FKI methodology
The FKI is a composite indicator of how severe the phytoplasma epidemic is: it expresses both the risk and the damage already done. This page describes what it is calculated from and how. It follows version 5.2 of the measurement rules.
The rules were built from the multi-round proposals and reviews of the three FD Finder Expert models, with fine-tuning by the founder. The parameters are expert estimates, not measured or validated values. The first scored check will take place in November 2027.
What the FKI is not
- not the share of infected vines,
- not the probability of an event occurring,
- not official data, but an expert estimate.
The formula
A = ∛(T × G × V) · Raw = A + (1 − A) × 0.20 × J · FKI = max(20%; 1 − (1 − 0.08) × (1 − Raw))
The base value (A) is the geometric mean of the three factors. All three lie between 0 and 100%, and for each of them a high value is bad. Epidemic burden (J), the damage already done, is added to this: it adds at most a fifth of the distance remaining to 100%. The weight is small because J is estimated data in most countries today. The 0.08 is the risk of return: where FD has once appeared, this risk remains even if the zones are lifted; there the value is at least 20%. Where FD has not yet been confirmed, FKI = Raw.
Each factor is made up of sub-indicators. A sub-indicator receives 0–100 points, and here too a high score is bad. A proportion-type sub-indicator is a measured proportion. For a scale-type one, five described levels (0, 25, 50, 75, 100) are the reference points; the experts give the exact value between the levels. I and J are converted from a percentage to points on a logarithmic scale.
Spread
How freely the phytoplasma can spread. There are two pathways: the leafhopper (the American grapevine leafhopper) and infected propagating material.
This factor's current value on the dashboard- Source: the surroundings of confirmed points (E1a) and the area of foci not yet found, weighted by infection density (I); or, if that is smaller, introduction (T9) with a multiplier. Latency (λ) adds the source that shows no symptoms yet.
- Leafhopper: the leafhopper’s distribution (T1) multiplied by its density (T2).
- Control brake: what share of the leafhopper population is reached by treatment (T3 in commercial vineyards, T5a in abandoned refuges, T5b in home gardens), how effective the treatment is (T4, T7), and whether there is money for it (T11, economic capacity, weakens it by at most μ_T). The brake can be at most k_max.
- Source removal: grubbing up infected vines by the deadline (T10) affects only known foci, so its effect weakens with the share of foci that stay hidden (U) (σ).
- Leafhopper pathway = source × leafhopper × (1 − brake) × (1 − source removal). Graft pathway = ω × T6 × the chance of a source of infected grafted vines (E1, T9, at least a base risk).
- Spread = 1 − (1 − leafhopper pathway) × (1 − graft pathway): each of the two pathways can cause infection independently of the other.
| Code | Sub-indicator | Measures |
|---|---|---|
| E1 | Extent of infection | The vineyard area of the settlements that contain an infected zone, relative to the country’s vineyard area. |
| E1a | Confirmed exposure | The vineyard area within 1 km of confirmed points of infection, relative to the country’s vineyard area. |
| I | Infection density | The share of vines in the exposed area that are confirmed infected or marked as symptomatic, on a logarithmic scale. |
| T9 | Introduction pressure | Infection in neighbouring countries, especially near the border. |
| T1 | Leafhopper distribution | The vineyard area of the counties where the leafhopper has been detected in the last two seasons. |
| T2 | Leafhopper density | Trap catches at untreated sites and pre-treatment nymph counts; catches after treatment do not count. |
| T3 | Lack of control in commercial vineyards | Where mandatory leafhopper treatment was not applied. |
| T5a | Lack of control in abandoned refuges | Abandoned and feral vines and feral rootstock where the leafhopper goes untreated. |
| T5b | Lack of control in home-garden vines | Home-garden, hobby and non-commercial vines where no treatment is applied against the leafhopper. |
| T4 | Shortfall in control quality | Forecasting, timing, choice of products and coordination of treatment. |
| T7 | Lack of awareness about control | What share of vine growers do not receive notifications in time. |
| T10 | Lack of source removal | How many of the vines confirmed as infected were not grubbed up by the deadline. |
| T11 | Lack of economic capacity for control | Whether there is money for thorough control and leafhopper monitoring, and whether new products against the leafhopper are developed. |
| T6 | Propagating-material pathway | How well certified and non-commercial movements of propagating material are controlled. |
| T8 | Weather | The effect of winter and spring on the leafhopper; only a forecast until the first density data. |
Unknown foci
How large the risk of a hidden focus is: the infection is present or may be introduced, and stays hidden with today’s detection. A focus counts as known once any of its symptomatic vines is spotted.
This factor's current value on the dashboard- We calculate three segments separately: commercial vineyards, home gardens and abandoned vineyards. Their weights express the estimated share of infected vines in each.
- Survey: the area covered (G1, G5a, G5b) multiplied by whether a focus in it is recognised (G2 survey, G4 automated pre-screening; at focus level, with n symptomatic vines).
- Notification by growers (commercial and home gardens): the grower recognises the symptom (π) and notifies the authority (G6 notification difficulty, G8 disincentives). In the commercial segment, the grower’s lack of economic interest (G11) weakens it by at most μ_G.
- Confirmation: the delay in lab confirmation (G3) is partly a loss (θ).
- Staying hidden (U): 1 − the weighted sum of the foci found in each segment.
- Presence (H): the infected zoned area (E1), and the hidden foci that can be assumed outside the zones, where the infection is already nationwide and the area has not been checked. Its lower limit is the value calculated from introduction pressure or 5%, whichever is larger.
- Unknown foci = U × H.
G is not a grade for the surveys. If detection works well, U falls; presence, however, shows the extent of the infection, which a survey does not change. No survey can find a fresh infection that shows no symptoms yet: this counts in Spread, through latency.
| Code | Sub-indicator | Measures |
|---|---|---|
| G1 | Unsurveyed share, commercial | What share of the commercial vineyard area was not surveyed in the symptom season. |
| G5a | Unsurveyed share, home gardens | What share of home-garden vines was not surveyed. |
| G5b | Unsurveyed share, abandoned vineyards | What share of abandoned vineyards was not surveyed or sampled. |
| G2 | Survey unreliability | Surveyors’ training, the protocol and the timing. |
| G4 | Weakness of automated pre-screening | The proven accuracy and actual coverage of drone and camera-based detection. |
| G3 | Lack of lab confirmation | How many suspicious vines did not receive a PCR result within 30 days. |
| G6 | Difficulty of notification | How easy it is to notify a suspicion, and whether the person notifying receives feedback. |
| G8 | Disincentives to notify | The rules on compensation and notification: how far notifying pays off for the grower. |
| G11 | Lack of economic interest in detection | Whether it is in the grower’s interest to search for and report infected vines, or whether they would rather exit. |
Vine loss
What share of the infected vines that have become known die, taking into account the resources for community treatment and research and the market for solutions.
This factor's current value on the dashboard- Biological loss: there is no cure (V1), no resistant variety (V2), and the vine does not recover (V3; poor condition, V4, reduces the chance).
- Under mandatory grubbing up, a grubbed-up vine cannot recover, and the lower limit of the factor is 90%.
- Community resources (V5): the available budget for research and treatment improves the chance of a solution; with a smaller weight (κ_k) under mandatory grubbing up, with full weight (κ) without it. Without a market able to pay (V6), fewer usable solutions come out of the resources (μ_V).
| Code | Sub-indicator | Measures |
|---|---|---|
| V1 | Lack of a cure | Whether there is an authorised treatment that also works in the vineyard. |
| V2 | Lack of resistant varieties | The share of the area planted with truly resistant varieties. |
| V3 | Lack of recovery | Whether lasting recovery of infected vines is proven. |
| V4 | General plant-health condition | Drought and heat stress in the current season. |
| V5 | Lack of resources for community treatment and research | Whether there is a dedicated, available budget for research and community treatment. |
| V6 | Lack of a market for development | Whether there is a market able to pay that makes it worth developing cures and resistant varieties. |
Epidemic burden
The damage already done. It is not a fourth factor: it raises the base value if the damage is large.
- Numerator: laboratory-confirmed vines or vines marked as symptomatic in one year; a symptom-free vine grubbed up together with its area does not count. Denominator: the vine population.
- Scale: below 0.05% it is 0 points, above 1% it is 100 points, logarithmic in between.
- Effect: Raw = A + (1 − A) × 0.20 × J. Grubbing up is itself the right measure; J shows the size of the damage. We also publish the value without J.
| Code | Sub-indicator | Measures |
|---|---|---|
| J | Epidemic burden | The annual share of laboratory-confirmed vines or vines marked as symptomatic, relative to the vine population. |
Calculated signals
- g – the growth in the number of settlements with confirmed infection over the previous 12 months.
- S – search ratio: by what factor the surveyed area grew between the two periods compared.
- G7 – surprise: the percentage of newly affected settlements that are not adjacent to a previously known zone. A check signal, not a term in the formula.
- G10 – lab positivity: the percentage of tested suspicious samples that are FD-positive. A check signal, not a term in the formula.
How a score becomes an effect
The scale levels are written descriptions of situations, not probabilities. Where the formula uses a scale as an effect, a fixed table converts the score, interpolating linearly between the levels. The other scales are converted linearly, because their levels are themselves proportions or they are scaled by a parameter of their own.
| Sub-indicator | Input to the formula | 0 | 25 | 50 | 75 | 100 |
|---|---|---|---|---|---|---|
| T2 | Leafhopper spreading power | 0 | 0.4 | 0.65 | 0.85 | 1 |
| T4 | Lost effectiveness of treatment | 0 | 0.12 | 0.3 | 0.6 | 1 |
| G2 | Lost sensitivity of the survey | 0 | 0.15 | 0.35 | 0.65 | 1 |
| G6 | Lost notifications (due to difficulty) | 0 | 0.1 | 0.3 | 0.55 | 1 |
| G8 | Lost notifications (due to disincentives) | 0 | 0.1 | 0.25 | 0.55 | 0.85 |
| V1 | Chance of losing an infected vine | 0.1 | 0.5 | 0.9 | 0.98 | 1 |
| V3 | Chance of recovery | 0.9 | 0.4 | 0.1 | 0.03 | 0 |
Parameters and weights
Every parameter is an expert estimate, not a measured value. When the estimate range is calculated, we vary them within their sensitivity ranges. We review them every year, before the November run.
| Symbol | Meaning | Value | Range |
|---|---|---|---|
| β_J | maximum effect of Epidemic burden | 0.2 | 0.16–0.24 |
| – | risk of return, where FD has ever been confirmed | 0.08 | 0.05–0.15 |
| – | lower limit of the FKI, where FD has ever been confirmed | 20% | – |
| – | multiplier for introduction in the source | 0.1 | 0.05–0.2 |
| – | base of infection density in the source | 0.5 | 0.3–0.7 |
| – | effect of latency in the source | 0.5 | 0.25–1 |
| – | weight of foci not yet found in the source | 0.85 | 0.7–1 |
| r_a | share of the leafhopper population in abandoned refuges | 0.25 | 0.15–0.35 |
| r_b | share of the leafhopper population in home-garden vines | 0.15 | 0.05–0.25 |
| τ | maximum effect of the lack of awareness | 0.25 | 0.1–0.35 |
| k_max | upper limit of the control brake | 0.85 | 0.75–0.95 |
| σ | maximum effect of source removal if every source is known | 0.6 | 0.4–0.8 |
| ω | upper limit of the graft pathway | 0.25 | 0.1–0.4 |
| – | base risk of the graft pathway | 0.25 | 0.1–0.35 |
| – | multiplier of introduction pressure in the graft pathway | 0.5 | 0.25–1 |
| – | maximum effect of the weather | ±10% | ±5%–±15% |
| w | segment weights: commercial / home garden / abandoned | 55% / 22% / 23% | – |
| m_max | upper limit of survey sensitivity | 0.75 | 0.6–0.9 |
| β | maximum improvement from automated pre-screening | 0.5 | 0.3–0.6 |
| π | grower recognition per vine: commercial / home garden | 0.4 / 0.2 | – |
| n | effective number of symptomatic vines in an undetected focus | 1.5 | 1–3 |
| θ | share of the lab delay counted as permanent loss | 0.5 | 0.3–0.8 |
| – | upper share of hidden presence outside the zones | 0.75 | 0.5–1 |
| – | E1 at which hidden presence outside the zones is complete | 0.65 | 0.5–0.8 |
| – | lower limit of presence | 5% | – |
| λ | transitional value of latency, until there is a detection rate | 0.25 | 0.1–0.5 |
| – | lower limit of Vine loss under mandatory grubbing up | 90% | – |
| – | maximum effect of poor condition on recovery | 0.5 | 0.3–0.7 |
| κ | maximum effect of community resources without mandatory grubbing up | 0.5 | 0.3–0.7 |
| κ_k | the same under mandatory grubbing up | 0.2 | 0.1–0.3 |
| μ_T | maximum effect of the lack of economic capacity (T11) on the control brake | 0.3 | 0.15–0.45 |
| μ_G | maximum effect of the lack of interest in detection (G11) on notification by growers | 0.7 | 0.5–0.9 |
| μ_V | maximum effect of the lack of a market for development (V6) on community resources | 0.5 | 0.3–0.7 |
How the estimate is made
- Machine data gathering: a fixed script downloads the official data and calculates the terms that can be calculated from them (E1, E1a, T1, T2, T8).
- Gathering: at least six and at most nine models gather the evidence independently of each other, with web search: the three experts and the research assistants. Nobody scores in this round.
- Merging: the aggregating model merges the findings and checks each one against its source. Anything that cannot be verified is left out.
- Scoring: three expert models (Google Gemini, OpenAI ChatGPT, Anthropic Claude) score from the verified list and FD Finder’s professional knowledge base, each with an exact value and its own estimate range. They do not see each other’s scores or earlier values. All three experts must complete a faultless run.
- Aggregation: for each sub-indicator, the central value of the scores, that is, the average of the median and the mean, with equal weight. The aggregator may depart from it only between the lowest and the highest score, with a reason.
- If the three scores differ widely, the aggregator examines them; with an even larger difference, a person decides. Then founder approval and publication.
This procedure is version 5.2 of the measurement rules (1 November 2026).
The latest evaluation: every score from the experts and FD Finder Expert’s assessmentFD Finder Expert’s professional knowledge base: what the experts score from, and how you can add to itThe estimate range
Alongside the headline figure, the range shows how uncertain the estimate is. Each expert states the range in which, in its view, the value lies with an 80% chance. In 2,000 random draws, each sub-indicator’s value comes from such a range of an expert chosen at random, and the parameters from their sensitivity ranges; the range is the 10th to 90th percentile of the resulting FKI values. It is not a statistical confidence interval.
Updates
Three values appear each year: the opening value in May, the preliminary value in August, and the closing value between the end of November and December. The time series shows each year’s closing value, and that is the one we put the most work into. The preliminary value is based on partial season data; the 2026 preliminary value on the data available up to October. In between, a monthly outlook is prepared for the factors.
Limitations
Many sub-indicators have no public figures; in such cases the experts estimate, and mark it as an estimate. For the other countries the experts also estimate the machine-calculated terms, so differences between countries are less certain than a change over time in one country. The past time series is an estimate reconstructed afterwards, not a measurement. The rules and the formula are versioned; after a change, the entire time series is recalculated.
Measurement rules: version 5.2.