[R-sig-Geo] How to automatically fit variogram in gstat

[Paul Hiemstra]

Hi Kai,

The note section of the documentation of fit.variogram (?fit.variogram)
tells you what is done in this function.

cheers,
Paul

Kai Zhang wrote:
> Hi Paul,
>
> Thanks for your reply. I have one more question on the fit.variogram function in gstat. Let's take an example from the gstat manual
>
> fit.variogram(vgm1, vgm(1,"Sph",300,1))
>
> My understanding is that the parameters like partial sill, range and nugget specified in vgm are just starting values and the key message is model type such as exponential model or spherical model. The fit.variogram function determines the optical parameters automatically. I am not sure if my understanding is right because I could not find the detailed description in the manual.
>
> I appreciate your response and best regards,
> Kai 	
>
> ____________________________________________________
> Kai Zhang
> PhD Candidate, Environmental Health Sciences
> School of Public Health                        
> University of Michigan                         
> 109 S. Observatory                       
> Ann Arbor, MI 48109-2029         
> ____________________________________________________
>
> ======= 2009-12-14 05:07:05 You wrote =======
>
>   
>> Hi Kai,
>>
>> If it improves depends on how you define improvement. If you want to use
>> IDW automatic you need to fit the distance power in some way. When you
>> fit the variogram automatically, you get this distance dependency of the
>> weights from the variogram. In regard to co-kriging, if there is a
>> strong correlation between the various pollutant measures (co-located)
>> co-kriging can be a good option.
>>
>> cheers,
>> Paul
>>
>> ps I included R-sig-geo again as a cc so the conversation is also on the
>> mailing list. It is best to use "Reply to all" when you reply to these
>> kind of e-mails.
>>
>> ZK wrote:
>>     
>>> Hi Paul,
>>>
>>> Many thanks for your suggestion. You are right and I would expect to Kriging improves a little compared to inverse distance method. Btw, do you think co-kriging is useful for 8 monitors?
>>>
>>> Thanks,
>>> Kai
>>> 	
>>>
>>> ____________________________________________________
>>> Kai Zhang
>>> PhD Candidate, Environmental Health Sciences
>>> School of Public Health                        
>>> University of Michigan                         
>>> 109 S. Observatory                       
>>> Ann Arbor, MI 48109-2029         
>>> ____________________________________________________
>>>
>>> ======= 2009-12-11 04:23:11 You wrote =======
>>>
>>>   
>>>       
>>>> Hi,
>>>>
>>>> One option is to use the automap package, this provides the function
>>>> autofitVariogram and autoKrige. Automap is available on CRAN. An
>>>> alternative is to use the fit.variogram function from gstat (which
>>>> automap also does).
>>>>
>>>> An additional comment, 8 stations is not a lot to base variogram models
>>>> on. Think carefully how you want to solve this problem. Maybe you can
>>>> make some kind of aggregated sample variogram that incorporates not only
>>>> the data from one day, but of one month. In this way you can increase
>>>> the amount of data you use to create the variogram. But still, just 8
>>>> stations looks to be quite a challenge.
>>>>
>>>> good luck!
>>>> Paul
>>>>
>>>> ZK wrote:
>>>>     
>>>>         
>>>>> Dear all,
>>>>> I have daily pollutant measurements (CO, NO2, PM2.5 and Ozone) at 8 monitors from 2005 to 2008.  My purpose is to estimate daily exposures for studying subjects using Kriging for three years.  A common procedure is to plot experimental variogram and then to choose a theoretical model (e.g., exponential model or Gaussian model) with some adjustment on parameters.  However, this manual way is not appropriate for my application. I am wondering how I can fit variogram in gstat automatically. 
>>>>>
>>>>> I appreciate your response and best regards,
>>>>> Kai
>>>>>
>>>>> 	
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> __________________________________________________
>>>>> ¸Ï¿ì×¢²áÑÅ»¢³¬´óÈÝÁ¿Ãâ·ÑÓÊÏä?
>>>>>
>>>>> _______________________________________________
>>>>> R-sig-Geo mailing list
>>>>> R-sig-Geo at stat.math.ethz.ch
>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>>>>>   
>>>>>       
>>>>>           
>>>> -- 
>>>> Drs. Paul Hiemstra
>>>> Department of Physical Geography
>>>> Faculty of Geosciences
>>>> University of Utrecht
>>>> Heidelberglaan 2
>>>> P.O. Box 80.115
>>>> 3508 TC Utrecht
>>>> Phone:  +3130 274 3113 Mon-Tue
>>>> Phone:  +3130 253 5773 Wed-Fri
>>>> http://intamap.geo.uu.nl/~paul
>>>>
>>>>     
>>>>         
>>> †Ûi??'™¨¥É¨h¡Ê&
>>>       
>> -- 
>> Drs. Paul Hiemstra
>> Department of Physical Geography
>> Faculty of Geosciences
>> University of Utrecht
>> Heidelberglaan 2
>> P.O. Box 80.115
>> 3508 TC Utrecht
>> Phone:  +3130 274 3113 Mon-Tue
>> Phone:  +3130 253 5773 Wed-Fri
>> http://intamap.geo.uu.nl/~paul
>>
>>     
>
>
> †Ûi'™¨¥É¨h¡Ê&


-- 
Drs. Paul Hiemstra
Department of Physical Geography
Faculty of Geosciences
University of Utrecht
Heidelberglaan 2
P.O. Box 80.115
3508 TC Utrecht
Phone:  +3130 274 3113 Mon-Tue
Phone:  +3130 253 5773 Wed-Fri
http://intamap.geo.uu.nl/~paul