Diet of moulting Swainson's Thrushes (Catharus ustulatus) and Tennessee Warblers (Leiothlypis peregrina) at a stopover site during fall migration measured with fecal DNA metabarcoding

Moult and migration are energetically demanding and require adequate nutrition. In some species, individuals may interrupt their fall migration to moult at discrete stopover locations outside of their breeding grounds (i.e., moult-migration) leading to competing nutritional demands for moult and migration. Here, we use DNA barcoding of fecal samples to compare the diet of moulting and actively migrating (post-moult) Swainson’s Thrushes (Catharus ustulatus) and Tennessee Warblers (Leiothlypis peregrina) during their fall migration stopover at a large urban greenspace in Montreal, Canada. Diet differed according to moult status, species, and seasonality. Swainson’s Thrushes had a broad diet with frequent detections of both insects and berry-producing shrubs; while detections in Tennessee Warblers’ diets were mainly arthropods. For both species, more actively migrating individuals consumed fleshy-fruiting plants than moulting individuals. A higher proportion of moulting birds consumed arthropods compared to active migrants, due to either arthropod availability or a dietary preference for proteinaceous foods to grow feathers. Both species and moult classes consumed more native plants than non-native plants later in the season. We show the importance of managing urban greenspaces with native plants and diverse food sources that can provide for the different dietary needs of migratory birds.


Arthropod detection
As DNA from each sample was amplified two times, there were a total of 112 replicates analyzed (e.g.1_Rep1; 1_Rep2).The S5 run resulted in over 3M raw reads (Figure 1).A total of 182,603 reads could not be assigned to a known IonCode molecular index and were excluded from all subsequent analyses.The remaining 3,488,127 reads were mapped to one of the wells with either sample DNA template or one of negative controls.The vast majority of the reads were detected in wells with sample DNA (3,264,001), while only 224,126 reads were scored from template-free wells.Arthropod detections were found in 84 replicates out of 112, or 45 out of 56 samples for replicate 1 and 39 out of 56 samples for replicate 2 (Appendix I).Detected arthropod items represented approximately 195 species belonging to 76 families and 13 orders.The most commonly detected orders were Lepidoptera, Diptera, and Araneae (Figure 3 and 4).Each detection was supported by an average of 1340 sequence reads.

Plant detection
The S5 run resulted in over 5M raw read (Figure 2).A total of 440,404 reads could not be assigned to a known IonCode molecular index and were excluded from all subsequent analyses.The remaining 4,577,170 reads were mapped to one of the wells with either sample DNA template or one of negative controls.The vast majority of the reads were detected in wells with sample DNA (4,107,793), while 469,377 reads were scored from template-free wells.Plant detections were found in 75 replicates out of 112, or 37 out of 56 samples for replicate 1 and 38 out of 56 samples for replicate 2 (Appendix II).Approximately 56% of the plant sequences recovered from all replicates showed no significant similarity to any of the publicly available records on BOLD and were thus assigned an identity of "unknown".Detected plant items (i.e.those that could be identified) represented approximately 62 species belonging to 28 families and 18 orders.The most commonly detected orders were Rosales, Ericales, and Fagales (Figure 3 and 4).Each detection was supported by an average of 1443 sequence reads.

INTERPRETATION
While many identities are listed to species (Appendix I and II), we recommend only considering identifications reliable to the level of genus due to the short length of the molecular markers used.We do not recommend using read counts as a measure for prey abundance, but rather as a guide for the reliability of detection (i.e., the more reads upon which a detection is based, the more confident you can be in that detection).The negative controls did not produce identifiable sequence reads, suggesting that the results presented here were not compromised by laboratory-introduced contamination.

Figure 3 -
Figure3-Proportion of the detected Arthropod orders across the ZBJ amplifications for plate NGSFA0135 compared to a comprehensive BOLD reference library (representing 99.4% of total detections).

Figure 4 -
Figure4-Proportion of the detected Arthropod families across the ZBJ amplifications for plate NGSFA0135 compared to a comprehensive BOLD reference library (representing 66.6% of total detections)

Table S2 .
The kept (Probable) and eliminated (Improbable) detections of arthropods in the bird's fecal samples

Table S3 .
Identification of the probable plant species detected in the bird's diet, including their status, fruiting season, and fruiting category.

Table S4 .
Results of two PERMANOVAs comparing songbird species (Tennessee Warbler and

Table S5 .
Moran et al, 2019.tected in fecal samples of molt and post-molt Tennessee Warblers and Swainson's Thrushes captured during their fall migration at the McGill Bird Observatory in 2021 and 2022.For sex, 'M' is male and 'F' is female.intounique taxonomic identifications per sample and identifications were only accepted as genuine if they were supported by at least 100 reads that matched a reference sequence with at least 95% identity across at least 100 bp (Appendices I&II).Further molecular and bioinformatic details are available upon request.All raw data will be privately and securely stored on CCDB servers for at least three months and until storage constraints require deletions of old data.All raw data is available upon request.For methodological reference, please citeMoran et al, 2019. aggregated