DELETE - Bird Logger Interpretation
The theory of light geolocation can be found in:
Hill, R.D. (1994): Theory of geolocation by light levels. Elephant Seals, Population, Ecology, Behaviour and
Physiology, ed. by B.J. Le Boeuf and R.M. Laws. Berkeley, Univ. of California Press, 227–236.
The BAS geolocator logger records light level in a compressed format. The decompression software we supply converts this compressed format into time stamped light values stored in a text file readable by any statistics or spreadsheet package. With the present algorithm, there is a light value for each 10 minute interval; the light value recorded is the maximum recorded within those ten minutes (actual light sampling occurs every minute). We believe this to be a sensible compromise between available memory and temporal resolution; inaccuracies due to the other factors mentioned previously (cloud, orientation, shading etc) outway this sampling error. Our actual light reading is only 6-bit, again, because we feel that more than this is unnecessary considering the other uncertainties; higher resolution would take up more memory space.
To convert the light data into usable location coordinates, the data must be processed. This has been done, sometimes by hand, using the Astronomical Almanac but software is available to do this for you. A program which will help you can be downloaded from the main geolocator page. Scientists at BAS use a package called MultiTrace. To obtain the most reliable results, each dawn and dusk can be visually analysed to obtain a confidence level with clearly bad readings being discarded. A confidence scale can be devised and a value attached to each location fix calculated. The example below shows quite a lot of intermittent shading by the bird in the hours before dusk (perhaps due to bobbing above and below the horizon on the water), but the dusk and dawn can clearly be seen. This example is actually 1 minute data; 10 minute logger data would show far less shading by the bird as it is the maximum within that 10 minute interval that is recorded.
It is possible to put the light level to location fix algorithm into the logger itself but, we believe that the inability to check on the original raw data is a disadvantage in determining confidence levels of the location fixes.
Many years ago, an albatross researcher named Nic Huin (now associated with Falkland Islands Conservation) put a geolocator on top of his caravan home in the Falklands as a test. The caravan was stationery for a year and, after the data was downloaded, Nic laboriously converted the light values into location fixes. The graphs below show the results with the Y-axis showing error in degrees of latitude and longitude respectively; note the different Y-axis scales.
A number of important features of light geolocation can be seen in these results. Looking at the Y-axis scales, it can be seen that the error associated with latitude is higher than with longitude. This is because the day length has less of a dependence on latitude than local noon does on longitude. The calculation for latitude is also a lot more complex. The result is that light geolocation usually gives more accurate results if studying the East-West coordinate than the North-South coordinate.
The equinoxes can clearly be seen on the latitude graph (near September and March 21st). The two or three weeks around the equinox produce very large uncertainties in latitude as, at equinox, the day length is the same wherever in the world.
The effect of weather can also be seen on the latitude graph. During the Falklands Winter there is a lot more cloud and mixed weather than in the Summer. This is probably why there is a bit more uncertainty in the latitude results of June and July (Winter) than in December and January (Summer).
The last feature to point out is the trend-line on the longitude graph. While latitude depends on day length, longitude depends on local noon with respect to a known clock. If the clock is not exact and drifts slightly over time, the longitude coordinate will also drift accordingly. The trend-line is a result of the clock drift inside the logger. If there were no clock drift, this line would be horizontal.
Clock drift depends on crystal manufacturing inaccuracies and, also, temperature. The study above was done at a location which had approx 30'C temperature range over the year. The straight line trend shows that most drift is due to manufacturing inaccuracies and not too much on temperature change. As can be seen, the longitude graph shows that it would not be so bad simply to average the drift over the whole deployment.
If the logger was subject to greater extremes of temperature, the trend-line would bend one way in the Summer and the other in the Winter as the effect of temperature has a greater effect on drift. For long periods of deployment it is very important to adjust the data for clock drift. This is why writing down the exact time of the start of logging is important, as the logger does not save this value internally.
An interesting astronomical feature can be clearly seen in the longitude graph as repeating curved lines. This is due to a phenomenon called analemma and will not be discussed here. With full astronomical calculations, this feature will not be seen but, because he did a lot of this by hand, Nic used simplified equations.
Thanks to Nic Huin for giving us permission to use this data.