Metabarcoding vs Direct Observation

Metabarcoding Versus Direct Observation in Wildlife Diet Studies

Direct observation is the oldest method in ecology.

Before sequencers and statistical ecology matured into their modern forms — there were notebooks, binoculars, and patient observers recording what animals ate, one bite at a time.

So how does dietary DNA metabarcoding compare to actually watching animals eat?

Well, it’s different, and sometimes it can be transformative. Both can be essential if your goal is to understand ecology or protect endangered species.

Direct observations of pronghorn antelope grazing in Yellowstone National Park

What Direct Observation Does Exceptionally Well

Direct observation can tell us:

• What plant parts an animal consumes or how it hunts its prey
• Whether dietary resources are targeted with specificity or opportunistically
• How animals interacted with their foods and one another while feeding
• How much feeding behavior depended on weather, predators, or competitors

It is challenging to scale, but it can be extraordinarily precise and enlightening.

One of the most remarkable examples comes from research led by Dr. Wilfred Odadi in Kenya. In experimental plots within the KLEE (Kenya Long-term Exclosure Experiment) system, Odadi watched cattle foraging and counted each bite they took in astonishing detail — literally counting and categorizing the number of bites taken of each forage species across manipulated herbivore treatments.

Because wildlife access was experimentally controlled, the system supported inferences about the effects of competition between livestock and wildlife on resource use. The manipulation made it possible to interpret outcomes mechanistically and they published their results in Science.

→ Read their paper: ​African Wild Ungulates Compete with or Facilitate Cattle Depending on Season

But even this extraordinarily detailed observational work had its limits. Researchers could record what cattle ate. They could measure how the presence of wildlife altered vegetation and livestock performance. But they could not be certain what wildlife were eating in the same way — wildlife would not have tolerated a researcher sitting down in the grass to monitor their feeding so closely. Inferences about foraging behavior and competition could only run in one direction.

Hero-Level Direct Observation of Wildlife Diets

Another stunning example comes from Elizabeth Kleynhans and colleagues, who observed wildlife foraging with incredible detail at Hluhluwe iMfolozi Park in South Africa.

Their wildlife-focused work pushed direct observation to its limits. After observing animals graze, researchers literally darted out into fields to inspect the stems and leaves of plants in the exact patch where feeding occurred. They identified grazed plants by eye, building species-level diet datasets for a diversity of large mammalian herbivores with extraordinary resolution.

→ Read their paper: Resource partitioning along multiple niche dimensions in differently sized African savanna grazers

Their dedication rivaled what dietary DNA metabarcoding can achieve.

We were inspired. It rivaled the best of what we were able to do with dietary DNA in Kenya. At one point, we considered trying to replicate their approach in Yellowstone. But before we began, we had to pause.

Yellowstone’s plant diversity presents an identification challenge even to botanists with decades of experience. Once grazed, many plant species become indistinguishable. Diagnostic floral features are gone. Leaves are torn. Stems are stripped. Morphology becomes ambiguity.

In that context, even heroic commitments to observation reach a ceiling. That is why turned to combining dietary DNA metabarcoding with our experimental manipulations in Yellowstone National Park.

What Changes with DNA Metabarcoding

Instead of watching bites happen, dietary DNA metabarcoding reconstructs diet based on genetic traces of material in animal feces or stomach contents. When reference libraries are strong and sampling is sufficient, it can detect dozens or hundreds of plant taxa from a single individual over time.

Its advantages emerge whenever:

• Resource communities are species-rich
• Morphological identification is unreliable
• Diets must be compared across many species simultaneously
• Sampling scales exceed what observers can reasonably monitor
• The wildlife that you are trying to study are rare, dangerous, shy, and forage at night or in habitats where you can’t safely follow.

At Yellowstone National Park, dietary DNA metabarcoding allowed us to resolve plant identities that even expert field botanists could not confidently separate once grazed.

Across East Africa, it allowed us to quantify wildlife diets at community scale — not just livestock — across diverse herbivore guilds.

In South America, it lets us integrate decades of experimental manipulation with modern genomic insight.

What Dietary DNA Metabarcoding Cannot Do

But DNA metabarcoding does not replace the need or benefit to engaging in direct observation.

It cannot tell you:

• Whether a bite was exploratory, accidental, or focused
• Whether feeding under the threat of predation
  
• How individuals competed at patch scales
• Which parts or life stages of a resource were consumed (e.g., caterpillar vs. butterfly; leaf vs. fruit)

It detects presence, not behavior. It reconstructs resource composition, not feeding activity. And it depends critically on well-curated plant DNA reference libraries — an issue particularly pressing in the Global South, where barcode coverage remains incomplete.

The Key to Comparison

 It is a question of scale, feasibility, and inference.

Direct observation excels when:

• Species richness is manageable
• Morphological traits remain diagnostic
• Behavioral context is central
• Experimental manipulation enables causal interpretation

Dietary DNA metabarcoding excels when:

• Resource diversity is high
• Food identifications are impossible without close study
• Scale exceeds the capacity of observers

In many systems, the strongest science combines both: observation generates behavioral hypotheses, experiments generate mechanistic insight, and DNA metabarcoding provides the resolution and scale required to be confident about our results.

→ See how we teach the combination in an HHMI module: Niche Partitioning & DNA metabarcoding

Why This Comparison Matters

Ecology increasingly asks large-scale, long-term questions:

• How does climate variability reshape food webs?
• How do livestock and wildlife share landscapes?
• How do diets shift before populations decline?

Answering these questions requires deep field experience together with scalable data.

The idea is not that we should abandon eyeballs in favor of sequencers.
​
It is to know when each approach reaches its limit — and how combining them can strengthen inference.

Explore More in This Series

• Metabarcoding vs Microhistology
• Metabarcoding vs Stable Isotopes
• Metabarcoding vs Metagenomics