1. Walk through a file Ilha_Grande 2016 10 23 CPOD1372 file01.CP1
Thanks to Renan Paitach
These data files are exclusively for training purposes. All rights are reserved.
Any use, disclosure and reproduction of data by third parties is strictly prohibited.

2. Location:

3. Species Franciscana Pontoporia blainvillei
NBHF – narrowband high frequency clicks
Guiana dolphin Sotalia guianensis
BBT – broad band transient clicks
Renan Paitach

4. 1. Open the file
This PowerPoint uses CPOD.EXE v it does the same things as earlier versions, but includes support for the C-POD-F and various minor changes.
View the PowerPoint 'Viewing and Reading POD data'
Open and process the file Ilha_Grande CPOD1372 file01.CP1 or, to save time, use the supplied CP3 file.
You then have a 'file set' consisting of the CP1 file and the corresponding CP3 file.

5. 2. Gain warnings (C-PODs, but not the C-POD-F)
When you 'Open a file set' on the Files page of the menu you will see 'Warning: registered gain is 161'. C-PODs (but not the C-POD-F) must be operated with their own SD card to set the gain of the main amplifier.
You can see the correct gain for this POD in the 'CPOD specific settings.txt' file. It is 152, which is actually very close (as the scale is not based on 0). Differences arise from using the wrong card, or recalibration of the POD.

6. 3. View the whole file: Species
Select these parameters to display:
Then click 'Sp class'.
The CP1 file will still show the Frequency, because it contains no information on Species class.
Click 'show whole time range' and then 'show from start'

7. 3.Species

8. 3. Species … and deployment
Deployment: Look at the record of angles at the top of each file panel. There is a gap when the POD lies flat and switches off - shown as red angle lines. This is useful information on the behaviour of the mooring.
Species: There are many detection of dolphins ('other cet') that are Guiana Dolphins (Sotalia guianensis) in this location, and of an NBHF species that is the Franciscana (Pontoporia blainvillei) in this location.
the 'unclassed' - grey - bars are higher when cetaceans are also identified, so they are likely to be due to cetacean sounds.
the black line for all clicks shows that most come from cetaceans, with a weaker continuous background.
Conclusion: Noise levels low, many cetaceans, some Broadband Transients (BBT), some NBHF.

9. 4. View the whole file: Q class
Switch to display Train Qclass, and include Lo quality trains.
again, click 'show whole time range' and then 'show from start'

10. 4. Q class
Look for periods when Lo or Mod Q trains are not associated with Hi Q trains.
Conclusion: Good correspondence between Q classes. No indication of serious noise issues.

11. Quality filters
You might find that Lo quality trains were also good in your data set, but it is unwise to use them in any project with other files in it. This is because various noise sources often generate Lo Q trains, and even though you haven’t seen that in the data you’ve looked at, it is a real risk, and would force you to re-analyse data that included them. Looking further at noise:

12. 5. Noise
Zooming in to 1min at the white arrow shows an isolated noise spike within a much lower background.
zooming further:

13. 5. Noise the background is typical of shrimp noise:
and the spike is a noise burst of uncertain origin:

14. 5. Noise
the noise burst appears to end abruptly, because the minute has 'maxed-out' at 4096 clicks.
Isolated noise bursts like this could be due to fish foraging in a sandy sea-bed nearby, or the rapid passage of a small boat.
Conclusion: no significant noise issues found so far.

15. 6. Species identification
With two species, one making broad-band transients and one making NBHF clicks, there is a risk of mis-identification in each direction, so this needs to be checked.
Use the settings shown. Click Sound pressure as this gives a clear picture of the amplitude envelope, with the colour coding showing frequency, so it's the best starting point.
Click Sp class to see what that is.
Click 'show from start' and 'show next screen' and you get:

16. 6. Species
evident train structure i.e. successively similar inter-click intervals.
amplitude envelope shows initial rise and final fall.
quiet background.
these are all positive train features so this is a real train, but it says nothing about species.
purple colour in CP3 shows that it was classed by KERNO as NBHF.

17. 6. Species click the Graphs button (which is then labelled kHz).
this shows a modal kHz of 129khz, typical of NBHF clicks.
in the CP1 file there are other frequencies that don't belong to the train.

18. 6. Species - hit F11 to show durations:
long clicks (for C-PODs the display typically jumps up for NBHF clicks and jumps down for BBTs, and this jump up/down property is helpfully quick)
some clicks are over 15 cycles long
which is unlike BBT clicks

19. 6. Species - hit Shift+F11 to show bandwidths:
Horizontal lines indicate minimum bandwidth.
Vertical lines get longer as the bandwidth rises.
Weak clicks tend to have higher bandwidth because of the background noise in the same time period.
The bandwidth here fits NBHF much better than BBT. So, species ID is good.
Next: Select Sound pressure (F10) and 'show next screen' (F1)

20. 6. Species - F10 for SPL, … F1 to move to next screen
the colours show these are trains with frequencies far below NBHF. Use F8 to toggle through train Q class and species. These are 'other cet' i.e. dolphin clicks, BBTs. Duration - F11 - shows they are mostly 5 cycles long.

21. 6. Species - Shift F11 for bandwidth
None of the clicks in the CP3 trains are minimum bandwidth:

22. 6. Species - Shift F10 for SPL - sound pressure level
Many typical dolphin trains follow. Check them out …

23. 7. Errors
To find false NBHF set this display - good quality NBHF only: and click 'show next screen' or F1 until you get to this:

24. 7. Errors
This is clearly not NB - narrowband as a wide range of frequencies area found in the train! … the algorithm found a train of high frequency components, and ignored the context.

25. 7. Errors
Zooming in to 1ms and viewing the frequencies - F9 - shows the big frequency spread in each multipath cluster as vertical lines of clicks:

26. 8. KERNO or GENENC?
There are other examples of that kind of error in this file. The GENENC classifier exists to reduce this kind of error, so let's have a quick overall look at what changes it makes. The 'Analysis' page does this for us, and if you have the excellent Windows 10 you can use the Snipping Tool to drop the results into a document or just view them against your next analysis. It's easiest to understand if we do one species class at a time:

27. 8. KERNO or GENENC? NBHF
This is KERNO, showing good quality NBHF trains.
File 2 is selected - that's the second panel up from the bottom of the display, and is normally the CP3 file.
The Clicks/sec graph is in two parts. For trains above 100clicks/s the Y scale has been reduced by x20 so that they can be seen.

28. 8. KERNO or GENENC? NBHF
This is GENENC: The number of detection positive minutes for NBHF trains has fallen by about 7% from the KERNO level. Examination of the file shows that the level of false positive NBHF trains is now very low. If dolphins were rare or absent the fall in NBHF detections would be less. The train quality classification has been used in GENENC so no Q class filter is active. Repeating this for dolphins:

29. 8. KERNO or GENENC? BBT
This is KERNO, good quality BBT / 'other cet' / dolphin trains.

30. 8. KERNO or GENENC? BBT
This is GENENC: The number of detection positive minutes for dolphins trains has increased by 119% from the KERNO level! This is because dolphin trains are much harder to classify individually, but in an encounter GENENC can see that the aggregated features are much more distinctive. These differences between KERNO and GENENC (which actually uses the KERNO output as an input) are typical.

31. 8. KERNO or GENENC? Conclusion
GENENC gives much better results on this data, by: Increasing dolphin sensitivity, without increasing false positives. Decreasing NBHF false positives coming from mis-classification of dolphin click trains. Examination of the GENENC 'Other cet?' and 'NBHF?' classifications shows that they do have significant false positive levels and should not be used here.

32. 9. Sonars
Set the display as here and > 'show from start' > 'show next screen'. This is supposed to be a sonar, but is very unimpressive - the clicks are short and there is no obvious regularity:

33. 9. Sonars
Click F7 to show 'Click rate' and a distinct horizontal line at just below 10clicks/second can be seen in the CP1 display, and matches the rate found for the first train in the CP3. That is a very distinctive and reliable indication of a boat sonar, but may only be visible as the sonar first becomes audible and as it fades because in between there are too many echoes.

34. 9. Sonars
Although the nominal upper limit of C-PODs is 160kHz, there is some sensitivity up to 200kHz and sonars at that frequency may be logged:

35. 9. Sonars
Sonars typically use more than one inter-pulse interval and this often shows up in the trains identified:

36. 9. Sonars
Sonars switch between different pulse rates, as 183s here. The pulse rate found by KERNO may be half or 1/3 or 1/4 of the true rate if pulses are missed and all these rate feature are seen here. However, for sonars as for cetaceans, the species classification may be good while the train intervals are not.

37. 9. Sonars
When 'ICI assessment quality' high only is selected, on the Files page of the menu, only those trains that are at the right rate are shown: Reviewing the whole file the sonars identified by KERNO are all correct.