AI-GEOSTATS: SUM: kriging and groundwater chemistry
- hello list,
following is the summary on my question from April 19. The
i have the following question regarding kriging and groundwater
i'm a grad student in GIS and i've chosen a topic
dealing with the interpolation of parameters for groundwater
chemistry by kriging (approx 3000 observation wells). i got tons of
additional data (landuse, hydrology, geology, soil parameters,
DGM , etc.....) and i'm looking for a kriging technique supporting
additional information (qualitative: landuse and aquifers as well as
quantitative: certain soil paramters, infiltration rate, etc). at first i
was thinking about Bayes-Markov-Kriging and/ or Simple Updating-
Kriging and finally compare the results of these with the results of
simulation. now after almost three weeks of intensive literature
review (Bardossy (Simple Updating), Goovaerts (Simple with
varying mean, Cokriging), Pebesma (Ordinary block), Lehmann
(Bayes-Markov and others), etc.....all dealing with kriging of
groundwater chemistry supporting additional information or kriging
of other z-values supporting additional information) and playing
around with some programs checking out their suitability and
capabilities (GSTAT, GSLIB, VarioWin, Geostatistical Analyst for
ArcInfo 8.2, SURFER 8) I'm a little bit confused about the
appropriate kriging technique and the program(s) to accomplish
this task. ...............
as a first step I want to analyse the data to distinct "real" outliers
that have to be removed from areas with exceptional high values
that shouldn't be removed (rather find the (hydro)geological sources
for their existence). as the data comes in ESRI's shape format I
really appreciated the interactive capabilites when testing their
Geostatistical Analyst for ArcInfo 8.2 (histogram, qq-plots, voronoi,
trend analysis, etc.) in combination with Moving Windows statistics
in ArcView 3.x and CrimeStat for some summary statistics (Moran
the Geostat Analyst also offers an interactive environment for
finding a suitable model variogram based on the experimental one
as well as several kriging techniques and crossvalidation. I'm still
not sure if these would be sufficient (in this case I was thinking
about Cokriging to handle the additional information as done by
Goovaerts for precipitation). Has anyone out there some
experiences with the GA (and cokriging with the second variable
being influential groundwater chemistry determining factors)?
working with gstat under W2K (or Linux....i got both running) would
have the advantages of being able to import/ export ASCIIgrid from
ArcInfo (running under W2K, that's whay I installed the Windows
verision of Gstat too) and do the simulation for comparison.
as this is a MSc thesis (limited time) and I have to mainly focus
on how to combine the most influental factors on groundwater
chemistry as additional
information for kriging, i don't have the time to focus on writing or
adapting routines (especially when considering my rudimentary
programming skills). so i'm looking for (an) existing program(s) to
support my results (the aforementioned appropriate groundwater
chemistry determing factors) with examples... which kriging
technique(s) and simulation technique(s) would be appropriate?
which of the aforementioned porgrams would be the best ("best" for
allowing quantitive and qualitative additional information for the
chosen kriging technique) option?
the three answers were (thanx alot to all of 'em!!!):
Hi I have not used geostatistics as you but.
I have applied Ordinary kriging in Idrisi 32 and ArcInfo geostatistical
Idrisi 32 provides a front end to Gstat with 3 modules and different
interfaces. Idrisi has a special option for inputing esri data formats.
ArcInfo had the same capabilities but I found it easier, with better
visualizing techniques, better tools for outputing the results
(perspective display ArcScene, 3d analyst etc.)But it takes only
values in the angles when you create the variograms while Idrisi
decimals as well
I do not know exactly about cokriging etc but as far as ordinary
the capabilities were almost the same but I suggest ArcInfo as the
wizard guides you and give you more options.
I hope this help.
Other references are those by Bob Hoeksema at Calvin College.
for modeling groundwater elevation, where surface elevation is a
and he developed
a program that implements an algorithm he and Peter Kitinidis
developed. You can probably do
something similar with Gstat. I'm sure others will discourage you
using Arc-Info, at
least for semivariogram modeling. A maximum likelihood program
estimating semivariogram parameters is better.
My MSc. thesis orientator forwarded your mail to me.
I am Chemical Engineer and I am working with groundwater, here in
Porto Alegre, Brazil.
I have only 170 monitoring wells with water table height values. I
will sample all or only some of them, looking for hydrochemical
parameters and then, apply kriging and cokriging for the water
table, and yet I have not decided wich technique to use with the
hydrochemical data, since I don't have it already.
I think that indicator kriging is a helpful tool when trying to use the
data in categories.
I work with GSLIB and with Variowin. Also using Surfer 8 and
AutoCad 2000. Variowin, is only for 2D analysis. GSLIB is the
software used here in Porto Alegre, and I like it. I don't know if
you're going to work in 2D or 3D.
I am a beginer in this subject, but I think that it is important for you
to look to your data set, in a GSLIB locmap. Then, there you will
see outliers, and see if they are outliers in the "neigborhood" being
considerated. If it is so, you can try to use the outlier, omit some of
the known data in the neigborhood, and interpolate using kriging.
Then, put it back again and repeat the routine. It might be helpful. If
not, it is a matter of decision, to use an outlier and obtain
overestimated values or maybe the other way.
* To post a message to the list, send it to ai-geostats@...
* As a general service to the users, please remember to post a summary of any useful responses to your questions.
* To unsubscribe, send an email to majordomo@... with no subject and "unsubscribe ai-geostats" followed by "end" on the next line in the message body. DO NOT SEND Subscribe/Unsubscribe requests to the list
* Support to the list is provided at http://www.ai-geostats.org