipitation input. In general, larger watersheds are likely to have more runoff and retention. If you want to control for area, you can calculate the volume per hectare.

Page 6

• The service built values = affected building values x runoff retention (m3) values. The runoff retention values are as high as 19,02, so it’s not surprising that the service built values are around 10 times larger. Given that the units are (currency x m3), I’d say that this model output should be used as an index only, so the values themselves are not used. Also remember that you are modeling watersheds that are much larger than the urban areas themselves, so you’re not only considering the urban area retention, but the retention of the whole (much larger) watershed, which the model wasn’t really built for.

• I’m not an economist, so don’t have much to say about costs. But I would say that if you need to calculate a more accurate cost of flooding, you’ll need to use a different model, since this one is not actually mapping flood depth and coverage.

Page 7: Since there are only urban areas in watersheds 3 and 7, it makes sense that only those watersheds have affected building and service built values.

~ Stacie