A probabilistic algorithm to process geolocation data

Table 3 Algorithm parameters used to compute locations for both assessment data sets

Model parameter	Description	Value used
particle.number	number of particles computed for each point cloud	10 000
iteration.number	number of track iterations	200
sunrise.sd & sunset.sd	shape, scale and delay values describing the assumed uncertainty structure for each twilight event following a log normal distribution	2.49/ 0.94/ 0^a
range.solar	range of solar angles used	-7° to -1°
boundary.box	the range of longitudes and latitudes likely to be used by tracked individuals	120 W to 40 E 90 S to 0
day.around.spring.equinox & days.around.fall.equinox	number of days before and after an equinox event in which a random latitude will be assigned	includes the entire wandering albatross tracking period
speed.dry	fastest most likely speed, speed standard deviation (sd) and maximum speed allowed when the logger is not submerged in sea water	12/ 6/ 45 m/s for black-browed albatross^b & 12/ 7/ 70 m/s for wandering albatross^b
speed.wet	fastest most likely speed, speed sd and maximum speed allowed when the logger is submerged in sea water	1/ 1.3/ 5 m/s^c
sst.sd	logger-derived sea surface temperature (SST) sd	0.5 °C^d
max.sst.diff	maximum tolerance in SST variation	3 °C
east.west.comp	compute longitudinal movement compensation for each set of twilight event [37]	used

^a The resulting uncertainty structure for both twilight events is illustrated in Additional file 1. These parameters are chosen as they resemble the twilight error structure of open habitat species in [20]
^binferred from GPS tracks (see Additional file 3 for details)
^c Antarctic circumpolar current speed up to fast current speeds (i.e. Malvinas current) [38] as the tagged animal is assumed to not actively move when the logger is immerged in seawater
^d logger temperature accuracy