Illuminating arthropod diversity in a tropical forest: Assessing biodiversity by automatic light trapping and DNA metabarcoding

Abstract

Although studies of insect decline have recently dominated headlines worldwide, their interpretation requires caution since for most species, we lack long-term population baselines. In the tropics, where most insect species thrive, our knowledge is even more limited and so reliable insect assessments must originate from well-established long-term monitoring efforts. Combining the extensive monitoring data from the Arthropod Program of the Smithsonian Tropical Research Institute (STRI) on Barro Colorado Island (BCI), Panama, we compare whether known arthropod diversity can be detected through metabarcoding of bulk insect samples obtained through automatic light-trapping. Our study detected 4402 species based on Barcode Index Numbers (BIN) and detected fine-scale differences between wet and dry seasons and sampling localities. We further refined our analysis to indicate which families and genera explained seasonal turnover. Using samples collected in parallel, but sorted manually as part of the ongoing arthropod monitoring program, we compared these methods. Out of 538 BINs recovered through manual sorting, there was a 70% overlap with the metabarcoding data; however, it represented 30% of the total BINs detected through metabarcoding. Expecting higher detection through metabarcoding, we also compare the results with the 14 years of sampling in BCI to better understand how well the monitoring program has captured the diversity of focal groups. Our results revealed a ~50% overlap between both methods and similar total catch. Barcode Index Numbers manually detected but not recovered by metabarcoding highlight some of the limitations of molecular detection methods such as primer bias. Contrastingly, BINs detected with metabarcoding, but not recovered by the traditional monitoring scheme, highlight the importance of local and regional barcode reference libraries.