topic Re: Returning names of rasters with maximum value in ArcGIS Spatial Analyst Questions

Returning names of rasters with maximum value

JeremyProville — Tue, 30 Jul 2013 18:07:27 GMT

Hi,

I have a pile of 23 different rasters, on which I'm performing various calculations. One of them is a simple Maximization, using Cell Statistics, and one thing I'd like to see is a raster showing me which of the 23 layers contained the maximum value that was chosen. I'm finding this is pretty hard to do.

In Arc 9.3, I solved this problem using a long string of Con() and Max() statements in the raster calculator, to assign an integer index value to each raster that was selected in the maximization. This is no longer possible in Arc 10 since Max() has been removed. I tried an approach in Python using the code below, but it does not really work. What I'm looking for is a way to return a raster with either index values from 1-23, or ideally the names of each input raster. Each name would be assigned to the cell if it was chosen as a maximum value.

Hopefully this makes sense, I can elaborate if not. Any ideas here would be hugely helpful. Here's some sample code I've been playing with, but there has to be a simpler, more elegant solution - perhaps an obvious tool I'm missing?

Thanks!

opt2 = Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "blank",-9999, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "CottonLB" ,1, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "FodderLB" ,2, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "FruitsLB" ,3, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "GroundnutLB" ,4, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "JatrophaLB" ,5, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "MaizeLB" ,6, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "MilletLB" ,7, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "MiscanthusLB" ,8, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "NarcostimulantsLB" ,9, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "OilpalmLB" ,10, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "PulsesLB" ,11, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "RapeseedLB" ,12, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "RiceLB" ,13, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "RootsLB" ,14, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "SorghumLB" ,15, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "SoybeanLB" ,16, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "SugarbeetLB" ,17, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "SugarcaneLB" ,18, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "SunflowerLB" ,19, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "SwitchgrassLB" ,20, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "TubersLB" ,21, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "VegetablesLB" ,22, Con(CellStatistics(["CottonLB","FodderLB","FruitsLB","GroundnutLB","JatrophaLB","MaizeLB","MilletLB","MiscanthusLB","NarcostimulantsLB","OilpalmLB","PulsesLB","RapeseedLB","RiceLB","RootsLB","SorghumLB","SoybeanLB","SugarbeetLB","SugarcaneLB","SunflowerLB","SwitchgrassLB","TubersLB","VegetablesLB","WheatLB","blank"],"MAXIMUM","DATA") == "WheatLB" ,23,-9999))))))))))))))))))))))))

Re: Returning names of rasters with maximum value

Luke_Pinner — Wed, 31 Jul 2013 02:13:54 GMT

Use the HighestPosition function/tool. Used to be called UPos in earlier versions.

Re: Returning names of rasters with maximum value

JeremyProville — Wed, 31 Jul 2013 15:22:52 GMT

Use the HighestPosition function/tool. Used to be called UPos in earlier versions.

Thanks lpinner, this was spot-on! In theory it is doing what I need, however in practice it's not working out exactly as planned. I have a follow up question I'm hoping I can get some help with:

Running the tool with three or more input datasets seems to include only results that are an intersection of all the inputs. All other non-intersecting areas are NoData. What I would like is to have a dataset represented in the results, even if it is the only one with a value in a specific cell.
I tried get around this by adding in a constant value as an input, or a blank raster with constant values everywhere, such that there would be a 'dummy' for comparison for every cell. This didn't seem to work and I'd love to hear about any ideas or further insight into how this tool works.

Much appreciated!

Re: Returning names of rasters with maximum value

TOMKONG — Wed, 31 Jul 2013 17:44:02 GMT

Thanks lpinner, this was spot-on! In theory it is doing what I need, however in practice it's not working out exactly as planned. I have a follow up question I'm hoping I can get some help with:
...
Much appreciated!

Did you check that HighestPosition (or Upos) function works well or not?
I never use this function, but it seems the illustration from ESRI book is not correct - the return is not always the cell with Max value. ig: the downside left conner cell values from input raster datasets are 4, 3, 1, but the return value for this cell is 1.
But the illustration for Max (9 version) function is correct - this cell return value is 4.

If the HighestPosition works and your source rasters does't include the 0 value, you can asign the NoData to 0 first;
After you get the HighestPosition output, you can asign the 0 value back to NoData (those 0 value celles indicates all input raster datasets are the 0 value in those locations. You don't need to have any value for those celles) .

Re: Returning names of rasters with maximum value

Luke_Pinner — Wed, 31 Jul 2013 21:08:27 GMT

Running the tool with three or more input datasets seems to include only results that are an intersection of all the inputs. All other non-intersecting areas are NoData. What I would like is to have a dataset represented in the results, even if it is the only one with a value in a specific cell.

It's working as expected. The default processing extent is the intersection of input extents, change this to union - see the help for more info. However, any cell with NoData will be NoData in the output. So change it to a low number first.

i.e. in the raster calculator:

HighestPosition(Con(IsNull(raster1), -999, raster1), Con(IsNull(raster2), -999, raster2), Con(IsNull(raster3), -999, raster3))

Did you check that HighestPosition (or Upos) function works well or not?
I never use this function, but it seems the illustration from ESRI book is not correct - the return is not always the cell with Max value. ig: the downside left conner cell values from input raster datasets are 4, 3, 1, but the return value for this cell is 1.
But the illustration for Max (9 version) function is correct - this cell return value is 4.

No, the return value of 1 is correct. HighestPosition/Upos returns the position of the raster with the highest value in the list, not the maximum value itself.

For your example of 4, 3, 1, 4 is the max and it is at position 1. For something like 9, 12, 7, the HighestPosition/Upos value would be 2 as the raster with the highest value is at the 2nd position in the list.

Re: Returning names of rasters with maximum value

TOMKONG — Wed, 31 Jul 2013 22:35:21 GMT

It's working as expected. The default processing extent is the intersection of input extents, change this to union - see the help for more info. However, any cell with NoData will be NoData in the output. So change it to a low number first.

i.e. in the raster calculator:
HighestPosition(Con(IsNull(raster1), -999, raster1), Con(IsNull(raster2), -999, raster2), Con(IsNull(raster3), -999, raster3))
No, the return value of 1 is correct. HighestPosition/Upos returns the position of the raster with the highest value in the list, not the maximum value itself.

For your example of 4, 3, 1, 4 is the max and it is at position 1. For something like 9, 12, 7, the HighestPosition/Upos value would be 2 as the raster with the highest value is at the 2nd position in the list.

Thanks for your explanation about HighestPosition output.

If the output of the HighestPosition is the number of raster dataset, this function is not suitable for Jproville's purpose, its return is still to be replaced by the ralated dataset cell original value, and to eliminate the "NoData " output value.

The Max function is fit this application, but why ESRI removed this tool in the 10 version?

Re: Returning names of rasters with maximum value

Luke_Pinner — Wed, 31 Jul 2013 23:47:45 GMT

If the output of the HighestPosition is the number of raster dataset, this function is not suitable for Jproville's purpose, its return is still to be replaced by the ralated dataset cell original value, and to eliminate the "NoData " output value.

I understood from jproville's question that he wanted to find out which raster had the max value:

...one thing I'd like to see is a raster showing me which of the 23 layers contained the maximum value that was chosen.

The Max function is fit this application, but why ESRI removed this tool in the 10 version?

The Max functionality is still there, it's just been moved into a new tool. ESRI changed a lot of things at version 10, names (i.e. UPos -> HighestPosition) and merged some tools (Max/Min/etc = CellStatistics, FocalMax/FocalMean/etc = FocalStatistics etc...).

Re: Returning names of rasters with maximum value

JeremyProville — Mon, 05 Aug 2013 18:40:11 GMT

Thanks to both for your responses.
TKONG, indeed lpinner's first answer was right, I was looking for the HighestPosition tool and not the max tool.

lpinner, appreciate your code suggestion - I tried a great deal of variations on it, but I continue to be baffled by the fact that I only keep getting the intersection of the inputs rather than the union. No matter what I seem to try (e.g. tinkering with the GP extent, con() isnull() and setnull() in raster calculator), I get the same result. I've also tried forcing -999 NoData values through Catalog.

In any case, thanks for all your help and I'll post back if I figure out the problem. Let me know if you have any further ideas.

Re: Returning names of rasters with maximum value

curtvprice — Mon, 05 Aug 2013 20:02:37 GMT

I tried a great deal of variations on it, but I continue to be baffled by the fact that I only keep getting the intersection of the inputs rather than the union.

Are you saying the output extent is wrong (this is what the environment controls) or just the particular cells are only getting a value if all the inputs from the stack are data cells? If the latter, Luke's suggestion,

Con(IsNull(raster), -999, raster)

should work.

Re: Returning names of rasters with maximum value

JeremyProville — Mon, 05 Aug 2013 20:56:31 GMT

Are you saying the output extent is wrong (this is what the environment controls) or just the particular cells are only getting a value if all the inputs from the stack are data cells? If the latter, Luke's suggestion,
Con(IsNull(raster), -999, raster)
should work.

It is the latter - I've tried the code you mentioned and variations of it, but even by forcing NoData values into all input datasets, it still only gives me the intersection of all inputs, and does not consider all areas that were previously NoData. It is very strange.

The closest I've gotten to understanding this is by going in ArcCatalog and forcing in NoData values to -9999. But it still is erratic, I'll illustrate with an example:

Rasters A and B contain data. I force NoData values in these rasters to be -9999, either in Catalog or by using raster calculator with the code above. If I now use HighestPosition on these, it still only returns intersecting data in areas where values were anything but -9999... i.e. it seems to know where those NoData areas were, and omits them.
However, if I put in Raster A, and a constant dummy value of e.g. -1 as the inputs to HighestPosition, it works and includes all '-9999' areas in the output. Putting in Rasters A, B and the dummy constant makes it fail again, and the output ignores -9999 areas in it's calculation.

Any insight much appreciated.

Re: Returning names of rasters with maximum value

Luke_Pinner — Mon, 05 Aug 2013 22:15:31 GMT

I'm sure setting the output extent to MAXOF/union of inputs used to work in Workstation GRID/older versions of Spatial Analyst when the multiple inputs were Con(IsNull(...),...) functions.

I think what is happening is that ArcGIS is processing the raster calculator statement from inside to out - i.e it processes the Con(IsNull(...),...) statements individually and as there is only one raster in each of those, the output extent for each Con... statement is set to that of the single input raster, instead of parsing the whole statement and working out the extent for all rasters before processing the Con statements.

Workaround:

Generate a dummy raster with output extent set to MAXOF/union of inputs using something like raster1+raster2+etc...
Then set output extent = dummy raster and run HighestPosition(Con(IsNull(raster1),-999,raster1),etc...)

Re: Returning names of rasters with maximum value

JeremyProville — Tue, 06 Aug 2013 16:11:50 GMT

Thanks Luke - that and a combination of the other workarounds seems to have done the trick. I ended up making a blank 'extent' raster with -9999 values in every cell, and using that to set the extent. I also had to toss in a constant (-9998) in the final raster calculator formula, for HighestPosition. I put this in last place so that I could filter out these NoData (now -9998) areas.

Appreciate everyone's help!