Activity 2: Visualizing and Refining Terrain Survey

Introduction:

This lab activity is a continuation of our previous lab activity, Activity 1: Creating a Coordinate System, during which we surveyed a study area using a grid system for the purpose of creating a digital elevation model. In this activity, we exported our data collected during Activity 1 into ArcMap 10.3.1 and created our 3D models using interpolation tools in ArcMap and importing them into ArcScene 10.3.1. We then revised the data to locate portions of our study area requiring resampling. We followed up our revision of data with additional data collection in the areas lacking clarity on the 3D models and recreated these models for a more accurate image.

Methods:

We began creation of our 3D models by first loading our excel data into ArcMap and displaying our XY data, where X and Y were representative of the location within the grid system and we defined a Z coordinate to contain the depth at each XY coordinate. We then exported this data to allow us to run our four interpolation tools: IDW, Kiging, Spline, and Natural Neighbor. These tools created a raster containing all elevation data we had previously collected and allowed us to observe areas within our data that did not adequately depict our study area surface.

Assessing the Interpolation Tools:

Inverse Distance Weighted (IDW):

This interpolation tool smooths a surface created by point values by calculating cell values based on an inverse distance function that weights the points surrounding the point of interest.

 

Figure 1: Inverse distance weighted (IDW) interpolation results. The left image depicts the raster file displayed in ArcMap and the right image depicts the 3D model displayed in ArcScene. These outputs depict our survey study area, however, the surface is not smooth enough to give an accurate image. There is also a loss of detail within the valley region of our study area.

Spline:
The Spline tool uses a mathematical function applied to nearest surrounding points of the point of interest and minimizes the curvature of the surface between each point. The result is a smooth surface that meets all point values.

 

Figure 2: Spline interpolation outputs. The left image shows the raster file in ArcMap while the right image shows the 3D model in ArcScene. These outputs depict an accurate image of our study area particularly in the 2 hills, however, there is significant loss of accuracy in the valley.

Kriging:
Kriging weights surrounding values to calculate an unmeasured location using a statistical formula. The weight of each value is based on both the distance between the points and the arrangement of the points. The result is an accurate and smoothed surface.

 

Figure 3: Kriging outputs. The left image shows the raster file in ArcMap and the right image shows the 3D model in ArcScene. These outputs display all landforms within our study area with significant lack of detail. Particular loss of accuracy is shown in the valley region of our study area.

Natural Neighbor:
The Natural Neighbor interpolation tool calculates values of points in close proximity. It uses values of points only within a close subset of values and weights the values based on proportionate area. It is a localized interpolation method.

 

Figure 4: Natural Neighbor outputs. The left image depicts the nearest neighbor raster file in ArcMap and the right image displays the 3D model in ArcScene. The outputs show detail in the 2 hills, the depression, and some detailing on the slope of the ridge, but lacks significant detailing of the valley.

Each interpolation tool showed a lack of accuracy in the detailing of the valley surface of our study area. For this reason, we decided to resample this portion of our study area.

Resampling 

On September 22, 2015 between 3pm and 6pm, we headed back out to our study area. Because there was little sign of erosion on our landscape, we decided to simply resample the area as it was.

Figure 5: A new grid was created only over the valley of our study
area and elevation data was only collected in this region.
The grid cells were smaller than the first data collection session.

We created a grid over the valley portion of our landscape with smaller cell sizes (4cm x 4cm) to provide more detail the surface of this region. Elevation data was only collected in this portion of the study area and recorded in excel as negative values. A total of 392 new data points were gathered and added to the data previously collected. This created an overall stratified grid for our study area.

Rerunning Interpolation Tools

After the new data was added to the previously collected data on excel, we then imported the data to ArcMap, displayed the XY data and defined the Z coordinate as our elevation data.

Figure 6: The metadata added to the XY data feature class.

We exported the data and added metadata to the resulting feature class to provide information on the data before running interpolation tools. In this manner, the metadata could carry over to each resulting file.










The resulting raster files and 3D models showed an improved level of accuracy in our valley landform of our study area. We then chose the interpolation tool that best depicted our study area. The kriging tool oversimplified most of our landforms, resulting in a great loss of detail while the IDW tool overexaggerated the them, resulting in inaccurate peaks and valleys. Spline, though it depicted our study area surface well, did not depict our landforms quite as accurately as Natural Neighbor appeared to do. Thus, natural neighbor was the best tool to display our elevation data.

Figure 7: Natural Neighbor Interpolation output in ArcMap (left) as a raster file and in ArcScene (right). These outputs show great detail in the valley region of our study area as well as all other landforms. This tool was chosen to depict our study area surface because it accurately smoothed our surface and connected the data points without leveling peaks and low areas. 

Discussion:

After running interpolation tools on our data collected from our Activity 1, we discovered all outputs showed a loss of detail in the valley landform within our study area. We proceeded to resample our study area using a stratified grid approach--only sampling within the portion that housed the valley landform and with smaller grid cell size. We then added the newly collected data to our data from the first activity to create our stratified grid and used this data in ArcMap and ArcScene to recreate our raster files and 3D models. The resulting outputs all showed an improved detailing of the valley landform in our study area and an accurate depiction of the remaining landscape. Of all the interpolation tool outputs, the Natural Neighbor provided the most accurate image of our study area as it connected our elevation data points smoothly while still providing a high level of detail on all landforms. 

The resulting Natural Neighbor output, though resembling our study area quite well and containing a notable improvement in the accuracy of the valley, still lacks some detailing in the slope of the ridge. Our stratified grid method worked well in gathering more data for the valley feature, however, I think it would be even more accurate if all features, save for the plain, had a smaller grid cell size applied to them. 

Conclusion:

During this activity, we applied our critical thinking skills to assess the results of interpolation tools and develop a method to better our output while still using previously collected data. We now know how to run interpolation tools in ArcMap 10.3.1 and import the output into ArcScene 10.3.1 to create digital evlevation models. We have also gained experience in interpreting the output of these tools and catching errors and areas of lesser clarity.

Activity 1: Creating a Coordinate System

Introduction

The goal of this activity was to create a coordinate system to evaluate a land surface in order to create a digital elevation model in later exercises. Creating coordinate systems and elevation models are used by people studying a range of different topics including bathymetry and island coastal mapping for planning safe entry and exit points for seafarers. We began this exercise by creating our own landscape within a 122cm x 125cm wooden box. We needed our landscape to contain a variety of landforms to practice accurately measuring differential elevations. For this reason, our landscape contained 2 hills, a valley, a ridge, a depression, and a plain.

Study area

We began our study within a sandy region of the floodplain of the Chippewa River on Friday, September  18. The study area was limited to the area inside our 122cm x 125cm wooden box. The study took place from 1:00pm-4:30pm. The weather was overall cloudy and dry, however, there was a slight sprinkle towards the later portion of the activity.

Methods

We began by laying down our 122cm x 125cm box in a relatively level portion of sandy floodplain. We then used a level to make sure both the wooden box and the sand within it was level to assure accurate elevation measurements.

Fig. 1: Our finished sculpted study area containing all land
features: hills, a valley, a ridge, and a depression.

Creating our Terrain

We then sculpted our terrain within the box to contain all of the landscape elements previously mentioned (2 hills, a valley, a ridge, a depression, and a plain).

Fig 2: The finished 8cm x 8cm grid created by wrapping
string around tacks set at the desired interval.

Setting up a Coordinate System

After sculpting our landscape, we divided the area into 8 cm x 8 cm squares using string thread around tacks placed on the wooden rim of the box at these set intervals. We labeled our coordinate system as X1, 2, 3, etc..., and Y1, 2, 3, etc... on either side of the wooden box edge.

Fig 3: Students (Ally Hillstrom and Casey Aumann) gathering
and recording elevation measurements

Our Z coordinate was measured as the depth  from the surface of the box to the sand surface beneath it and thus recorded in negative values. 

Fig 4: A sample of 25 points within the data recorded on excel.

Data Collection

The data was collected from the upper right-hand corner of each 8cm x 8 cm square using a meter stick and all depths were measured in centimeters. A total of 201 values were collected and entered into an excel spread sheet. 

Discussion

This activity allowed us to engage our critical thinking skills and gain a better understanding of how to create a coordinate system and elevation model.  Some difficulties we had during this project was deciding upon where to designate sea level, or z-value of "0cm." We decided to make "0cm" flesh with the top of our wooden box because it allowed us to make accurate measurements as it eliminated any error that may have existed in leveling the sand within the box before sculpting. We also had difficulty deciding how large to make the grid squares of the x- and y- coordinates as the length and width of the study area was irregular and the box corners were held together by pieces of wood from the inside of the study area, making an inconsistent study area shape. We eventually decided to make the squares 8cm x 8cm because we felt it would allow for enough measurements to make an accurate digital elevation model without having too many measurements. We also decided any fraction of an 8cm x 8cm square would not be included within the measurements. This made data collection consistent, but did negate portions of total study area. It would be interesting to see how other groups dealt with this problem of an irregular study area.

Conclusion

In conclusion, this activity did help us to develop a better understanding for the processes that occur behind developing coordinate systems and elevation models. Because we were given no instructions on how to gather data, we had to use our critical thinking skills to develop our own methods.