ArcPad Data Collection II

Introduction

 This weeks exercise is a continuation of a previous exercise in which we set up a microclimate geodatabase in ArcMap 10.2.2. This week we will be taking that previously created geodatabase and employing it onto a handheld Trimble Juno GPS unit (Figure 1). We are then going to be going throughout the University of Wisconsin-Eau Claire campus and collecting different climate variables using the Kestrel 3000 Wind Meter (Figure 2).

 



Figure 1  Juno Trimble GPS unit that was used to collect each of the weather variables for every point.

Figure 2  Kestrel 3000 Wind Meter that was the source of the microclimate data. This Kestrel
provides temperature, wind speed, wind chill, humidity, and dew point.

     As previously stated, we will be using a geodatabase that was created in a previous exercise. Creating a geodatabase before going into the field is a standard protocol that helps ensure accurate data collection. This geodatabase has eight different variables that will be collected at each point. The variables include temperature at the surface, temperature 2 meters above the surface, dew point, wind chill, wind speed, wind direction, and relative humidity. In order to ensure each variable is recorded in the GPS correctly we set up domains for each. All temperature related variables (temp, wind chill, dew point) were set with a range between -30 and 60 degrees Fahrenheit. The wind speed attribute ranges from 0 to 50 mph. Wind direction ranges from 0 to 360 degrees. Ground cover is set up with coded values for each expected ground cover that we will see in our study area. These range values can help prevent the user from incorrectly inputting the data, for example adding an extra 0 to temperature values. Incorrect data entry can be a huge problem while in the field, especially if environmental conditions aren't favorable.

Study Area

     In this exercise we are going to be collecting microclimate data across the UW-Eau Claire campus (Figure 3). UWEC sits at the heart of the city of Eau Claire, WI on the banks of the Chippewa River. UWEC contains 28 major buildings on 333 acres. For this exercise we will be mainly staying within the main courtyard of the UWEC campus. March temperatures usually range between 0 and 50 degrees Fahrenheit. Relative humidity values commonly range between 20 to 60 percent during this the spring months, which is significantly lower than during the summer months. UWEC seems to always be really windy, especially on the walking bridge over the Chippewa River. Depending on the day wind speeds usually range from 0mph to 20mph, but 50mph gusts are frequent during large storms. The majority of the ground cover at UWEC is grass, blacktop parking lots, or concrete sidewalks. However, since it's March, there is a large amount of campus that still has snow covering the ground.










Figure 3  The UW-Eau Claire campus that where we collected the microclimate data. This map shows the
different zones that each group was assigned and the location of each point that was collected.

 

Methods

     Since this exercise was a direct continuation of last weeks exercise we used very similar methods to obtain our climate data. We started by creating a uniform geodatabase with a set of domains that everyone will deploy to their group's Juno Trimble GPS unit. We then split up the UWEC campus into five different sections and each group was tasked with collecting as many data points as possible. For each of the points we were to collect six climate variables:

• Temperature at surface
• Temperature two meters above surface
• Wind speed
• Wind chill
• Relative humidity
• Dew point
• Ground Cover
• Any associated notes that would be helpful.

     When each group finished collected points we all met back in the lab to combine our data points into one feature class. To do this we opened the file manager and navigated to the folder in the GPS unit that contained the microclimate data. We then copy that folder and past it into the desired folder in ArcMap. We then use the ArcPad Data Manager, similarly to last week, and use the Get Data From ArcPad feature. This feature converts the .apm file, collected in ArcPad, to a .mxd file that can be used in ArcMap. After each of the groups imported their data into an ArcMap friendly format we used the Merge tool to combine all the individual feature classes into one feature class that can be used to create interpolations for each variable.

     After the feature class with every point was created we ran the Inverse Distance Weighted spatial interpolation to create a raster surface for each variable that covered the entire study area. I decided to use the IDW interpolation because it is exact, meaning that the surface is given the value of each point that it passes through. IDW also sets each of the points as the highest values in the interpolated surface, which can give certain values a "bulls eye" look to it.

Results

     After the individual feature classes were merged together we were then able to run a geostatistical interpolation to create surface maps for each of the variables. Figure 4, below, shows the table of the main feature class after being merged.








Figure 4  Screenshot of the table after each individual feature class was merged together to form the main feature class that will be used to create the interpolated surface. As you can see there were some values for temperature at two meters that were left null, which had an impact on the interpolated surface.

     If there were any ground cover types that were not included in the original geodatabase, such as the canoe pictured below (Figure 5), we were to select the ground cover type of "other" and input a note that provided more detail as to what the exact ground cover was.







Figure 5  An example of a ground cover that was not originally set in the domains of the geodatabase. For this specific point we set the ground cover to "other" and wrote entered more information in the notes field.

 

 

     After running the IDW interpolation we were left with a series of surface rasters. Figure 6 is the raster surface that was created based on the temperature values collected two meters above the land surface. As you can see, there are two really cold points located along the western edge of the study area and randomly located hot spots spread throughout the study area.

 



Figure 6  IDW interpolated temperature two meters above the surface. Also shown are the ground covers associated with each point.

 

     We also created a surface for the surface temperature values (Figure 7). This surface seems to have a lot more variation than the map above where a very small area of the surface is given an intermediate temperature. 

 



Figure 7  Surface temperature surface created using the IDW interpolation. The temperature surface is compared to the ground cover located at each point.

 

     Figure 8 shows the surface created based on relative humidity values collected around campus. There is a large area with high relative humidity located along Putnam Drive. Compared to the surface temperature map above, the areas with high relative humidity correlate to areas that have colder surface temperatures. 

 



Figure 8  Relative humidity raster created using IDW interpolation.

 

     Figure 9 is a raster surface created based on dew point values collected by the Kestrel 3000. Dew point is the temperature the air must fall to in order for precipitation to occur. The dew point map below seems to have a strong correlation with the relative humidity map above. This means that areas with high relative humidity also have a higher dew point temperature. This makes sense because air that has more moisture requires the temperature to drop a lot less before it starts precipitating.









Figure 9 Dew point raster created using the IDW interpolation. Dew point is the temperature at which the air must drop to in order for precipitation to occur. The dew point raster and the relative humidity raster seem to have a direct correlation.

 

     Figure 10 shows the values of wind speed collected around the UWEC campus. As expected the area along Putnam Drive had really low wind speeds since it is located right at the bottom of the campus hill. Areas on top of the campus hill had an overall higher wind speed than areas on lower campus. The area on the walking bridge that spans the Chippewa River also has higher wind speeds, which is expected, as well. 

 



Figure 10  Wind speed surface created using IDW interpolation. This map is an accurate representation of the different areas of campus. The areas in the southwestern portion of the map that have high wind speeds are located on upper campus. There is also high wind speeds located on the walking bridge that goes across the Chippewa River, which as everybody who has ever walked across the walking bridge knows is very accurate.

 

     Figure 11 is a surface raster of the wind chill values collected from the Kestrel 3000. Since wind chill is directly correlated to temperature and wind speed I combined the wind chill raster with the associated wind speed points. One would think that areas with higher wind speed would have lower wind chills. However, this was not always the case. Certain points had higher wind chill values than surface temperature values.  

 



Figure 11 Wind chill surface combined with wind speed values at each location.

 

Discussion 

     There were many sources of error in this microclimate modeling project. First off, since the data was collected over a two hour period late in the evening the temperature related variables were significantly different from the time we started collected data points to the time we finished collecting the data. For a more accurate microclimate model we would need the data to be collected at each point simultaneously to prevent environmental conditions from changing. This is possible over larger areas that have multiple weather stations running, but not on such a small scale microclimate project like we are doing for this exercise.

 

     The instruments used for collecting the data also seemed rather inaccurate which provided incorrect climate data. There were several occurrences in which the wind chill was actually warmer than the surface temperature values at that specific location, which is not possible. I also noticed that some areas had surface temperatures that were warmer than the temperature at two meters and some areas where the surface temperature was colder than the temperature at two meters. Although one would expect surface temps to be colder when there is snow on the ground, I noticed that there was no real pattern associated with ground cover.

 

     Another source of error that directly impacted the interpolations was incorrect data entry into the Trimble Juno GPS unit. For instance, as you can see in Figure 12 several temp at 2 meter locations were not collected. This error gave the central part of the study area a more generalize interpolation because there was less data points to influence the model. There was also two points that were given a value of 0 degrees. This error had a huge impact with the interpolation. Since these two points were huge outliers they minimized the effect of the other extremes, meaning that the lowest temperature that was correctly collected did not seem like it was as cold as it actually was.

 

Figure 12  Errors that had impacts of the raster surface created. Several temperature at two meter values were left null. These points did not  have a significant impact on the overall surface raster. However, there were two points that had temperature values of zero. These points had huge impacts on the overall surface raster.

 

      Since these two errors greatly reduced the accuracy of the interpolation, I removed the erroneous data points and ran the interpolation a second time. Figure 13 appears to provide a much more accurate representation of the actual temperatures that were collected. As you can see there is a string of cold points that were collected in the southern portion of the study area. These points coincide with Putnam Drive, located at the bottom of the UWEC campus hill. Since this area was located on the north face of a steep slope it does not recieve much sunlight in the winter months and was still covered with several inches of snow, which could have impacted the temperature in that area.

 

Figure 13  After the erroneous values were removed we were left with a much more uniform looking raster surface that better models the actual temperature around UWEC.

 

 

     Similarly to Figure 12 above, there was a data entry for the Wind Chill field as well (Figure 14). One data point was given a value of zero for the wind chill even though that point had a 52 degree surface temp and only 2 mph winds. Since there was an abvious data entry error I went into ArcMap and removed the point from the dataset and ran the IDW interpolation a second time Figure 15. 



Figure 14  Erroneous wind chill value that was set to zero degrees. This point had a huge impact on the overall raster surface.

 

 

     As you can see, there is a significant difference between the two Wind Chill interpolations. Figure 14, above, has a cold spot that makes everywhere else appear much warmer than it actually is. After removing the erroneous data point we can see that more of the variation in wind chill throughout the area. After removing that point, the surface better represents the actual wind chill values experienced around campus (Figure 15). 

 

Figure 15  After the erroneous wind chill value was removed we were left with a much more accurate surface model.

 

Conclusion

     In conclusion, this exercise was very similar to last weeks exercise. The only main difference was that we covered a larger area and we were required to merge each group's data into one main feature class, which is a very simple process. Although we set the domains before collecting the data to prevent erroneous data, we still experienced several while in the field collecting the data. These errors can have huge impacts on the surface rasters created. Therefore, it is very important to be as careful as possible while inputting the data into the Trimble Juno.