Purdue Wildlife Area "Green Up" Operation
Purdue Wildlife Area "Green Up" Operation
UAS Mission Background:
- Your company GIS specialist has worked with the PWA researchers and has created a mission area to be flown over the designated plots. The location of the GCP markers are also on this map. This map has been shared with you via arcgis.com where it is accessible online. This map is also available for you to access via ESRI Field Maps
- Please fly the mission by drawing your mission boundaries as close as possible to what you were assigned in the Fields Application. Other flight parameters are:
- Two missions need to be flown, with at least 7 days separating each mission
- Altitude: 122m
- Lateral and Frontal Overlap: 80%
Introduction:
Flight crew #6, which is myself and Conner Klinkhamer were tasked with collecting data of burned vegetation out at Purdue Wildlife Area (PWA). Our goal was to provide data to a group of researchers that are interested in quantifying what areas, and how fast, portions of a recently burned prairie plot “green up” as we move from early to mid-spring. We are going to do this by taking UAS data and putting it into the context of a geographic information system package for further analysis. Our job was to collect two different sets of the same data over the burned plots at least 7 days apart to compare the difference of vegetation growth. The first set of data was collected on 4/15/22 and the second set was collected on 4/22/22.
Method:
We utilized a DJI Mavic 2 Pro for our data collection and 5 Aeropoint GCP markers to ensure the quality of our data. We then used Pix4D Mapper to generate and orthomosaic and digital surface model (DSM) of each flight operation while incorporating the Aeropoint GCP’s. Next, we used ArcGIS Pro to run the analysis and create our cartographic maps. There were multiple different crews working on this same mission and we were each tasked with marking the location for each GCP point, but the actual data was provided to us by our GIS specialist.
The study area for this operation was at Purdue Wildlife Area, located just off of state road 26, North West of Purdue University campus. Figure 146 shows the mission area at PWA along with the locations of where the GCP data points were collected. PWA has a large open field with many different plots where they do research on all different kinds of wetland, savannah, and prairie ecosystems. The 4 plots that we focused on are marked by the red outline signaling the mission area. The two plots at the bottom are areas that were recently burned and the two on top were the control areas. The blue pin marks show the location of the actual GCP data that was provided to us by our GIS specialist and is the data that was used to generate our orthomosaic and digital surface models.
For figure 147, I was able to export it before all the GCP points were removed from ESRI Field Map so you can see the location variability between all the points collected by each member of the different crews. You can see that the GCP points are numbered diagonally starting with GCP #1 in the bottom right hand corner with #2 in the middle and #3 in the top left. Then #4 in the bottom left and #5 in the top right. I added insets for each GCP so you can easily see how wide spread and inaccurate data collection via mobile devices actually is. You can see within the insets the highlighted cyan points that I mentioned before which represent the data points that I collected. I also have the metadata for the data collection. Another aspect that I think it is important to have is the actual photographs of the location for each GCP which I have on the left side of the map. It is important because it helps to visualize the location, and if there’s an error, it’s easier to go back and see where that specific GCP was located. I would like to note that the image for GCP #1 was spray painted with the number 5, that was an error as it is actually GCP #1.
Results:
Once you launch Pix4D, add the images, set your coordinate system, run the initial processing, add the GCP’s with the raycloud editor, re-optimize, and finish the processing, we have our orthomosaic and DSM for each flight. Figure 148 shows the quality report from the first flight processing and Figure 149 shows the quality report from the second flight processing.
Figure 150 shows the orthomosaic and DSM from our first flight operation that we collected on 4-15-22. The image on the right is the orthomosaic which shows the RGB colors of the operation area. The map on the left shows the DSM which shows the elevation of the terrain. I believe the uneven shading on the DSM is due to the cloud cover casting a shadow messing with my elevation model. Below figure 150 the shows the metadata sheet from the first flight operation. The metadata is also presented in the map itself.
Location: Purdue Wildlife Area
Date:4-15-22
Vehicle: DJI Mavic 2 Pro
Sensor: 1” CMOS
Battery: 1
Flight Number: 1
Takeoff Time: 12:58pm
Landing Time: 1:04pm
Altitude (m): 122m
Sensor Angle: NADIR
Overlap: 80%
Sidelap: 80%
Cloud Cover: Scattered high altitude cover
Wind Direction: Out of South West
Wind Speed: 16 mph
Temp: 63F
PIC: Connor Klinkhamer
VO: Hunter Donaldson
Notes: Winds were strong upon arrival but slowed down before takeoff.
Figure 151 shows the orthomosaic and DSM from our second flight operation that we collected on 4-22-22. The image on the right is the orthomosaic which shows the RGB colors of the operation area. The map on the left shows the DSM which shows the elevation of the terrain. Below figure 151 the shows the metadata sheet from the second flight operation. The metadata is also presented in the map itself.
Location: Purdue Wildlife Area
Date:4-22-22
Vehicle: DJI Mavic 2 Pro
Sensor: 1” CMOS
Battery: 1
Flight Number: 2
Takeoff Time: 12:36pm
Landing Time: 12:42pm
Altitude (m): 122m
Sensor Angle: NADIR
Overlap: 80%
Sidelap: 80%
Cloud Cover: Heavy cover upon arrival, cleared up mid/post flight
Wind Direction: Out of East
Wind Speed: 9 mph
Temp: 62F
PIC: Connor Klinkhamer
VO: Hunter Donaldson
Submitter:
Notes: GCP locations were hard to locate due to spray paint fading over time
Figure 152 shows a side-by-side comparison between the first flight operation and the second flight operation. An easy observation to make is that the ground on the second flight is much darker than the ground on the first flight. This is due to recent rainfall that occurred just before our flight operation. It is difficult to see, but if you look close enough you can see that the second flight has more green areas in the burned area. Just from a week, there is a notable change in vegetation growth in the burned area.
Figure 153 shows the first flight operation after it has been reclassified. Reclassification is the interpretation of raster data by changing a single value to a new value, or grouping ranges of values together. This allows us to compare the different features and allows us to better interpret the burned and unburned areas. I go into more detail on the reclassification process in another post that you can find here. The 5 different classes I used is: Burned Area, Dead Grass, Thorn Bush/Sticks, Green grass, and Trees.
Figure 154 shows the second flight operation after it has been reclassified. Once again, the 5 different classes I used is: Burned Area, Dead Grass, Thorn Bush/Sticks, Green grass, and Trees.
The final map, figure 155 shows a side-by-side comparison of just the burned/ unburned area. The map on the left is the first flight operation and the map on the right is the second flight operation. These are the two parent classes that I created which all the other subcategories fall into. The nice part about this type of classification is that we can calculate the area for both burned and unburned on each flight operation. As you can see, there is more unburned spots on the second flight in the burned region compared to the first flight. Although, the area calculations do not tell the same story. This is due to the fact that during the reclassification, ArcGIS Pro mistakenly grouped some of the “Thorn Bush/Sticks” into the burned category. This is most likely due to the fact that it just rained so the ground was darker and resembled the same color as the burned area.
Conclusion:
Overall, I think this was a successful operation and Conner and I would like to thank the researchers out at Purdue Wildlife Life Area for allowing us to come out and collect some data and run analysis for them. I think the findings are significant and show just how fast the vegetation can come in after a recent burn in the early to mid-spring. If I were to change anything for the next time to improve this sort of operation, I would like to collect more than two sets of data so we can better visually assess the time span of how quickly this vegetation can grow. I would also try not to collect data right after rainfall so that it doesn’t mess with our reclassification and area measurement.
Comments
Post a Comment