One of the first questions asked when presenting eDNA data is: “It is great that you found eDNA from beaver in your water sample, but where was the actual beaver?” Unfortunately, the answer is usually unsatisfying as it is either extremely simplified (and therefore probably not quite right and should be taken with a big grain of salt) or extremely complicated…
The problem with this question is that answering it requires considering every other eDNA process to answer it, many of which we still don’t know much about. We need to know 1) how much eDNA an animal sheds (which can vary several orders of magnitude, even within the same individual), 2) at what rate the eDNA is decaying (which can vary with lots of biotic and abiotic factors like temperature – which can also change drastically in a given environment like night versus day or different depths in the water column), 3) what form the eDNA is in and how that will impact its physical transport (think a single cell versus a fecal pellet), and 4) the decisions we make in the lab (more on this later). Harrison et al. 2019 summarizes these processes and their interactions particularly well.
Of course, there is a spectrum of complexity with perhaps a stagnant pond on one end and a deep, dynamic oceanic environment on the other. And regardless of the physical environment, it would be great if we knew that the thing we were studying was stationary, but more likely we have an animal swimming around and shedding eDNA at different points in space and time (and at different rates!).
Ok, so it can get overwhelming quickly. But we can try to isolate things and determine what we can maybe ignore and what we definitely can’t ignore. After eDNA is shed from an animal, theoretically, it can move in the water column by advection (“going with the flow”) and dispersion (aka mixing). We traditionally think in the horizontal direction, but there can also be vertical advection and dispersion (although usually much lower values than their horizontal counterparts), and we also need to consider if eDNA is settling. Finally, in shallow or benthic environments, we should consider if previously settled eDNA is resuspended. So eDNA can theoretically move in a lot of ways, but which ones are important?
We can generally use two approaches. One is to model eDNA as we do with other particles like fine particulate organic matter (FPOM), microplastics, fish larvae, or marine snow. This will give us some basic information about what how far we might expect eDNA to travel based on what we know of how water moves (aka use ocean models). But, it requires a lot of assumptions, and some of which we just don’t know yet (see Andruszkiewicz et al. 2019). Our estimates of transport distances given what we know about the timescale of eDNA decay can be on the order of 10s to 100s of kilometers. So, it’s somewhere to start but maybe not the best approach.
Another method is to conduct field sampling in a very specific way to isolate some of the elements of the problem. Going back to our simplest case of a pond, a recent study (Littlefair et al. 2020) looked at how eDNA is vertically distributed in ponds, and how those distributions change as the lake is stratified (or not stratified) by sampling during different seasons. The authors found that when the lake was stratified (discrete layers of different temperatures), the eDNA is also stratified, and thus not moving up and down vertically. And importantly, different fish resided at different depths, with some preferring warmer or colder water, and the eDNA signals matched the fish occupancy as determined by acoustic telemetry (sound signals). This implies that the depth at which the eDNA is located is the depth at which the animal is located, which is great! But, the water itself was not mixing in that case. When the lake was well mixed (as indicated by the same temperature throughout the water column), the fish occupied many depths and the eDNA was found to be at many depths. But now we can’t separate the fish moving from the eDNA moving after it has left the fish. So not perfect, but this is important and a good place to start – and we obviously need more data and modeling to explore more complicated scenarios.
Moving to a more dynamic habitat, two different studies have now put cages with fish in rivers and measured how far downstream eDNA can still be detected (and how spread out the signal is across the river as you move downstream). Laporte et al. 2020 put fish cages in a very large river (the St. Lawrence River) and used a hydrodynamic model to simulate eDNA concentrations and compare results to field samples. The authors found that eDNA was transported up to 5 km downstream and that lateral mixing (side to side) was minimal. In a much smaller alpine stream, Thalinger et al. 2020 assessed lateral (side to side) and longitudinal (downstream) eDNA transport over different seasons (and thus, drastically different flow rates). The authors found similar results where eDNA was detected up to 1.3 km downstream (the farthest sampling point) and importantly that on small spatial scales (within the first 20 m downstream of the cages), eDNA lateral distribution mostly matched the location of the cages, indicating low lateral mixing. These studies are really important for advancing our understanding of how far downstream we can detect eDNA (also see Shogren et al. 2017, Civade et al. 2016, Deiner and Altermatt 2014, and Pont et al. 2018 for more on transport in rivers).
As far as I know, similar studies in the ocean have not been conducted, but aquaculture facilities could be a good way to sample the “radius” at which eDNA can be detected around a stationary point source. However, in the literature, there are several studies that have found that fish communities determined by eDNA vary on very small spatial scales (Port et al. 2015, Kelly et al. 2018, etc.)? Those imply that eDNA actually can’t move that far from the animal after it has been shed. How can we reconcile the different estimates of how far eDNA can move with these studies? We have been thinking about this a lot and have to continue reminding ourselves of what we are trying to measure and how we are measuring it. Or in other words, it depends on your perspective (and methods).
Basically, there are two ways to ask the question about eDNA transport:
1) How far does a single eDNA particle move?
2) At what spatial scale can you detect differences in communities by eDNA?
Conveniently, these correspond well with the two different ways we currently use molecular tools to find eDNA in water. It makes more sense to answer question 1 with qPCR for just a single species at a time. Question 2, however, requires looking at many species at once and implies the use of eDNA metabarcoding. But eDNA metabarcoding does not do a great job at giving quantitative results for a given species (for many of the reasons Ryan described in a previous blog post – namely, PCR bias and other errors introduced in laboratory processing). And we have to be particularly careful about false negatives in metabarcoding data. If something is not detected via metabarcoding, it might not mean that the DNA (and thus the animal) wasn’t there. Going back to reconciling the long transport distances but small scale spatial differences, we have to remember that they have primarily been answering two separate questions using two different tools (a notable exception is Civade et al. 2016, which looked at eDNA transport using metabarcoding rather than qPCR). So, there is a world in which coho salmon eDNA canbe transported 100s of meters but we can still determine that Site A is different than Site B which is only 20 m away because we found coho salmon at both sites, but other fish and invertebrates and diatoms unique to each site.
So… how does this apply to field samples where we found beaver eDNA? First we need to ask how was the beaver identified – using qPCR or metabarcoding? And then we probably need to think about the other processes I mentioned before as it relates to beavers and the water we sampled – like how much eDNA does a beaver shed and how long do we expect eDNA to last in this water temperature? And a final complication to leave you with: what if there were multiple beavers? One might have been 5 m upstream a day before collecting our water sample and another one 75 m upstream and hour before collecting our water sample.
Like I said, it can get complicated! Probably the beaver was less than a kilometer upstream. But who really knows.