a) Cladogram of all species and genera detected by 12S rRNA metabarcoding of Whenua Hou soil samples. b) Relative taxon abundances of sampled locations averaged across replicates (from left to right: kākāpō display sites, feeding stations, abandoned nests, and Nestor parrot aviaries). Two different sites per location were sampled (top and bottom) at three different distances, and two aviaries of the Nestor species kea and kākā. For feeding station 2 (4m), both replicates resulted in dropouts.

Number of passed nanopore reads and bases (Q-score > 7), number of reads and bases mapping to the kākāpō reference genome, and relative amount [%] of mapped reads and bases per soil sample. For samples 3 and 35, the results of the selective and the non-selective nanopore sequencing runs are shown.

Resulting read length distribution (log10 scale) of nanopore sequencing of three exemplary soil samples (samples 3, 11 and 35; Supplementary Table 1). Left: Distribution of all passed (Q-score > 7) reads; right: Distribution of all passed (Q-score > 7) reads that map to the kākāpō reference genome using minimap2. The subset of mapped reads that have been accepted by selective sequencing (not ‘unblock’ reads; Methods) is highlighted in orange. The selective sequencing results are shown by a) and b) (Sample 3), c) and d) (Sample 11), e) and f) (Sample 35). The non-selective nanopore sequencing data is shown by g) and h) (Sample 3) and i) and j) (Sample 35). The selective runs result in many reads of ~ 500 bp length, which is the average sequencing length at which reads are long enough to be taken a decision upon and to be rejected.

(a-c) Distribution of haplotype agreement scores between all Whenua Hou kākāpō and a) soil sample 3 (Moss’ display site), b) soil sample 11 (Merv’s display site), and c) soil sample 35 (Nora’s feeding station). d) Mixing proportions [%; log10 scale] and e) posterior means of individual assignment per sample (y-axis) assessed through Bayesian inference of individual assignments (see Methods). The heatmaps show Sinbad’s omnipresent signal in the first column, the best hit when disregarding Sinbad in the second column, the second-best hit in the third column, and the mean values of all remaining Whenua Hou kākāpō in the last column.

a) A kākāpō (picture credit: Lydia Uddstrom). b) Map enhancement of sampling locations on Whenua Hou, New Zealand (service layer credit: Esri, Maxar, GeoEye, Earthstar, Geographics, CNES/Airbus DS, USDA, USGS, AeroGRID, IGN, and the GIS User Community).

Details of the three soil samples subjected to nanopore sequencing: DNA concentration after bead clean-up [ng/ul]; volume used as input for library preparation [ul] to achieve a DNA input amount of 1 ug per library preparation; amount of DNA in the final library [ng] used as input for sequencing; number of active pores per nanopore flow cell; and metadata of each sample.

Stamen terrain map of the sampling sites of the three soil samples 3 (Moss’ site), 11 (Merv’s site) and 35 (Nora’s site), and of the radio transmitter signal receiver that recorded Sinbad’s presence far away from his home range (‘Sinbad’s bowl’) and in the middle of the sampling sites, on February 24th 2019, three days before our sampling efforts. A terrain map with blurred contour lines is used to not disclose the exact home ranges of the individuals of this critically endangered species.