Geese

Data previously gathered comparing with acceleration data for animals during activity on a treadmill at Buenos Aires Zoo [12] were reanalyzed to supplement the work on humans. Species used were; coypu (Myocastor coypus) (4 individuals), larger hairy armadillo (Chaetophractus villosus) (1 individual), Muscovy duck (Cairina moschata) (1 individual), greylag goose (Anser anser) (2 individuals), Magellanic penguin (Spheniscus magellanicus) (2 individuals) and rockhopper penguin (Eudyptes chrysocome) (1 individual). Briefly, animals were equipped with acceleration data loggers, attached variously, before being exposed to a treadmill with the tread moving at a range of speeds between 0 and 2.52 km/h, the upper limit dependent on their capacities. The animals were given rests between the higher speeds where the predominant behavior was locomotion however at the lower speeds the animal typically exhibited a range of behaviors including searching, scratching and lying. An open circuitry respirometry system was used to measure . Full details of the protocol are given in Halsey et al. [12] .

The procedure builds on the method developed by Wint and Robinson [16] for disaggregating livestock statistics, based on environmental and other spatial data. Spatially stratified, statistical regression models are developed using data from a series of sample points within each training data polygon and these models are then applied to the entire one-kilometre resolution set of predictor variables in order to estimate livestock densities, disaggregated over a defined study area. For GLW 2, this basic methodology has been revised and improved in a number of ways and was first assessed by Van Boeckel et al. [19] (link) and Prosser et al. [18] (link) in a detailed analysis of its performances for modelling domestic ducks in Asia and poultry in China respectively. Van Boeckel et al. (2011) looked at how downscaling performance was influenced by the aggregation level of input domestic duck data in Thailand (no data for the country, only one value for the country, administrative level 1 data, and administrative level 2 data) and comparing the predictions to actual admin level 3 data. The result was that downscaling based on the method outlined below was giving relatively good results provided that the training data were available at administrative level 1, and were degraded with coarser (national-level data) input data. In a separate study, Prosser et al. (2011) compared land-use based downscaling with the GLW 2 methodology to predict chicken ducks and geese in China, and found land-use based methods to give lower performances. In another previous study, though based on a non-stratified implementation of regression methods, Newmann et al. (2009) found comparable results between land-use and regression based downscaling [17] . In human population mapping, the AfriPop and AsiaPop databases are still largely based on land-use weighting of human population density per land use class [35] (link)–[36] (link), whereas new developments of the WorldPop consortium (www.worldpop.org.uk) are moving toward the use of machine learning methods such as random forest or boosted regression trees (Andy Tatem, comm. pers).
The stratified regressions are repeated a specified number of times using random selections of pixels from which to extract dependent and independent variables (bootstraps). This produces multiple models from which the variability as well as the mean values of the model predictions can be calculated for each pixel.
Because of the sheer volume of input and covariate data, the modelling must be broken down into continental tiles. Whilst bespoke geographic tiles can be created for specific tasks or projects, there are six continental tiles that are processed independently, and the global dataset is updated every time a new continental tile is processed (see file S1).

To test for universality of primers and cycling conditions, we performed parallel experiments in three different laboratories (Berkeley, Cologne, Konstanz) using the same primers but different biochemical products and thermocyclers, and slightly different protocols.
The selected primers for 16S [30 ] amplify a fragment of ca. 550 bp (in amphibians) that has been used in many phylogenetic and phylogeographic studies in this and other vertebrate classes: 16SA-L, 5' - CGC CTG TTT ATC AAA AAC AT - 3'; 16SB-H, 5' - CCG GTC TGA ACT CAG ATC ACG T - 3'.
For COI we tested (1) three primers designed for birds [7 (link)] that amplify a 749 bp region near the 5'-terminus of this gene: BirdF1, 5' - TTC TCC AAC CAC AAA GAC ATT GGC AC - 3', BirdR1, 5' - ACG TGG GAG ATA ATT CCA AAT CCT G - 3', and BirdR2, 5' - ACT ACA TGT GAG ATG ATT CCG AAT CCA G - 3'; and (2) one pair of primers designed for arthropods [2 (link)] that amplify a 658 bp fragment in the same region: LCO1490, 5' - GGT CAA CAA ATC ATA AAG ATA TTG G - 3', and HCO2198, 5'-TAA ACT TCA GGG TGA CCA AAA AAT CA-3'. Sequences of additional primers for COI that had performed well in mammals and fishes were kindly made available by P. D. N. Hebert (personal communication in 2004) and these primers yielded similar results (not shown).
The optimal annealing temperatures for the COI primers were determined using a gradient thermocycler and were found to be 49–50°C; the 16S annealing temperature was 55°C. Successfully amplified fragments were sequenced using various automated sequencers and deposited in Genbank. Accession numbers for the complete data set of adult mantellid sequences used for the assessment of intra- and interspecific divergences (e.g. in Fig. 5) are AY847959–AY848683. Accession numbers of the obtained COI sequences are AY883978–AY883995.
Nucleotide variability was scored using the software DNAsp [31 (link)] at COI and 16S priming sites of the following complete mitochondrial genomes of nine amphibians and 59 other vertebrates: Cephalochordata: AF098298, Branchiostoma. Myxiniformes: AJ404477, Myxine. Petromyzontiformes: U11880, Petromyzon. Chondrichthyes: AJ310140, Chimaera; AF106038, Raja; Y16067, Scyliorhinus; Y18134, Squalus. Actinopterygii: AY442347, Amia; AB038556, Anguilla; AB034824, Coregonus; M91245, Crossostoma; AP002944, Gasterosteus; AB047553, Plecoglossus; U62532, Polypterus; U12143, Salmo. Dipnoi: L42813, Protopterus. Coelacanthiformes: U82228, Latimeria. Amphibia, Gymnophiona: AF154051, Typhlonectes. Amphibia, Urodela: AJ584639, Ambystoma; AJ492192, Andrias; AF154053, Mertensiella; AJ419960, Ranodon. Amphibia, Anura: AB127977, Buergeria; NC_005794, Bufo; AY158705; Fejervarya; AB043889, Rana; M10217, Xenopus. Testudines: AF069423, NC_000886, Chelonia; Chrysemys; AF366350, Dogania; AY687385, Pelodiscus; AF039066, Pelomedusa. Squamata: NC_005958, Abronia; AB079613, Cordylus; AB008539, Dinodon; AJ278511, Iguana; AB079597, Leptotyphlops; AB079242, Sceloporus; AB080274, Shinisaurus. Crocodilia: AJ404872, Caiman. Aves: AF363031, Anser; AY074885, Arenaria; AF090337, Aythya; AF380305, Buteo; AB026818, Ciconia; AF362763, Eudyptula; AF090338, Falco; AY235571, Gallus; AY074886, Haematopus; AF090339, Rhea; Y12025, Struthio. Mammalia: X83427, Ornithorhynchus; Y10524, Macropus; AJ304826, Vombatus; AF061340, Artibeus; U96639, Canis; AJ222767, Cavia ; AY075116, Dugong; AB099484, Echinops; Y19184, Lama; AJ224821, Loxodonta; AB042432, Mus; AJ001562, Myoxus; AJ001588, Oryctolagus; AF321050, Pteropus; AB061527, Sorex; AF348159, Tarsius; AF217811, Tupaia; AF303111, Ursus (for species names, see Genbank under the respective accession numbers).
16S sequences of a large sample of Madagascan frogs were used to build a database in Bioedit [32 ]. Tadpole sequences were compared with this database using local BLAST searches [33 (link)] as implemented in Bioedit.
The performance of COI and 16S in assigning taxa to inclusive major clades was tested based on gene fragments homologous to those amplified by the primers used herein (see above), extracted from the complete mitochondrial sequences of 68 vertebrate taxa. Sequences were aligned in Sequence Navigator (Applied Biosystems) by a Clustal algorithm with a gap penalty of 50, a gap extend penalty of 10 and a setting of the ktup parameter at 2. PAUP* [34 ] was used with the neighbor-joining algorithm and LogDet distances and excluding pairwise comparisons for gapped sites. We chose these simple phenetic methods instead of maximum likelihood or maximum parsimony approaches because they are computationally more demanding and because the aim of DNA barcoding is a robust and fast identification of taxa rather than an accurate determination of their phylogenetic relationships.