Sandbox Survey

Introduction

 Our task was to build a model representation of various geomorphological processes in a sandbox environment and then to capture those processes using spatial sampling, in the hopes that this exercise will begin to explain the techniques and dilemmas that may occur as field surveys are conducted, and to help us understand sampling and spatial thinking. The terrain that was to be included is, as follows; a ridge, hill, depression, valley and a plain. Ultimately, this data will be imported into ArcGIS in order to assess the accuracy of our chosen survey technique.

Sampling, in the broadest sense, is an attempt to capture or categorize data from a sample or population such that the sample can be representative of the population from which it was taken. Similarly to sampling that is often used in statistics to capture attributes of populations, spatial sampling seeks to capture attributes of spatial locations such that a model or representation can be assessed at a later time. As noted by Gao, Ge and Wang, the main aim of spatial sampling is to collect samples in a 1-, 2-, or 3-dimensional space (2012). It is typically used to estimate the total or mean for the parameter in an area, to optimize parameter estimations for sample locations, or to predict the location of a movable object (Gao, Ge and Wang 2012). Spatial sampling can be categorized by the statistical sampling models that are present in all sampling techniques, such as random sampling, systematic sampling and stratified sampling. In simple random sampling, samples are chosen from the population with the intention that all samples are capable of being chosen with equal probability, such that any sample chosen has no inherent bias (Gao, Ge and Wang 2012). Though random sampling may be used in spatial sampling techniques, major issues can occur, such as large portions of data not captured as some features may not be chosen or poorly represented, and thus a representative model could not be accurately or reliably reconstructed from the representative data that was captured using the technique.

Systematic sampling is a highly regulated process as compared to random sampling, such that every measurement or capture of data is done in a preset order with regular intervals. Once a sample is taken all subsequent samples are measured in the same way to ensure uniformity in samples taken. Systematic sampling also has drawbacks such as under or over representation of the population due to the pattern of systematic sampling chosen (Royal Geographical Society).

Stratified sampling attempts to break up a population into smaller sections that are then captured and used to create a representation of the whole. But in order for stratified sampling to be truly effective, it must be assumed that the subgroups that are being sampled are representative of the whole. After determine the subgroups to be sampled, random or systematic sampling is applied to subgroups in order to capture the representative information. Drawbacks to stratified sampling include difficulty in determining the criteria of each subgroup such that each subgroup is an accurate representation of the whole population, and that the subgroups must be proportional in order to quantify the population (Royal Geographical Society).

Methods

 The location of our “sandbox” was a garden planter in Eau Claire Wisconsin, at the University of Wisconsin Eau Claire, in the Phillips, science building, courtyard. The Garden box was roughly two meters long and a little over a meter wide. For the purposes of this experiment, exact measurements of the box were not taken as we had chosen to do a systematic sample and restricted our modeling to an area exactly one meter square. As this exercise was designed to show us scale representation of the modeling in the field we choose to further limit our area, scaling it down, such that a grid of uniform size would allow us to build our model and systematically sample data with ease for us. Systematic sampling was chosen because we determined that an area one meter square would not require a large effort or long time to capture as many data points as possible, increasing the accuracy of the spatial sample that we would gather. Additionally, all of the members of our group are currently studying Geographical information systems, and we know that a coordinate system that is made up by a gird is often a accurate and precise way to capture geographic and spatial data.

 Our first task was to fill the planter box full with snow in for the area of interest we would be conducting our survey. Using a collapsible car shovel we filled half of the planter box to the top with snow, then leveled off the top of the snow using the meter stick.

Fig 1. Packing the planter box with snow and leveling the top with the meter stick. 01/21/16

We then sectioned off a one meter square area with flags to set our stage, and determine where we would construct our terrain features.

Fig 2. After leveling the snow, we section off the area of interest. 01/21/16

Then we proceeded to construct a hill, using the dirt found to be frozen in the shape of planter. A valley that encompasses the hill. The surface of the snow that was level with the box became our plain, and the depression was made using my knee as I kneeled into the box, and the ridge was created by purposely building up only half of the box. The area that was filled with snow was purposely cut short in order to create a sharp drop and a ridge on the far end of the sample area.

Fig 3.Construction of the terrain is nearing completion, the planter dirt that makes up the hill can still be       seen. 01/21/16

Fig 4. A different angle of the construction of the terrain, showing the valley (middle), depression (top left). Ridge (foreground) and hill (top right corner). 01/21/2016

We then wanted to create a grid using pushpins as anchor points along the side boards of the box, in 5cm intervals such that a 20x20 grid would be created, however we did run into an issue of having no forth side in which to anchor the pins, so we endeavored to improvise. Finding a discarded board which had nails protruding from the bottom, we laid the board across our area of interest and pinned in place to create our forth side (carefully so that we did not stab ourselves on the nails).

Fig 5. The found board, and fourth side put in place, complete with thumbtacks to keep the board in place. 01/21/2016

Fig 6. The found board, and fourth side put in place,  as shown from the perspective of the entire area of interest.  01/21/2016

We then proceeded to construct the backbone of the grid, the thumbtacks that would hold down the grid at 5 centimeter intervals for one meter.

Fig 7. Setting up the grid “backbone”. 01/21/2016

After setting up the pins around the entire planter, and on our forth side board, we constructed the grid by wrapping sting around the thumbtacks and passing it back and forth.

Fig 8. Constructing the gird!. 01/21/2016

After completing one pass of the string we then did the cross side, creating a grid pattern across the entire area of interest.

Fig 8. Constructing the grid! (Part 2)

Fig 10. The finished product, the flag in the top of the hill is part triumph in completion and part functional as the string would not stay in a manner that would create a 5 cm² grid section. 01/21/2016

We then determined that sea level would be the top of the snow that was level with the box, and would be a zero measurement for height. The grid began in the upper right hand corner of last photograph which was marked with a fifth flag, on the north side of the box. From the (0,0) XY coordinates the top of the box was marked with number designations while the North side (right side in the last photo) was marked with letter designations. Sampling of the height occurred at the same corner of every 5 cm² grid square, the lower right hand corner of the last picture. As the imposed grid varied in height we had to come up with a way to measure sea level that did not rely on the grid we had laid. In order to ensure accuracy of the Z axis, we bound multiple meter sticks together to give them some rigidity then set them level with one side of the box, marking sea level across that row of the grid, and measured the Z axis perpendicularly to the inferred sea level.

Fig 11. Measuring the Z axis across the grid, as noted the grid did not lay flat across the created terrain features. So a imposed sea level was inferred by a bound rigid object that could lay flat across the surface, and could represent sea level. 01/21/2016

To measure objects that would were above sea level another rigid object (another board found in the shed) was placed flat on top to capture the maximum height of the object. Then the sea level meter stick was pushed as close as possible to the base of the object in order to capture its height.

Fig 12. Measuring the Z axis across the grid (Part 2). Measuring the height of objects above “sea level” l. 01/21/2016

To capture the data we had one person designated as a data recorder with a drawn grid in a notebook with each square noted with a number and letter. One inserted the rigid sea level meter stick into that row of the grid and also held the meter stick upright such that the holder of the meter stick could relay the accuracy of the third person who was doing the actual Z access measurements. As the measurements were taken, the recorder repeated the grid number and letter designation and the measurement. We choose this method as it made sure that the measurements that were recorded had the consensus of two people and was therefore more accurate or at least double checked when recorded, and that the recorder of the data would be able to double check the grid location of each measurement.


Results/Discussion


We recorded one value for each square in the one square meter area, with each square being 5 cm2, this resulted in a 20 x 20 square grid and a total sample of 400 values. Our sampling technique did remain constant throughout our survey, with measurements taken consistently with the way that was described in the above methods section, a Z axis measurement in comparison with a sea level plain which was created by a second meter stick. Each sample was taken from the same corner of each square.
While we may have been able to employ another sampling technique such as a stratified technique as we did know that other than the features we created, the plain of most of the surface in the area of interest would be at sea level do to us packing the snow that way when we made the initial setup of the box, at that point we could have taken measurements of each of the features Z axis measurements or a sub sample of Z axis measurements and extrapolated the data across the full area of the feature. This technique would have taken less time, but would have been a lot of extrapolation and more guess work then the sampling technique we employed.
We did run into a few issues during sampling, such as the first measurements of the Z axis had to be changed on the fly, as it was discovered as we began moving towards the hill that the sea level measurement would not be consistent across the entire grid. So we had to figure out what we could do to impose a sea level that would be consistent across the entire grid, and we determined using a rigid level object to impose the sea level would be the best way to get accurate measurements of the Z axis. Additionally, we had issues with the grid remaining at a constant taught position, especially when the wind picked up, we had to readjust how taught the string of the grid was while we were measuring.

Conclusion

 The sampling technique we employed was truly a systematic sample, which did serve our needs well for this experiment. We consistently and systematically sampled from the same relative location in each square of the grid, using the same sample method each time. Knowing that this was a model of scale-able techniques that could be used on sites and locations that are much larger than the artificial site we had created, I would choose a different sampling technique if we had to do this same experiment in a life sized situation, all the features being equal.

 At the end of this experiment two major concepts have been reinforced in my head, one from GIS and is the reason why a projected coordinate system develops a false origin and uses only positive numbers to relay positions. It is much easier to record, organize, and maintain the integrity of the data that you are using if all of the data you are using is positive and localized to your specific area of interest. And the other concept is just how scale-able sampling of spatial areas is, and how much information it can give. We could easily have implemented this survey over lager areas (with varying techniques, and different technologies) but the core of the survey would have remained the same, measurements of a X,Y, and Z axis. Spatial sampling can take something that is not on a scale that we are used to thinking of it and use that data in a way that makes the terrain, landscape, or feature being surveyed tangible. By using the data to build a picture or model or a topographical map we can really look at any area of interest or feature and have that information on hand in such a way where we can make informed decisions. These decisions have been used in countless ways in the past and will be continued to be used in the future.

Citations

Wang, Jin-Feng, Gao, Bin-Bo, Ge, Yong. A Review of Spatial Sampling. Spatial Statistics Volume 2, December 2012, Pages 1-14.

Royal Geographical Society. Royal Geographical Society, Sampling Techniques