Skip to content

the volcano orientation

7 messages · Michael Sumner, Ben Tupper, Abby Spurdle +2 more

Michael Sumner
# 
Does anyone know why 'volcano' is oriented as it is?

image(volcano)  ## filled.contour is the same

I know it's all arbitrary, but north-up is a pretty solid convention. Is
there any reason why the classic 'image()' example data set would not
default to this orientation?

A Google map of the site (in Web Mercator):

https://www.google.com/maps/place/Maungawhau+%2F+Mount+Eden/@-36.8763271,174.7619561,856m/data=!3m1!1e3!4m8!1m2!2m1!1smaungawhau!3m4!1s0x6d0d47db8d7bd1ff:0x8bcffe2a5c7360d2!8m2!3d-36.8666667!4d174.7666667


For image(), the north-up orientation is 't(volcano[,ncol(volcano):1])'.

If you are interested in a roughly georeferenced version I have code here:

https://gist.github.com/mdsumner/20fe3ffa04421bf8e0517c19085e5fd8

(Also see fortunes::fortune("conventions") )

Best, Mike
Ben Tupper
# 
Hi,

Hmmm.  The only place I have ever seen a georeferenced version of 'volcano'
is here...

https://waterdata.usgs.gov/blog/inlmiscmaps/

It was on the internet so I assumed it was true. Now, I suspect that, since
the original survey by Ross Ihaka, continental drift is happening waaaay
faster than anyone guessed. Could be a decent grant proposal somewhere in
all this.

Seriously, though, I haven't any idea why 'volcano' is the way it is shown,
nor was I awake enough to actually look at a map as you have done. I would
love-love-love to see a georeferenced version be part of the stars package
as example data. It's small enough to be lightweight but has enough
information in it to be handy for meaningful demonstrations. Maybe along
the lines of ...

https://gist.github.com/btupper/8e8eb8c0ebf4402a3f87b5638eca954a

... but with the correct spatial info.

Cheers,
Ben
On Sat, May 9, 2020 at 8:44 PM Michael Sumner <mdsumner at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Abby Spurdle
    

    

      
      #
    

  


  

          

      
      
      
    

        
Great question!

graphics::image produces a "plot".
It follows the same x y conventions as other plots in the graphics package.
It's *defaults* are not designed to display photos, etc.

However, the format of the volcano data is not consistent with either
the defaults of graphics::image or what I (personally) would expect in
a photographic data, with top-left value at top-left of matrix and
bottom-right point at bottom-right of matrix, but that's debatable...

According, the documentation of the volcano data:

    A matrix with 87 rows and 61 columns,
    rows corresponding to grid lines running east to west
    and columns to grid lines running south to north.

Perhaps that could be improved slightly...?

And one more thing that caught me out.
My initial expectation (using a simple interpretation) was the data
would need to be transposed and then either the ylim reversed or the
rows reversed.
But when I tried to plot the volcano data using my own function (which
does just that), I got the wrong result.

But in the documentation for graphics::image we have:

    "Need to transpose and flip"
    image(t(volcano)[ncol(volcano):1,])

Which produces the right result.

I had to think about this for a while...

The example for graphics::image above is actually transposing the
matrix *twice*.
First in the input to the function, and then again (implicitly), where
rows (going down the data) are interpreted as x (going right across
the plot).

  
  


  


      

    

    

      
      

        


  

        

      Abby Spurdle
    

    

      
      #
    

  


  

      
Sorry, one more thing.
My response didn't really answer your question.
But I would say that the formats of most datasets used in statistics
are reflective of the preferences of the people that collected or
published them, at the time...

Also, I've found the older publications quite often have considerable merit...

        
On Sun, May 10, 2020 at 12:44 PM Michael Sumner <mdsumner at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Richard O'Keefe
    

    

      
      #
    

  


  

      
Hey, I know that volcano!  It's walking distance from the Intermediate
school I attended.
To you it's a plot; to me it's a place.
So I offer you four scenarios.

1. You think of it as a place you know and have been.
    In that case the "right" orientation is the one that best matches
what you are used to seeing.
    For me, that would put the peak on the right of the plot.

2. You think of it as a patch in a map.
    In that case the "right:" orientation is the one that matches the map.
    That would put the peak at the bottom of the plot.

3. You think of it as a product of geological processes, and are
perhaps interested in
    whether there is any connection between the orientation of the
volcano and the
    direction the Auckland hot-spot (currently at White Island) was moving.
    In that case you'd choose south-west -> north-east as the primary axis.
    (I think.  Not really sure.)

4. You think of it as a picture, an illustration in a textbook.  It
might need to be cropped
    vertically so you can fit another illustration on the same page.
For that and
    perceptual reasons you want the major linear axis of the image to
be  horizontal.
    In that case, what we have now is a perfectly reasonable choice.

"Quality is fitness for use."

        
On Sun, 10 May 2020 at 12:44, Michael Sumner <mdsumner at gmail.com> wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      John
    

    

      
      #
    

  


  

      
Out of curiosity, and considering the bewildering array of projections
and grids in use for various mapping purposes, you seem to be saying in your
example 2 that the grid coordinates number south to north and east to
west.  Given scale of the coordinate numbers, would that be a national
grid system employed in New Zealnd?

J. W. Dougherty

On Mon, 11 May 2020 13:56:49 +1200

        
"Richard O'Keefe" <raoknz at gmail.com> wrote:

        

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Richard O'Keefe
    

    

      
      #
    

  


  

      
Like other countries, New Zealand revises its maps and its coordinate
system from time to time.  The one in use at the time that image was
digitised is probably the one described here:
https://www.linz.govt.nz/data/geodetic-system/datums-projections-heights/projections/new-zealand-map-grid-nzmg

        
On Tue, 12 May 2020 at 04:58, John via R-help <r-help at r-project.org> wrote: