Coherence of dengue incidence and climate in the wet and dry zones of Sri Lanka

https://doi.org/10.1016/j.scitotenv.2020.138269Get rights and content

Highlights

  • Dynamics of dengue disease in wet and dry regions of Sri Lanka are different

  • Different periodicities of 6 and 12 months in the epidemic patterns of two regions

  • Periodicities in rainfall and dengue incidence are in agreement

  • Lag weather variables make little contribution in predicting dengue outbreaks

  • Lag times in epidemic response differ in wet and dry regions

Abstract

We studied the dynamics of dengue disease in two epidemic regions in Sri Lanka, the densely populated Colombo district representing the wet zone and the relatively less populated Batticaloa district representing the dry zone. Regional differences in disease dynamics were analysed against regional weather factors. Wavelets, Granger causality and regression methods were used. The difference between the dynamical features of these two regions may be explained by the differences in the climatic characteristics of the two regions. Wavelet analysis revealed that Colombo dengue incidence has 6 months periodicity while Batticaloa dengue incidence has 1 year periodicity. This is well explained by the dominant 6 months periodicity in Colombo rainfall and 1 year periodicity in Batticaloa rainfall. The association between dengue incidence and temperature was negative in dry Batticaloa and was insignificant in wet Colombo. Granger causality results indicated that rainfall, rainy days, relative humidity and wind speed can be used to predict Colombo dengue incidence while only rainfall and relative humidity were significant in Batticaloa. Negative binomial and linear regression models were used to identify the weather variables which best explain the variations in dengue incidence. Most recent available incidence data performed as best explanatory variables, outweighing the importance of past weather data. Therefore we recommend the health authorities to closely monitor the number of cases and to streamline recording procedures so that most recent data are available for early detection of epidemics. We also noted that epidemic responses to weather changes appear quickly in densely populated Colombo compared to less populated Batticaloa. The past dengue incidence and weather variables explain the dengue incidence better in Batticaloa than in Colombo and thus other exogenous factors such as population density and human mobility may be affecting Colombo dengue incidence.

Introduction

Dengue is a widely spread vector-borne virus infection which has no explicit treatment. The virus is spread by female mosquitos of Aedes species (Wilder-Smith and Gubler, 2008). There are four serotypes of the dengue virus known as DENV1, DENV2, DENV3 and DENV4. Infection with one serotype provides immunity against that particular serotype but not against others, and therefore humans can be infected more than once. The most common symptom is high fever from which many patients recover within about a week, but occasionally the disease can develop into severe stages of dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS) to which a few victims succumb (WHO, 2019). About 3.9 billion people in 128 counties around the world including all the tropical and subtropical regions are exposed to the risk of dengue disease (Brady et al., 2012). And Sri Lanka being a tropical island in the Indian Ocean, has had no escape from the escalating tide. In 2017, Sri Lanka experienced the worst ever dengue disease outbreak in the country's history with >175,000 reported infections and 400 fatalities (Gunatilleke and Mudugamuwa, 2017). Emergence of DENV 2 serotype which was previously uncommon in Sri Lanka and heavy monsoon rains and floods have contributed to the severe outbreak in 2017 (Ali et al., 2018). It was brought under control by early 2018 through community and government initiated interventions such as eradication of mosquito breeding sites, use of mosquito repellents during daytime and fumigation of high risk areas (Ali et al., 2018). In 2018, the number of reported cases was approximately 50,000, which was still a substantial 0.23% of the country's 21.5 million population.

The tropical climate in Sri Lanka is favourable to dengue transmission and dengue cases are reported from almost all parts of the country throughout the year. However the regional climate varies widely. Mean annual temperature ranges from 27 °C in coastal lowlands to 16 °C in NuwaraEliya at 1900 m altitude and average annual rainfall varies from 900 mm in the driest areas to over 5000 mm in the wettest areas. Two monsoons coming from south-west during May to September and from north-east during December to February are the main sources of rain. Previous research has shown that the country has two epidemic peaks which coincide with the two monsoons and that the outbreaks occur when the local environmental factors are favourable for mosquito breeding and survival (Sirisena and Noordeen, 2014; Sun et al., 2017; Prabodanie et al., 2020).

The impact of weather conditions on dengue disease prevalence has been studied at the regional scale around the world and different results have been observed. In Davao, Philippines the relative risk of dengue was high with moderate rainfall, peaked at 32 mm rainfall, and decreased with further increase in rainfall (Iguchi et al., 2018). Lowe et al. (2018) found that excess rainfall lagged 1–2 months had a positive impact on dengue risk in Barbados. Rainfall lagged 0–3 months had a significant influence on dengue incidence in Cambodia (Choi et al., 2016). In India, the highest risk of dengue occurred at 60 mm rainfall lagged 12 weeks (Kakarla et al., 2019). Rainfall lagged 2 and 5 months had significant positive correlations with dengue incidence in Sri Lanka (Sirisena et al., 2017).

Mordecai et al. (2017) studied the general effect of temperature on the transmission of dengue, chikungunya and zika viruses using mechanistic models. They estimated that transmission of all three viruses occurs between 18 and 34 °C and peaks in the range 26–29 °C. In Davao, Philippines, the relative risk of dengue was high at 26 °C and lower within the range 27 °C to 31 °C (Iguchi et al., 2018). In Guangzhou, China, temperatures below 28 °C lagged 7–14 days had a significant positive impact on dengue incidence (Wu et al., 2018). Kakarla et al. (2019) found that dengue risk in India tends to rise when the mean temperature exceeds 24 °C and peaks at 30 °C with 0–3 weeks lag. Severe dengue cases in north-eastern and southern parts of Thailand peaked between 24 °C and 30 °C (Xu et al., 2019a).

Relative humidity also has a significant impact on dengue (Sirisena et al., 2017; Tuladhar et al., 2019; Xu et al., 2019a; Xu et al., 2019b). In Thailand, dengue risk was high between relative humidity 75% and 85% (Xu et al., 2019a). In Guangzhou, China, relative humidity up to 76% lagged 7–14 days had a significant positive impact on dengue incidence (Wu et al., 2018). Relative humidity was found to be the key factor associated with the second epidemic peak in Philippines (Xu et al., 2019b). The time delay in humidity-dengue association is usually low (0–2 weeks) (Wu et al., 2018; Tuladhar e al., 2019), but Sirisena et al. (2017) observed that the delay may be 9 months in Sri Lanka. Wind speed has also been considered as a factor affecting dengue incidence, particularly in monsoon-fed regions (Sedda et al., 2018; Mala and Jat, 2019). Mala and Jat (2019) observed that dengue incidence in Delhi, India increased as wind speed increased from 4 km/h to 6 km/h, and decreased with further increase in wind speed. Sedda et al. (2018) and Withanage et al. (2018) found negative associations between wind speed and dengue incidence in Southeast Brazil and in Sri Lanka, respectively.

Previous research has provided evidence of significant spatial variations in weather-dengue associations even within Sri Lanka. In Gampaha district, rainfall and rainy days lagged 3 months, relative humidity lagged 2 months and dengue cases squared lagged 1 month were the best predictors of monthly dengue incidence (Withanage et al., 2018). Liyanage et al. (2016) found that dengue risk in Kalurara district of Sri Lanka is strongly associated with 300 mm or above rainfall lagged 6–10 weeks. Wickramaarachchi et al. (2015) observed 25 weeks cycles in the time series of dengue cases form Colombo Municipal Council area and concluded that the epidemic cycles are driven by rainfall. Those regional studies have some limitations. First, the validity of their outcomes is limited by monthly averaged data sets of relatively small size. Second, no recent study has focused on Colombo district in entirety, which is the most affected district with the highest number of dengue cases reported every year. Third, most regional studies have focused on the wet climatic zone and no study has compared the weather-dengue associations in wet and dry zone districts. Hence potential differences between the epidemic patterns and weather-dengue associations in wet and dry districts are not known.

In this paper, we investigate the periodic patterns of dengue epidemics and the impact of local weather conditions on the dengue epidemic patterns in two districts in Sri Lanka, Colombo and Batticaloa. Colombo is the economic and political capital of the country, providing residence to approximately 12% of the island's population (Department of Census and Statistics, 2019) and about two million people, i.e. approx. 10% of the entire population, commute to Colombo daily (Kantharvel, 2017). An acceleration of dengue cases has been reported from eastern district Batticaloa since 2016; and in 2018, third highest number of dengue cases have been reported from Batticaloa preceded by Colombo and its close neighboured Gampaha district (Epidemiology Unit, 2019). Among these three hot spots, densely populated Colombo and Gampaha are in the wet climatic zone receiving 2000–5000 mm annual rainfall while relatively less populated Batticaloa is in the dry climatic zone receiving 700–1000 mm annual rainfall (IUCN, 2011). Owing to these factors, we found it worth comparing the dynamics of dengue disease in Colombo and Batticaloa districts which represent the wet and dry zones in the country.

The main objective of this paper is to reveal (potentially different) dengue disease dynamics in the two regions Colombo and Batticaloa and to explain the differences in the dynamics. In addition, identifying the variables which best explain and predict the dengue incidence in each district was a secondary objective. We used weekly dengue incidence and weather data over a period of 9 years from 2009 to 2017.We first studied the periodicities and associations in dengue and weather data in the two districts using wavelet analysis. Statistical methods, mainly Granger causality, negative binomial regression and linear regression analysis methods were used to explain the variations in dengue dynamics across the two regions.

Section snippets

Methods and data

This study is focused on two districts in Sri Lanka, Colombo and Batticaloa. Colombo district occupies an urban landscape of 676 km2 in the western coastal belt of Sri Lanka. Approximately 2.5 million residents live in Colombo (Department of Census and Statistics, 2019). Colombo belongs to the wet climate zone and receives about 2400 mm annual rainfall (Department of Census and Statistics, 2018). Batticaloa is a relatively rural district occupying 2610 km2 area in the eastern coastal area. The

Wavelet analysis

The wavelet power spectrum of a time series illustrates the distribution of its power over various frequency (or periodic) bands. The wavelet spectrum of Colombo dengue incidence series (Fig. 2-a) shows that Colombo dengue incidence has a significant 6 months periodicity (epidemic waves of 6 months length; the red band at period 0.5 in the figure). The wavelet spectrum of rainfall series (Fig. 2-b) also displays 6 months periodicity from 2009 to 2013 and from 2015 to 2017. Wavelet coherence

Discussion

The periodic pattern of dengue incidence in each district is in agreement with the periodic pattern of rainfall and rainy days. While a 6 month periodic component was dominant in the Colombo rainfall pattern, a 1 year component dominated the Batticaloa rainfall pattern. This is reasonable as Colombo usually receives heavy rainfall from May to September (south western monsoon rains) and from November to December (second inter-monsoon rains), while Batticaloa, on the other hand, receives most of

Conclusions

We studied the dynamics of dengue disease in two epidemic regions in Sri Lanka, Colombo and Batticaloa. The two regions have different climates and demographic characteristics and we studied how such differences explain the differences in disease dynamics. Wavelets and statistical analysis methods were used.

The 6 month periodicity in Colombo dengue incidence and 1 year periodicity in Batticaloa dengue incidence are well explained by the similar periodic patterns in rainfall and rainy days.

Acknowledgements

This work was carried out at RMIT University. The first author was supported by the Endeavour Research Fellowship awarded by the Government of Australia. Lewi Stone was supported by the Australian Research Council [Grant number: DP 150102472].

References (43)

  • B. Cazelles et al.

    Wavelet analysis of ecological time series

    Oecologia

    (2008)
  • B. Cazelles et al.

    Wavelet analysis in ecology and epidemiology: impact of statistical tests

    J. R. Soc. Interface

    (2014)
  • Y. Choi et al.

    Effects of weather factors on dengue fever incidence and implications for interventions in Cambodia

    BMC Public Health

    (2016)
  • T. Craciunescu et al.

    Causality detection methods applied to the investigation of malaria epidemics

    Entropy

    (2019)
  • D.A. da Cruz Ferreira et al.

    Meteorological variables and mosquito monitoring are good predictors for infestation trends of Aedes aegypti, the vector of dengue, chikungunya and Zika

    Parasit. Vectors

    (2017)
  • E.A.P. de Almeida Costa et al.

    Impact of small variations in temperature and humidity on the reproductive activity and survival of Aedes aegypti (Diptera, Culicidae)

    Revista Brasileira de Entomologia

    (2010)
  • Department of Census and Statistics

    Census of Population and Housing - Final Report

    (2012)
  • Department of Census and Statistics

    Annual and monthly rainfall at observation stations 2012–2017

  • Department of Census and Statistics

    Midyear population estimates by district and sex

  • Epidemiology Unit

    Weekly Epidemiological Report

  • C.W.J. Granger

    Investigating causal relations by econometric models and cross-spectral methods

    Econometrica

    (1969)
  • Cited by (7)

    • Impact of environmental factors on the spread of dengue fever in Sri Lanka

      2022, International Journal of Environmental Science and Technology
    View all citing articles on Scopus
    View full text