Table 1 Previous analyses of long-term datasets that have used automated algorithms to detect calls and report spatial distribution and/or temporal occupancy of Antarctic blue whales.

From: An open access dataset for developing automated detectors of Antarctic baleen whale sounds and performance evaluation of two commonly used detectors

Study

Detection method (software used)

Noise pre-processing

True positive rate

False positive rate

False positive removal

Characterisation/validation summary

Širović et al. 200417

Spectrogram correlation (Ishmael)

Energy sum (Ishmael)

Not reported

Not reported

 < 1%

No

Threshold was iteratively adjusted until false positive rate was < 1%. Calls on days with fewer than 50 detections were inspected

Širović et al. 200918

Spectrogram correlation (Ishmael)

Not reported

Not reported

Not reported

All

Visual inspection of all detections to remove false positives

Samaran et al. 201330

Spectrogram correlation (XBAT)

Not reported

Not reported

6%a

Some

Months with fewer than 50 detections: visual inspection of detections to remove all false positives. Otherwise 10% of randomly selected detections inspected

Tripovich et al. 201520

Energy detection (Ishmael)

Not reported

Not reported

14.6%

All

Visual inspection of all detections to remove false positives

Thomisch et al. 201619

Spectrogram correlation (custom developed)

Not reported

Not reported

Nominally < 1%

No

False detection rates and thresholds determined via detection function quantiles from 100

randomly selected detectionsb

Leroy et al. 201626

Subspace projection detection36

Noise-adaptive threshold

Not reportedc

Nominally < 3%

Some

Detections deemed false if the frequency at maximum amplitude was different than that of unit-A

Balcazar et al. 201745

Energy sum (Ishmael)

Not reported

93.3–97.3%

14.6–98.9%

All

Comparison against expert human observer who annotated 1 randomly selected day each month for each site. Visual inspection of all detections to remove false positives

Buchan et al. 201742

Spectrogram correlation (Ishmael)

Not reported

99.998%d

Not reported

All

20% subset of days with no automated detections visually inspected to determine false negative rate. Visual inspection of all detections to remove false positives

Shabangu et al. 201716

Spectrogram correlation (XBAT)

Not reported

42–83%

Not reported

All

Visual inspection of entire dataset (1518 h) to assess remove false positives and include missed detections

  1. aFalse positive rate reported only for months when there were more than 500 calls detected.
  2. bFalse positive detections were from a different detector operating in an adjacent frequency band with a similar, frequency-adjusted, spectrogram correlation kernel.
  3. cTrue positive rates for this detector for high, medium, and low signal to noise ratio (SNR) calls and a variety interfering noises reported by Socheleau et al. 2015, but the prevalence of these conditions within the full dataset is not indicated.
  4. dFalse negative rate reported as a percentage of total uncorrected detections for a 20% subset of days without automated detections.
Back to article page