Exercise 2: Visualizing and Refining the Terrain Survey

Introduction

Last week our group created a landscape using a large sandbox filled with snow, created a coordinate system to survey the landscape, measured the depths, and created a digital table filled with coordinates and depths for each of our 1056 zones. This week we are going to use that data to create a digital representation of our landscape. We will use a variety of different interpolation methods and analyze which ones resemble our landscape with the most precision. 

Methodology

Our data from last week was recorded in a large table with an X and Y axis which resembled our sandbox. In order for the GIS to use this data, it needed to be reformatted into an XYZ columned format. 

A portion of the original large table

A portion of the reformatted XYZ table

With the new GIS friendly table, 5 different interpolation methods were executed with the data: inverse distance weighted (IDW), kriging, natural neighbors, spline, and a triangulated irregular network (TIN). IDW is an interpolation technique that estimates cell values based on the parameter that the further a sampled point is from the cell being evaluated, the less weight it has in the calculation of the cell's value. Kriging an interpolation technique in which the surrounding measured values are weighted to derive a predicted value for an unmeasured location. Weights are based on the distance between the measured points, the prediction locations, and the overall spatial arrangement among the measured points. Natural neighbors is a method that calculates the value for an interpolation point by estimating using weighted values of the closest surrounding points in the triangulation. Spline is technique in which cell values are estimated using a mathematical function that minimizes overall surface curvature, resulting in a smooth surface that passes exactly through the input points. The final method is known as TIN and it uses the XYZ data to connect a series of triangles between each data point. This process forms geographic space into contiguous, nonoverlapping triangles. 

IDW

Kriging

Natural Neighbors

Spline

TIN

After examining the various interpolation techniques, we decided that it would be best to resurvey some portions of our landscape. The walls of our valley appear to be separated near the center of the landscape, and the lake to the east of the mountain seems to be half the size and not as deep as it should be. 

Once again, we ventured back out into the frigid courtyard for another round of data collection, focusing on the valley and lake where the data seemed a bit inaccurate. We agreed that the best method to do this would be to split each of the existing zones into 4 smaller zones. This was done by using another set of string and measuring all of the affected areas in increments of 2.5cm. 

Pink string was used for the 2.5cm increment resurvey

We updated the table to include the newly surveyed data zones and began the interpolation process once more with more accurate data, bringing our zone total to 1442. 

Dot density of the original survey

Dot density of the resurvey locations

Week 2 IDW

Week 2 Kriging

Week 2 Natural Neighbor

Week 2 Spline

Week 2 TIN

After examining the various interpolation techniques for the newly resurveyed data, I believe that kriging and natural neighbor techniques were most effective, with kriging producing the most similar landscape. These two processes replicated our data the best out of the five, even though none of them were perfect. We then took all of the rasters we created and viewed them in a 3D modeling software called ArcScene. By setting the base height to the surface elevation values assigned to each cell, we could create stunning 3D models of our landscape and view the survey from a completely different angle. 

3D model of the IDW interpolation method

3D model of the Kriging interpolation method 

3D model of the natural neighbor interpolation method 

3D model of the spline interpolation method

3D model of the TIN interpolation method

Discussion

I am quite pleased with the results of our resurvey. We located areas in which quality data was lacking, we brainstormed a resurvey process, carried the process through, and created several 3D models of what was once a pile of snow. Our group continued to function as a fantastic team in which everyone played a crucial role. Collecting data outside in subzero temperatures was not easy, so we managed to divide the workload throughout the week very nicely.

It was challenging to try and explain why each interpolation method was different from one another. Besides the triangulated irregular network, I was unfamiliar with all of the interpolation techniques. While it was easy to just plug in the right values and create different rasters, it would have been nice to have some background knowledge of each of the different interpolation methods before this exercise. 


Conclusion

In conclusion, kriging appears to be the best method in recreating our landscape. Kriging preserved both the sharp edges and the rounded hills of our landscape. The IDW process lifted each data point too high and created spikes all over the landscape. The natural neighbor method was great for shallow elevation change, but it created too many sharp peaks when the elevation rose. The spline method was more smooth compared to the IDW, but it created the same tall spikes throughout the landscape. Finally, the TIN showed elevation very well, but the triangulated network couldn't replicate the same results in flatter areas of the landscape. 

If there was a contest for which method replicated the snow landscape the best, I believe that kriging would win for our purposes. Maybe the different methods exist because each is better at displaying a certain characteristic of the landscape, such as flatness, roundness, or rapid changes in elevation.