Jellyfish Monitoring UAV Imagery - Pruth Bay - Calvert Island

UAV imagery of jellyfish blooms in Pruth Bay, Calvert Island, BC. Imagery is stitched and georeferenced in order to extract jellyfish numbers and physical orientation. Data collection and analysis is being conducted by the Hakai Institute with further research at the University of British Columbia under the supervision of Dr. Brian Hunt.

Low level flights were conducted simultaneously with oceanographic sampling in order to get high spatial resolution imagery of the jellyfish groupings in Pruth Bay. In the methodology below we detail "sets" - each "set" represents an imagery dataset / jellyfish column tow data / Topcon DGPS point / and oceanographic sampling data.

Methodology:

1) Use the UAV to search for jellyfish masses. Success for finding jellyfish depends on weather conditions and spatial distribution in the area. Generally on clear and calm days the jellyfish are very easy to spot in the ipad. Under poor conditions and low jellyfish numbers some image processing and filtering may be required.

2) Once a jellyfish mass is found plan a transect that include as many georefereing targets as possible. Referencing to land features works well if the mass is close to land, but open waters could require water targets. We developed visually complex anchored devices (Lund boats with lots of targets and material) which will assist image stitching software merge imagery together accurately. If it is calm and the boats are anchored well use the topcon to RTK the boat targets. The boat targets are to be spread out to contain the area of interest - triangle / diamond pattern depending on the number of targets.

3) Transects are to be flown quickly to avoid jellyfish blur in the imagery. Overexpose the imagery so the jellyfish are observed easily. Aim for image overlap of 80% - some images may be removed in the stitching process. Maintain a constant exposure as this will greatly speed up processing time in the lab. Remote sensing techniques can be used more efficiently if the imagery has a homogenous exposure. Single line transects are more preferable, but if conditions are calm a rectangle / return gridded transect can be attempted. Ensure to include land features wherever possible for referencing. Image capture should be limited in time for any one area a continuous straight transect is ideal to ensure no movement or blur is recorded in the mosaiced imagery.

4) Flight elevation depends on a number of factors like weather / water quality / size of jellyfish mass / distance from land. Typically 90 meters is recommended. Swath size can range from 200 m to 500 m (depending on flight elevation) and stretch over 2 kilometers in distance.

5) Use Autostitch software over PIX4D due to the poor outputs observed in marine habitats. Autostitch outputs should be at maximum resolution and include fewer photos if possible. Aim for 75% at most and less than 20 images is ideal.

6) Once the imagery is exported from Autostitch and has been quality checked it must be georeferenced. Use existing georeferenced orthophotos to target known points on land and if available ocean targets gathered in step 2. A minimum of 10 targets should be used throughout the imagery and a spline transformation is recommended. Save the georeferencing TXT file for future reference and possible corrections. Export the final product and it is ready for GIS analysis. 7) GIS analysis
Still being developed at UBC under the supervision of Brian Hunt and Jessica Schaub

Notes from the field:

September 11th, 2016

9:00 am Recon flight to search for jellyfish smack. Smack found in loose patch at the opening to Pruth Bay. Secci disc testing showed visibility to 9+ meters from the drone viewing on the ipad.

Set 1 Net drop 1) 9:25 am - ground control point #2 Fog was thick and flight elevation is ~ 50 meters.

Set 1 Net drop 2) 9:37 am - ground control point #3 Fog. Flight elevation ~ 60 meters. Transect should be good for processing.

Set 1 Net drop 3) 9:53 am - ground control point #4 Fog. Return transect at 50 meter elevation. Poor imagery for processing.

Flight recon from the dock at 11:30am. Sun and mixed clouds - light winds. 3 lunds have been moved to a smack observed at the end of the bay to act as photo stitching targets.

Set 2 Net drop 4) 11:39 am - ground control point #5. Flight elevation 117 meters. Good conditions. Lots of glare in the imagery.

Set 2 Net drop 5) 11:49 am - ground control point #6. Flight elevation 120 meters. Set 2 Net drop 6) 11:59 am - ground control points #7 + #8. Return transect at 120 meters. Secci visible at 14.5 meters.

Set 3 Net drop 7) New group of jellyfish. North end of Pruth Bay. 12:20 pm. Ground control point #9. Flight elevation 150 meters.

Set 3 Net drop 8) 12:32 pm. Return transect. Ground control points #10 + #11. Flight elevation 150 meters.

Set 3 Net drop 9) 12:45 pm. Ground control point #13 + 14. Lots of glare in the imagery. Return transect. Flight elevation 150 meters.

Set 4 Net drop 10) 3:15 pm. Single transect near parks log. Elevation 100 meters. Ground control point #15.

Set 4 Net drop 11) 3:25pm. Flight elevation 90 meters. Ground control point #16.

Set 4 Net drop 12) 3:34pm. Ground control point #17. Tricky first pass. Just process the last pass. Elevation 100 meters. Jellyfish are dispersed more and are much deeper. Long line of them heading to the North East bay (Goose Grass Bay).

Set 5 Net drop 13) 4:00 pm. 120 meter elevation. South Pruth Islet. Ground control point #18 and 19. Return transect. Use last (GCP 19) transect.

Set 5 Net drop 14) 4:12 pm. 120 meter flight elevation. Deeper jellyfish at mid - density. Delete first few photos - false start. Ground control point # 20.

Set 5 Net drop 15) 4:18 pm. 120 meter flight elevation. Ground control point #21.

Finished work at 4: 25 pm.

Data is stored as 15 georeferenced datasets and is stored in the Hakai server - UAV Files - Calvert - September 11, 2016 Folder. All uav flights, data processing, mapping, and reporting conducted by Keith Holmes.

Contact data@hakai.org for more information

Access and Use

Licence: Appropriate credit must be given to Hakai Institute and the authors of the dataset.

Data and Resources

Dates

Metadata Created October 26, 2018, 22:48 (UTC)
Metadata Updated October 26, 2018, 22:48 (UTC)
Reference Date(s) 2016-08-28 (Creation)
2016-08-28 (Publication)
2016-08-28 (Revision)
Frequency of Update asNeeded
Metadata Date October 17, 2018, 23:21 (UTC)

Graphic Preview

Dataset extent

Map data © OpenStreetMap contributors

Additional Info

Field Value
Contact Email data@hakai.org
encoding utf8
metadata-language eng
progress completed
Responsible Party Hakai Institute (Point of Contact, Processor)
spatial-reference-system 26910