Adaptation Pathways¶
URBADAPT-HEAT evaluates three distinct policy pathways and their interactions. All policies share a common mortality baseline defined as the current-AC world (observed AC penetration in place, no additional policy) rather than a hypothetical no-adaptation counterfactual.
1. Air Conditioning (AC)¶
AC is modelled as a private adaptation option that attenuates the heat-mortality response of covered households. In CLIMADA terms, AC acts on the impact function: on a given day the effective mortality dose for cells with AC coverage c and age-specific efficacy ε_a is scaled by (1 − ε_a · c).
Coverage mapping¶
- City-level AC penetration rates are obtained from NUTS3 regional statistics.
- These are downscaled to local administrative zones (postal-code or sub-municipal) using a logistic income-rank curve, calibrated so the population-weighted zone average reproduces the NUTS3 input while the within-city income gradient is preserved.
- The calibrated zone coverage is rasterised to the 100 m reference grid.
Policy scenarios (equal-users constraint)¶
Both policies add the same total number of AC users citywide, so differences in mortality benefit arise from spatial allocation, not scale:
| Scenario | Description |
|---|---|
| Income-targeted (catch-up) | Coverage in zones of the lower half of income-ranked population is raised to a minimum prescribed threshold |
| Uniform | Equal percentage-point uplift to every zone, calibrated to match total new users |
Costs (incremental only)¶
- Per-unit capital expenditure (CAPEX)
- Annual maintenance (fixed share of CAPEX per user)
- Electricity costs: city-level tariff × kWh per user × number of users
- kWh per user downscaled from NUTS-level inputs using local summer heat exposure as dispersal weight
2. Urban Street Trees¶
Urban vegetation is modelled as a public adaptation pathway that cools the outdoor thermal environment, reducing hazard intensity before it reaches the exposed population. Trees act on the hazard, not on exposure or vulnerability.
Baseline greenness¶
Each grid cell's greenness is described by the Green View Index (GVI) — a point-based measure of street-level visible greenery from street-view imagery (Huisman et al. 2025; 356-city dataset).
GVI observations are associated with Local Climate Zone (LCZ) classes and rasterised onto the reference grid.
Two-step cooling pathway¶
The translation from vegetation to temperature change operates in two steps:
Step 1: ΔGVI → ΔLST
A change in GVI is converted to a change in Land Surface Temperature using city-, month-, and LCZ-specific linear regression coefficients derived from paired satellite LST and GVI data across European cities. Positive (warming) coefficients are clipped to zero — the model is cooling-only.
Step 2: ΔLST → ΔT2M
ΔLST is translated into a change in daily mean air temperature using a quadratic emulator (Falchetta & Hammad 2025) that approximates the local LST–T2M slope as a function of the daily temperature state and LCZ context. The emulator bundle (pooled or city-specific "safe" holdout bundle) is set per city via trees.veg_emulator_bundle. The direct ΔGVI→ΔT2M coefficient table is retained only as a bridge diagnostic, not as the main pathway.
Default policy: regional Q3 catch-up¶
Each policy region whose mean GVI falls below the 75th percentile (Q3) of the city-wide regional distribution receives a ΔGVI equal to its shortfall. The policy is restricted to built LCZ classes (1–10); natural, water, and agricultural pixels are excluded.
An equity-weighted variant reallocates ΔGVI toward the most SVI-vulnerable regions, improving distributional outcomes at the same total investment.
Tree costs¶
CAPEX and O&M are city-specific (PLI purchasing-power-adjusted from a common D5.1 base):
| Item | Config key | Rome | Athens | Lisbon | Copenhagen |
|---|---|---|---|---|---|
| CAPEX per tree | capex_per_tree_eur |
€680 | €612 | €606 | €978 |
| Annual O&M per tree | om_per_tree_per_year_eur |
€63 | €57 | €56 | €91 |
| CAPEX per +1 GVI point | capex_per_index_pt_eur |
€10M | €10M | €10M | €10M |
Common timing parameters: ramp_years: 12 (zero benefit at installation, full at year 12; central case assumes pre-grown trees with a ~5-year head start), and lifetime_years: 25 (CAPEX distributed linearly as a gradual planting rollout). The tree allocation geography is set by veg_regions (regions or zones).
3. Early Warning Systems (EWS)¶
EWS are modelled as a public, city-level mortality-reduction policy layered on top of the current-AC baseline. EWS do not modify the hazard or exposure layers — they reduce the conditional probability of a heat-attributable death on days when a warning is declared.
Warning activation modes¶
| Mode | Trigger | Used for |
|---|---|---|
| Standard (deaths-threshold) | threshold_method: epi_deaths_file; threshold calibrated from CLIMADA baseline mortality; recalibrated every N years (threshold_recalib_interval, default 5) |
Rome, Athens, Lisbon |
| Track-B (event-mask fallback) | Primary: deaths-based threshold (same as standard); fallback to extreme event-day mask (nonpositive_threshold_fallback: event_mask) when baseline heat deaths are too low to calibrate a threshold |
Copenhagen |
Efficacy interpretation (city-specific)¶
The interpretation config key governs whether EWS efficacy is marginal (benefit on top of an existing mature system) or counterfactual (full benefit from establishing a new system):
| City | Archetype | Interpretation |
|---|---|---|
| Rome | mature_hhwws |
Marginal |
| Athens | weak_threshold |
Counterfactual — Greece lacks a comprehensive HHAP; the full benefit of establishing one is modelled |
| Lisbon | mature_hybrid_icaro |
Marginal — ICARO system operational since 1999 |
| Copenhagen | cold_city_extreme_focus |
Counterfactual |
Key parameters¶
| Parameter | Description |
|---|---|
| Warning season | Default May 1–September 30 for Rome/Athens; May 15–September 30 for Lisbon/Copenhagen (configurable via start_md/end_md) |
| Efficacy interpretation | marginal or counterfactual; city-specific (see table above) |
| AC–EWS overlap | EWS efficacy scaled down as function of population-weighted AC coverage |
| Mortality displacement | Gross → net avoided deaths via configured short-term displacement factor |
| Ramp-up period | 3 years (ramp_years; institutional uptake and behavioural adaptation) |
Cost models¶
| Model | Basis |
|---|---|
| Pavanello et al. (2025) (default) | Per-capita per-warning-day |
| Chiabai et al. (2018) | Per-warning-day city-level |
Interactions among adaptation options¶
Three interaction channels are explicitly represented:
AC waste-heat externality¶
Higher AC penetration increases outdoor heat rejection, raising ambient temperatures and generating a partial mortality penalty. This is quantified in two steps:
1. Nighttime ΔT: empirical lookup table (Salamanca et al. 2014) maps citywide AC penetration to nighttime ambient warming (config key: waste_heat.lut).
2. COP degradation: temperature-dependent efficiency loss (Kou et al. 2026) increases electricity demand and thus waste heat as ambient temperature rises (config key: waste_heat.cop_degradation; sensitivity range: 4–9% per °C).
The daily-mean warming is derived from nighttime ΔT via a configurable conversion factor (dailymean_from_night_default: 0.50). A thermally-weighted activation function (kou_cdd_share) scales the waste-heat effect by cooling-degree-day share so it is only active when cooling demand is positive. Only the incremental waste heat from policy-added users is charged to the AC policy in the CBA.
EWS–AC overlap¶
Higher AC penetration reduces the residual mortality pool addressable by warnings. EWS efficacy is scaled downward as a function of population-weighted current-AC coverage on warning days.
Vegetation–AC electricity co-benefit¶
Vegetation cooling reduces peak ambient temperatures, lowering AC energy demand (Falchetta, De Cian & Lunghi 2026; config block electricity_feedback, default ~0.8% AC-electricity reduction per GVI point). This co-benefit is now integrated directly into NB08's AC cost computation — trees reduce kwh_per_user in the combined AC + trees scenario — and is reported as an explicit tree-owned interaction term, not silently subtracted from either policy's standalone cost (see Cost-Benefit Analysis).
This page is maintained in the URBADAPT-HEAT wiki and synced automatically. Edit it there, not in the website repository.