Pages

Sunday, February 10, 2013

Field Activity #2: Surface Terrain Survey Redo

Introduction

Today’s activity is a follow up from last week’s activity. As I had mentioned in my last technical report, the X, Y, and Z points we acquired from our survey were going to be imported into Arc Map to create a real-life replica of the actual landscape we created in our planter boxes. To do this we used different types of interpolation methods, and then observed those methods in a 3-D version in Arc Scene. We were then instructed to evaluate the different methods and decide which one we felt best matched our survey. From there were asked to resurvey our landscape terrain and use our critical thinking skills to improve our survey in ways in which it may have been lacking from the first survey. Again, we were to import our points into Arc Map and create an interpolation method that we felt replicated our actual terrain the best.

Methods
To start the second part of this project we began by importing our X, Y, and Z coordinates into Arc Map (Image 1).

Image 1: Survey points imported into Arc Map from our first survey

Then we attempted to run the different types of interpolation methods. An interpolation predicts values for cells in a raster from a limited number of sample data points. In turn, this creates a continuous surface that depicts a real-life surface. However, the tool to run these methods wouldn’t execute the task. We realized that we needed to make the shape file, which is initially created by importing X, Y, and Z values into Arc Map, into a feature class in order for this to work. First, we created a file geodatabase in Arc Catalog in our personal folders on the W drive. We then exported our survey points feature class into the geodatabase as a feature class. Again, we tried to run different tools to execute interpolation methods; and again, it didn’t work. We then thought that the decimal points for the Z values (image 2) were throwing something off while the interpolation tools were trying to execute. We thought this because past experience has taught us that Arc Map is very picky with the layout of Excel files.

Image 2: Initial X, Y, and Z values imported into Arc Map from our Microsoft Excel file from our first survey
 

The Z values have decimal points in them since as we measured to the nearest half centimeter
So, we decided we needed to get rid of them. We did this by multiplying all our X, Y, and Z values by 10 in Excel (image 3). This works because it essentially moves the decimal place over to the right once. This, in turn, left our values with no decimals.  We also titled the initial values columns with a 1, example X1, Y1, and Z1, and titled the new columns X, Y, and Z. Again, we did this to keep everything simple as to prevent Arc Map and Excel from not working properly with one another.

Image 3: Microsoft Excel after all values were multiplied by 10 and column titles were changed

The function equation located above columns D and E depicts the equation that was used to change the numbers for the Z values. This same method was used for the X and Y values but with their appropriate equation.
We then imported the new values into Arc Map. Again, we exported the shape file into our geodatabases as feature classes, and brought the new feature class of our points into Arc Map (image 4) to run the interpolation tools.

Image 4: Survey points imported into Arc Map from our first survey

The points are the same as the ones in image 1, the only difference between them is that image is a shape file and this image is a feature class.
Finally, the tools executed the different types of interpolation methods properly. Interpolation methods we used included inverse distance weighted (IDW), Kriging, Natural Neighbor, Spline, and triangulated irregular network (TIN).

The first method we ran was IDW (image 5). IDW creates a continuous surface by estimating cell values by averaging the values of sample data points in the neighborhood of each processing cell. IDW is considered a deterministic interpolation method because it is directly based on the surrounding measured values or on specified mathematical formulas that determine the smoothness of the resulting surface. This method assumes that the variable being mapped decreases in influence with distance from the sample location.

Image 5: IDW

The next method we ran was kriging (image 6). Kriging works by generating an estimated surface from a scattered set of points with z values. Unlike IDW, kriging is considered a geostatistical method. A geostatistical method is based on statistical models that include autocorrelation; that is the statistical relationship among the measured points. Because of this, geostatistical techniques not only have the capability of producing a prediction surface but also provide some measure of the certainty or accuracy of the predictions.
Image 6: Kriging
 

Then we ran natural neighbor (image 7). Natural Neighbor finds the closest subset of input samples to a query point and applies weights to them based on proportionate areas to interpolate a value. Its basic properties are that it’s local, using only a subset of samples that surround a query point, and interpolated heights are guaranteed to be within the range of the samples used. It does not infer trends and will not produce peaks, pits, ridges, or valleys that are not already represented by the input samples. The surface passes through the input samples and is smooth everywhere except at locations of the input samples.
Image 7: Natural Neighbor
 

Next we ran spline (image 8). Like IDW, spline is considered a deterministic method. Spline uses an interpolation method that estimates values using a mathematical function that minimizes overall surface curvature. This results in a smooth surface that passes exactly through the input points. Conceptually, the sample points are extruded to the height of their magnitude; spline bends a sheet of rubber that passes through the input points while minimizing the total curvature of the surface. It fits a mathematical function to a specified number of nearest input points while passing through the sample points. This method is best for generating gently varying surfaces such as elevation.
Image 8: Spline
 

Finally, we ran TIN (image9). TIN creates a triangulated surface that does not deviate from the input raster by more than the specified Z tolerance. First, a candidate TIN is generated using sufficient input raster points (cell centers) to fully cover the perimeter of the surface. It then incrementally improves the TIN surface until it meets the specified Z tolerance. It does so by adding more cell centers on an as-needed basis during an iterative process.
Image 8: TIN
 

After all the different types of interpolation methods were ran we brought them into Arc Scene. Arc Scene allows it’s user to observe a raster in 3-D (the continuous surface created by interpolation methods results in a raster file) (image 9-13). Here we were able to observe all our surfaces and decide which one depicted the true surface terrain we were trying to capture.
Image 9: 3-D view of IDW

Image 10: 3-D view of Kriging

Image 11: 3-D view of Natural Neighbor

Image 12: 3-D view of Spline

Image 13: 3-D view of TIN

Overall, I thought all the different methods did a great job at depicting our terrain. However, I thought that spline did the best job. This is probably because this specific technique works well for elevation surfaces and creates such a smooth looking surface. As we discussed earlier, spline is considered a deterministic interpolation method because it is directly based on the surrounding measured values or on specified mathematical formulas that determine the smoothness of the resulting surface. Since we sampled every 5 or 10 centimeter increments our sample points were very close together. Since this method is based on surrounding measured values I feel cell values were calculated appropriately for our surface. Also, since our land terrain was made of snow and was created with us wearing gloves and mittens, there were no elongated or sharp objects used in the creation of the surface, it makes sense that a smooth looking surface worked well for this. This technique allowed the smooth surface of the raster to depict the smooth surface of the actual terrain.
The next step in this project was to resurvey our surface terrain. After examining our 3-D surfaces we needed to determine areas where more survey points were needed to make our 3-D models more of an accurate description of our terrain. Together, we brain stormed on what we could do differently to make our survey better. We decided that we really liked our first survey method, but came up with a few cool ideas on what to do differently and thought of a few extra tools to bring with us this time.

Like the first survey, we agreed to keep the coordinate system the same with the short end of the planter box be X axis and the long edge be Y axis. We also agreed to do our survey at 5 and 10 centimeter increments on the X and Y axes. If there was a lot of terrain feature to capture, such as a ridge or depression, we did 5 centimeter increments; this is so we would capture a more accurate, real-life surface elevation. If there wasn't much terrain feature to capture, such as a plain, did 10 centimeter increments. We also decided to keep taking our Z value measurements to the nearest half centimeter.
We also used a lot of the same tools we had the first time, such as measuring tapes, meter sticks, and tape. However, this time around we agreed to use string and thumbtacks to assist us with making our survey more accurate. Again, we printed off more Excel spreadsheets to record our X, Y, and Z values and thought to bring a clipboard with to make recording our values easier.

We also remembered to dress very warm. The first time we did our survey it was 15F outside and we needed to keep coming inside to warm up, which wasted a lot of time. The day of our second survey it was about 25F outside. It was a little warmer, but we knew it wouldn’t take long to get cold if we didn’t dress appropriately for the weather.
To begin our second survey we needed to reshape our surface terrain since it had snowed since the last time we were out there (images 14 and 15).

Image 14: Our surface terrain 3 days after our first survey

Image 15: Reshaping the landscape

Then we established our X and Y axes. We started this by creating X axes through the interior of the box. We did this by first laying two measuring tapes on both X axes. Then, we used the thumbtacks and placed them into the wood on the X axis at every 10 centimeter increment (image 16 and 17).
Image 16: Placing the thumbtacks in the wood at every 10 centimeter increment on the X axis
 

Image 17: The X axis complete with thumbtacks at every 10 centimeter increment

Once all the thumbtacks were in place at their 10 centimeter increments we removed the measuring tape. Then we used the string to create X axis throughout the center of the box. Creating 10 centimeter X axes all through the center of the box provided us with a higher degree of accuracy. We only did 10 centimeter increments instead of 5 because we didn’t intend on taking every 5X5 centimeter measurement, as discussed earlier. We started this by tying one end of the string to the end thumbtack on the planter box (image 18).
Image 18: Tying the end of the string to the thumbtack 
 

Then we laced the string around the thumbtack straight down on the opposite end, and then over to the thumbtack next to it (image 19). We laced the string around the thumbtacks instead of tying it off and starting over to save on time.
Image 19: Lacing the string around the thumbtack
 

Next, we placed the two measuring tapes on the Y axis of the planter box (the long edges) and taped them down (image 20). We taped them down so they wouldn’t move; this would also create greater accuracy. We could have tacked them into place but the measuring tapes didn’t belong to us and we didn’t want to vandalize property that wasn’t ours. Having a measuring tape on both Y axes allowed us to measure in both 5 and 10 centimeter increments.
Image 20: Placing the measuring tapes down on the Y axis
 

Since the end of the measuring tape didn’t exactly start at 0, we started the measurement at 10 (image 21). So, 10 was our 0 in reality.
Image 21: Starting our Y axis measurements at 10
 

Finally, our the strings were in place for our 10 centimeter increments on the X axis and the Y axis were in place (image 22)
Image 22: Our 10 centimeter increments on the X axis with string and our Y axis completed
 

Next, we had to finish our X axis. Like the first time, we decided to use a mobile X axis again. A mobile X axis allowed us to find the exact spot where the X and Y axes lined up. Since a meter stick alone was too short to lay across the planter box we used a measuring stick that was in inches that laid all the way across.
As was discussed earlier we agreed to take 5X5 increment measurements if there was a lot of terrain feature to capture. However, the string only provided us with 10 centimeter increments, so we still needed a solution for 5 centimeter increments. We could have just eye-balled it up between two 10 centimeter strings and estimated where a 5 centimeter increment would’ve been, since 10 centimeters isn’t very large to begin with, but we knew that wasn’t an accurate way to do it. So, we taped a meter stick to the top of the other stick (image 23). This worked out perfect for measuring increments. The length of the meter stick was short by 5 centimeters on each end in comparison to the planter box. By keeping the meter stick at the 5 centimeter increment lined up with the actual 10 centimeter increment of the first string (same went for the opposite end of the stick and last string), we were able to keep  the 5 centimeter increments lined up where they were supposed to be to keep our measurements accurate. So here 0 centimeters marked 5 centimeters.

Image 23: Meter stick taped to the top of another measuring stick that provided a mobile X axis for a 5 centimeter increment

This coordinated system worked out very well, once again. By having measuring tapes on both Y axes we were able to make sure our mobile X axis was straight across and not at an angle, which again lead to greater accuracy.
Finally, we were physically able to take the survey. To do this we laid the mobile X axis across the box making sure it lined up at the evenly on both Y axes. Then one team member held the meter stick, used for measuring elevation, down to the top of the land surface in the appropriate spot of the coordinate system. Elevation was measured from the distance of the bottom of the long measuring stick of the X axis to the top of the land surface (image 24). Once the stick was in place another team member read the measurement off to the recorder, and final team member recorder the elevation and coordinates onto the Excel spreadsheet. To be the most time efficient for taking all of the elevation measurements we began our survey at X=0 and Y=0. We continued all the way down the X axis before moving the mobile X axis up 1 increment on the Y axis, and again worked our way down the axis.

Image 24: Depicts how measurements of land elevation were measure using our coordinate system

This measurement would’ve been recorded as having an elevation of -13 centimeters
After all the elevation measurements were taken and our box parameters were taken measured and recorded, which came out to X=112.5cm, and Y=224cm, we were done with the physical part of the second survey.

Next, we had to repeat the steps outlined at the beginning. First, we had manually enter our values into Excel and multiply our columns by 10 to get rid of the decimal points (image 25).

Image 25: Microsoft Excel file for the second survey once all the values were entered and multiplied by 10
 

Then, we had to import our values into Arc Map, export the shape file of them into our geodatabase as feature classes, and add the feature class file to Arc Map (image 26). (image 27 is being provided for comparison purposes of survey 1)
Image 26: Feature class file of the values from our second survey
 

Image 27: Feature class file of the values from our first survey
 
Finally, we were asked to create the interpolation method that we felt best worked for our data (image 28). We were instructed to compare it to the first survey we took (image 29) to see how it compared and if more surveying was needed.
 
Image 28: Interpolation method we felt best represented our surface terrain
 

Image 29: Interpolation from first survey for comparison

Discussion
This project helped me realized the importance of critical thinking and team work. It is amazing the ideas you can come up with when you collaborate with fellow colleagues, and keep an open mind and positive attitude. Things aren't always going to go the way image them inside your head. This project definitely proved to have its challenges. I honestly wasn't expecting to come up with so many problems while taking this survey. One factor can influence the path you have to take to get something done. It can be as simple as not having enough people to get the job done as you anticipated to, or something on much larger scale such as the weather. The first day we did our survey it was about 15F outside and was very cold. For the second survey it was about 25F, so it was still cold, but a little warmer. However, we knew to come prepared. This time around we had a better idea of what to expect for the survey as well. This was helpful because we were able to come prepared with more tools to get the survey done. One tool that would’ve been useful and provided a more accurate survey would’ve been some sort of a measuring tool with a smaller, or pointier, end for measuring elevation. This is because the meter stick is approximately 1 inch in length, so it doesn’t provide an accurate reading where there is a lot of relief, such as going up a hill or ridge.

Conclusion
Overall I feel as though I learned a lot from this project.  I was really surprised at how much I felt like the second survey was done much more professional by using more accurate measuring techniques. This exercise challenged my ability to think spatially, along with utilizing my critical thinking and team work skills. I thought that our team worked great together! The first time around it was just Phil and me, and I was surprised at how well we collaborated our ideas together to get the survey done. For the second survey it was even more surprising to me all the ways we thought to improve our survey. Overall, I couldn’t tell much of a difference between the two surveys while examining them in the 3-D view, but felt confident that by doing a second survey I learned much more. I also learned the importance of coming prepared. For the second survey our team split duties for bringing in more tools. Things as simple as tape, thumbtacks, string, and a clipboard can make a world of difference for whatever challenges or obstacles that may come. I also learned the importance of coming prepared for the weather. If you’re cold and wet it’s hard to think and stay focused when in the back of your mind you’re really just thinking about how cold you are and how you just want to be done. That type of mind set can have a big effect on your overall work ethic and results. The only thing I can think to improve this activity would be for it to be warm. However, that’s just a personal preference. The fact that it was cold, I feel, challenged me to learn more and take more away from this activity.




No comments:

Post a Comment