EDP Foz Tua: Bat Habitats - Ecological & Compensatory Measures Monitoring Programs [2011-2017]

Study Extent

The data were collected within a 5-km radius buffer around the lower valley of River Tua (Northeast Portugal), up to 45 km from the mouth, in the municipalities of Alijó, Carrazeda de Ansiães, Vila Flor, Mirandela and Murça.

Sampling

Sampling was designed to cover all suitable foraging habitats for bats in the study area. A stratified sampling procedure was used, with the number of samples allocated to each habitat in proportion to the area occupied and/or its significance for bats. A total of 57 sampling points was set in natural woodlands (10), forest plantations (mostly Eucalyptus and Pinus), (9) river margins (10), shrublands (9), orchards/olive orchards (6), vineyards (6) and dryland crops (7). Each point was visited 2-4 times per year in 2011-2014, and monthly (March - October) from 2014 to 2017. Acoustic sampling was made within ± 3 hours of sunset, beginning 30 minutes after it. Each point was sampled for 10 consecutive minutes. The detection of the specimens in flight was made with a bat detector (Pettersson Elektronik, D240X model), with a frequency range of 10-120 kHz. The detector was connected to an automatic digital recorder (model Sony ICD-PX312) which saved the recordings in a “Wave” format, at a frequency of 44 kHz and at an expanded mode of 10x, making them audible to the human ears while, at the same time, preserving the original acoustic characteristics. From 2011 to 2013 the analyses of bat sound records used the manual/traditional methodology based on the the software “Batsound Pro – Sound Analysis”, from Pettersson Elektronik. This software generates graphics (oscillograms, sonograms and power spectrums) that allow the identification of the detected species by comparing them with those available in reference bat sound libraries. The acoustic variables usually employed to manually identify the sounds are (1) the structure of the pulse, (2) the frequency of maximum energy (kHz), (3) the range of frequencies (kHz), (4) the pulse duration (ms), (5) the interval between pulses (ms), and (6) the pulse repetition rate (Hz). From 2014 onwards, a more automated and expeditious method based in a model developed by Silva et al. (2013) was employed for species identification. A semi-automatic program provided 19 diagnostic parameters of the pulses (16 spectral and 3 temporal parameters). Subsequently, a probabilistic model (Artificial Neural Networks, ANN) was run to compare these parameters with those of a reference database of more than 16000 individual pulses from 1400 individuals of 9 genera recorded in Continental Portugal. The ANN attributed the pulses to a species or group of species (phonic group or “species complex”), with an associated probability of certainty. The results of the classification model showed an average sensitivity of 95% and an average error of 4%. Only the classifications with a probability greater than 70% were considered correct. When the above threshold was not reached by the model, the records were analyzed by the manual/traditional method described above. The exact identification of the bat species by sound analysis is not always possible, due to problems in detectability and/or identifiability. In the context of this work, whenever it was not possible to perform the identification until the species, the binomials, multinomials or genera to which the identification was possible were considered to be present. Thus, the presence of all species constituting these phonic groups (or “species complexes”) was considered possible.

Quality Control

All identifications were performed by qualified technicians

Method steps

1. To be completed

The taxonomic coverage of this dataset spans 1 phylum, 1 class, 1 order, 4 families, 10 genus and 15 species positively identified. Additionally, a set of 9 potential species may be present in the 23 “species complexes” identified.

1. Chordata
   rank: phylum
2. Mammalia
   rank: class
3. Chiroptera
   rank: order
4. Vespertilionidae
   rank: family
5. Miniopteridae
   rank: family
6. Rhinolophidae
   rank: family
7. Molossidae
   rank: family

The data were collected within a 5-km radius buffer around the lower valley of River Tua (Northeast Portugal), up to 45 km from the mouth, in the municipalities of Alijó, Carrazeda de Ansiães, Vila Flor, Mirandela and Murça.