METHODOLOGY · VERSIONED
Scoring, growth and yield, and fair value: how EMPI works
EMPI is governed by a single rule: methodology drives code, not the other way around. Scoring weights and sub-dimensions are versioned configuration, and a methodology entry must precede any change in the software. This page sets out the full chain, from raw Estonian data to a fair-value figure an investment committee can interrogate.
G&Y v1.0.0 · calibration tier formis_emu_ee · last methodology update 2026-05-11
What is land productivity? Three layers, never one number
Land productivity index
In EMPI, land productivity is not one index but a panel: strategy-specific suitability scores on a 0–100 scale, reported on three separate layers, biophysical capability, legal usability, and economic usefulness. No single composite number is ever published for a parcel.
The same hectare can be excellent ground and a poor investment. Soil and climate may support a productive stand, while a protection regime forbids harvesting it, and a missing access road makes whatever remains uneconomical. Collapsing those three facts into one score destroys the information a buyer pays for.
Biophysical
What the land can grow: soil, moisture, climate and terrain, the ceiling set by nature, before any law or market enters the calculation.
Legally usable
What the law lets you do: protection regimes, zoning, designations and encumbrances that restrict or remove uses the land could physically support.
Economically useful
What is worth doing: the productive capability that survives restrictions, access and market reality, the layer investment decisions actually rest on.
Eight scores and a verification status, never one composite
Each parcel receives a panel of eight strategy-specific scores on a [0,100] scale: agricultural use, grassland, existing forestry, afforestation, restoration and conservation, climate-resilience baseline, restriction severity, and uncertainty. The panel is reported as-is, EMPI never averages it into a single verdict, because the average would be wrong for every strategy at once.
Verification before scoring
Scoring is never shown without a verification status. The status is a function of data coverage, legal-data consistency and access, and it tells the analyst what kind of work the number still needs.
Each score panel ships inside what EMPI computes per parcel
Never averaged into a composite, the panel is the output.
Illustrative example, not a live parcel.
| Status | Meaning |
|---|---|
Verified, high confidence VERIFIED_HIGH_CONFIDENCE | All critical layers loaded and consistent. Decision-ready. |
Verified, medium confidence VERIFIED_MEDIUM_CONFIDENCE | Core layers in place; one important layer approximate. |
Low confidence, GIS review needed LOW_CONFIDENCE_REQUIRES_GIS_REVIEW | Important spatial data missing, conflicting or stale. |
Legal review required LEGAL_REVIEW_REQUIRED | Co-ownership, enforcement, mortgage or designation mismatch flagged. |
Field validation required FIELD_VALIDATION_REQUIRED | High potential, but soil, forest, water or access needs an on-site check. |
Rejected, insufficient data REJECT_DATA_INSUFFICIENT | Not enough data to score. Excluded from ranking. |
Rejected, heavily restricted REJECT_HIGH_CONSTRAINTS | Protected or restricted land dominates the productive area. |
Rejected, access blocked REJECT_ACCESS_BLOCKED | No public or legal access. |
How EMPI values forest: Estonian-empirical growth and yield
Site index
A measure of how good the ground is for a given species, expressed as expected stand height at a reference age. EMPI fits Hossfeld IV site-index curves calibrated on Estonian permanent sample plots (Tammiste 2021).
Growth-and-yield (G&Y) model
A model of how a stand’s volume develops with age on a given site. EMPI’s G&Y v1.0.0 is Estonian-empirical: Kangur 2007 for Scots pine and an Estonian allometric stack for six further species, not generic European curves.
Rotation age
The stand age at which harvest is legally and silviculturally due. EMPI anchors rotation by species and site index on the Estonian Forest Management Regulation and Forest Act §29, the model harvests when the law allows.
The valuation chain runs in five steps. First, the stand: species, age and density from the Forest Register. Second, the site: a Hossfeld IV site index says how fast that stand grows here, not in an average European forest. Third, the harvest schedule: rotation ages by species and site index, anchored in Estonian regulation. Fourth, restrictions: where protection regimes overlap the stand, a harvest-restriction factor scales the timber cashflow, and only the timber cashflow. Fifth, the DCF itself: discounted timber cashflows plus a land terminal value.
The empirical grounding is checked, not asserted: the G&Y tables are cross-checked against Estonia’s National Forest Inventory, and 5 of 6 species pass within ±25% (EMPI G&Y v1 validation report, 2026). When literature and empirical data disagree, EMPI ships both numbers with confidence bounds rather than silently picking one, that is the platform’s standing validation pattern.
How to value forest land: a growth-and-yield DCF walkthrough
| Protection regime | Harvest cashflow factor |
|---|---|
| Strict protection | 0.00 |
| Heritage meadow | 0.00 |
| Water-protection zone | 0.50 |
| Natura 2000 | 0.85 |
The factor scales timber cashflows only. Land-resale terminal value is never scaled, restricted forest can keep land value after timber value reaches zero.
Protected forest in Estonia: how restrictions change parcel value
Kangur, A. (2007) · Tammiste, L. (2021) · National Forest Inventory cross-check: 5/6 species within ±25% (EMPI growth-and-yield validation, 2026)
From NOI anchors to fair value per target return
Fair-value-per-ROE matrix
A table of the maximum fair price per square meter for each use strategy at each target return, 3, 5, 8 and 12%, with a decision label in every cell. It shows what a parcel is worth to you, given your required return.
The income side starts from NOI anchors calibrated to Estonian references, Maa-amet’s valuation methodology and Statistics Estonia 2024 lease data. Arable land anchors at 60 / 115 / 160 / 220 €/ha per year for weak, medium, good and premium ground; grassland at 25 to ~100 €/ha; bare forest land at 10–150 €/ha; and nursery operating modes substantially higher (500–1,500 €/ha open-field; 3,000–8,000 €/ha container, flagged explicitly as operating businesses, not passive land).
Each cell of the resulting matrix carries a plain-language decision label, from pass, through pass if timber is verified or marginal, to overpriced or fail. The matrix rolls up to a top-level verdict, for example fair at 3% only (fair at a 3% target return, overpriced at 5% and above), overpriced for productive use, or nursery upside required. The label states the condition under which a price works; it never tells anyone to buy.
| Strategy | 3% | 5% | 8% | 12% |
|---|---|---|---|---|
| Agricultural use | Marginal | Overpriced 2× | Fail | Fail |
| Grassland | Pass | Marginal | Overpriced 2× | Fail |
| Existing forestry | Pass | Pass if timber verified | Marginal | Overpriced 2× |
| Afforestation | Pass | Marginal | Overpriced 2× | Fail unless special upside |
| Restoration & conservation | Marginal | Fail | Fail | Fail |
| Roll-up verdict | Fair at 3% only fair at a 3% target return, overpriced at 5% and above | |||
Illustrative example, not a live parcel.
Two theses, honestly separated
Every parcel carries two independent valuation passports: a productive thesis (what the land earns in use) and a market/conversion thesis (what comparable sales support), with a pointer naming the stronger one, the two are never merged. The market thesis draws on Maa- ja Ruumiamet’s official transactions register, where detailed access is granted case by case; until that access is in place, or when the comparable set is too thin, EMPI flags the market thesis as insufficient instead of valuing off bad data.
screening and provenance in detail
Methodology questions, answered
Through a discounted-cash-flow model on Estonian-empirical growth-and-yield curves: Hossfeld IV site indices (Tammiste 2021), Kangur 2007 pine growth, rotation ages per the Estonian Forest Act, and a harvest-restriction factor for protected forest. Results are cross-checked against the National Forest Inventory, with 5 of 6 species within ±25%.
Because one number hides the decision. A parcel can rank high biophysically and fail legally or economically. EMPI reports three layers and eight strategy scores separately, so an investment committee sees exactly which use works, at which return, and under which verification status.
Yes. Every output carries a provenance envelope, dataset versions, model-weights hash, code commit, run timestamp and config versions, and outputs without it fail closed. Re-running the same parcel surfaces a delta panel showing exactly what changed since the prior analysis.
No. EMPI computes fair value under explicit target returns and scenario assumptions, and benchmarks it against comparable sales from the official Maa- ja Ruumiamet transactions register, reporting the comparable set as insufficient when access or coverage is thin. It is scenario analysis with confidence bounds and decision labels, not a price forecast, and never a recommendation to buy or sell.
Rotation ages by species and site index anchored on the Estonian Forest Management Regulation and Forest Act §29. The DCF schedules harvests only when regulation permits them, which is one of the main reasons generic, non-Estonian yield curves mislead buyers here.
The methodology is versioned, cited and built to be interrogated. Walk through it with us on a real parcel.
Request the methodology overview
A methodology briefing call, scoring decomposition, Estonian-empirical G&Y, verification taxonomy.