Singapore Researchers Develop Climate-Virus Model to Forecast Dengue Risk Two Months Ahead

The Growing Challenge of Dengue in a Tropical Hub

Singapore, a bustling city-state in Southeast Asia, faces ongoing battles with dengue fever, a mosquito-borne viral illness transmitted primarily by the Aedes aegypti mosquito. Despite robust public health measures, the disease remains a persistent threat due to the country's warm, humid climate ideal for mosquito breeding. In recent years, outbreaks have surged, with over 32,000 cases reported in 2022 alone, straining healthcare resources and prompting heightened vigilance from authorities like the National Environment Agency (NEA).

Understanding dengue's dynamics requires grasping its four distinct serotypes—DENV-1, DENV-2, DENV-3, and DENV-4. Infection with one serotype offers lifelong immunity to that strain but only short-term protection against others, leading to risks of severe secondary infections known as dengue hemorrhagic fever or dengue shock syndrome. Singapore's hyperendemic status, where all serotypes circulate year-round, amplifies outbreak potential, especially when a dominant serotype shifts, exposing populations to new strains.

Breakthrough Model from Singapore's Research Powerhouses

Researchers affiliated with Singapore's premier institutions, including the National University of Singapore (NUS) and Nanyang Technological University (NTU), have spearheaded a groundbreaking climate-virus model to forecast dengue outbreaks up to two months in advance. Published in Nature Communications, this study integrates over two decades of weekly case data with climate variables and serotype surveillance, offering unprecedented predictive power.

Led by international collaborators but anchored by local experts like Associate Professor Lee-Ching Ng from NEA, NTU's School of Biological Sciences, and NUS's Saw Swee Hock School of Public Health, the model represents a pinnacle of Singapore's higher education contributions to public health. Chia-chen Chang from NUS's Department of Biological Sciences and Shuzhen Sim from NEA also played pivotal roles, blending academic rigor with practical surveillance data.

Decoding Climate's Role in Dengue Transmission

The model's foundation lies in dissecting climate's non-linear influence on dengue. Key drivers include maximum temperatures peaking around 32°C, which optimize mosquito development; intermediate dryness (about 30 rain-free days over three months), fostering breeding sites in water-holding containers; and negative Niño 3.4 Sea Surface Temperature Anomalies (SSTA), signaling La Niña conditions that boost transmission. These factors interact with lagged effects—up to 20 weeks—creating windows of heightened risk.

Singapore's urban landscape exacerbates this: high-rise housing, dense populations, and imported cases via travel sustain cycles. The model employs a Bayesian hierarchical mixed-effects framework with a negative binomial likelihood, accounting for overdispersion and random effects for weekly seasonality and yearly trends. High-performance computing from partners like Barcelona Supercomputing Center enables rapid simulations.

Serotype Competition: The Immunity Puzzle Solved

Beyond climate, serotype dynamics are crucial. The model uses 'time since dominant serotype switch' as a proxy for population immunity shifts. Risk spikes in the first two years post-switch (novel exposure), dips between years 2-6 (building cross-protection), and rises again after six years (waning immunity). This captures interannual variability better than climate alone, reducing unexplained variance by incorporating surveillance data from NEA, where ~27% of cases are genotyped since 2006.

Photo by K8 on Unsplash

• DENV-2 dominated early 2000s outbreaks (e.g., 2004-2005 peaks).
• Shifts to DENV-1/3 in 2013-2016 and DENV-3 resurgence fueled recent surges.
• Serotype inclusion boosts model skill by 6% over climate-only (60% vs. 54% improvement over baseline).

Forecasting Mechanics: From Data to Actionable Alerts

Predictions use time-series cross-validation: trained on data up to forecast issue minus lead time (up to 8 weeks), employing posterior simulations for probabilistic outputs. Covariates are lagged and scaled—e.g., 8-week averages for 4-week horizons. Continuous Ranked Probability Score (CRPS) shows 32% improvement at 8 weeks; outbreak detection hits 92% true positives with 2.1% false alarms (AUC 98%).

This outperforms baselines and rivals NEA's LASSO model, enabling ensemble use. For instance, it flagged 2020/2022 peaks early, despite underpredicting amplitudes due to unmodeled vectors.

Real-World Validation and Averted Cases

Retrospective analysis (2009-2022) confirms fidelity: median predictions track observed cases closely. Counterfactuals for Project Wolbachia—releasing Wolbachia-infected mosquitoes to suppress Aedes—reveal impacts. Retraining to mid-2022, the model estimates 13,491 median cases without intervention vs. actual lower numbers, averting 3,789 (28%, 95% PI: 994-18,659) in 2023 amid expanded releases.

Details in the full study.

Singapore Universities: Hubs of Infectious Disease Innovation

NUS and NTU exemplify Singapore's investment in higher education for health security. NUS's Saw Swee Hock School of Public Health trains epidemiologists, fostering talents like Lee-Ching Ng, who bridges academia and policy. NTU's Biological Sciences supports serotype genomics, aligning with national priorities like Research, Innovation and Enterprise 2025 (RIE2025).

These institutions collaborate via A*STAR and NEA, producing tools like this model. Their work extends to Wolbachia trials (77% efficacy) and genomic surveillance, positioning Singapore as a dengue control leader.

Challenges Ahead: Climate Change and Urban Pressures

Despite successes, hurdles persist. Low population immunity to all serotypes heightens vulnerability; 2026 cases remain low (348 YTD as of mid-March) but vigilance is key. Climate projections warn of fewer mid-year cases with hotter temperatures but intensified peaks. Model limitations—no vector data, smoothed serotypes—suggest enhancements like susceptibility mapping.

Photo by Angelyn Sanjorjo on Unsplash

• Integrate mosquito indices for finer resolution.
• Ensemble with machine learning for hybrid forecasts.
• Expand to neighborhood scales via spatial models.

Global Lessons from Singapore's Academic Excellence

This model offers a blueprint for dengue-endemic regions. Southeast Asia reports millions annually; similar approaches could avert outbreaks elsewhere. Singapore's universities demonstrate how targeted research translates to policy—NEA integrates forecasts for vector control, public alerts.

Future: AI enhancements, real-time serotyping. For more on Singapore's dengue efforts, visit NEA's dashboard.

The Road Forward: Sustaining Momentum in Research

Singapore's higher education sector must continue prioritizing interdisciplinary training. Programs in computational epidemiology at NUS/NTU equip graduates for global challenges. As climate shifts intensify dengue risks, such innovations ensure resilience, blending academia, government, and tech for healthier futures.