Posted Mon, 17 Dec 2007 11:36:22 GMT by Colin Smith
I have a 3-layer 4 slice model.  I have always assumed that the hydraulic conductivity used by the model is that defined on the the slice with the same number as the layer.  So that, for example, slice 1 defines the hydraulic conductivity for layer 1 and so on.  I have assumed that the values on the bottom slice do nothing.  In the FEM file there are data for three sets of elements for the three layers (defined in m/d rather than m/s x E-4).  I have just output the values of K (using the Special option) so that I can plot a map of transmissivity by multiplying the layer thickness by K.  I find that the values in slices 1 and 4 are indeed those in slices 1 and 4 (without the factor of 1E-4) but those in slices 2 and 3 are averages of slices 1 and 2, and 2 and 3 respectively.  Is this just a feature of the interpolation, i.e does the model use the values in the element file, or does Feflow itself determine values of layer K by this interpolation method?  Also is this general to the special option, e.g. will the slice elevations for slices 2 and 3 come out as averages in the same way?

Regards William Faulkner
Posted Tue, 18 Dec 2007 18:52:49 GMT by Denim Umeshkumar Anajwala

Hey William!

You just got me really confused about layers and slices and data attached to them :D Here is how I think about it - and I'd say that's how FeFlow stores and uses it:

Some data are specified at nodes, that is on the slices. Other data however are specified for elements, which make up the layers. To give some examples: Heads and elevations are nodal data; they are specified for each node on each slice. Material properties, such as Ks for example, are elemental values and specified for each element on each layer. So you're really dealing with data at different "locations in (a 3D) space".

For calculating transmissivity, you need to combine an elemental property (K) with a nodal property (thickness) - I call thickness a nodal property here, since it's calculated as the difference between nodes on two slices. In order to accomplish that, you need indeed some averaging somewhere. The way you described in you post uses average Ks at nodes.

I didn't even know one can export Ks at nodes. But I just tried it (Special - Save nodal materials as Points) and find that very "interesting", because of the way the averaging is done. If Ks vary only from one layer to the next, it's simple the average between the 2, which is fine I guess. However, if Ks vary within a layer itself, then the average is basically
  the sum of Ks of all attached elements
  divided by
  the number of attached element of that node.
Is that a proper average??? I'm  not sure  :-\

Personally, I prefer to think about transmissivity as another elemental property. Therefore I compute transmissivity by multiplying the elemental K with an average element thickness, which I approximate by taking the average of the thickness at the (3) nodes defining the element. Whether that average is better might be questioned, but I find it definitely more intuitive.

Here is how I do it:
1) Export Ks as a "mesh" polygons (Flow Materials - Special - K - Isolines - Elements)
2) Export layer thickness as points (3 D slice elevation - Special - Z - select Thickness - Points)
3) Overlay both in my GIS
4) Calculate the average elemental thickness from the nodes attached to an element and
    propagate the result to the elements
5) Calculate the transmissivity = K x averageElementThickness

You have to do that for each layer separately and step 4) depends on your GIS (in Manifold it's a simple spatial overlay with transfer rule set to n->1:average). You can show the transmissivity then by applying a colour scheme to the polygons. And voila! You're done :)

A lot easier and faster yet is using an IFM. But that's a completely different story  :)

Hope that helps, Chris.
Posted Wed, 09 Jan 2008 12:28:39 GMT by Colin Smith
Thanks for that Chris.  Alas I am not a GIS person I just use excel, text pad, and surfer but I will see what I can do.

Best regards - William
Posted Thu, 10 Jan 2008 16:01:52 GMT by Denim Umeshkumar Anajwala

Hey William,

just out of curiousity: How do you plot your maps (without GIS)? In Surfer? What types of input format do you use for that?

Posted Tue, 15 Jan 2008 21:57:01 GMT by Colin Smith

I import jpg etc files or shp/dxf etc (from other providers), or digitize bna/blns using the mouse or a digitizer from background or paper maps in Didger.  When importing jpgs you have to make sure that you can get the limits defined with a reasonable accuracy bottom left and top right of the image.  I convert polygons into bnas.  For contouring I use Surfer in the normal fashion. 


Posted Tue, 15 Jan 2008 23:38:03 GMT by Denim Umeshkumar Anajwala

I'm afraid some of those formats I've never heard about. I had no experience with GIS myself when I started at AquaResource a couple of years ago, but can't imagine a work life without it anymore. Should you ever consider looking into it, I strongly recommend Manifold. It's extremely inexpensive (starting at $245), comes with powerful capabilities, and is one of the most straightforward software packages I've ever seen.

So, going back to your initial question about transmissivities. I take it, you generate nodal transmissivities and then contour those in Surfer? Would it help to have a shape file with polygons showing elemental transmissivities (as described in my previous post)?

Cheers, Chris

Posted Sun, 03 Feb 2008 11:39:12 GMT by Colin Smith

I will look into Manifold - thanks.

I create polygons using the digitizer in Surfer, edit them into Felfow ply format and load them as maps.  I then use these to import hydraulic conductivity.

One thing that I have noticed in editing the files is that Feflow (5.1 at any rate) is inconsistent in how it reads the end of data marker of various types of file.  I have found that it always accepts 'END' but may not accept 'end' or 'End'.

Another point that might be of interest to Surfer users is that with large arrays of data Feflow (5.1 at any rate) has difficulty contouring large regular grids (i.e. those exported from a Surfer grid).  My array has about 19000 xyz points in a regular grid which I use for importing slice contours.  The way I have found to over come this is to add a random digit - in my case between -1 and 1 to the X and Y coordinates.  It took me a long time to work that out!  I use the Akima extrapolation method as Kriging always gives an out of memory error even though it's not - I have 2Gbyte most of which is free and remains free.

I have also found that 'Triangle' has problems working on my grid: it usually gives a floating point error.  I have found that I have to import all the polygons as a single file in order for it to have some chance of working - the basic geology is simple but the area is shot through with faults and dykes which I import as polygons.  Currently I am getting a floating point error which I can't resolve.

Regards - William

You must be signed in to post in this forum.