Lab Three: Creation of a Navigation Map

Figure 1. Finished hill shaded navigation map.  

All members of the class were required to create a navigation map of the area around the UWEC Priory. The Priory was a monastery at one point but it has been converted to a multiuse property that, amongst other things, houses students and offers a small slice of outdoor access to the college population. It is located south of the city of Eau Claire and encompasses an area of about half a square kilometer.

                It was decided that the best approach to the creation of a simple, uncluttered map, was to place a hill shade raster under a transparent image to give the effect of elevation without the mess of isometric contour lines. Figure two is the workflow model for the creation of the hill shade raster. The elevation data came from a Lidar dataset where six tracts were used to cover the area of interest. The first step was to collate the six .LAS files into a single multipoint feature class. I chose to only include the final return from the Lidar data as it typically signifies the ground level.  

Figure 2. Flow model showing the process used to create a hill shade raster from a Lidar dataset. 

The multipoint feature class was then re-projected to the same coordinate system (GCS North American 1983 HARN) and projection (NAD83 WISCRS Eau Claire County) as my dataframe. This projection can be defined as lambert conformal conic. It works well at portraying areas in the multitudes that stretch from East to West. I chose it because it was the most specific predefined projection available from Arcmap, meaning that any levels of distortion at this scale will be minimal.

After the multipoint feature was projected it was then converted to a raster file. This conversion included the option to set the cell size for each pixel and the values to be calculated. Each cell was set to have a spatial resolution of two meters with the mean of the elevation data set for each pixel. This gives a very good estimation for the elevation while producing a moderately smooth image.

From here it was a simple process to finish the operation by running the new raster image through a hill shade tool. The resulting image was placed under a true color satellite which was set to have 50 percent transparency. This creates a three dimensional effect on the image that makes it easy for a map reader to discern the location of hills, valleys and their respective slopes on the map.

Lab Two: Survey of a Sandbox

Introduction

The previous lab was comprised of collecting survey points from a planter plot. In this exercise we learned about different survey techniques. These techniques are: random, systematic, and stratified. For my group’s survey we used a systematic sampling technique with 8 cm intervals. In total there were 180 points measured in a 122 cm by 88 cm rectangular grid. For this lab I went through the process of importing the survey data to ArcMap and editing it for a three dimensional display in ArcScene.

After importing the data, we cleaned it up via normalization. This entailed organizing and analyzing the data to remove redundancies and correct errors.

Each data point collected has three data fields associated with them: an X value, Y value, and a vertical Z value. In this lab we were required to import the Excel data into ESRI ArcMap. This was a simple process of adding the coordinates to map and exporting as a point feature. From here the newly exported feature was opened in ArcScene, a program capable of showing the point data in three dimensions. ArcScene was further utilized as an interpolation tool to improve the display of data via the creation of a 3D raster file. Below I have summarized and given an example of several different interpolation methods available within ArcMap.

For each display I decided to use an oblique angle to best show differences in relief, while maintaining some aspect of scale. One will notice the two scale bars on each map, one labeled “88CM” and the other “122CM.” I have also included a north arrow in each map. It may seem silly to label the distances and orientation for such a small survey area but if someone were unaware of the nature of these past two projects they may not be able to tell the difference between the display of these features in a sandbox to large, kilometer long, features. 

Methods

Figure 1. IDW Interpolation

Inverse distance weighted. This interpolation technique runs on the assumption that features close together are more similar than ones that are further apart which makes this interpolation method useful for large data sets. Due to how the technique weights distances the average number cannot be above or below the highest and lowest points respectively. This makes the display of ridges and valleys a far cry from reality unless the valley bottoms and crests have also been sampled.

Figure 2. Nearest Neighbor Interpolation

Natural Neighbors interpolation, much like IDW, weights the values survey data to create a surface which passes through each point. IDW also will not display un-surveyed valleys or ridges as the displayed surface cannot be more or less that the highest or lowest points.

Figure 3. Kriging Interpolation

Kriging interpolation uses both weighted values and the overall spatial arrangement survey points. The result is a surface that was created from predicted values that strive to emulate unknown values. These values are calculated with the use of clusters and trends in the data. Compared to the other interpolation types Kriging appears to be the most true to form.

Figure 4. Spline Interpolation

Spline interpolation, also known as “thin plate interpolation,” bends a sheet through each point and strives to keep that sheets curvature to a minimum. This creates a smooth, continuous surface.

Figure 5. TIN (Triangular Irregular Network) Interpolation

TIN or triangular irregular network interpolation, works by using each survey point as a node. These nodes are then used to create triangle polygon features that effectively model elevation. Tin files are put to use best with data sets that have a high density of points in areas that have large variations in elevation.

After running all the interpolation methods it was obvious to me that several areas of the landscape needed to be surveyed to better create an accurate representation of the real world sandbox. For example, there was a ridge line on the south edge of the map that appears much lower in the model than it is in the reality. This is due to the spacing of the original survey and how the crest of this ridge line lined up approximately between the survey distances, meaning that only the lower sides were measured.

Figure 6. Locations of new survey points (North is down).

My group did a second, much less extensive, survey of the micro landscape that included 17 new points that were collected in the most problematic areas of the original survey (Figure 6).

Figure 7. TIN interpolation showing the effects of additional survey points to underrepresented areas. Note that the ridge crest on the southern edge of the map is displayed with a higher altitude and appears more peaked than the previous TIN image. 

I added the new survey points in the same fashion as the first set of data. From here it was a simple process to merge both data sets, open them in ArcScene, and create a TIN file using the same interpolation method as before (Figure 7).

Summary and Conclusions

When compared to other field based surveys I would have to say that this example is relatively realistic, aside from how we had the ability to resurvey the landscape. In a full scale survey of a large landscape it is often likely that resurveying an area is not possible. 

It is not always possible to perform such a detailed grid-based survey. In many cases the terrain is far too rough and inaccessible to perform such a survey on the ground. With remote sensing technology as it is today it is fairly easy, if expensive, to get very accurate measurements of landscapes.

Interpolation methods such as those used here are commonly used to display temperature data, rainfall amounts, and, in the field of remote sensing, image rectification.

Lab One: Survey of a Sandbox

Lab One: Survey of a Sandbox 

Introduction

The first lab of Geography 336, Geospatial Spatial Field Methods, strove to introduce students to field work through the collection of data points on a relatively small scale. Students were tasked with creating a miniature landscape contained within an approximate four by eight foot (~120x240cm) planter box located on the UWEC campus. Students were tasked to create a landscape containing features such as valleys, hills, and ridges with snow or dirt, depending on the season (Snow in this case). Once the scenery was created we were required to survey the area in the hopes of eventually creating a digital elevation model of the scene with the use of ESRI ArcMap software.

A short definition of sampling is a technique of measurement that doesn’t require the data collection of all aspects of a study area. Yet it still merits significant results. Sampling measures a representative group of factors instead of recording all factors. This saves time and resources. And, when done correctly, will typically give results very close to those that would come from taking all possible measurements.

There are three major types of sampling techniques used for the quantification of spatial data:

Random sampling is used when the area of study in relatively uniform. One can assume that any sample location will have many of the same attributes as the surrounding area. Random sampling locations should not be determined by the person collecting samples; this has the potential to introduce biases into the experiment. Instead it is typically acceptable to use a random number generating computer program to create coordinates for the data collection locations.  

Systematic sampling is the collection of sample data with equally spaced distance between every point. As far as sampling techniques goes this is debatable the quickest and easiest to do for surveying spatial data. It is very useful for recording data that undergoes gradual change over the study area. One needs to be careful with the use of this technique as it is common to entirely miss areas of rapid change when point intervals are set too wide.

Stratified Sampling is best used for surveying a study area with multiple feature types. It is a technique developed to minimize the exclusion of data that helps to demonstrate variation in the area of study. This stratified sampling works at the discretion of the researcher and is reliant on their background knowledge of the area of study. This is because it focusses on the collection of points from areas that are most important to the research question. Many points are simply replicates of previous points collected in that area and when put together statically represent the feature. Fewer points are collected from surrounding areas, saving resources, but limiting background data.

Methods

For this activity my group used a systematic sampling technique so as to give every aspect of our study area equal representation. In hindsight I believe a stratified sampling technique would have been a better choice for the environment seeing as there were many areas of rapid change missed by the systematic measuring system that could have been focused on more in depth. But due to the time constraints on the experiment (daylight was fading as the study drew to a close) a systematic sampling technique worked out well as results for the entire study area were quickly quantified.

Figure 1. Planter’s box containing area of study. 

The study area itself was a planters plot located on the south side of the Philips Science Hall courtyard. This is an educational building part of the University of Wisconsin Eau Claire Campus, a public postsecondary school located in the Midwest region of the United States of America. This planter’s box was approximately four by eight foot (~120x240cm) in size but for this experiment my group utilized a little less than half of it (pictured above in Figure 1). 

Figure 2. 8cm spacing of lines to delineate y coordinates for systematic survey. 

Figure 3. Altitude measurements taken with the use of meter sticks and a small level.

With the use of lengths of string, stainless steel meter sticks and thumb tacks my group was able to create a sampling scheme of 12 strings running east to west across the plot. Each string was placed 8cm apart to form a length covering 96 cm across. For the north to south axis of the grid 14 markers were placed along the short edge of the plot (Figure 2). These markers were use as references for the x,y location for every elevation measurement. Zero elevation (Z coordinate) for each point was determined by its relative position to the height of the planter box frame. Negative vales were given for measurements lower than the frame while positive values were given for points above it. These altitude measurements were taken by a two person team with the use of two stainless steel meter sticks, one as a straight edge and one for recording measurements, and a small level. A third group member recorded the results of each measurement on a paper spreadsheet. For each point an x, y and z value were assigned. After the measurements were recorded in the field the values were entered into an excel spread sheet for the easy transfer of information to ESRI ArcMap.

Results/Discussion

 The plot measured 12x14 rows. Measurements were taken at the intersection point of each row and the immediate area adjacent to the sides of the plot. This resulted in 168 interior points and 38 border points totaling 206 points of measurement. As the experiment was conducted my group stuck to the original plan for the most part. The exception being that we reduced the number of points we were planning to measure. This choice was made in the interest of completing the project while there was still enough daylight to see what we were doing.  One far more major issue was found while laying the grid lines (Figure 4). The plot that we were using had features that were much higher than the sides of the plot (the reference level for the zero elevation value). This not only meant we had to come up with two different elevation measuring techniques but, due to the overlay of strings across features, the x,y values recorded would be very difficult to verify for their accuracy. A proper remedy for this problem would be to elevate the grid lines above the tallest features so that there would be no interference with the lines, but due the lack of obvious materials to do this, my group used an “eyeballing” approach to get the approximate location for x,y values.

Figure 4. Strings laid across the plot. Notice the displacement caused by tall features. 

Conclusion

In short our sampling exercise was a sloppy version of the systematic sampling technique. I will concede that a systematic sampling technique is useful in the collection of a representative amount of data for a large area. This makes it especially advantageous as it saves resources and is far quicker/easier than measuring every aspect of a study area while still producing useful results. Unfortunately it is common to miss areas of rapid change as the measurement intervals may be spaced to far apart to accurately represent the study area. The same techniques used in this lab are small scale versions of what have been used in the field. While the tools and materials may be different at larger scales the techniques and practices are quite similar. Yet despite this I do not believe that my group’s survey will create an accurate representation of the survey area. There simply were not enough points generated.

Sources used for background information:

http://www.rgs.org/OurWork/Schools/Fieldwork+and+local+learning/Fieldwork+techniques/Sampling+techniques.htm

https://www.geography-fieldwork.org/geographical-enquiry/before-you-start/2-fieldwork.aspx