Jump to content

 
Photo

Correlation between rasters

- - - - -

  • Please log in to reply
6 replies to this topic

#1
NeverEasy

NeverEasy

    Contributor

  • Validated Member
  • PipPip
  • 28 posts
  • Finland

Hello!

I am a student studying environmental planning, and at the moment I am in a very ArcGIS-centered internship. I am still quite new to ArcGIS, so even relatively easy tasks are often hard for me.

What I would like to know is: If I have one raster describing the interpolated depth of a seabed, and another raster describing where bay mussel's can be found (in percent, i.e. a cell showing a value of 5, would mean that at that spot 5% of the seabed is covered in bay mussel), how can I with ArcGIS calculate and visualize (preferably via a new raster) a correlation between depth and abundance of bay mussel? Hope that's clear enough :)

Thanks!

Oh and our school has the 3D analyst and Spatial analyst extensions, I am not able to download any new ones.

Edited by NeverEasy, 06 May 2010 - 08:28 AM.


#2
Keith Map Service

Keith Map Service

    Contributor

  • Validated Member
  • PipPip
  • 18 posts
  • Gender:Male
  • Location:Mobile, AL
  • Interests:Geography, mapping, GIS, Fishing, Outdoors, Mountain Biking
  • United States

Try clicking on the 3D analyst toolbar and click on Raster Calculator.
From here choose the two corresponding layers and multiply them together with
a statement. [layer]*[layer] click calculate. This should do the trick. To make
layer permanent right click then click 'Make Permanent'.
Keith Map Service, Inc.
http://keithmaps.com

#3
frax

frax

    Hall of Fame

  • Associate Admin
  • PipPipPipPipPipPipPip
  • 2,303 posts
  • Gender:Male
  • Location:Stockholm, Sweden
  • Interests:music, hiking, friends, nature, photography, traveling. and maps!
  • Sweden

How does a multiplication of the layers say anything about the correlation?

What I am thinking a bit is that you might want to classify the depths into e.g. equal interval classes (50m? don't know what span we are talking about) and do the same with mussels (5 classes?). Then extract the data so that you get a chart with depth classes on the x axis and number of cells in the y axis. The data would be a stacked chart with number of cells in each mussel class. I hope you follow what I mean!

After the classification of both layers, you can run a combine to get a value table that you can use in e.g. excel.
Hugo Ahlenius
Nordpil - custom maps and GIS
http://nordpil.com/
Twitter

#4
dsl

dsl

    Master Contributor

  • Validated Member
  • PipPipPipPip
  • 225 posts
  • Gender:Male
  • Location:Denver
  • United States

You can also calculate the correlation coefficient to create a correlation coefficient surface. Here is an explanation on how to do a moving average correlation coefficient spatial analyst. The author of the response in that post goes through it pretty quickly. Here is the wikipedia page about correlation, if you don't already know. Your X variable will be seabed depth, and your dependent Y variable will be mussel abundance.

Hope that helps,
David

#5
NeverEasy

NeverEasy

    Contributor

  • Validated Member
  • PipPip
  • 28 posts
  • Finland

Thanks for the comments! I actually found my own way of vizualising the correlation. I made a thiessen layer, and via "multiple attributes" i symbolized the squares to match the depth and the amount of blue mussel is shown via different sized cirkels, within the squares. Frax's way wasn't really what I was looking for and dsl's way was a little too advanced for me :) But again, thanks for helping!

#6
frax

frax

    Hall of Fame

  • Associate Admin
  • PipPipPipPipPipPipPip
  • 2,303 posts
  • Gender:Male
  • Location:Stockholm, Sweden
  • Interests:music, hiking, friends, nature, photography, traveling. and maps!
  • Sweden

Hi - I'd be curious how easy that map is to read, and how visible the relationship is. I would expect it to bet quite busy!
Hugo Ahlenius
Nordpil - custom maps and GIS
http://nordpil.com/
Twitter

#7
NeverEasy

NeverEasy

    Contributor

  • Validated Member
  • PipPip
  • 28 posts
  • Finland

The area in question is about 4*2 kilometers and there are 560 squares in the thiessen layer. This might (or might not) sound like a lot, but since the korrelation is pretty strong, you only need to look at the darkest (ie. deepest) squares to notice that they rarely have much bay mussel. So it's really not as busy as one might expect. I scratched the idea of using interpolation and rasters, since the point really was to visualize correlation - not predict depth.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users

-->