Assignment 2: Visualizing and Refining Our Terrain Survey

Introduction

This week’s assignment was to do a follow-up on the assignment from last week.  The intent was to examine the methods we used the week prior, and improve upon them to increase the accuracy of the terrain survey.  This second report is to be more thorough and detailed, providing a better understanding to the methods we used and a more accurate picture of what we accomplished.

Methods

To begin this week’s assignment, we loaded our x,y, and z coordinates from last week’s data collection into an Excel file, which was then loaded into ArcMap. With the points in ArcMap, we used the 3D Analyst toolbox to create terrain maps using five different interpolation methods, and can be seen in Figures 1-5. After the terrain maps were created, we then made a 3D rendering of them using ArcScene.

Figure 1:
IDW is the inverse distance weighted interpolation method.  IDW allows you to control how far away measured points are able to affect the interpolated point, allowing for smother or rougher surfaces depending on the power you set.  We used the default values during interpolation and there was no vertical exaggertaion in ArcScene.



Figure 2:
Kriging is a geostatistical procedure that estimates the surface based on a set of points.  Kriging looks at the statistical relationship of points, meaning it will predict a location’s values based on the measured values surrounding it. We used the default values during interpolation and there was no vertical exaggeration in ArcScene.




Figure 3:
The natural neighbors method interpolates points by drawing Voronoi polygons around all of the measured points.  A Voronoi polygon is then created around the interpolation point, and assigns a weight based on the percentage of overlap.  We used the default values during interpolation and there was no vertical exaggeration in ArcScene.

Figure 4:

The spline method minimizes the overall surface curvature.  It, essentially, takes a sheet of rubber and bends it through each of the measured points.  This provides a very smooth surface that mimicked our snow terrain effectively.  We used the default values for regularized spline during interpolation and there was no vertical exaggeration in ArcScene.

Figure 5:

A TIN is a triangular irregular network.  A TIN takes the input points (nodes) and draws lines to all the nearest neighboring points.  This creates many triangles, which also results in the terrain being very angular.  Before loading this into ArcScene, we had to convert it to a raster in ArcMap using ArcToolbox.

After completing the terrain maps and seeing the renditions of our prior week’s effort, it was obvious our data was not even close to the accuracy required for a good terrain map.  Whereas last week we used 10 cm x 10 cm grid system for our 100 cm x 230 cm flowerbed, we decided this week we would construct a 5 cm x 5 cm grid system.  It was cold last week, and it was even colder this week, which caused us to seek additional processes that would decrease our time outside.  One such creation was the use of a slide rule, as seen in Figure 6.  On top of the cold, our flower bed terrain was also covered in substantially more snow.  So, we had to do our best to create the same terrain as we had the week prior, which we believe we did relatively well.  Our etch marks from last week were still present, so it did not take us long to add the 5cm increment etches and pin up the y-lines.  With the grid setup complete, it was time to start collecting data.

Figure 6:
The slide rule laying across the flowerbed, with lengths of twine making up the y-axis.  Each twine is 5cm apart.

Rather than writing down the data points on paper and then having to type them into excel, we decided to have one person inside on a computer entering the data immediately.  I was the lucky person to be inside entering the points, while Amy took the x, y, and z measurements, and Joel relayed that information to me over the phone (Figures 7 & 8).


Figure 7:
Kent (myself) entering the coordinates into an
Excel spreadsheet.  Amy is taking the picture
while her and Joel were on a short break from
the cold.

Figure 8:
Amy (left) collecting the coordinates as Joel (right) relays
them to me via phone.

When the data collection was finally complete, much to Amy and Joel’s delight, we were able to immediately load the Excel file into ArcMap, run our preferred interpolation method (Spline), and render it in ArcScene.  Figures 9 shows the significant increase in accuracy of the terrain map.

Figure 9:
3D rendering of the spline interpolation method of our collected data. As can be seen, the terrain is
substantially more accurate and detailed (fine).  The same default values were used during
interpolation and there is no vertical exageration in ArcScene.

Discussion

Our methods during this resurvey of the terrain were substantially improved.  From the beginning we set ourselves up better than we did last week.  We made the decision that, given nature of Wisconsin weather at this time of year, we needed to remake the terrain, setup our grid system, and collect all of the data in a single session.  This can be difficult for our group, as all of us are both taking many courses and working at the same time.  The only day that we are really able to meet for two hours or more, outside of class, is Tuesday.  Last week we made the terrain during class on Monday, and returned on Tuesday to collect the data.  As can be seen in Figure 10, rain reduced our terrain in elevation considerable, and the frozen flowerbed made it very difficult to dig.  The use of the meter sticks as a moveable x-axis also reduced the amount of time it would take, as we did not need to place the 46 strands of string across the flowerbed.

Figure 10:
This image is from last week's data collection day.  It rained throughout the night,
substantially reducing and changing the elelevation of our terrain.

 

Deciding to have one person inside recording the data in Excel as it was being collected was also an excellent decision.  Last week it took me almost half an hour to enter the 291 points into Excel, and there were a multiple errors from typing mistakes, which required me to go back and correct them.  This time we had 987 data points, and the only issue was there were about 5 points that appeared to be missing. All I did was take a quick look at the attribute table, went to where the points should be, and found that I had copied the x coordinate too far down.  So, those 5 points were merely shifted over 5 cm and was an easy fix. 

The effect of our decision to use a 5 cm x 5 cm grid system is even apparent in just the data points when initially added to ArcMap (see Figures 13 & 14).  Figure 15 shows the end result of our improvised methods.

Figure 11:
Map export from ArcMap of our first week's data points.  These
were taken from a 10cm x 10cm grid system. There was a total
of 291 points.

Figure 12:
Map export from ArcMap of this week's data points. These
were taken from a 5cm x 5cm grid system.  There was a total
of 987 points.

Conclusion

Doing this assignment a second time was an excellent learning experience.  We were able to look back at what we did the week prior, improve upon our methods, and, in the end, create a significantly better map of our terrain.  Not only did we learn and improve our ability to collect data, import them into ArcMap, create maps of the terrain, and create 3D renditions of them in ArcScene, but we also came together as a group.