[R-sig-Geo] predict a map from point data

[Tom Gottfried]

Hi Sadz,

Am 01.03.2011 15:59, schrieb Sadz A:
> Hi Tom, 
> 
> For the krigging I used variogram and modelling through arcGIS. the
> predictions are bad in the sense that If predictions are plotted against
> real data (where you would get a 1:1 line if it was a perfect model) I
> get strait horizontal line. 
> 
> (what is 'a Gaußian model and doing block kriging', and how does it
> work/differ from the Arcgis approach (which was called ordinary
> krigging)?

You can consult any textbook on geostatistic to learn about it (e.g. Goovaerts, 1997. Geostatistics
for Natural Resources Evaluation). For R-specific things the ASDAR-book (Bivand et al., 2008,
Applied Spatial Data Analysis with R) is appropriate.

> I think the problem with using the krigging is that it
> assumes that things that are closer are more alike, 

... respectively more "similar". Depending on the semivariogram-model, this is often assumed.

> but with the volume
> this does not necessarily seem to hold true...

Thinking of steres of wood standing on a unit area, I would intuitively suggest that it does hold
true. But this is beyond my (and probably the lists) scope.

regards,
Tom

> no idea how to get around this.
> 
> Thank you for your response
> sadz
> 
>   
> 
> ------------------------------------------------------------------------
> *From:* Tom Gottfried <tom.gottfried at wzw.tum.de>
> *To:* r-sig-geo at r-project.org
> *Sent:* Mon, 28 February, 2011 9:41:25
> *Subject:* Re: [R-sig-Geo] predict a map from point data
> 
> Hi Sadz,
> 
> Am 28.02.2011 01:54, schrieb Sadz A:
>> Hi,
>>
>> I'm trying to predict the distribution of timber over an area, I have
> point
>> location data- so it would make sense to use a krigging to interpolate
> the data
>> over the whole map. Unfortunately the krigging predictions are pretty
> bad.
> 
> What does "pretty bad" mean? What did you actually do? There are many
> "flavours" of doing kriging
> including variogram estimation and modelling. E.g. you seem to
> interpolate volume data, so I suppose
> your "point location data" actually represent a volume too. Sounds like
> fitting a Gaußian model and
> doing block kriging could be appropriate.
> 
>>
>> I have now got environmental data for my study site and have used the
> r package
>> 'quantreg' to make a model that I can predict the volume from
>> R code:
>> fit <- rqss(sum_vol~qss(env.factor1,lambda=1)+
> qss(env.factor2,lambda=1), tau =
>> 0.9)
>> predi<-predict(fit, new, interval = "none", level = 0.9)
> 
> Do you mean you have any covariates? Maybe multivariable geostatistics
> (the gstat package) are your
> friend.
> 
>>
>> unfortunately this is not working either (a problem with the predict
> function).
>> I have also tried linear modelling, Gams and inverse interpolations.
> 
> Any errors, traceback()?
> 
> regards,
> Tom
> 
>>
>> Does anyone have any ideas on how I could get a spatial map of volume
>> distribution from the point data?
>>
>> Any help is appreciated,
>> thank you
>> sadz
>>
>> ps- if anything is unclear I would be happy to clarify
>>
>>
>>
>>     
>>     [[alternative HTML version deleted]]
>>
>> _______________________________________________
>> R-sig-Geo mailing list
>> R-sig-Geo at r-project.org <mailto:R-sig-Geo at r-project.org>
>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 
> -- 
> Technische Universität München
> Department für Pflanzenwissenschaften
> Lehrstuhl für Grünlandlehre
> Alte Akademie 12
> 85350 Freising / Germany
> Phone: ++49 (0)8161 715324
> Fax:  ++49 (0)8161 713243
> email: tom.gottfried at wzw.tum.de <mailto:tom.gottfried at wzw.tum.de>
> http://www.wzw.tum.de/gruenland
> 
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org <mailto:R-sig-Geo at r-project.org>
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
> 

-- 
Technische Universität München
Department für Pflanzenwissenschaften
Lehrstuhl für Grünlandlehre
Alte Akademie 12
85350 Freising / Germany
Phone: ++49 (0)8161 715324
Fax:   ++49 (0)8161 713243
email: tom.gottfried at wzw.tum.de
http://www.wzw.tum.de/gruenland