Geostatistics Workshop

FIA Tutorial Videos for Conducting Spatial Analyses using Common Tools

All of these videos are useful if you're not familiar with the procedure; I would recommend looking at them all at one point or another.  They can be rather large, unzipped (5-45 meg!), so I would not necessarily unzip them all at once; rather, I would store them zipped, and unzip them as needed.  Check out the sizes of these videos:  video size.

All Windows 95 computers should have some sort of a movie viewer on them.  If you double click the .avi files you get after unzipping these videos, it should start automatically.  Your viewer will have a pause, stop, and rewind capability; just experiment to see how it works.  There is audio with the videos.  If you can't hear the sound, go to start-settings-control panel-multimedia, and adjust the speaker volume, etc.  You don't need external speakers to hear the audio.  If you still have trouble, ask a computer person, or feel free to give me a call.

If you don't have zip software on your computer, you really should (better sooner than later).  Go to www.winzip.com for Winzip.

Listing of zipped videos in order of use:

xyztoshape.zip:  This video shows you how to go from text file, like you get out of a database like Oracle or elsewhere, to a lat/long point shapefile.  It takes you from importing a text file into Excel, using a formula to change seconds to decimal degrees, and saving as an Arcview compatible file.

tableaddevent.zip:  This video will show you how to bring the file you just saved into Arcview and create a temporary shapefile out of it using the "event theme" procedure.

project.zip:  This video shows how to take your "event theme" from the above video and turn it into a correctly projected shapefile.  You need to have either the projector extension loaded
(see how to load extensions by viewing loadext.zip) or have arcview 3.1 and use the projection wizard tool (it's the same principle, the only difference is there's an interactive series of screens).  Call me if you have problems with this.

addcoor.zip:  This video will show you how to load a script in Arcview, and add the correct x y coordinates in meters so that you can correctly perform variogram analysis, and export the table, ready for importing into Surfer.

variomodel1.zip:  This video shows how to import the table from the previous step into Surfer, and begin variogram modelling, so that you can get a variogram model to use in kriging or sequential gaussian conditional simulation.

variomodel2.zip:  This video continues with variogram modelling, showing how to model the omnidirectional variogram.  This variogram can be used for the autofit procedure and subsequent kriging (see the next few videos).

variomodel3.zip:  This video continues further with variogram modelling, showing how to make more complex structures (nested structures), or combinations of two or more acceptable variogram models.  This technique is common, and allows you to more closely model your data.  People normally recommend using your intuition as to how to best model your data.  For example, if you feel like there might just be a gradual change in spatial continuity over distance, you might use one model, say, and exponential.  However, if you feel like there might be a rapid change in continuity over short distances, say because of small scale disturbance effects, but then a gradual change over larger distance, due to geologic or climatic phenomena, you might consider a combination of a spherical (steeper slope) and an exponential model.  This part of the modelling process is something you need to experiment with until you feel  comfortable; there's no "right" way to do this.  Many people use only one structure until they feel comfortable branching out into more complex models.

anisotropy.zip:  This video goes into more depth about modelling anisotropy, and shows an example of how it is done.  You can either model it manually, and calculate your own anisotropy ratio for the Surfer kriging (see Surfer help r.e. this), or model it using autofit (below).  If using gslib, you should model it manually, and fill in the parameters from sgsim following the example in the gslib book (a straightforward 2-d example).

autofit.zip:  This videos deals in a very minor way with modelling anisotropy manually (this should be done when using gslib (see the video above).  It then discusses in depth the use of the Surfer autofit function, which basically takes the eyeballed omnidirectional variogram you've made in the previous 3 steps, and then refines it and models its anisotropy.  Warning!  The autofit is a "black box" technique, which might lead to erroneous results.  Few "super expert" people use black box techniques; they seem to use manual fitting of a model, including anisotropy, and then maybe use a statistical technique to refine their choice, and then enter these parameters into the kriging program.

kriging.zip:  This video shows how to do simple kriging in Surfer using the variogram model you've either manually fitted, or manually fitted and refined using autofit.  It mentions some of the parameter settings in Surfer, as well.

mse.zip:  This video shows how to evaluate your kriging/simulation results in Surfer, specifically how to calculate the MSE, or the closeness of the original point values to the estimated values.  The goal is to have a map whose residuals are small, normally distributed with a mean of zero, and fairly evenly distributed throughout the map.  The video shows how to assess this.

finalmapshort.zip or finalmaplong.zip:  These videos show a short (smaller) and long (larger) version of how to go from your kriged (or simulated) map from the previous step to a final publication quality map.  The short version is a quick overview, and the long version is very detailed, but is giant.

Below are some extra steps you might want to consider if you want to use Sequential Gaussian Conditional Simulation.  They are examples of one way to do these things, but there are many other ways to approach the problems.

zonedefinitionarcview.zip:  This video shows you how to subdivide your population of points in arcview into areas with a stable mean and variance.  My understanding of why this might be advantageous is that you will not be violating assumptions of the stationarity of the mean (which is apparently a theoretical assumption of kriging) if you do this.  For a discussion of "stationarity", see any introductory geostatistics text, or one of the webpages in the reference section.  The video details the use of the xtools extension, an extension which presumably you know how to load from viewing loadext.zip.  Remember, all of the aml's and avenue extensions and scripts can be downloaded from the software section on the main page.

allstatsprep.zip:  This videos shows how to prepare the output allstats.txt file from running the postmegrid aml in Excel 2000.  It shows a great example of how to run an Excel Macro, or a small program which automates simple tasks.  It's an incredibly useful tool to use, and can be created to do things like bootstrapping simulations, and lots of other statistical operations.  The nice thing is, you don't need to be a fancy programmer to create one.  It records your key strokes, and writes its own code!   You do have to remember to be very economical with key strokes, and fool around with it.  It often will malfunction if you hit the wrong sequence of keys, or if you try something too fancy.  I'd recommend trying, and trying again with different key strokes, to get what you want to work.   Hopefully, this video will get you started!

avsumstats.zip:  This video shows how to use the DNR's tools_dd extension to summarize a point shapefile in arcview.  The resulting summary table will be based on some geographic area, e.g., a county shapefile, where for each county, you calculate the mean, variance, etc. of the forest inventory data inside of that county.  You can also use this to summarize the output from the postmepoint aml.  Creating the summary tables, such as the allstats.txt (above), or the arcview summary table, allows you to compare your original data's local summary statistics with those of the estimated map.  For some applications (e.g., Sequential Gaussian Conditional Simulation), you might want your estimates to have the same local summary statistics as the original data.  You can choose a simulation percentile to report based on how well that percentile compares with the original data.  One way to do this is to join summary tables produced based on the original data, and those produced from the estimates, and evaluate the differences.  The percentile with the smallest difference might be the one you choose to report.  The video should give you some good ideas on how to approach this.

exceljoin.zip:  This video shows how to join the "allstatsprep" step text file to a summary table from arcview, created in the above step.  Once this is done, you can calculate a summary row, and use a pivot table to find the smallest difference between the the original data and the percentile maps.  The point of this step is to pick a final map from one of the percentile maps you created using "postme".  Joining files in excel is a really nice trick, as well, and is highly worth knowing!

pivottable.zip:  This video shows how to go from a joined excel output table (above) to a summary of the differences between the original data and a bunch of simulated outputs using a pivot table.  The pivot table is a remarkable tool, one that you should definitely learn how to use if you end up summarizing data across various fields.  Once you've decided on your postme output file that you might want to report, you can make your final map by various means, one of which is by using the postmegridpoint, postmegrid, or postmesurfer aml's, and then using one of the below videos, or using the "finalmap" videos, above.

postmegridpoint.zip:  This video shows how to get your chosen postme or other output files into a form that you can create a final map in arcview, as well as conduct summary analyses as described above in avsumstats.zip.  You should be familiar with ftp software to be able to do this.  Also, be aware that you should navigate in unix to your working directory, and then type "arc" at the unix prompt to start arc info.  NOTE!  This is just one way to make a final map in arcview.  You might also want to try the below method, or the "finalmapshort.zip" or "finalmaplong.zip" videos, above.

grid2aviewlayout.zip:  This video shows how to take an arc-info grid file format and turn it into a final map product, right in arcview.  It uses an arc info grid, and allows you to basically load the final map as an image, and create a fake legend using a special arcview tool, and then produce your final map.  Use of this tool would be facilitated by some knowledge of arcview.  If you're not comfortable with arcview, you might want to run postmesurfer.aml on your postme outputs, and create your final map in Surfer, after having used the arc grid command "clip" to clip out the final output map.

Previous: Example & Downloads | Next: Arc Scripts