[SOLVED] general help

If you have problems using RGeostats package or if you simply need some help on an unlisted subject.

[SOLVED] general help

Postby radillon » Wed Dec 06, 2017 5:19 pm

I am having some trouble with basic understanding of geostatistical analyses. I am interested in figuring out at what bin size (or EDSU unit) my fish density estimates from hydroacoustics data is autcorrelated and at what distance it is not correlated. Eventually, I will also want to compare variance estimates from geostastical methods to arithmatic methods, as well as krig my density estimates over the entire reservoir. I know I need to calculate an experimental variogram and then fit a model to that variogram, and based on that plot I can tell at what distance my data is autocorrelated vs not. However, when I run my data through a script I put together to do just this, the variogram plots I get back are nothing like I've seen in any manuscript or help page online.

I am working in very small reservoirs, and our survey design is zig-zag transects. The data I feed into the script are density estimates calculated in 50-m vertical bins, summed over the water column depth. Perhaps my variogram looks odd because the bin size I use to feed my analyses is too small? Should I try out different bin sizes? Why is my nugget effect so large (4000)?
Rplot.jpeg
Rplot.jpeg (41.52 KiB) Viewed 2764 times


Running a simple correlation (density vs northing, and density vs easting) to determine if there is anisotropy, it does not appear that there is any. Therefore, I calculated an experimental omni-directional variogram.

Are there any steps I am missing? Does my variogram look ok? Should I be feeding in different data?

If you could provide an email address so that I can send my script, along with csv data file, and shape file of the lake, that would be great.

Thank you.
radillon
 
Posts: 5
Joined: Wed Oct 05, 2016 5:38 pm

Re: general help

Postby Didier Renard » Wed Dec 06, 2017 11:23 pm

Hello

Your problem is a reasonable one and should deserve a correct answer.
Obviously you face a problem of data sampling design which may corrupt any standard varigoram calculation: the variogram calculation parameters should therefore be tuned specifically.
Regarding your variogram, yes its shape is quite unconventional. But is must be looked at paying attention to the number of pairs per lag. It may then happen that the first lag (the one that could drive your nugget effect) could simply correspond to a much too small number of pairs. Then either you neglect it (at least during the variogram fit ... which an automatic procedure will do anyway as it takes the number of pairs as a weighting function) or change the lag value slightly: again if the number pairs is small, the variogram value is very erratice and this slight change may affect drastically its value.
As you mentionned the best is to communicate the data set at didier.renard @mines-paristech.fr. However, for the sake of completeness of this forum, I would like the answer to be added to this post at end...


Talk to you directly then.
Didier Renard
 
Posts: 337
Joined: Thu Sep 20, 2012 4:22 pm

Re: general help

Postby radillon » Fri Dec 08, 2017 2:54 pm

Thank you, I have sent an email with my data.
radillon
 
Posts: 5
Joined: Wed Oct 05, 2016 5:38 pm


Return to Other Troubleshooting

Who is online

Users browsing this forum: No registered users and 4 guests

cron