Over the past decade the number of reported cases of dengue fever and dengue haemorrhagic fever (DHF), worldwide, has increased. This has been widely attributed to an expansion in the geographic distribution of the two main vectors - Ae. aegypti and Ae. albopictus. A link between climate and dengue/DHF has been firmly established.
However, most studies have been limited in at least one of three ways - examining the relationship between dengue/DHF incidence and one climate variable, being restricted to the analysis of one epidemic event, or adopting an aggregated national or global perspective which allows for no regional or temporal variation. This study addresses these three perceived shortfalls.
Focusing on eight provinces within Indonesia, data for temperature, rainfall, rainfall anomalies, relative humidity and the SOI have been examined in relation to provincial dengue/DHF incidence from the period 1992 to 2001, which includes two ENSO events. The provinces selected were representative of different groups of provinces defined using the 10-year dengue/DHF incidence profiles of all 27 provinces. Mean and peak yearly provincial incidence rates, provincial population densities and climate data were used to ensure that the selection procedure included a range of incidence, demographic and climatic situations.
Significant Pearson correlations were observed between dengue/DHF incidence and at least one climate variable in each province (r= +/−0.2 to +/−0.43; p<0.05). Multiple regression analyses showed that between 10% and 25% of the variance in dengue/DHF incidence was explained by two or three climate variables in each province (p<0.1 to 0.01). Temperature showed the greatest number of significant Pearson correlations (consistently positive) and generally explained the greatest amount of variance in most provinces (R2=up to 45.4%; p<0.01 to p<0.1). However, in those provinces with a strongly seasonal climatic regime, and during ENSO events in other provinces, rainfall exceeded temperature in the number of significant correlations and variance explained (R2=up to 49.7%; p<0.01 to p<0.1). Thus, although temperature was critical for the intensity of outbreaks, regional rainfall was more closely associated with different provincial patterns and was, therefore, crucial in determining their timing.
Three groups were identified which represented very different regional rainfall regimes and their corresponding dengue/DHF dynamics. A strongly seasonal incidence pattern was associated with regular seasonal increases in rainfall and little climate variability. A highly epidemic pattern of incidence was associated with increased rainfall and high susceptibility to climate variability under ENSO. A third group, exhibiting both seasonal and epidemic incidence, was associated with decreased rainfall and little or no relation of epidemics to ENSO events. This information may assist in predicting high-risk periods for increased dengue/DHF incidence in different regions, thereby providing the opportunity to implement effective control strategies at the appropriate time and place.