Dengue Waves: What Links the Fever to the Ocean and the Foliage
Context
Dengue has been endemic in Costa Rica since 1993, and outbreaks have recurred every year since, but their rhythm varies across the country. Alongside this work, other studies view the city as an organism and distinguish epidemics not by place but by time. The present study fits into this vein but takes not one city or one variable, but 32 cantons and six satellite indices, to see how dengue breathes on a national scale.
Methods
From January 2000 to December 2019, weekly dengue incidence data were collected for 32 cantons, aggregated monthly, square-root transformed, and standardized. For each series, a wavelet power spectrum was computed using the Morlet wavelet to identify dominant periods. To link with climate, wavelet coherence and phase shifts were used between incidence and indices: EVI, NDVI, NDWI, daytime and nighttime land surface temperature, TNA, and four El Niño indices. Significance was assessed using 1000 bootstrap series.
Results
Dominant periods of dengue across all 32 cantons fall within bands of 1, 2, and 3 years, with 1- and 2-year cycles prevailing before 2008, and the 3-year cycle strengthening after 2008. Vegetation indices EVI, NDVI, and NDWI correlate significantly with incidence on the annual cycle in cantons of the central and northern Pacific regions, with a phase shift of about three months. Surface temperature correlates with dengue in all cantons, but signals are in antiphase: the incidence peak lags the temperature minimum by 3–6 months. TNA is linked to dengue at periods of 1 and 3 years, especially after 2011. El Niño indices 3, 3.4, and 4 show strong correlation on the 3-year period in most cantons.
Implications
This work stands out from neighboring studies by being the first for Costa Rica to combine six disparate satellite and climatic variables and cover all cantons relevant to health authorities. While adjacent studies focus on a single city or on temporal structure, here it is shown that even within one country, significant variables differ by region: vegetation matters on the Pacific coast, while oceanic indices act everywhere. This underscores the need for local, not national, forecasting approaches.
Future development
The topic could evolve by adding socioeconomic variables and cluster analysis of cantons to identify groups with similar conditions. A shift from correlations to mechanistic models that include explicit biological processes is possible: mosquito reproduction, virus incubation period, population immunity. Wavelet results set the time scales that such models must reproduce.
Impact
The results directly affect epidemiological surveillance and resource allocation in Costa Rica, and serve as a template for other tropical countries with non-stationary dengue dynamics.
Next steps
The immediate next step is to include precipitation and humidity in the analysis, and to test the robustness of the associations on longer series after 2019.
Key open problems
In dengue epidemiology, an open question remains: why do some regions show a dominant annual cycle while others show multi-year cycles, and how exactly do climatic variables interact with population immunity? This work does not solve the problem but provides a map of local correlations that narrows the hypothesis space for future mechanistic models.
Wavelet analysis, originally created for processing seismic signals, now helps epidemiologists see how dengue fever pulses in time with ocean currents.
Key numbers
- cantons: 32
- years of observation: 20
- dominant periods: 1, 2, 3 years
- vegetation phase shift: 3 months
- temperature phase shift: 3–6 months