Teil C - Annex 
65 
2008) in a marked contrast to the ambient noise. Only the calls recognised on basis of these 
peaks will be automatically saved as audio files, whereas disturbance noise caused by wind 
and waves is largely ignored. Compared to recording technology that is triggered only by level 
(cf. Frommolt et al. 2012), the filtering function of this system reduces the data volume col 
lected for evaluation to a manageable amount (Hill & Hüppop 2008). However, this is not a 
system for automatic call identification. Tests with different algorithms ultimately rendered no 
satisfactory results for the expected broad range of species. In about 10 years, for instance, 
FIN01 registered 112 species purely sound-based. Therefore, software for automatic identifi 
cation of species must not be used. 
The stored data should subsequently be listened to by qualified personnel for identification 
down to species level, if possible. To this end, a closed headphone system must be used, 
while at the same time, to simplify analysis, the spectrum of the audio files should be illus 
trated in sufficient resolution by means of suitable software 
(e. g. RAVEN http://www.birds.cornell.edu/brp/raven/ravenoverview.html). In times of strong 
bird call activity, several files per minute might be generated. Individual birds or flocks are 
saved on several consecutive files, if the intervals between detected calls are longer than 1.5 s 
or the maximum file size of 5 s has been reached (both this is pre-set in AROMA and should 
not be changed). Each call-positive file is documented as one data set together with the time 
of recording and the detected species. If more than one species can be identified per file, 
several data sets are created accordingly. Calls (or songs) of individuals or flocks that evidently 
happen to rest at the recording site - recognisable by the temporal accumulation of birds call 
ing obviously in always the same distance - are marked according to file. Even so, the data 
sets should be further identified with the remark “rest” and recorded. Quantification is con 
sciously abstained from, since it must be assumed that the system features a varying range 
depending on the weather, bird species and many other factors. General migration intensity is 
recorded by radar at night-time, yet only call recording provides the additional insight into the 
involved species, in spite of all the limitations of the procedure. Even without direct quantita 
tive reference, the analysis of data in the relative unit “call-positive files/h” provides sufficient 
information to identify certain call concentrations. The presentation of results should be by 
species per migration night and hour under due registration of duration of night. 
Regarding further discussion of the method, the data thus obtained and the evaluation op 
tions see Hüppop et al. (2012). They processed the calls with the help of only a few experts 
and expressed them quantitatively in estimated individuals/h. Since the number of future re 
visers and their qualitative comparability is not known, we abstain here from estimation of 
individuals/h (requiring a high degree of experience) in favour of the simpler unit “call-positive 
files/h”. To ensure quality, a randomised sample of 5% of call-positive files should be evalu 
ated by a second reviser. In the event of significant deviations, the revisers should receive cor 
responding training or be replaced. 
More information on the method and potential applications of the AROMA software are avail 
able upon request to: 
Institute for Avian Research “Vogelwarte Helgoland” 
Dr Ommo Hüppop 
An der Vogelwarte 21 
26386 Wilhelmshaven 
Germany 
e-mail: ommo.hueppop@ifv-vogelwarte.de