Project Data Flow SAMBAH a case study Is

Project Data Flow

SAMBAH - a case study Is the Baltic harbour porpoise extinct? • One Data Manager - Daniel Wennerberg • 9 countries and languages. • Multiple Lat/Long formats from ships (see freeware on Chelonia website) • 400 years of data from >300 C-PODs. • 29 billion clicks, 4 billion porpoise-like clicks • 6 million clicks in trains from the Hel 1 classifier. <1 false positive sec/year • Outcome: ? designation of 3 marine conservation zones.

File naming Island SE 2017 3 10 FPOD_0004 file 0. FP 1 ! t s e b e h t s i e m e h c s s i h t Location date of file start POD identifier file number type this deployment • • Keep all files in one directory Use stable location names If you want to use the array viewer there are rules on the first 3 characters Let CPOD. exe construct the name then • File lists in Windows can be sorted by place, deployment date, POD number, or by file type, or file date • Batch exports are quick and easy

Archiving • Check length of CP 1 or FP 1 after reading SD card. Does it match the deployment? • More checks if it doesn't match… • Archive all CP 1 / FP 1 files, plus metadata spreadsheet. That's the minimum. • CP 3 / FP 3 files with detections can be regenerated later if needed.

File listing File. Name POD starts ends start. Min end. Min Logdays Muara Pahu POD 1600 file 01. CP 1 1600 25/11/2016 15: 55 29/01/2017 07: 16 61487515 61580596 64. 64 Muara Muntai 2016 11 25 POD 2840 file 01. CP 1 2840 25/11/2016 09: 32 03/02/2017 19: 17 61487132 61588517 70. 41 Muara Kedang Kepala 2016 11 21 POD 2837 file 02. CP 1 2837 29/12/2016 05: 22 31/01/2017 23: 18 61535842 61584438 33. 75 Muara Kedang Kepala 2016 11 21 POD 2837 file 01. CP 1 2837 21/11/2016 17: 24 29/12/2016 22: 44 61481844 61536884 38. 22 Muara Kaman 2017 11 23 POD 2836 file 01. CP 1 2836 23/11/2016 17: 30 29/01/2017 20: 58 61484730 61581418 67. 15 Pela River Mouth 2016 11 23 POD 2839 file 01. CP 1 2839 23/11/2016 17: 14 30/01/2017 14: 49 61484714 61582489 67. 9 Muara Belayan 2016 11 2 POD 2838 file 02. CP 1 2838 31/12/2016 23: 27 26/01/2017 18: 29 61539807 61576949 25. 79 Muara Belayan 2016 11 2 POD 2838 file 01. CP 1 2838 21/11/2016 17: 26 31/12/2016 23: 26 61481846 61539806 40. 25 Mahakam POD 1600 file 02. CP 1 1600 25/09/2016 01: 29 08/10/2016 06: 16 61398809 61417816 13. 2 Mahakam POD 1600 file 01. CP 1 1600 04/08/2016 17: 22 25/09/2016 01: 28 61324882 61398808 51. 34

Metadata • Keep a spreadsheet of all deployments: place, POD start time, deployment time, mooring type, local conditions. • Use the data fields in the files. Can be exported. to change embedded data

Checking files • after reading the file from the SD card, look at the whole file: for discontinuities in sound, temperature, angles

ackup Backup B …if you don't have 3 copies of your data, you don't have data Aaargh

Cropping files • after reading file from SD card

Cropping files • Select point after the boat noise has gone • Right-click pop-up menu • Set selection start go to end … Set selection end or • use automated cropping … crops to midnight

Cropping files copy & crop Kawda, Sarjekot 2016 05 10 POD 1592 file 01 PART 206 d 5 h 28 m. CP 1 then save cropped file in 'Useful Files' folder for project

• Kawda, Sarjekot 2016 05 10 POD 1592 file 01 PART 206 d 5 h 28 m. CP 1 • Kawda, Sarjekot 2016 05 10 POD 1592. CP 3

Folders and processing • Archive folder : all CP 1 files. • Working folder : 'Useful files': all useful CP 1, CP 3, cm 1, cm 3 files • cm 1, cm 3 files are essential map files. They can be safely deleted, but will be rebuilt when required! • Time selections can now be used for other purposes • Re-run train detection

Train detection • The big one! Does multiple hypothesis testing. Slow. • Encounter classifiers can be run and rerun without repeating KERNO • … if there cannot be any NBHF species • e. g. if there are VEMCO fish tags

Decide on encounter classifier • KERNO - widely used for porpoise studies. Often gives low sensitivity for dolphins. • GENENC - often gives better ( x 2) sensitivity for dolphins and reduces false NBHF detections from dolphins at a cost of slightly reducing NBHF sensitivity. • Hel 1 - only detects porpoises. Assumes no dolphins occur, so it may give false positives from dolphins. Developed specifically for the Baltic using training data from the Hel Marine Station, Poland. • KERNO-F - under development. Uses F-POD data. Wide range of training data needed. Please give generously! • Choice based on validation of a sample of detections.

Validation tools • Inspect trains - the key test - see 'Validating detections' • Analysis page - shows the impact over the whole file of different criteria

DATA issues see Choice of Filters and Stats Difficult issues: • How much of the time are dolphins silent? • Can dolphin group size be determined from acoustics? • Is bottom feeding missed in behavioural studies? • Diel patterns of detectability - all species. Seasonal change in this? • Deployment depth effects. • Halocline, thermocline ( known together as the pycnocline) effects? • Few parameters are normally distributed. • Do you need a detection function? Or can you use a trend?

than k you for y our a ttent ion