Sample identification

 In contrast to the above, sometimes sufficient information on the potential origin of the samples in question may already be available, and it is therefore just necessary to assign the samples to the correct subspecies from a few options. This is a very different issue, and a far less demanding task. It relates to fairly common situations: for example, somebody simply wants to know which subspecies, ecotype or possibly even breeder-line a given set of bee samples belongs to. Mostly, the question can even be asked more narrowly, as only a restricted set of possibilities needs to be considered, e.g., the allocation of a given sample to two, three or four subspecies.

Accordingly, few characters would already suffice to distinguish these possibilities, thus considerably reducing the labour involved. In specific cases, as in the replacement of A. m. mellifera by A. m. carnica in Germany during the past century, diagnosing a sample into one of these two subspecies could be based on hair length and cubital index alone, and for current characterization of A. m. carnica the addition of two further characters, tergite colour and tomentum width, is sufficient. Similar simple methods have been applied in other countries, e.g. the three characters width of tergite IV, length of proboscis, and cubital index are used in Poland, and a similar set of length of proboscis, and cubital index together with wing length and width is applied in Slovakia (Bouga et al., 2011). In particular, the derived characters, such as composed measures and indices which are useless in statistical evaluations, can be very useful to construct simple character sets for simplified distinctions which can be applied by bee breeders. In Switzerland, distinctions between A.m. mellifera and A. m. carnica are predominantly based on the cubital and dumb-cell indices.

Because the requirements for subspecies identification are predominantly designed by local demands (see Bouga et al., 2011), together with the need for economizing the effort of character measuring and the lack of a complete catalogue of honey bee morphometric variation in Europe, there is no accredited identification system for European bees with general validity. Instead, several independent national systems exist, each considering a limited numbers of alternatives. In some countries (Italy, Greece) formal accreditations are given to laboratories for relevant identification (Bouga et al., 2011).

However, there is no need to economize on the number of wing characters or the precision of shape analysis, once our suggestion has been followed and the standardized and generally accepted method has shifted to storing the point coordinates of the wing venation junctions. Presumably, automated geometric wing analysis may then prove to be the most efficient identification method in the future. However, it is not yet clear how reliably geometric wing morphometry alone will discriminate all European subspecies and known ecotypes from each other. Additional morphometric characters from other body parts may need to be included in such an envisioned all-European system still to be worked out.

The necessity to recognize Africanized bees in the New World may be regarded as a special case of sample identification, for instance when investigating if a swarm originates from an Africanized colony, or for certifying a potential breeder colony as free from African genes. In principle, the same criteria as above apply to this situation, but to facilitate achieving a fast result, specialized applications have been developed for this specific use, based on only few morphometric characters and the size differences between honey bees of African and European origin (e.g. FABIS, Rinderer et al., 1986, 1987). However, as the purpose of the method is focused on discriminating “Africanized” from “European”, the resolution is rather low and it is not suitable for further discrimination between subspecies. The USDA-ARS (http://www.ars.usda.gov/Research/docs.htm?docid=11053) uses FABIS for a preliminary test, followed by a full morphometric analysis, if possible Africanization is detected.

The most relevant and most powerful method to establish an allocation of "unknown" samples to a number of reference groups is discriminant analysis, based on the full set of characters. In a first step of this process, a large set of reference samples would be included, confirming their correct reallocation to their own groups. By entering the measured characters stepwise, the character subset relevant to separate the groups in question can be determined, and for further identification only these characters need to be considered. In the next step, the unknown samples would be entered into the analysis and allocated to the reference groups, including the probability of these allocations. It needs to be stressed again that this method is unsuitable to detect any new variation or ecotypes, as it forces a choice between a limited and predefined number of possible group allocations. Non-fitting types may not be allocated with high probability to any of these, but hybrids could not be told apart from genuinely new types.