Viral diseases in potato crops:
recognising PVY and Leafroll

Back to Field Notes

Potato farming in the Netherlands — one of the world's most productive potato-growing regions — faces a persistent challenge that no amount of agronomic excellence can fully prevent: viral disease. Two viruses dominate the concern: Potato Virus Y (PVY) and Potato Leafroll Virus (PLRV). Together they account for the majority of yield losses in commercial potato production worldwide.

H2L Robotics BV built the POTECTOR300 specifically to address this problem — an autonomous machine that scans potato fields row by row, using computer vision to identify symptomatic plants before their virus can spread. The annotation teams at H2L Robotics India train the AI models that power this detection. Understanding what these diseases look like, how they behave, and why visual detection is difficult is central to understanding the work.

Why viral disease in potatoes is a particular problem

Unlike fungal diseases that can be treated with fungicides after the fact, viral infections in plants have no cure. Once a plant is infected, it remains infected. The only interventions are preventative (controlling aphid vectors, using certified virus-free seed) or roguing — physically removing and destroying infected plants before they become a source of further spread.

The Dutch potato farming model relies heavily on seed potato production: growing tubers specifically to be planted in subsequent seasons. Viral disease in a seed crop is particularly damaging because the virus is carried into the tuber and passed to next season's plants. A single infected seed tuber produces an infected plant that becomes a source of infection for its neighbours, via aphids, across the entire field.

The economic stakes are significant. PVY can cause yield losses of between 10% and 80% depending on variety, virus strain, and timing of infection. Leafroll has historically been the most damaging single viral cause of yield loss in ware potato crops in Europe.

Potato Virus Y (PVY)

How PVY spreads

PVY is transmitted by aphids in a non-persistent manner. This means the aphid acquires the virus rapidly — within seconds to a few minutes of feeding on an infected plant — and can transmit it equally quickly when it probes a new plant. Importantly, the aphid does not need to remain on the plant for long to spread the disease. Many aphid species that are not even typical potato pests can act as vectors during their brief landing and probing behaviour.

This non-persistent transmission makes PVY particularly difficult to control through aphicide sprays alone: the aphid causes infection before systemic insecticide can take effect. Mineral oil sprays (which interfere with aphid probing) and early-season scouting are more effective management tools.

Visual symptoms on the plant

PVY symptoms vary significantly by strain, potato variety, and growing conditions — some virus-variety combinations produce mild or no visible symptoms, while others produce severe, easily visible disease. The following symptoms represent those most likely to be encountered and annotated:

• Mosaic pattern: Irregular yellow-green mottling across the leaf surface, creating a map-like or jigsaw-puzzle appearance. The contrast between lighter and darker leaf areas is the defining visual feature of PVYO infection.
• Necrotic spots: Dark brown or black patches of dead tissue on the leaf surface, typically distributed along or between leaf veins. These indicate the necrotic strains (PVYN, PVYNTN).
• Necrotic stripes: Brown-black linear streaks running along leaf veins or down the stem. Stem necrosis is a strong indicator of PVYN infection and can lead to premature stem death.
• Leaf distortion: Crinkling, puckering, or slightly curled leaf edges, particularly along the margins. The leaf surface may appear uneven or bumpy.
• Loss of uniform colour: Healthy potato leaves are a consistent deep green. PVY-affected leaves show uneven colour — some areas brighter yellow-green, others darker — breaking the uniform pattern of the healthy canopy.

Potato Leafroll Virus (PLRV)

How PLRV spreads

Unlike PVY, PLRV is transmitted by aphids in a persistent manner. The green peach aphid (Myzus persicae) is the primary and most efficient vector. For persistent transmission, the aphid must feed on an infected plant for several hours before it acquires the virus, and the virus must then replicate within the aphid before it can be transmitted. Once infective, the aphid remains so for the rest of its life.

This persistent mode means that individual flying aphid arrivals are less immediately dangerous than with PVY, but established aphid colonies on infected plants are highly effective spreaders. The virus can also be transmitted from mother plant to offspring through vegetative propagation — meaning infected seed tubers produce infected plants.

Visual symptoms: primary vs secondary infection

PLRV symptoms differ significantly depending on whether a plant was infected during the current growing season (primary infection) or grew from an infected seed tuber (secondary infection). Secondary infections are more severe.

Primary infection (virus acquired from aphids in the current season):

• Initial symptoms can be subtle: a slight upward rolling of the upper leaves along the central vein
• A red or orange tinge to the upper foliage — often noticeable before rolling is obvious
• Symptoms typically appear in the upper canopy first and may be easy to miss at early stages

Secondary infection (plant grown from an infected tuber — more severe):

• Leaf rolling: Leaves roll sharply upward along the midrib, forming a tube or cup shape. This is the most recognisable symptom. Lower leaves are affected first, progressing up the plant.
• Stiffness and brittleness: Rolled leaves become rigid and dry, with a papery, crinkly texture when touched. Unlike healthy leaves, which are soft and flexible, PLRV-affected leaves feel stiff.
• Pale or yellowish colour: Leaves may show light green, pale yellow, or yellowish-white colouring, particularly at the margins.
• Upright, closed posture: The whole plant appears narrower and more vertical, as leaves no longer spread outward. From a distance, an infected plant looks distinctly different in shape from its healthy neighbours.
• Stunting: Ranging from slight to severe depending on variety and timing of infection. Severely stunted plants may be only a fraction of the height of healthy neighbours.
• Net necrosis in tubers: When tubers from PLRV-infected plants are cut, a brown discolouration of the vascular tissue (phloem) is often visible — a pattern called net necrosis. This renders tubers unmarketable.

Why visual detection is hard — and where AI helps

Despite the distinctive symptoms described above, field disease detection has historically depended on trained human scouts walking or driving through fields. This approach has inherent limitations: scout fatigue, inconsistent detection across hundreds of hectares, difficulty in distinguishing mild symptoms from normal plant variation, and the impossibility of evaluating every plant in large-scale commercial fields.

Several factors make visual detection genuinely difficult even for experienced scouts:

• Symptom variability: Different virus strains on different potato varieties produce widely different visual symptoms. Some variety-strain combinations are virtually asymptomatic even when the plant is infected and spreading virus to neighbours.
• Overlap with normal variation: Natural variation in plant size, age-related yellowing at the end of the season, nutrient stress, and water stress can all produce symptoms that visually overlap with viral disease.
• Scale: A commercial Dutch potato field covers hundreds of rows with thousands of individual plants. Walking every row carefully is not economically viable within the time window when detection leads to effective roguing.

The POTECTOR300 addresses all three. Its cameras capture every plant in every row, at consistent speed and height, during the optimal detection window. Its AI model — trained on thousands of annotated images where human experts have identified and labelled diseased plants — classifies what the cameras see and marks infected plants with a chalk spray. The farmer removes marked plants; the rest of the crop is protected.

The quality of that AI model depends entirely on the quality of the training annotations. A model trained on precisely labelled images learns the subtle texture, colour, and shape cues that distinguish a mildly sick plant from a healthy neighbour. A model trained on imprecise or inconsistent annotations learns noise. This is why the annotation work at H2L Robotics India sits at the foundation of what the POTECTOR300 can do in a Dutch potato field.