Methodology

Overview

The Kew Reach Climate Resilient Tree Tool integrates multiple strands of cutting-edge research into a single, actionable assessment framework. Our approach combines Species Distribution Modelling (SDM) with functional trait scoring — adapted from the methodology established by Martin et al. (2025) — to produce a composite suitability rating for each of 291 tree species across 12 UK regions.

This methodology is grounded in the work of Martin, Sjöman, Hirons, and their collaborators, who have pioneered the use of the Climatic Moisture Index (CMI) as a more ecologically meaningful metric for assessing tree suitability than temperature alone. The tool also incorporates site suitability data from the TDAG Tree Species Selection Guide (Hirons & Sjöman, 2019), providing expert-curated environmental tolerance ratings, use potential classifications, and crown characteristics.

The tool is designed to support local authorities, landscape architects, developers, and urban foresters in making evidence-based planting decisions.

The Climatic Moisture Index (CMI)

At the heart of our assessment is the Climatic Moisture Index (CMI), using the formulation developed by Hogg (1997):

CMI = (P / PET) − 1

Where P is annual precipitation (mm) and PET is potential evapotranspiration (mm). The resulting CMI ranges from −1 (extreme drought, no precipitation) through 0 (precipitation exactly equals PET) to positive values (precipitation surplus). This formulation is consistent with the TreeGOER database (Kindt, 2023), which provides species climate envelopes derived from GBIF occurrence data — confirmed by the presence of species with CMI maxima exceeding +1.0 in the database.

For the UK, current CMI values range from approximately +0.8 (Scottish Highlands) to +0.05 (South East England). Under the SSP3-7.0 emissions scenario, London's CMI is projected to shift from +0.05 to approximately −0.2 by 2090, crossing from a marginal surplus into a moisture deficit — a significant ecological threshold.

The Growing Season CMI (GS-CMI) is also calculated for informational display, but the suitability scoring uses annual CMI exclusively — consistent with how TreeGOER species envelopes are derived.

Climate Envelope Modelling (Martin/Sjöman Method)

Species climate envelopes are derived from the TreeGOER database (Kindt, 2023), which provides environmental range data for 48,129 tree species worldwide. TreeGOER computes the 5th and 95th percentile of CHELSA bioclimatic variables at verified GBIF occurrence locations for each species, with outlier detection and cleaning. This gives us empirically-derived MAT and CMI ranges representing the realised climatic niche of each species.

For each assessment, the target location's projected climate (MAT and CMI at 2050 or 2090) is plotted as a single point in this MAT × CMI space. The Euclidean distance from that point to the nearest edge of the species' climate envelope determines climate suitability:

• Distance = 0 (inside envelope): The location's future climate falls within the species' natural range — high climate suitability. • Small distance (0–2 normalised units): The location is near the edge of the envelope — moderate suitability with some climate stress. • Large distance (>3 normalised units): The location's climate is far outside the species' natural range — low suitability or vulnerable.

Critically, this is a two-dimensional assessment. A Mediterranean species like Quercus ilex may tolerate the projected CMI of London in 2090, but its MAT envelope (10–17°C) also matters. A species adapted to hot, dry conditions will score poorly in a cool, wet region like the Scottish Highlands — not because of moisture stress, but because the temperature falls below its natural range.

This approach follows the methodology established by Martin et al. (2025), who evaluated the entire publicly owned tree population of London (approximately 1.1 million trees) against projected MAT and CMI values for 2050 and 2090. For hybrid species and cultivars not present in TreeGOER, we average the climate envelopes of the parent species.

Temperature Metrics: MAT vs MTGS

The tool offers two temperature metrics for climate envelope assessment, selectable via the Temperature Metric toggle in the Coordinate Assessment:

Mean Annual Temperature (MAT) is the default temperature metric for both Regional and Coordinate (Site) Assessments. MAT represents the average temperature across all 12 months and is the most widely used temperature variable in species distribution modelling. It is the primary metric used in the TreeGOER database (Kindt, 2023) and provides a broad measure of thermal suitability. Note: Martin et al. (2025) use MTGS exclusively in their published assessments; our use of MAT as the default is a deliberate departure that prioritises consistency with the TreeGOER species envelope data.

Mean Temperature of the Growing Season (MTGS) is available as an alternative metric via the Temperature Metric toggle in the Coordinate Assessment. MTGS is defined as the mean temperature for the fixed April–October period (7 months), following the convention used by Martin et al. (2025) in their assessment of London's urban tree population. This fixed-window definition ensures consistency across all sites and climate scenarios, rather than varying the growing season length based on a temperature threshold.

MTGS is arguably more ecologically meaningful for deciduous species, as it captures the thermal conditions during the period of active transpiration, photosynthesis, and growth. Species envelopes for MTGS are derived from the TreeGOER_CHELSA dataset (Kindt, 2023), which provides growing season temperature percentiles (Q05/Q95).

The MTGS approach follows the rationale of Sjöman et al. (2025) and Martin et al. (2025), who argue that growing season conditions are more relevant than annual averages for predicting tree performance in urban environments. For example, a species from a continental climate with cold winters but warm summers may have a moderate MAT but a high MTGS — and its urban performance is better predicted by the latter.

Users can switch between MAT and MTGS to compare how the choice of temperature metric affects suitability ratings for their site.

Drought Tolerance Assessment

Beyond climatic envelope matching, we assess each species' physiological capacity to withstand drought stress using the turgor loss point (ΨP0). This metric, measured in megapascals (MPa), represents the leaf water potential at which a plant's cells lose turgor pressure — essentially, the point at which the plant begins to wilt irreversibly.

Species with more negative ΨP0 values can maintain cell function at lower water potentials, indicating greater drought tolerance. For example, Quercus ilex (Holm Oak) has a ΨP0 of approximately −3.5 MPa, while Betula pendula (Silver Birch) has a ΨP0 of only −1.7 MPa — meaning the Holm Oak can continue functioning under much more severe drought conditions.

This data is drawn from the comprehensive evaluation by Hirons et al. (2021), who measured ΨP0 across 96 taxa commonly used in UK urban forestry, and from the TRY Plant Trait Database.

Leaf Dry Matter Content (LDMC)

Leaf Dry Matter Content (LDMC) is a key functional trait that indicates a species' position on the leaf economics spectrum. LDMC is defined as the ratio of leaf dry mass to leaf fresh mass (g/g), and higher values indicate more structurally robust, stress-tolerant leaves — a conservative resource strategy associated with drought and heat tolerance.

LDMC data is available for all 291 species in the database, sourced from three complementary datasets:

• The LEDA Traitbase (Kleyer et al., 2008): A European plant trait database providing direct species-level LDMC measurements. We obtained 20 direct species matches and 121 genus-level median values from LEDA.

• Published literature: For genera not covered by LEDA, we compiled LDMC values from Niinemets (2001), Kattge et al. (2011), and other peer-reviewed sources covering 95 additional genera.

• Genus-level proxies: Where species-level data was unavailable, we used the median LDMC of congeneric species from the same database, following the approach of Wright et al. (2004) who demonstrated that leaf economic traits are strongly conserved at the genus level.

LDMC is scored on a 1–4 scale: very stress-tolerant (≥ 0.40 g/g) = 4, stress-tolerant (0.33–0.40) = 3, moderate (0.27–0.33) = 2, acquisitive (< 0.27) = 1. This score is incorporated as the fourth component of the composite suitability score, alongside climate distance, turgor loss point, and wood density.

Botanic Garden Screening

A complementary strand of evidence comes from the systematic evaluation of tree species growing in botanic garden collections, as described by Sjöman et al. (2025). Botanic gardens serve as natural "stress-test" environments, where species from diverse climatic origins are grown side-by-side under local conditions.

By evaluating growth performance, winter hardiness, drought response, and overall vitality of species in collections such as those at Kew, Gothenburg, and other European botanic gardens, researchers can identify species that demonstrate real-world adaptability beyond what climatic envelope modelling alone would predict.

This screening approach is particularly valuable for identifying "sleeper species" — taxa that are rarely used in urban forestry but show exceptional promise when evaluated under controlled conditions.

Composite Suitability Scoring

The Composite Suitability Score is adapted from the Martin et al. (2025) methodology, combining five independent assessments for each species–location combination. All components are weighted equally:

Climate Rank (1–4): Determined by the species-relative distance from the site's projected climate to the species' GBIF-derived envelope. The assessment uses a two-dimensional climate space of temperature (MAT or MTGS) and moisture (annual CMI). Annual CMI is calculated as (P / PET) − 1 using the Hogg (1997) formulation. Species envelopes are derived from TreeGOER_CHELSA (Kindt, 2023) using 5th/95th percentile bounds. Distance is normalised by the species' own envelope width. Inside envelope = 4, marginal (<0.3 envelope widths) = 3, outside (0.3–0.7) = 2, far outside (>0.7) = 1. For future periods, suitability is assessed under both SSP3-7.0 and SSP5-8.5 scenarios and the results are blended (arithmetic mean).

Turgor Loss Point Score (1–4): Scored from the species' measured ΨP0 value using fixed boundaries derived from k-means clustering on the global TRY database (Martin et al., 2025): score 4 if ΨP0 ≤ −3.0 MPa, score 3 if ≤ −2.5, score 2 if ≤ −2.0, score 1 otherwise. Species with more negative ΨP0 values (greater drought tolerance) receive higher scores.

Wood Density Score (1–4): Scored from the species' wood density using fixed boundaries from the global TRY database (Martin et al., 2025): score 4 if WD ≥ 0.78 g/cm³, score 3 if ≥ 0.55, score 2 if ≥ 0.45, score 1 otherwise. Higher wood density indicates greater structural resilience.

Leaf Dry Matter Content Score (1–4): LDMC (g/g) is scored using fixed boundaries from the global TRY database (Martin et al., 2025): score 4 if LDMC ≥ 0.40, score 3 if ≥ 0.33, score 2 if ≥ 0.27, score 1 otherwise. Higher LDMC values indicate more stress-tolerant, conservative leaf strategies. LDMC data is sourced from the LEDA Traitbase (Kleyer et al., 2008), TRY database, published literature, and genus-level proxies.

The Composite Score is the arithmetic mean of all component ranks (range 1.0–4.0). The final suitability category applies a dual-gate rule requiring both a minimum composite score AND climate envelope overlap:

• High: Composite mean ≥ 3.0 AND climate distance ≤ 0.3 (species envelope overlap) • Moderate: Composite mean ≥ 2.0 AND climate distance ≤ 0.3, OR composite ≥ 3.0 but climate distance > 0.3 • Low: Composite mean ≥ 2.0 but climate distance > 0.3 • Vulnerable: Composite mean < 2.0

Critically, a species cannot achieve a "High" rating on traits alone — it must also demonstrate clear climate overlap (climate distance ≤ 0.3, meaning the site's projected climate falls within or very near the species' natural envelope). A species scoring ≥ 3.0 on the composite but outside the climate envelope is capped at Moderate. Conversely, a species inside the climate envelope but with weak drought tolerance traits will be downgraded.

Ecosystem services and biosecurity risk scores are displayed separately as expert filters but do not affect the suitability rating.

TDAG Site Suitability Data

In addition to climate-based suitability, the tool incorporates expert-curated site suitability data from the TDAG Tree Species Selection for Green Infrastructure guide (Hirons & Sjöman, 2019). This guide, produced for the Trees and Design Action Group, provides detailed environmental tolerance ratings and use potential classifications for 285 tree species commonly used in UK urban green infrastructure.

The TDAG data enriches our assessment with the following dimensions:

Use Potential: Each species is classified for suitability in specific planting contexts — Paved areas, SuDS (Sustainable Drainage Systems), Coastal environments, Transport corridors, and Small gardens. These classifications are based on expert evaluation of root architecture, salt tolerance, canopy spread, and other site-specific factors that cannot be captured by climate modelling alone.

Environmental Tolerances: Expert-assessed ratings for drought tolerance, shade tolerance, and waterlogging tolerance, each on a four-level scale from "tolerant" to "intolerant" (or "sensitive"). These complement the physiological drought assessment (turgor loss point) by incorporating field observations and horticultural experience. The TDAG drought tolerance rating reflects the species' overall capacity to cope with water stress in urban settings, including soil compaction and restricted rooting volume — factors not captured by the leaf-level ΨP0 measurement.

Crown Characteristics: Crown form (columnar, conical, globular, irregular, ovoid, spreading, vase) and crown density (dense, moderate, light) inform spatial planning decisions and ecosystem service estimation. For example, a dense, spreading crown provides superior shade and cooling, while a light, columnar crown may be more appropriate for narrow streets.

Successional Status: Classification as pioneer, early-successional, late-successional, or climax species. Pioneer species (e.g., Betula, Alnus) are fast-growing and tolerant of open, exposed conditions, making them suitable for new developments and restoration sites. Late-successional species (e.g., Fagus, Tilia) are shade-tolerant and long-lived, better suited to established landscapes and woodland planting.

Phenology: Peak flowering and fruiting periods, supporting biodiversity planning and seasonal interest considerations.

The TDAG data is available for 200 of the 291 species in our database (69% coverage). Site Filters based on these attributes are available to Professional and Consultancy tier subscribers across all three assessment views (Species Explorer, Coordinate Assessment, and Region Assessment).

Climate Data Sources

The tool uses high-resolution climate data from the CHELSA V2.1 dataset (Karger et al., 2017; 2022), which provides downscaled climate variables at approximately 1 km resolution globally.

Baseline climate (1981–2010): Monthly temperature and precipitation are sourced from CHELSA V2.1 climatologies. Potential Evapotranspiration (PET) for the baseline period is sourced directly from CHELSA’s Penman-Monteith PET product, which incorporates radiation, wind speed, humidity, and temperature — providing a physically-based estimate of atmospheric evaporative demand.

Future projections (2011–2040, 2071–2100): Monthly temperature (tas), precipitation (pr), maximum temperature (tasmax), and minimum temperature (tasmin) are sourced from CHELSA CMIP6 projections under both the SSP3-7.0 and SSP5-8.5 emissions scenarios, using the GFDL-ESM4 global climate model. Final suitability ratings are the arithmetic mean of assessments under both scenarios. Since CHELSA does not provide PET for future periods, we calculate future PET using the Hargreaves-Samani (1985) method, which requires only tasmax and tasmin. The Hargreaves PET is calibrated against the CHELSA baseline Penman-Monteith PET to ensure consistency across time periods.

Urban Heat Island adjustment: When UHI is applied, PET is recalculated by scaling the CHELSA PET proportionally. The ratio of Thornthwaite PET at the adjusted temperature to Thornthwaite PET at the original temperature is applied to the CHELSA PET, preserving the accuracy of the Penman-Monteith baseline while correctly increasing PET with the urban temperature increment.

Additional climate indicators displayed in the assessment include the Aridity Index (AI = PET / P, expressed as a percentage, where values above 100 indicate PET exceeds precipitation) and the Maximum Climatic Water Deficit (MCWD), which represents the most negative cumulative monthly water balance (P − PET) during the year — a measure of the severity of the driest period.

Urban Heat Island Detection

The tool provides automatic Urban Heat Island (UHI) estimation for any coordinate using the Global Human Settlement Model (GHS-SMOD) R2023A dataset, produced by the European Commission Joint Research Centre (EC JRC).

GHS-SMOD classifies the Earth’s surface into settlement types at 1 km resolution based on population density and built-up area, using data from the Global Human Settlement Layer (GHSL). When a user selects a coordinate for assessment, the tool reads the settlement classification from the GHS-SMOD GeoTIFF and maps it to a suggested UHI temperature offset:

• Urban centre (class 30): +3.0°C • Dense urban cluster (class 23): +2.0°C • Semi-dense urban cluster (class 22): +1.5°C • Suburban or peri-urban (class 21): +1.0°C • Rural cluster (class 13): +0.5°C • Low density rural (classes 11–12): 0°C

These estimates represent mean annual UHI intensity, based on UK urban climate literature (Levermore & Parkinson, 2019; Heaviside et al., 2016; Wilby, 2003). London’s UHI intensity averages 3–4°C in summer with peaks exceeding 5°C during heatwaves; mean annual UHI is typically 60–70% of peak summer values.

The auto-detected settlement type and suggested UHI are displayed in the assessment sidebar. Users can accept the suggestion or manually adjust the UHI slider (range 0.5–5.0°C) based on local knowledge. The UHI offset is applied uniformly to all monthly temperatures and cascades through to PET and CMI calculations, affecting the final suitability assessment.

Known Departures from Martin et al. (2025)

This tool is adapted from the methodology of Martin et al. (2025) but differs in several important respects. These departures are deliberate design decisions, documented here for transparency:

CMI formulation: Martin's published papers use the Thornthwaite Moisture Index (scale approximately −200 to +200). This tool uses the Hogg (1997) formulation (P/PET − 1, scale −1 to +∞) for consistency with the TreeGOER database (Kindt, 2023), from which species climate envelopes are derived. The two formulations rank locations in the same order but produce different absolute values.

PET calculation: Martin uses Thornthwaite (1948) PET exclusively. This tool uses CHELSA Penman-Monteith PET for the baseline period and Hargreaves-Samani (1985) PET for future projections. Penman-Monteith is the FAO-recommended standard and is generally considered more physically accurate, but it produces systematically higher PET values than Thornthwaite.

Default temperature axis: Martin uses MTGS (Mean Temperature of the Growing Season, April–October) exclusively. This tool defaults to MAT (Mean Annual Temperature) for consistency with the TreeGOER species envelope data, with MTGS available as a user-selectable alternative.

Temporal periods: Martin uses 2011–2040 (near-term) and 2061–2090 (end-century). CHELSA CMIP6 provides 2011–2040 but not 2061–2090; we use 2071–2100 as the closest available end-century period.

LDMC data coverage: Like Martin, this tool uses genus-level and family-level averages to fill gaps where species-specific LDMC measurements are unavailable. The estimation hierarchy draws on the TRY database, LEDA Traitbase, published literature, and genus/family-level proxies.

Climate envelope overlap: Martin's SDM assigns species to categorical zones (optimal, within range, marginal, outside range) based on where the site falls relative to the species' distribution. This tool uses a continuous Euclidean distance metric normalised by envelope width, with thresholds at 0, 0.3, and 0.7 mapping to the same four categories. The combined distance approach may occasionally differ from per-axis categorical assignment for borderline species.

Limitations and Caveats

This tool provides a strategic-level assessment to guide species selection decisions. Users should be aware of several important limitations:

Microclimate variation: Urban environments contain significant microclimate variation (heat islands, wind tunnels, shade corridors) that cannot be captured at the regional scale. The Coordinate Assessment with Urban Heat Island (UHI) adjustment provides a first-order correction based on GHS-SMOD settlement classification, but site-specific assessment remains essential.

Soil conditions: The tool does not account for local soil type, pH, compaction, contamination, or drainage — all of which significantly affect tree performance. Species recommendations should be cross-referenced with soil suitability data. The TDAG use potential classifications (particularly SuDS and Paved) provide some guidance on species tolerance of challenging soil conditions.

Provenance matters: Within a species, different provenances (geographic origins) can show markedly different climate tolerances. Where possible, sourcing from southern or drought-adapted provenances is recommended.

TDAG data coverage: TDAG site suitability data is available for 200 of 291 species (69%). Species without TDAG data are not excluded from climate-based assessments but will not appear in Site Filter results.

Climate model uncertainty: Future projections use a single GCM (GFDL-ESM4) under both SSP3-7.0 and SSP5-8.5 scenarios, with results blended across scenarios. Different GCMs would produce different projections. The tool currently does not provide multi-model ensemble ranges.

PET methodology: Future PET is calculated using the Hargreaves method (which requires only temperature extremes) rather than the full Penman-Monteith method (which also requires radiation, wind, and humidity data not available in CHELSA future projections). This may introduce some systematic bias in future PET estimates.

Evolving science: Climate projections, species distribution data, and physiological trait databases are continuously updated. This tool will be regularly revised to incorporate the latest evidence.

The tool is designed to complement, not replace, professional arboricultural advice. We recommend using it as a starting point for species selection, to be refined through site-specific assessment and professional consultation.