Every number, and how it was produced

Wood Intelligence service-life screening method (WI-SL) ranks candidate wood species for a described exterior exposure and estimates service life, maintenance interval and weathering behaviour. It is deterministic: identical inputs always produce identical outputs, with no model call and no randomness. It is a screening instrument for early design decisions, not a substitute for project-specific testing or a structural verification per EN 1995.

Identifier

WI-SL v1.2.0

Released

2026-08-01

Licence

CC BY 4.0

Cite as: Wood Intelligence. Wood Intelligence service-life screening method (WI-SL) v1.2.0. 2026-08-01. woodintel.io/method

1. E1

   UV load

   L_uv = clamp(UV_climate · f_orientation.uv · f_exposure, 0, 100)

   UV_climate = normalised annual global horizontal irradiance for the Köppen zone; f_orientation.uv and f_exposure from the parameter tables.

   Basis: Orientation weighting after EN 927-1 exposure severity; climate term from ERA5 reanalysis.

2. E2

   Moisture load

   L_m = clamp(((rain + RH) / 2) · f_orientation.wet · f_exposure, 0, 100)

   rain = driving-rain index; RH = mean relative humidity, both normalised 0–100.

   Basis: Driving-rain exposure per EN ISO 15927-3; wetting/drying logic per EN 335 use class 3.

3. E3

   Chloride and freeze-thaw load

   L_salt = clamp(salt · f_exposure, 0, 100);  L_ft = clamp(FT · f_exposure, 0, 100)

   salt = coastal chloride deposition index; FT = annual count of 0 °C crossings, normalised.

   Basis: ERA5 2 m temperature series; distance-to-coast chloride gradient.

4. E4

   Biological load

   L_bio = clamp(((RH + rain) / 2) · f_orientation.wet · k_north, 0, 100), k_north = 1.15 for north, else 1.0

   North correction accounts for reduced drying and algal colonisation.

   Basis: Decay-hazard logic after EN 335 and Scheffer-type climate indices.

5. E5

   Weighted performance

   P = 100 − Σ_i (L_i / ΣL) · max(0, L_i − R_i) · 1.35

   R_i = species resistance for stressor i, including the treatment credit; each stressor penalises in proportion to its share of the total load.

   Basis: Wood Intelligence formulation. Ordinal, not a physical damage model.

6. E6

   Service life

   L_service = clamp(L_ref(DC) · (0.55 + P/160) · t, 5, 80) years,  t = 1.22 if treated else 1.0

   L_ref(DC) = reference life for the EN 350 durability class.

   Basis: Factor-method structure after ISO 15686-8, with a single aggregated performance factor.

7. E7

   Maintenance interval

   M = max(2, round(M_base · (0.55 + P/150) · s)),  s = 0.8 for south, else 1.0

   M_base = 11 years treated, 7 years untreated.

   Basis: Recoat intervals after EN 927-1 thin-film system classes.

8. E8

   Silvering

   S = clamp(46 − 0.33 · L_uv, 8, 60) months

   Time to visually uniform grey on an untreated or clear-finished surface.

   Basis: Photodegradation rate as a function of cumulative UV dose.

9. E9

   Confidence

   C = clamp(96 − min(40, 0.8 · max_i(L_i − R_i)) − p_exposure − p_treatment − p_profile, 25, 96)

   p_exposure = 3 / 6 / 10; p_treatment = 8 when a system is assumed but not declared; p_profile = 12 when the species has no tabulated resistance profile.

   Basis: Wood Intelligence uncertainty heuristic. Not a statistical confidence interval.

10. E10

    Heuristic service-life range

    [L_service − k·L_service, L_service + k·L_service],  k = 0.15 / 0.25 / 0.40 by confidence band

    Band: High ≥ 80, Moderate ≥ 62, otherwise Indicative. Retained for backwards comparison only.

    Basis: Deterministic spread. Superseded by the Monte Carlo percentiles in E11–E13.

11. E11

    Input sampling

    X_j^(n) = X̂_j · (1 + z·σ_j^rel·m_zone)  for relative terms;  X_j^(n) = X̂_j + z·σ_j^abs  for absolute terms,  z ~ N(0,1) truncated at ±3

    X̂_j = nominal parameter value; σ_j from the declared uncertainty table; m_zone = Köppen main-class reliability multiplier (climate terms only). Draws use a mulberry32 PRNG seeded from a hash of the brief, so a run is exactly reproducible.

    Basis: Standard Monte Carlo propagation, GUM Supplement 1 (JCGM 101) structure.

12. E12

    Output percentiles

    P10, P50, P90 = empirical quantiles of {y^(n)} over N = 1200 draws, y ∈ {P, L_service, M, S}

    Every draw re-evaluates E1–E8 in full; percentiles are the sorted empirical quantiles, not a normal fit.

    Basis: Non-parametric quantiles; no distributional assumption on the output.

13. E13

    Per-parameter uncertainty share

    s_g = Var_g(L_service) / Σ_g Var_g(L_service),  Var_g from a one-at-a-time run of 400 draws per group g

    Group g varies alone with every other input held at its nominal value. Shares are first-order contributions and do not capture interaction terms.

    Basis: One-at-a-time variance decomposition. An approximation of a Sobol first-order index.

14. E14

    Empirical calibration

    L_final = L_model · k_zone,  k_zone = exp( (n/(n+m))·mean(ln(obs/pred))_zone + (m/(n+m))·ln k_global )

    obs = observed service life in the WI-VAL benchmark; pred = uncalibrated model P50; n = benchmark cases in the Köppen main class; m = 2 shrinkage pseudo-observations.

    Basis: Fitted against WI-VAL v1.0.0. Applied to service life only, as a published layer on top of E1–E8, so the physics and the correction can be audited separately.

15. E15

    Residual σ inflation

    σ_residual,calibrated = σ_residual · 4.5

    σ_residual is the declared model residual on service life (12 %).

    Basis: Chosen so that leave-one-out P10–P90 coverage on the benchmark reaches the 80 % target. Declared input uncertainty alone understated real scatter roughly fourfold.

Resistance profiles are ordinal expert scores anchored on EN 350 durability class, extractive chemistry, cell-wall modification and dimensional movement. They are not measured quantities, and they are the part of this method most in need of field calibration.

Every reported output is propagated through 1,200 draws (E11–E13). Each draw perturbs every uncertain input at its declared 1σ, re-evaluates E1–E8 in full, and the reported P10/P50/P90 are the empirical quantiles of the resulting distribution — no normal fit. Draws are gaussian, truncated at ±3σ; mulberry32 PRNG seeded from an FNV-1a hash of the input brief, so the percentiles are exactly reproducible. The declared σ values below are modelling assumptions, published so that they can be challenged and replaced by measured field variance.

• EN 350:2016Durability of wood and wood-based products — Testing and classification of durability to biological agentsUsed for: Durability class → reference service life (E6)
• EN 335:2013Durability of wood and wood-based products — Use classesUsed for: Use class assignment and biological load (E4)
• EN 927-1:2013Paints and varnishes — Coating materials for exterior wood — Classification and selectionUsed for: Exposure severity and maintenance interval (E1, E7)
• EN ISO 15927-3Hygrothermal performance of buildings — Driving rain indexUsed for: Moisture load (E2)
• ISO 15686-8Buildings and constructed assets — Service life planning: reference service life and service-life estimationUsed for: Factor-method structure (E6)
• ERA5Copernicus Climate Change Service, ERA5 hourly reanalysis on single levelsUsed for: Temperature, humidity, precipitation and irradiance inputs (E1–E4)
• Köppen-GeigerBeck et al. (2018), Present and future Köppen-Geiger climate classification mapsUsed for: Climate zone assignment

• —Service life is empirically calibrated (E14) against a 12-case benchmark. The correction removes the average bias; it does not make the model a validated damage model, and the zone factors for tropical, arid and polar climates rest on two cases or fewer.
• —The calibration is fitted on the same benchmark used to report accuracy. Leave-one-out metrics are published on /validation and are the correct figures to judge the model by.
• —Resistance profiles are expert-assigned ordinal scores, not laboratory measurements. Only the EN 350 durability class behind them is a standardised value.
• —The confidence band (E9) is a rule-based heuristic, not a statistical interval. The P10/P50/P90 figures come from Monte Carlo propagation (E11–E13) of declared input uncertainty — they are as good as those declared σ values, which are modelling assumptions until replaced by measured field variance.
• —Input uncertainties are treated as independent. Real correlation between rain, humidity and biological load will make the true spread wider than modelled at the tails.
• —Detailing quality — ventilated cavity, end-grain sealing, fall, ground clearance — dominates real outcomes and is only partially represented.
• —Treatment performance assumes a correctly applied system maintained to its stated interval. Product-specific data is not used.
• —Climate inputs are zone-level, not site-level. Local topography, shading and coastal microclimate are not resolved.

Every evaluation in the decision engine can be exported as a calculation record: the brief, the climate source values, each applied factor, every intermediate, the per-species trace of penalties and scores, the outputs — and this complete method specification embedded inside the same file. No access to our code is required to check our arithmetic.

1. 01 Open the decision engine and describe the exposure.
2. 02 Choose “Download calculation (JSON)”.
3. 03 Apply equations E1–E10 to the recorded inputs using the embedded parameter tables. The outputs must match exactly.