[Qgis-user] Stratified stats of Raster map by vector map in QGIS?

[Agustin Lobo]

Carson,

Sorry I did not answer to this.

Yes, this would be a workaround (just have to figure out how to get
the centers of the pixels, probably an R function will be the best),
but just in the case pixels are a lot smaller than polygons,
which is not always the case.

Note that many thematic maps are developed
from RS images in which pixels are (at least, should be) very small 
compared to the
resulting polygons (which are produced either by digitizing and 
interpretation, as the
Habitat maps in Europe, which are developed from 1:25M and 1:5M 
orthoimages with sub-meter resolution; or by (best segment-based) 
classification and subsequent raster to vector conversion).

But an important additional use of RS imagery
is to actually get information from other sources of imagery for each 
type of polygon, and these RS images can be much coarser than the
images used to generate the polygons. For example, I use time series
of NDVI, which are 300m to 1000m resolution only, to derive phenologic
attributes. In this application, you get that the boundaries
of polygons intersecting pixels very often, and the program must provide 
the user the choice of (i) weighted mean, (ii) ignoring any not fully 
included pixel
or (iii) count any partially included pixel as if it were fully included.

The application of transferring information from raster images to vector
thematic maps is crucial for RS applications in the GIS world and, IMO, 
cannot be left
to v.rasts.stats, we need a native tool within QGIS itself. It would be
great if you could consider including this functionality in fTools.

Thanks a lot for your plugins, they are very useful.

Agus


Carson Farmer wrote:
> Agus,
>> Given a vector map (i.e., a land cover map)
>> and a raster of a continuous variable (i.e., height),
>> is there any plugin calculating a table of average
>> (and perhaps other stats) of the raster for
>> each category in the vector?
> I don't know of any plugins that will do this directly, but here is one 
> possible solution:
> 
> If you are able to convert your raster to points (i.e. extract the 
> centre of each raster cell as a point), then you could use the 'Join 
> attributes by location' tool in fTools to extract the info that you need 
> (and you can opt to calculate the mean values).
> 
> Hope this helps,
> 
> Carson
> 
-- 
Dr. Agustin Lobo
Institut de Ciencies de la Terra "Jaume Almera" (CSIC)
LLuis Sole Sabaris s/n
08028 Barcelona
Spain
Tel. 34 934095410
Fax. 34 934110012
email: Agustin.Lobo at ija.csic.es
http://www.ija.csic.es/gt/obster