topic How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel in ArcGIS Spatial Analyst Questions

How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Wed, 31 May 2017 23:13:41 GMT

Hi,

I am using the 'range' statistics type within the 'focal statistics' tool. For each central pixel ArcGIS calculates the difference between the highest and lowest values in the defined neighborhood area and assigns it to the central pixel. What I am now looking for is a way to identify the highest and lowest pixels within that defined neighborhood area.

Any help is much appreciated.

Regards,

Fidel

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

DanPatterson_Retired — Thu, 01 Jun 2017 00:18:40 GMT

focal statistics minimum and focal statistics max plus all the other tools in the focal statistics tools in ArcToolbox will cover all combinations and permutations of gleaning other information

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Thu, 01 Jun 2017 10:37:04 GMT

Hi Dan,

Thanks for the reply.

Yes, but what I am trying to find out is to actually locate the max and min pixels. The focal statistics tools assign the results of the operation to the central pixel and do not give the actual location of the max/min pixels as an output.

Regards,

Fidel

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

XanderBakker — Thu, 01 Jun 2017 12:04:51 GMT

You can use Focal Flow—Help | ArcGIS Desktop to determine the location of the lowest value. To determine the maximum value, invert your raster (highest value - original raster) and run again.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

DanPatterson_Retired — Thu, 01 Jun 2017 12:10:44 GMT

so if you are using a focalmin, for example, it will determine the minimum value and assign it to the middle value. Try subtracting the original raster from that raster and values of 0 should identify the cell location where the minimum occurred.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

XanderBakker — Thu, 01 Jun 2017 13:53:42 GMT

If the OP needs to identify the lowest and highest position in a 3x3 window for each pixel I think it will need the Focal Flow to obtain that. Each pixel will obtain a value to points to the relative position where the highest or lowest pixel is located.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

DanPatterson_Retired — Sat, 11 Dec 2021 22:51:37 GMT

From the link on focal flow or even flow direction, they do use the surrounding cells in their calculation. In the case of focal flow, the cells that contribute the core cell. they don't tell you the exact location of the cells represent the minimum. In the example below, we should get a value of -5 for the minimum, 15 for the maximum and 20 for the range assigned to the core cell. If we use the focal minimum raster and subtracted the input raster from it, we should get a 0 for the top left cell. Values of 0 would be the minimum of one of the moving windows, assuming raster processing is from top left, then the value of 0 would correspond to core cells in standard row/column format.

From Focal Flow

Now having said that, if the raster were exported to a numpy array, there are other ways of doing this even using rolling statistics or queries.

a
array([[-5, 11,  3],
       [ 6,  9, 15],
       [10,  8,  8]])

m = np.min(a, (0,1)) # along the row, column... returns -5

np.argwhere(a == m)
array([[0, 0]], dtype=int64) # row 0, column 0‍‍‍‍‍‍‍‍‍

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

curtvprice — Sat, 11 Dec 2021 22:51:40 GMT

Here is how to locate each cell that is the highest or lowest value in its own 3x3 neighborhood. Not sure this is exactly what you want, but I thought I'd throw it out there.

The expressions below are formatted for the Raster Calculator tool. BTW, I used "" for neighborhood to use the default, NbrRectangle(3,3).

# Cells that are the lowest of the 3x3 around that cell
Con(FocalStatistics("in_grid", "", "MINIMUM") == "in_grid"), "in_grid")
# Cells that are the highest of the 3x3 around that cell
Con(FocalStatistics("in_grid", "", "MAXIMUM") == "in_grid", "in_grid")‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍‍

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Fri, 02 Jun 2017 15:28:44 GMT

Dan_Patterson‌, xander_bakker‌, curtvprice‌

Thank you all for the help so far.

The suggested methods are unfortunately not doing what I want to do. This is mainly because the neighborhood zone is overlapping as the operation moves from one cell to another cell. The minimum cell for the first central cell will change its value to the minimum value for its own neighborhood as the operation moves from cell to cell. As a result subtracting the minimum raster from the original raster (or the method suggested by Curtis) will not give me the locations of all the minimum cells for each neighborhood zone.

Any more suggestions are appreciated.

Cheers

Fidel

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

XanderBakker — Fri, 02 Jun 2017 16:09:12 GMT

Have you tried the Flow Direction tool yet? It will return the lowest value for each pixel in a 3x3 neighborhood.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

curtvprice — Sat, 11 Dec 2021 22:51:43 GMT

How about Block Statistics? This works with neighborhoods that do not overlap. I suspect this still isn't what you want though, since this is around the central cell again. Can you explain your use case -- what exactly are trying to do?

# Cells that are the lowest of the 3x3 around that cell
Con(BlockStatistics("in_grid", "", "MINIMUM") == "in_grid"), "in_grid")
# Cells that are the highest of the 3x3 around that cell
Con(BlockStatistics("in_grid", "", "MAXIMUM") == "in_grid", "in_grid")‍‍‍‍‍‍‍‍‍‍‍‍

All these Spatial Analyst raster operators output values of the central cell of a window, if the procedure Xander Bakker suggests (messing with FocalFlow) won't do it, it seems likely Dan Patterson's suggestion of numpy arrays is the only way to do what you want. There is a useful cookbook in the help on how to do cell by cell neighborhood stuff in numpy:

NumPyArrayToRaster—Help | ArcGIS Desktop

Back in the day, Spatial Analyst (ArcInfo Workstation GRID) supported neighorhood notation, but it was pretty slow and limited, this functionality is now covered by working with numpy arrays.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Sat, 03 Jun 2017 21:25:55 GMT

Unfortunately Focal flow and Flow Direction are not working for me.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Sat, 03 Jun 2017 21:41:53 GMT

Dan_Patterson‌, curtvprice‌, xander_bakker‌

I am trying to find upper and lower reservoir locations for potential pumped storage hydropower sites in a large area. My input is a Digital Elevation Model. So if I use a non-overlapping neighborhood, I will miss many possible upper and lower reservoir combinations in which one of the reservoirs lies in one neighborhood and the other reservoir in the adjacent neighborhood.

I am not familiar with NumPy Arrays and would really appreciate any help.

Thanks

Fidel

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

DanPatterson_Retired — Sat, 03 Jun 2017 21:56:22 GMT

Image of what you are working with would help....

We are either not talking about the same thing, since both my suggestions will work, or your error in location would be one whole cell size width.

If need be, zip and save a portion... not the whole thing... to a raster maybe in *.tif format, so I can demonstrate how to do it with real data as either a raster/grid or numpy array.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Sat, 03 Jun 2017 22:07:38 GMT

Do you mean image of a typical pumped storage hydropower site?

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Sat, 03 Jun 2017 22:09:44 GMT

Sorry didn't read your whole message when I when I sent the previous reply. Will send it.

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

DanPatterson_Retired — Sat, 03 Jun 2017 22:10:26 GMT

Whatever this is ....

What I am now looking for is a way to identify the highest and lowest pixels within that defined neighborhood area.

It is a raster I presume... or the question doesn't make sense

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Sat, 03 Jun 2017 22:25:13 GMT

Here you are

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

DanPatterson_Retired — Sat, 03 Jun 2017 22:45:29 GMT

Is this what you are looking for?

Over the DEM with transparency

Re: How to identify the highest and lowest pixels within a defined neighborhood area of a central pixel

FidelTedla — Sat, 03 Jun 2017 23:12:20 GMT

This is not what I am looking for. This is the original raster minus the Focal Min raster. As explained in one of my posts above this is not giving me what I am looking for. I am looking for the location of the min or max for each central cell. That means the number of min or max locations to be identified must be equal to the number of cells in the original raster.