Combining Fuzzy membership values in ArcGIS 10 - Problems

[KarenBlakey]

Hello,  I have produced a number of raster layers that contain fuzzy membership values for soil parameters.  I believe that I have carried out the calculations correctly.  I have ensured that I have set the correct extent too.   I have calculated the convex combination of the raster layers that contain the fuzzy values using a number of different methods.  I have used the raster calculator, the fuzzy overlay and the weighted sum tools and they all produce the same raster layer with a lower range that is set the same as the value of my NoDataValue (I looked in the properties o the layer in the source tab). I was expecting the range to be withing 0 and 1. The range runs from +3,40282346639e+038 through to -3,40282346639e+038 .  I dont know if it is just coincidence but NoDataValue is set at -3,40282346639e+038  in all of my layers.  I cant see anything when i zoom to the layer either. Hence, it looks even more probable that the layer is full of cells with noData Values. Can anyone help me please? A thousand apologies if I havent given enough information for you to solve this problem.  I am not very experienced with this aspect of ArcGIS and also new to the forum.  Thanks in advance of a speedy reply. Karen  🙂

[KarenBlakey]

Hello,  I have produced a number of raster layers that contain fuzzy membership values for soil parameters.  I believe that I have carried out the calculations correctly.  I have ensured that I have set the correct extent too.   I have calculated the convex combination of the raster layers that contain the fuzzy values using a number of different methods.  I have used the raster calculator, the fuzzy overlay and the weighted sum tools and they all produce the same raster layer with a lower range that is set the same as the value of my NoDataValue (I looked in the properties o the layer in the source tab). I was expecting the range to be withing 0 and 1. The range runs from +3,40282346639e+038 through to -3,40282346639e+038 .  I dont know if it is just coincidence but NoDataValue is set at -3,40282346639e+038  in all of my layers.  I cant see anything when i zoom to the layer either. Hence, it looks even more probable that the layer is full of cells with noData Values. Can anyone help me please? A thousand apologies if I havent given enough information for you to solve this problem.  I am not very experienced with this aspect of ArcGIS and also new to the forum.  Thanks in advance of a speedy reply. Karen  🙂

If I haven't provided sufficient information to allow anyone to help me solve this I would appreciate guidelines on how I set out my question? Many thanks Karen

[KarenBlakey]

Has anyone any suggestions why the rater layer I have produced when I calculated the convex combination of the raster layers that contain the fuzzy values for some soil parameters such as pH, texture etc have such a strange range as stated above?  I am absolutely lost with this. I have tried all different ways of calculating the convex combination (I have used raster calculator,  ArcTool Box > Overlay > weighted sum; and I have also used ArcTool Box > Overlay > fuzzy overlay) and I get exactly the same result.  So I am assuming it is something about the layers I have included although I have checked the ranges of the individual raster layers and they appear to be correct.  Please help or let me know what additional information I need to include in my message so that I can get help.  Thanks, Karen

[MalcolmNunn]

Hi Karen

Not sure I can provide you with any definitive solutions here but I will attempt to help a little.

You would be right about the data range matching up with the NoData values.
Could you please tell me what the formats of your input and output rasters are, and what pixel depth they are in? This may provide a clue as to what is happening. Is your value range 3,40282346639e+038 or 3.40282346639e+038 by the way?

Also, are there any sensible values in your raster at all? You can perhaps check this by changing the symbology to Unique Values and seeing what values are actually present.

Also, you might like to include a specific example of the tool you are using (your raster calculator operation for example).

Cheers
Malcolm

[KarenBlakey]

Hi Karen

Not sure I can provide you with any definitive solutions here but I will attempt to help a little.

You would be right about the data range matching up with the NoData values.
Could you please tell me what the formats of your input and output rasters are, and what pixel depth they are in? This may provide a clue as to what is happening. Is your value range 3,40282346639e+038 or 3.40282346639e+038 by the way?

Also, are there any sensible values in your raster at all? You can perhaps check this by changing the symbology to Unique Values and seeing what values are actually present.

Also, you might like to include a specific example of the tool you are using (your raster calculator operation for example).

Cheers
Malcolm

Hi Malcolm,  thanks very much for your reply.  I think I have solved it.  I will explain how I resolved things in case anyone has the same problems. I decided to start everything again but just before I did, I zoomed to full extent.  I then noticed that one of the layers was in a completely different position on the canvas.  I then checked the projected co-ordinate system associated with the layers and I found the problem.  I thought I had used the same projected co-ordinate system for all layers BUT the far eastings and far northings were different in the layer that was mapped in the wrong location.  The original shape file was the one that was affected and so every other raster layer I had created from it was also misplaced on the map.  Once the correct co-ordinate was associated with the file all of the layers were arranged as was expected.  However, when I zoomed to full extent again it still appeared that there was a lot of empty space and my layers appeared like a small stamp on a larger envelope.  I could not change this and if you have any idea why this occurred even when I corrected all of the layers I would love to hear from you again.  Instead, added all of the raster layers to a new project and then used the weighted sum tool to create the new raster layers of the convex combination fuzzy membership values. Apologies if I haven't explained myself well enough.  I am very new to GIS as an spatial analysis tool.  Once again, thank you so much for your reply.  It is greatly appreciated.  Karen

[MalcolmNunn]

It sounds like the data frame retained the full extent of all the layers (including the previous erroneous one that you fixed). Did you remove all the layers that had the erroneous extent? I just tested clipping a raster to a smaller extent, removing the original, then selecting Full Extent - the data frame extent was automatically adjusted to the largest of all the layers present (now the smaller one).
Perhaps you could try going into the Data Frame properties -> Extent and change it to something else then back to automatic to see if it re-adjusts, just to double-check. If the previous extent remains, there must be a layer present that has the old extent - eg when a tool was used to produce one of your current layers that used the old geoprocessing extent.

In any case, it sounds like you found your solution. When your output raster looks like it only contains NoData values as in your case, it suggests there was an issue with either your input rasters (eg wrong extent ) or the operation itself (eg a bad Con statement that produces NoData for every pixel).

Hope this has helped.

Cheers
Malcolm

[KarenBlakey]

It sounds like the data frame retained the full extent of all the layers (including the previous erroneous one that you fixed). Did you remove all the layers that had the erroneous extent? I just tested clipping a raster to a smaller extent, removing the original, then selecting Full Extent - the data frame extent was automatically adjusted to the largest of all the layers present (now the smaller one).
Perhaps you could try going into the Data Frame properties -> Extent and change it to something else then back to automatic to see if it re-adjusts, just to double-check. If the previous extent remains, there must be a layer present that has the old extent - eg when a tool was used to produce one of your current layers that used the old geoprocessing extent.

In any case, it sounds like you found your solution. When your output raster looks like it only contains NoData values as in your case, it suggests there was an issue with either your input rasters (eg wrong extent ) or the operation itself (eg a bad Con statement that produces NoData for every pixel).

Hope this has helped.

Cheers
Malcolm

Hello again Malcolm, once again thanks very much for your reply. You are exactly right.  the whole thing baffled me because I know I painstakingly made sure the extent was set to what I required for every operation. But I do recall it looking as though it was changing as I clicked the OK button.  I went back to the original project because I had been thinking on these lines and I found one of the layers that had the error in the extent values.  I could not work out how to change the individual extent value (the left value had been set at 0) so i re-ran the polygon to raster operation that I had used previously and removed the other layer from the project.  I was delighted when everything snapped to the correct position and no more unwanted white space!

So once again thanks for your reply.  I will try changing the extent in data frame properties too.  When I tried to change the extent in the layer that had the error it wouldn't let me.  I could have moved on and left this but I feel that trying to understand out how ArcGIS doesn't work has allowed me to learn so much more about how IT does 🙂

Cheers! Karen