Accelerating invasion potential of disease vector *Aedes aegypti* under climate change

The study addresses how climate change affects the development intensity and invasion potential of Aedes aegypti, a key vector of arboviral diseases. Because ectothermic vectors are highly sensitive to environmental conditions, quantifying temperature- and precipitation-driven development is crucial for anticipating shifts in vector distributions and associated disease risk. Prior work has often used correlative species distribution models or mechanistic population models, but few have been applied at large spatiotemporal scales to evaluate long-term environmental change for invasive vectors. The authors propose using a mechanistic phenology model to estimate the number of life-cycle completions (LCC) as a process-based index of environmental suitability. Objectives are to build and validate a phenology model across Ae. aegypti life stages, quantify historical (1950–2000) and projected (to 2050; RCP 4.5 and 8.5) changes in LCC globally, and assess acceleration of invasion fronts in key regions (Europe, USA, China) and seasonal changes in suitability.

The paper contrasts correlative suitability models, which rely on associations between occurrences and environmental covariates, with mechanistic models that link physiological responses to environmental drivers. Mechanistic models have advantages for predicting into novel climates and uninvaded areas, reducing biases from observation effort and extrapolation. Prior mechanistic models for Ae. albopictus and Ae. aegypti focused on temperature-sensitive, stage-structured population dynamics but have rarely been scaled to global distribution or long-term climate change assessments. Phenology models, widely used in agriculture to predict invasive pest establishment, explicitly model development rates and critical thresholds from lab-derived responses and can compute generations (LCC) per period from simple temperature-dependent development rates. This approach has not previously been applied to invasive human disease vectors at global scales.

Study species: Aedes aegypti, primary vector of dengue, zika, yellow fever and chikungunya.

Model overview: A spatially explicit phenology model calculates the number of life-cycle completions (LCC; generations) per grid cell and time period using daily climate inputs and empirically derived developmental responses. LCC serves as a development intensity index linked to occurrence, abundance, and establishment potential given dispersal.

Life-cycle representation: Two periods—aquatic (immature: egg, larva, pupa) and aerial (adult: mating, blood feeding, gestation, oviposition)—were simplified into four sub-stages: (1) egg hatching, (2) immature development (larvae+pupae), (3) pre-bloodmeal (blood feeding), and (4) oviposition. Only females modeled (assume sufficient males). Development depends on growing degree days (GDD) plus threshold constraints.

GDD framework: Daily mean temperature capped at 35°C; accumulation above a lower threshold (T_hr); negative GDD not allowed. Progression occurs when accumulated GDD reaches stage-specific requirements.

Stage parameters (from literature):

• Egg hatching: requires minimum daily temperature ≥14.59°C and cumulative 42.4 GDD since oviposition.
• Immature development (larva+pupa): baseline 11.78°C and 126.38 GDD to adult emergence.
• Blood feeding (pre-blood meal interval): temperature-dependent duration: <20°C: 4 days; >20°C: 2 days; >26°C: 1 day; 35°C: 2 days.
• Oviposition (gonotrophic cycle): temperature-dependent: <26°C: 8 days; >26°C: 3 days; >30°C: 2 days; >35°C: 4 days.

Thermal mortality thresholds:

• Cold-kill on eggs: if average daily temperature <0°C for 152 consecutive days, eggs die (limits overwintering).
• Heat-kill at any stage: any daily temperature >38°C for 1 day causes failure of development.

Precipitation constraint: Areas with annual rainfall <200 mm are excluded as too dry to support Ae. aegypti; sensitivity tested with 900 mm threshold.

Climate data: NASA NEX-GDDP daily minimum/maximum temperature and precipitation at 0.25° resolution, 1950–2100. Projections under RCP 4.5 and RCP 8.5. Ensemble mean of four GCMs (BCC-CSM1.1, MIROC-ESM-CHEM, IPSL-CM5A-LR, CCSM4) to capture inter-model variability. Sensitivity comparisons with ERA5 for overlapping period are provided in supplement.

Validation: Assumed ≥1 LCC necessary for establishment; higher LCC correlates with higher occurrence probability and abundance. Validation against ~40,000 global georeferenced occurrence records (2001–2010) using AUC (with pseudo-absences from other mosquito observations), Kappa (presence threshold 10 LCC), and country-level AUC in nations with >150 records. Also validated against 2011 abundance data from central Mexico across 10 villages spanning 0–2000 m elevation; correlated site-level LCC with adult abundance.

Analyses:

• Global spatial-temporal LCC trends for 1950–2050; summarized by continents, climate zones, and latitudinal bands using 5-year averages for the 1950s, 2000s, 2050s.
• Invasion frontiers: defined as the 2.5th percentile of LCC among occurrences (≥10 LCC/year). Tracked decadal frontier contours (1950, 1970, 1990, 2010, 2030, 2050) and estimated invasion speeds by minimum distances between successive contours along leading edges in China, USA, and Europe.
• Seasonality: Monthly LCC estimated per grid cell and summarized by 10° latitudinal bands from 40°S to 40°N. Seasonal Kendall trend tests and Sen slopes assessed temporal trends accounting for seasonality.

Software: R 3.6.1; code available on GitHub and Zenodo.

Model validation:

• Global AUC = 0.92; Kappa = 0.80 using presence threshold >10 LCC.
• 99.9% of occurrence records fall in locations with LCC ≥1.
• Country-level AUCs (n≥150 records) ranged from 0.63 (Malaysia) to 0.99 (Taiwan); lower in countries with homogeneous high suitability or sparse/biased observations (e.g., Cuba 0.47, Indonesia 0.49, India 0.55, Brazil 0.35).
• Abundance validation (Mexico 2011): Pearson r = 0.752, r^2 = 0.571, p = 0.011 between LCC and adult abundance.

Global trends in LCC:

• Mean global LCC increased from 7.08 (95% CI across GCMs 6.96–7.19) in the 1950s to 7.62 (7.42–7.82) in the 2000s, a ~7.0% (3.1–12.4%) rise.
• Projected additional increases by the 2050s: +17.1% (12.4–21.8%) under RCP 4.5 and +24.3% (18.5–30.0%) under RCP 8.5 relative to 2000s.
• Acceleration in suitability increase: from 1.5% (0.6–2.4%) per decade (1950s–2000s) to 3.2% (2.4–4.0%) per decade (2000s–2050s, RCP 4.5) and 4.4% (3.5–5.4%) per decade (RCP 8.5).
• Total change 1950–2050: +26.0% (22.1–29.9%) RCP 4.5; +33.8% (28.8–38.8%) RCP 8.5.
• Strongest gains in tropical regions (South East Asia, South America, West Central Africa); temperate regions show marked gains; arid, polar, boreal remain low.

Invasion frontiers (threshold ≥10 LCC):

• China: historical (1950–2000) frontier expansion ~1.58 km/yr (CI 1.41–1.75); projected (2000–2050) ~5.59 km/yr (5.20–5.98) with northeastward shift; by 2050 much of southeastern China ≥10 LCC under both RCPs.
• North America: historical ~2.29 km/yr (2.12–2.46); projected ~5.52 km/yr (5.23–5.81) by 2050, with expansion in southeastern and parts of western USA (especially under RCP 8.5).
• Europe: suitability remains patchy; southern margins (Iberian Peninsula, Italy, Greece) reach ≥10 LCC by 2030s (particularly under RCP 8.5); no continuous pan-European suitability by 2050.

Seasonality:

• LCC increases across all months with stronger effects under RCP 8.5.
• Higher latitudes (20–40° N/S) show large increases in duration of favorable periods (more favorable months per year), implying enhanced establishment potential outside the tropics.
• Equatorial regions experience strongest increases in peak LCC; middle latitudes (10–30° N/S) show more pronounced extension of favorable seasons.
• Seasonal Kendall trends show significant increases at all latitudes between 40°N and 40°S since 1950, with future rates accelerating toward the tropics and under higher emissions (e.g., at 0–10°S, Sen slope increases 2.5× under RCP 4.5 and 3.9× under RCP 8.5 for 2000–2050 vs 1950–2000).

The mechanistic phenology approach translates daily, lab-derived thermal performance and threshold responses into global predictions of Ae. aegypti development intensity over decades. Findings indicate that climate change has already increased LCC since 1950 and will further accelerate increases through 2050, especially under RCP 8.5. These changes enhance suitability within current ranges and expand it into historically marginal or uninvaded regions. The predicted acceleration of invasion frontiers in China and the USA (2.4–3.5× faster than historically) implies greater risk of establishment, including in areas with recent dengue activity (e.g., Guangdong/Guangzhou). Europe remains patchier but southern margins increasingly meet establishment thresholds. Seasonal analyses show both prolonged favorable periods at higher latitudes and increased peak intensities in the tropics, suggesting higher potential mosquito abundance and extended seasons conducive to establishment and, potentially, transmission. Together, results provide process-based evidence that ongoing and future warming will likely intensify Ae. aegypti invasion potential and development-driven components of arboviral risk, with trajectories diverging by emissions scenario.

This study introduces and validates a global, daily-resolution phenology model that estimates Ae. aegypti life-cycle completions as a mechanistic suitability index. It demonstrates significant historical increases and projected accelerations in development intensity, expanding seasonal windows, and advancing invasion fronts, particularly under higher emissions. These insights can support biosecurity, public health preparedness, and climate mitigation/adaptation planning. Future work should refine thresholds with additional experimental data (including responses to fluctuating temperatures), incorporate humidity and microclimate effects, account for dispersal and species interactions, and integrate phenology outputs into transmission models to link vector development to realized disease risk. Deeper emissions reductions would likely avert part of the projected increases in suitability.

The model may be conservative in some regions where historical presence was widespread (e.g., Mediterranean, parts of USA), potentially overestimating the LCC needed for establishment. LCC captures climatic development suitability but not all determinants of presence or transmission. Discrepancies can arise from factors not explicitly modeled: dispersal limitations; microclimates and behavioral thermoregulation; control interventions and environmental stochasticity; species interactions (competition/predation); additional environmental constraints (e.g., humidity); population or lineage variation in thermal tolerance/acclimation; and differences between fluctuating and mean thermal regimes affecting development and mortality. Precipitation is represented with a simple annual threshold, and heat/cold-kill thresholds are stylized. Suitability for vector development does not equate directly to transmission suitability; a disease transmission component is not included. Validation is strongest at broad scales; country-level performance varies with data coverage and environmental homogeneity.