Skip to content

Question about derivative work - what is the license for map derived using e.g. spatial "predict" function?

7 messages · Forrest Stevens, Edzer Pebesma, Barry Rowlingson +2 more

Original
Tomislav Hengl
# 
Dear list,

I have a question about licensing the data that is produced by spatial 
prediction from point data. My ideas is that a map produced by using 
e.g. geostatistics from point data is a new data product and as such 
does not falls under the regulations of the original license used for 
the point data (so if the license for the point data is restrictive, the 
license for the output maps does not have to respect this). Consider for 
example:

R> library(gstat)
R> library(sp)
R> demo(meuse, echo=FALSE)
R> m <- vgm(.59, "Sph", 874, .04)
R> x <- krige(log(zinc)~1, meuse, meuse.grid, model = m)

The produced map "x" can be considered a new data product. There is 
absolutely no way that one could reproduce the original input points 
from this map, hence it should be considered "a non-derivative work". 
Only if we would derive a map using interpolation technique that allows 
re-constructions of points (e.g. Thiessen polygons) the license would 
need to be respected.

Or am I mistaken? (I know this is a type of a question for lawyers in 
fact, but any experience / opinion you have is welcome)

http://www.publicdomainsherpa.com/derivative-work.html

"To qualify as a derivative work, the derivative must use a substantial 
amount of the prior work?s expression. How much? Enough so that the 
average person would conclude that it had been based on or adapted from 
the prior work"

thank you,
Forrest Stevens
# 
This is indeed a sticky question, and probably hard to generalize across
specific legal situations, but I'm sure the original data holder's view
would be that without access to the original points layer there's no way
you could produce your interpolation. Therefore, your output constitutes a
derivative work.

I think of it like classifying data that you extract in some way from
Google Earth, which is against their Terms of Service. Even if you produced
a binary, classified land cover dataset from a screen capture, yielding a
situation like you describe, impossible to recreate the original
aerial/satellite imagery from, you'd still legally be violating the TOS and
couldn't distribute your classified data.

Interested to hear other's thoughts on it though,
Forrest


On Thu, Nov 27, 2014 at 8:37 AM Tomislav Hengl <hengl at spatial-analyst.net>
wrote:

  
  


  


      

    

    

      
      

        


  

        

      Edzer Pebesma
    

    

      
      #
    

  


  

      
Tom, in your example below, x contains the kriging variance; points with
zero kriging variance must be observation locations, with the predicted
value equal to the observation.

In case the nugget in m would have been replaced by a measurement error
component (Err in m and m1 below), you would not have this effect, and
also have no discontinuity in the interpolated surface at observation
locations:

            

      
      
      
    

        
model      psill    range
1   Err 0.05066243   0.0000
2   Sph 0.59060780 897.0209

            

      
      
      
    

        
model      psill    range
1   Nug 0.05066243   0.0000
2   Sph 0.59060780 897.0209

            

      
      
      
    

        
[using ordinary kriging]
       x      y var1.pred     var1.var
1 181072 333611  6.929517 1.110223e-16

            

      
      
      
    

        
[using ordinary kriging]
       x      y var1.pred   var1.var
1 181072 333611  6.884401 0.03648868
# note identical predictions but different variances
# for other locations:

            

      
      
      
    

        
[using ordinary kriging]
       x      y var1.pred  var1.var
1 181180 333740  6.499624 0.3198084

            

      
      
      
    

        
[using ordinary kriging]
       x      y var1.pred var1.var
1 181180 333740  6.499624 0.269146

        
On 11/27/2014 02:35 PM, Tomislav Hengl wrote:

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Tomislav Hengl
    

    

      
      #
    

  


  

      
Haa! Thanks for spotting this Edzer. But I am only talking about the 
predictions in this example. The predictions from any sp prediction 
method do not go through points (plus you do not have the coordinates of 
the points). So the question is - is it a derivative work or not?

        
On 27-11-2014 15:44, Edzer Pebesma wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Barry Rowlingson
    

    

      
      #
    

  


  

      
On Thu, Nov 27, 2014 at 2:44 PM, Edzer Pebesma

        
<edzer.pebesma at uni-muenster.de> wrote:

            

      
      
      
    

        
Now all that is going to be fun to explain to a judge and jury.

 Suppose you took all the notes of a Bach fugue as X=time, Y=pitch,
and interpolated them in time, then created a new piece using the
interpolation prediction at time points between the notes, would this
be a derivative work?

Yes, I think: "A ?derivative work? is a work based upon one or more
preexisting works, such as a translation, musical arrangement,
dramatization, fictionalization, motion picture version, sound
recording, art reproduction, abridgment, condensation, or any other
form in which a work may be recast, transformed, or adapted."
[Wikipedia, where I get all my legal advice from] - they key word
being "transformed".

But does a dataset count as a "work" here? It is supposedly a set of
measurements of "the truth", rather than something that is a creative
work. A random australian web site that comes up tops in a google
search says:

Use of a Copyright Licence

A dataset may attract copyright protection (as a literary work), if it
meets certain threshold criteria of human authorship, originality, or
creativity, for example. On that basis, significant quantities of
research data will attract copyright protection. As such, it may not
be reused by researchers (or anyone else) without permission.
[http://ands.org.au/guides/copyright-and-data-awareness.html]

 I suspect the permission to make derivative works has to be stated
when the dataset gives usage permission.

Minefield.

Barry

  
  


  


      

    

    

      
      

        


  

        

      Tomislav Hengl
    

    

      
      #
    

  


  

      
Hi Barry,

Thanks for reminding me about the "you can not copyright facts" 
principle 
(http://www.newmediarights.org/business_models/artist/are_facts_copyrighted). 
So the point measured values can be considered to be, in fact, - facts.

So here is one more time the question to everyone (I am sure some of you 
have experienced these issues in practice):

1. A point data set has a restrictive license.
2. From this point data you can derive predictions - these predictions 
can not be used to re-produce point locations and values, hence the 
source is untraceable.
3. Does this derivative work (spatial prediction) has to follow the same 
data license as used for the point data, or this a new data sets for 
which you can set license freely.

Some basic principles to consider:

1. One can not copyright facts 
(http://www.newmediarights.org/business_models/artist/are_facts_copyrighted) 
but facts can be protected using e.g. the "trade secret" laws.
2. If one would digitize content from Google Earth / maps imagery i.e. 
digitize polygon maps representing geomorphology without a permission 
from Google this would mean breaking of the Legal Notices 
(http://www.google.com/help/legalnotices_maps.html) - has anyone bee 
prosecuted for this ever?
3. "The transformation, modification or adaptation of the work must be 
substantial and bear its author's personality to be original and thus 
protected by copyright" (http://en.wikipedia.org/wiki/Derivative_work) 
hence it is OK to create derivative works as long as one can prove that 
there are "substantial" new elements. Do spatial predictions from points 
fall into this category?

thank you!

T. Hengl

        
On 27-11-2014 16:38, Barry Rowlingson wrote:

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Jochen Albrecht
    

    

      
      #
    

  


  

      
What makes Tomislav's question really hard to answer is that there is no
*one* copyright law. Just look at the long list of exceptions in the Google
legal notices that he provided. My experience is only with the United
States and here the matter is relatively clear (in spite of Hollywood's
extreme provisions).
1. As mentioned, facts are not copyrightable; so if you have just plain
data without any create packaging/mapping/analysis then these are not
protected - regardless of what the vendor may claim (and some vendor are
very creative with their claims).
2. Your derivation is your work and does therefore not inherit any
copyright either- even if the original was copyrighted, say because of the
packaging. The Google Earth digitization is a case here. Possibly not in
Israel, but here in the U.S. you may freely digitize , i.e., create new
data - the boundaries that you are digitizing are not part of the source
material but your (creative) interpretation. You may, however, not copy;
that is, you may not re-assemble the tiles.
3. The notions of original vs. derivative are every once in a while
contested in court but so far, the courts have usually sided with the
defense.
There are only two authors in our field that have published on this topic
that I am aware off. Check out either Onsrud or Cho for a long list of
cases.
Unless you find a way to make millions of your work (which would interest
me immensely) you are unlikely to be prosecuted across national boundaries
because the claimant's effort would not be worth it.
Cheers,
     Jochen

Dr. Jochen Albrecht
Computational and Theoretical Geography
HN 1030
Hunter College CUNY
695 Park Avenue
New York, NY 10065

On Thu, Nov 27, 2014 at 4:50 PM, Tomislav Hengl <hengl at spatial-analyst.net>
wrote: