484 Zenbakia 2009-05-01 / 2009-05-08


Location accuracy of emissions gathered via satellite

TOBAR-ARBULU, Joseba Felix

This work is part of a series of different papers done by a team composed by Ibon Telletxea, Mikel Arrazola, Zarbo Ibarrola, Raúl Migueliz, Joakin Anso, Izaskun Ajuriagerra, Ruben Ibáñez, Roberto Gogeaskoetxea, Felipe Diez and Joseba Felix Tobar-Arbulu.

We have been tracking woodcocks via satellite with PTTs developed by Microwave Telemetry Inc (MTI) since 2006.

According to Argos:1

1. “How accurate is the error estimation? Our experiments with different sets of transmitters at fixed positions or moving slowly have shown proper matching with the error estimations. Yet, in a few cases users have reported significant discrepancies. The error estimation process is not totally independent of the transmitter frequency stability, or of the platform motion which also translates into a frequency shift. It assumes that the frequency is “approximately” stable during the satellite pass. As a consequence, oscillator instability or a fast moving platform may lead to underestimation of the error.”

2. “How accurate are class A & B locations (empirical data on class A & B)? Class A & B location may be accurate. We cannot specify the accuracy since more messages are needed to estimate the error. We can just say that class A locations are usually more accurate than class B because the transmit frequency has been computed thus enhancing the process. Some users have experimentally tried to answer this question and provided interesting hints. We are confident that approaches of this type will help provide interesting guidelines to enhance practical use of this type of locations.”

As to the accuracy ofclass A and B locations, here what can be read in Hays et al. (2001):

According to C. Vincent et al. (2002):

“Both filtered and unfiltered LC2 A locations were of a similar accuracy to LC 1 locations, and considerably better than LC 0 locations.”

Michael Kaatz (2004) has taken into account the real positions following the storks and the positions given by Argos. When dealing with the accuracy of the emissions, Kaatz underlines the following:

Using the 68th percentile3 to define the accuracy of locations estimates, observed accuracy was 3.19 km for Location Class (LC) 1.

Let’s take a look at our experiment in 20074:

3. Methodology

Here the accuracy of LC 2 and LC 1 locations given by Argos:

Taking as good the LC 2 and LC 15 locations given by our bird (Navarre), we have calculated the discrepancy of LC B locations and LC A locations from the good ones (LC 2 and LC 1) in the following way:

Assuming as in March 5th that we have a LC 2 location followed by a LC B location, in a very short period of time when the bird is supposed not to move, the accuracy of LC B location is, more or less:

0.350 km + distance between LC 2 and LC B (measured through Google Earth):

0.350 + 0.6 = 0.950 km

Studying all the data given by our birds, we have found that in the worst cases taking into account locations in very short periods of time, where it is supposed that the bird doesn’t move:

(a) The accuracy of the LC A and LC B locations depend on the pair PTT/bird.

(b) For Navarre, ID 73388, the LC A locations, in the studied cases, haveanerror between 1.77 and 5.65 Km6.

(c) For Navarre, ID 73388, the LC B locations, in the studies cases, haveanerror between 0.95 and 10.67 Km7.

We have very few samplesof data from the Asturianbird to compare LCA and LC B locations; even worse, we do not have any emission with LC 2 or LC 18 location to measure the discrepancies. We can only say the following:

(d) For the Asturian woodcock, ID 72452,the distance between LC A locations has an error between0.42and 3.66 Km.

(e) For the Asturian woodcock, ID 72452, the distance between locations Bs has an error between3.12 and 4.36 Km.

In this case, we needed at least a LC 3, a LC 2 or a LC 19 location to measure the error of these locations.

All in all, one can say that, whatever the real errors of LC A locations and LC B locations for this bird, the LC A locations do not have a big difference among their errors (0.42-3.66), neither LC B locations among theirs (3.12-4-36).

(f) Using the same methodology, our LC 0 locations for Navarre can be considered as pretty goods... although in our experiment one LC B location and some LC A locations had more accuracy than LC 0 locations, in line with the works of Hays et al. and Vincent et al., quoted before.

Now, let’s see this year’s (2008) data:

(g) This year Navarre, ID 73388, has given very few good positions to make such kind of calculus.

We have only a datum, on May 16th, when we have a LC 2 location (at 3:27 UTC) followed by an LC A location (at 3:22). Five minutes between both locations.

If we apply the above mentioned methodology (as an exercise, since, the time between both locations is too long):

Distance (through Google Earth) between LC 2 and LC A locations: 4.5 km.

Accuracy of LC 2 location according to Argos: between 150 and 350 meters.

So, in the worst of all cases, accuracy of that LC A location: 4.5 + 0.35 = 4.85 km.

But, five minutes is too long. So, let’s try a new methodology.

4. New methodology

Velocity of fly10:

When Araba was flying at nigh, above the sea, we calculated his velocity: 66km/h, in tune with the literature.

Assuming a velocity of 60 km/h during the day, high enough, we can perform the following calculus.

1m/s = 3.6 Km/h, so 60 km/h = 16 m/s, again enough high, but let’s take that very extreme hypothesis as good.

(h) Let’s use this new methodology with Navarre:

As mentioned above, in the worst of all cases, accuracy of Navarre’s LC A location, on May 16th: 4.5 + 0.35 = 4.85 km.

Now we have to add the possible space traveled in those five (5) minutes (a very strong hypothesis).

5 x 60 x 16 ---> 4,800 meters ---> 4.8 km.

So the accuracy of the LC A location given by Navarre is 4.8 + 4.85 = 9.65 Km.

(Again, the hypothesis is really strong.)

(i) With Laguna (ID 83297) and Araba (ID 83300) we don’t have the possibility to know the accuracy of LC A and LC B locations, since the locations given in a single emission are not very near in time.

Assuming, as in March 16th, that we have, for Araba (ID 83300), a LC 111 location followed by a LC A location, in a period of time of 4.5 minutes, the accuracy of LC A location is, more or less:

1.0 km + distance between LC 1 and LC A locations (measured through Google Earth) + distance traveled by the bird (at 16 m/s) in those 4,5 minutes:

1.0+ 0.5 + 4.32 = 5.82 Km.

We don’t have any more data to apply this new methodology.


The point to be underlined is that for our birds these LC B and LC A locations are very important: furthermore, they are real, objective and witha littlework one canget their accuracy.


Hays, G. C., S. Åkesson, B. J. Godley, P. Luschi and P. Santidrian (2001) The implications of location accuracy for the interpretation of satellite-tacking data, Animal Behaviour 61; 1035-1040.

Kaatz, M. (2004) Mit Prinzesschen unterwegs, 25 Jahre Storchenhof Loburg, 25 Jahre NABU-Bundesarbeitsgruppe Weißstorchschutz. Thesis’ title: Der Zug des Weißstorchs Ciconia ciconia auf der europäischen Ostroute über den Nahen Osten nach Afrika.

Tobar-Arbulu, J.F. et al. (2009) Scolopax Rusticola without frontiers, paper presented to the MTI 2009 Conference, Maryland.

Vincent, C., B. J. McConnel, M. A. Fedak and V. Ridoux (2002) Assessment of Argos location accuracy from satellite tags deployed on captive grey seals, Marine Mammal Science 18:301–322 (see).

1 For points 1 and 2.

2 LC: location class.

3 With more percentile the accuracy is even less. According to Kaatz (2004), the accuracy depends on the radiated power and on oscillators: lower oscillators minimize the frequency drift.

4In 2006, with Trasgu (ID 62540), we had no possibility to make this kind of calculus. The periods of time between the different emissions given in the same day were not short at all. (About the first project, see http://www.euskonews.com/0380zbk/gaia38002eu.html and http://www.euskonews.com/0383zbk/gaia38302eu.html in Basque; and http://www.euskonews.com/0380zbk/gaia38002es.html and http://www.euskonews.com/0383zbk/gaia38302es.html in Spanish. About the second project, see http://www.euskonews.com/0431zbk/gaia43102eu.html and http://www.euskonews.com/0433zbk/gaia43302eu.html in Basque; and http://www.euskonews.com/0431zbk/gaia43102es.html and http://www.euskonews.com/0433zbk/gaia43302es.html in Spanish.)

5One should take into account what Kaatz says in his thesis about class 1 locations. For the time being, let’s assume that LC 1 locations have the accuracy that Argos says.

6 Taking into account Kaatz’s thesis, when the calculus is done after LC 1 locations, one should add, in the worst cases, 3.19 km to the related distances.

7 See note 5.

8 When dealing with LC 1 locations, we should take into account Kaatz’s thesis.

9 See note 8.

10 See Scolopax Rusticola without frontiers: New technolgy, paper presented to the MTI 2009 Conference, Maryland. See http://www.microwavetelemetry.com/conf09/index.php and then click on the title:“Scolopax rusticola without frontiers: New technology”.

11 See what is said in note 8. (We would like to thank Dr. Michael Kaatz for the sending of his Ph.D. dissertation.)