Biology professor needs a 3D map of the US, where the z-values represent particular species richness as opposed to the actual elevation of the terrain. The professor (and her GRA) came during my office hours with a delimited text file containing sample lat/long points along with the associated richness values.
Here's a quick shot of the result:
How many hours I spent on this:
Too many. At least 10 hours over the last month.
How I justify spending so much time on this project:
This is my first interaction with this biology professor, and I really, really want the professor to leave with a favorable view of the library's GIS services. I also worked closely with the professor's GRA, and I try to always take advantage of those teachable moments.
How I finally accomplished this task:
My first thought was that this would be a super-easy project. Unfortunately, this was not the case. This is probably due to my lack of experience working with the 3D functions of ArcGIS 3D Analyst/ArcGlobe/ArcScene. After this project, this should no longer be the case.
Here are the steps (from my notes):
- First Using ArcMap
- Create point feature class from sample data provided by the professor.
- Obtain U.S. state boundary shapefile from NationalAtlas.gov.
- Dissolved the continental (48 states) U.S.
- Converted the dissolved feature class to raster.
- Interpolated the sample points using kriging with the following parameters: (100 km radius kriging limit / 1 km cell size). The 100 km radius limit was requested by the professor.
- Merged the kriging layer with the raster U.S. layer.
- Next Using ArcGlobe
- Add merged raster layer
- Redefine raster layer as elevation.
- Exaggerate ArcGlobe by 1000. This was essential as this equated 1 richness factor = 1 km of elevation.
- Add the states layer, with hollow polygon symbology, to drape on top.
Why, develop a workshop of course. Keep your eyes peeled for a good workshop on these techniques planned for Fall 2006 semester. If the professor allows me to use the data points, this might really make some inroads into closer connections with he Biology department.
What about Google Earth?:
I would have to create a Tin (which is no problem) and then convert the Tin to KML using Arc2Earth. If I can find a free morning this summer (after finals...), I will compare how the elevation renders compared to ArcGlobe.
5 comments:
Hey I have done something similar for professors in Economics (the graphic on top). I am not sure why you merged rasters for the state boundaries on top.
Use the natural neighbors interpolation, it produces better "looking" surfaces than kriging. Make sure you use cubic convolution for the display of the elevation surface. You can also get 3d labels which really adds a nice touch to the images. Ping me @ scitronpousty@gmail.com if you have any other questions.
Steve, wow. Thanks for the help.
I read that the nearest neighbor provides a smoother surface than kriging, but the professor was adamant that the interpolation be limited to 1 km around each sample point. I played around with clipping (extracting) the nearest neighbor using a polygon mask, but was unsure if this would generate flaws.
I will have an opportunity to work on this later in the afternoon today, and will repost here the results using the method you suggested.
Thank you immensely.
one more hint is to border interpolated raster with some 0s, this will allow the surface to touch the earth at the edges. It looks much nicer.
Hi Joshua,
Not that this script is related to what you're doing here but that it might be a free alternative to arc2earth...I heard about this from a list. Maybe you already use it:
"For those who have ArcGIS - I have used the plugin below to project shapefiles from ArcMap which opens up the possibility to add your own data through the attribute table and you don't have to pay for arc2earth. I was impressed about how well that worked. The script adds a little GE icon to your ArcMap menu. The wizard lets you select the layer you'd like to export, you can even choose a variable for height to make it 3D. It creates a kml file on the fly, starts up GE directly and projects it."
http://arcscripts.esri.com/details.asp?dbid=14273
from: Academic Technology Specialist, Claudia Engel, at Stanford U.
The important reason for using kriging is that it also gives you a standard error of prediction map.
This will tell you where there are "holes" in your sampling network or that the kriging parameters need to be changed.
If you have Geostatistical Analyst then most of this is straight forward.
Steve Lynch
ESRI
Post a Comment