This activity is based off of last weeks creation of a digital elevation surface activity. The reason why we are refining the survey is because we have become privy to new and improved methods and therefore should go back through all the work we did so that we can make a better end product. After last weeks activity we heard from every group and how they went about survey their sandbox and from that discussion we were suppose to refine ours. Our group decided the best way to refine our survey would be to go back and re survey the points on our map that had sharp changes in elevation so that we could get better detailed data about those structures. Once we collected all the data again we would go through the same process as before and make maps in ArcMap as well as exporting them to ArcScene so we can get a 3D representation of the surfaces as well.
Study Area:
We conducted our second study on September 23rd, 2015 from 8:45am to 10:00am. Our study area was a 4x4 area once again located on the point bar underneath the footbridge on our campus here at UW- Eau Claire. We determined this location to be suitable for this activity due to the sand being deposited by the meandering Chippewa River. The sand allowed us to manipulate it into our own rendition of the City of Eau Claire so that we could survey it for this activity and create the required landforms; a ridge, hill, depression, valley, and plain. Since we were trying to create our own rendition of Eau Claire, we thought it would be appropriate to include the following features found inside Eau Claire; the "Campus Hill", Chippewa River, Half Moon Lake, Carson Park, and the Putnam Park/Third Ward area.
 |
| Figure 1: Aerial of Eau Claire showing the location of Campus. |
 |
| Figure 2: Aerial of Campus showing the study area underneath the footbridge. |
As stated in the Introduction, we are keeping all of the data previously surveyed and just going back through and resurveying the areas that we thought were of upmost importance. The reasoning behind this is because many groups used a survey spacing that was too large and did not pick up their features accurately. For example, if your group is using 10cm spacing and on of your features is only 10cm across, how can you expect to get any detail out of that data? We determined that the areas of upmost importance were the Chippewa River and Half Moon Lake. When comparing the XY data we had last week to the data collected this week we can see the areas that we focused on by the intensity of data points on Figure 3 below.
 |
| Figure 3: Our XY data showing where we collected more data points. |
 |
| Table 1: Data flow model showing the creation of the five interpolation maps |
TIN (Triangular Irregular Network)
A TIN uses vector data points and from that data, creates many triangles that are connected at the vertices and are bordered by each others edges. The more points that you have, the more triangles there will be and therefore the more accurate your map will be become. I like TIN because it is easy to understand but the straight edges look unnatural and therefore throw off the realism of the maps(Figure 3).
 |
| Figure 3: Side by side comparison of the first 2D map vs. the first 3D map |
Kriging
The Kriging interpolation method is a geostatistical method that uses autocorrelation. This method is often used in soil maps and is a multistep process that can only be explained by those of us with a mathematics degree. I like Kriging because it gives a more smooth representation of the surface which to me looks more realistic instead of a bunch of random points sticking out (Figure 4).
 |
| Figure 4: Side by side comparison of the first 2D map vs. the first 3D map |
Spline
Spline is another method that utilizes a mathematical function. This method is used heavily with gently varying surfaces such as elevation and water table heights. I do not care for spline because at least in my map, it does not give in my opinion a good representation of the features (Figure 5).
 |
| Figure 5: Side by side comparison of the first 2D map vs. the first 3D map |
Natural Neighbors
This method takes advantage of Thiessen polygons and uses the neighbors of all given points and creates new polygons that are then weighted out and smoothed. I like Natural Neighbors because it gives a nice smooth representation of the model as well as hitting all the appropriate points of elevation change (Figure 6).
 |
| Figure 6: Side by side comparison of the first 2D map vs. the first 3D map |
IDW (Inverse distance weighted)
IDW uses a linearly weighted set of sample points and assumes that the variable decreases in influence from the sampled location around the data points. I do not like IDW for this data set because its result is way too clustered with point data sticking out and it doesn't do a good job of giving the illusion that there is no points there (Figure 7).
 |
| Figure 7: Side by side comparison of the first 2D map vs. the first 3D map |
Now that I have produced maps of the different interpolation methods, I can go through and decide which method I believe was the best for our study. All these methods have their own applications but for our study, I like the Kriging method the best. Knowing this I can address the issues my first kriging map had and think about how I can resample so that I get an even better end product for my second map.
To resurvey we basically had the same grid system as before but since we are going to be taking data points in between our former points, we had to renumber the x and y coordinates. Beforehand we had label each point by 1,1 or 1,2 or 1,3. For each point we moved over we just added another number in that direction as spaced out by 5 centimeters. After thinking through how to add numbers between 1,1 and 1,2 we decided to label them by how many centimeters away they were from the axis. So with 5 centimeter spacing, instead of 1,1 and 1,2 we switched them to 5,5 and 5,10. This allowed us to survey points in the middle such as 5, 7. With this new system we were able to get accurate measurements of our most prominent features.
Once we surveyed, we uploaded the excel table to ArcMap and went through the entire process again, except this time we only produced kriging maps.
Results/Discussion:
Looking at the results from the last 3D kriging map (Figure 8), I am content with the content and accuracy of the data. Assuming that we could have resurveyed without any disturbance of our study area from all the rain, we would have ended up with a better product. For any discrepancies that we had I would blame most of them on the site being disturbed by the rain and then some on user error because we are never perfect. As a matter of fact, one point that I know is user error is the point on the map where you can see we had our depression but then in the middle of that depression, there was a big spike, which I associate with a mistake in the data collection process. That is genuine user error and the result of just being sloppy, but it is also a good indicator of if we did that on other points as well, to my knowledge I don't see any other huge spikes like that again, so we must have done quite well.
 |
| Figure 8: The second 3D rendition of the kriging interpolation method. |
If I had to do this all over again I would first start with a plan to minimalize the amount of time it took to record the data because in all honestly, it was not efficient for my personal sake, since the first collection took so long, it could have been better if we would have just assumed that some of the flat plains were truly flat and just did not measure them since it did not provide any meaningful data. I would also stand my ground with my group when I said that we should put our point of origin in the bottom left, because we didn't all of our resulting maps made it difficult to pick out the features in Eau Claire because the axis was completely flipped horizontally. I was able to remedy this flipping effect in Adobe Photoshop as shown in Figure 9 below.
 |
| Figure 9: Final map of Eau Claire! |
I will also say that there is not a huge change in our maps because we did such small spacing the first time (5 cm) as compared to other groups (10cm).
Conclusion:
It is truly amazing what we are capable of when technology breaks down and all we have is string and a ruler. By creating a simple grid system we were able to create our own terrain and then duplicate that once we got back to the technology and then create some pretty neat maps from that. The biggest part of this lab and the most exciting was not having someone hold our hand through it, working as a team to finish the task, and being truly proud of the fact that once we created the resulting maps, we could say that we made these from scratch.
Sources: ESRI, "Comparing Interpolation Methods." n.d. Digital file, ArcGIS 10.3.1 Help.
Metadata:
Created by Nik Anderson
Created at UW-Eau Claire Geography and Anthropology Department
Created for Geospatial Fields Methods
Created on September 25th 2015
Created with ESRI ArcMap and ArcScene, Microsoft Excel
No comments:
Post a Comment