TL;DR
This study provides the first national-scale assessment of Great Crested Newt (GCN) occupancy in England using environmental DNA (eDNA). By applying multiscale occupancy models to a massive dataset of nearly 5,000 ponds, the researchers quantified the "hidden" error rates of eDNA surveys, finding that while high laboratory replication virtually eliminates false negatives, it significantly inflates the risk of false positives.
What Problem Does This Paper Address?
The Great Crested Newt (Triturus cristatus) is a protected species in the UK, making its presence a major factor in land development and conservation planning. Traditional survey methods—like setting traps or counting newts by torchlight at night—are labor-intensive and slow.
Environmental DNA (eDNA) has revolutionized this process by allowing researchers to detect "biological breadcrumbs" (skin cells, waste, etc.) in water samples. However, eDNA isn't perfect. You might fail to collect DNA even if the newt is there (false negative), or a sample might get contaminated in the field or lab (false positive). Most standard models assume false positives don't happen, which can lead to overestimating species distribution. This paper seeks to quantify these errors and identify which environmental factors actually drive newt occupancy across England.
Methods
The authors utilized a multiscale occupancy model (specifically the Griffin et al. 2020 model) via the eDNAShinyApp R package. This approach is sophisticated because it separates detection into two stages:
- Stage 1 (Field): Did the water sample actually contain DNA?
- Stage 2 (Laboratory): Did the specific qPCR (quantitative Polymerase Chain Reaction) replicate detect that DNA?
Standard protocols use 12 qPCR replicates per pond sample. The researchers used Bayesian variable selection (Posterior Inclusion Probability or PIP) to determine which of 19 environmental, climatic, and landscape variables—ranging from water quality to "Northing" (latitude)—were true predictors of newt presence.
Datasets
The study analyzed a massive open-access dataset from Natural England, consisting of eDNA surveys from 4,925 ponds across England collected between 2017 and 2019.
To enrich this data, the authors integrated several other high-resolution datasets:
- UKCEH Land Cover Maps 2019: For surrounding habitat types.
- BGS Geology Layer: To account for bedrock and soil drainage.
- HadUK-Grid: For 20-year climate averages (temperature, rainfall, ground frost).
- Standard HSI (Habitat Suitability Index) data: Collected by field surveyors (e.g., pond area, shade, fish presence).
Key Insights and Ideas
- The Replication Paradox: The standard protocol of 12 qPCR replicates is a double-edged sword. While it makes the laboratory false negative rate negligible, it compounds the false positive risk. With a 2% false positive rate per replicate, a sample with no DNA actually has a 24% chance of returning at least one positive result just by chance.
- Thresholding Matters: Currently, a single positive qPCR replicate is often enough to declare a pond "occupied." The authors argue this is unwise. Moving the threshold to three positive replicates would drastically reduce false positives without significantly increasing false negatives.
- HSI is Over-Simplified: The traditional Habitat Suitability Index (HSI) used by ecologists weights ten variables equally. This study shows that's ecologically unrealistic. Factors like fish presence and water quality are heavy hitters, while pond area and terrestrial habitat quality were surprisingly poor predictors at this national scale.
- Climate and Geography: Newts are "southerners" at heart; occupancy probability drops as you move North. They also prefer areas with more ground frost and less rainfall, likely due to how these factors influence pond hydroperiods (how long a pond stays wet).
Key Results
- National Occupancy: Approximately 20% of ponds in England are occupied by Great Crested Newts.
- Field Errors: The false negative rate at the sampling stage was 5.2%, while the false positive rate was 1.5%.
- Laboratory Errors: The false positive rate was 2% per qPCR replicate.
- Top Predictors: Water quality, shade, rainfall, fish presence, and macrophyte (aquatic plant) cover were the most reliable predictors of occupancy.
- The "Fish" Factor: "Possible" fish presence (when a surveyor isn't sure) was found in equal proportions in both occupied and unoccupied ponds, suggesting this specific metric in the HSI needs more rigorous measurement.
Limitations and Future Work
The authors acknowledge that they couldn't account for variation between different commercial laboratories because that metadata wasn't available. Furthermore, some categorical variables (like specific bedrock types) had very small, unbalanced sample sizes, making them hard to interpret.
For future surveys, the authors suggest a "redistribution of effort." Instead of running 12 replicates on a single water sample, it might be more statistically robust to collect two separate water samples from a pond and run 6 replicates on each. This would better account for the 5.2% chance of failing to catch DNA in the bucket during the initial field visit.
Summary updated