Table 1.
Comparisons of survey methods.
Fig 1.
Map of the study region showing mainland sites, Northern Channel Islands sites, and Catalina Island sites on the coast of southern California, USA. Black dots and numbers correspond to site names. 1– Dangermond, 2 –R Beach, 3 –Santa Claus, 4 –Santa Monica, 5 –Cuyler Harbor, 6 –Sandy Point, 7 –Soledad, 8 –Bechers Bay, 9 –Water Canyon, 10 –Southeast Anchorage, 11 –Ford Point, 12 –China Camp, 13 –Forney Cove, 14 –Christy Beach, 15 –Coches Prietos, 16 –Emerald Bay, 17 –Little Harbor, 18 –Ben Weston. The base map was created with Natural Earth Dataset (http://www.naturalearthdata.com/).
Fig 2.
Venn diagram of eDNA, seine, and BRUV species detections.
Environmental DNA methods captured the majority (30/36) of fish species detected by both BRUV and seine surveys, only failing to identify six fish species found in the other two survey methods. In addition eDNA identified 58 additional fish species missed by seine and BRUV methods. In contrast, BRUV and seine surveys only captured 50% of species detected by both methods, showing strong difference in the species detected by each method. This was largely driven by the unique detection of elasmobranchs as well as nearshore pelagic and rocky reef carnivorous fishes in BRUV surveys compared to the unique detection of surfperches (Family Embiotocidae), grunts (Family Sciaenidae), and planktivorous nearshore pelagic species in seine surveys.
Fig 3.
Constrained analysis of principal components.
Constrained Analysis of Principal Components (CAP) analysis was conducted on Jaccard binary dissimilarities of fish assemblages of all species detected across surveys. Survey method explained 42.6% of the total variation observed in the composition of detected taxa while site explained an additional 28.3% (PERMANOVA p <0.0001). We found no significant difference in homogeneity of dispersions across sites (betadisper > 0.05). BRUV and eDNA approaches more frequently detected leopard sharks (Triakis semifasciata) and California bat ray (Myliobatis californica) compared to seine surveys. Both seine and BRUV surveys detected Barred surfperch (Amphistichs argenteus) while eDNA methods could only achieve family level resolution for this taxon. eDNA approaches more consistently detected opaleye (Girella nigricans), northern anchovy (Engraulis mordax), giant kelpfish (Heterostichus rostratus), and dwarf perch (Micrometrus minimus).
Fig 4.
Heatmap of surf zone fishes jointly detected between surveys.
Teleost species in black font and elasmobranch species in blue font. Environmental DNA approaches more frequently detected 26 of 27 known surf zone species detected by either BRUV or seine surveys. Only Leopard shark Triakis semifasciata was more frequently detected by BRUV surveys. Aggregating surfperches with identical 12S barcodes to family level assignment (surfperch), eDNA detected surfperches at the same rate at seines. Aggregating to genus level assignment, eDNA detected Syngathus sp. More frequently than either BRUV of seine surveys.
Fig 5.
Probability of occupancy, specificity, and sensitivity of eDNA, seine, and BRUV surveys.
Environmental DNA surveys had higher probability of occupancy and sensitivity than BRUV and seine surveys. eDNA had significantly higher specificity than seine surveys. We found no difference in specificities between BRUV and eDNA and seine surveys. Probability of occurrence is a measure of how likely a species is present at a site as a function of the commonness of the species as well as the true positive and false positive detection rates of the method surveyed. Sensitivity is the proportion of true positive species detections correctly identified as true positive detections. Specificity is the proportion of true negative species detections identified as negative detections.
Fig 6.
Sample coverage estimates of eDNA, seine, and BRUV surveys.
Across all sites, environmental DNA surveys had an estimate sample coverage of 98.9%, higher than the sample coverage of BRUV (89.6%) and seine (85.2%) surveys. Shaded area represents 95% confidence intervals. Sample rarefaction curves across sites suggest BRUV and seine surveys would have to be conducted at more than 100 sites to achieve similar sample coverage estimates to eDNA surveys conducted at the 18 sites surveyed here.