Geophysical Research Letters

Dual peak cholera transmission in Bengal Delta: A hydroclimatological explanation

Authors


Abstract

[1] Cholera has reemerged as a global killer with the world witnessing an unprecedented rise in cholera infection and transmission since the 1990s. Cholera outbreaks across most affected areas show infection patterns with a single annual peak. However, cholera incidences in the Bengal Delta region, the native homeland of cholera, show bi-annual peaks. The mechanisms behind this unique seasonal dual peak phenomenon in cholera dynamics, especially the role of climatic and hydrologic variables, are not fully understood. Here, we show that low flow in the Brahmaputra and the Ganges during spring is associated with the first outbreaks of cholera in Bangladesh; elevated spring cholera outbreaks are seen in low discharge years. Peak streamflow of these rivers, on the other hand, create a different cholera transmission environment; peak flood volumes and extent of flood-affected areas during monsoon are responsible for autumn cholera outbreaks. Our results demonstrate how regional hydroclimatology may explain the seasonality and dual peaks of cholera incidence in the Bengal Delta region. A quantitative understanding of the relationships among the hydroclimatological drivers and seasonal cholera outbreaks will help early cholera detection and prevention efforts.

1. Introduction

[2] Cholera, an acute water-borne diarrheal disease caused by the bacterium Vibrio cholerae, has reemerged as a global killer in recent decades. The world has seen an unprecedented rise in cholera infection and transmission since the 1990s, which have become a major public health concern for the World Health Organization (WHO) [Collins, 2003]. Cholera records typically show single annual peaks in most affected areas, such as parts of Africa, South-east Asia and Latin America [Hashizume et al., 2008] (Figure 1a). However, cholera incidence records for the Ganges-Brahmaputra-Meghna (GBM) delta region of the Indian subcontinent show distinct bi-annual peaks [Lipp et al., 2002; Akanda et al., 2007]. In this outbreak pattern, the first peak of the year occurs in the spring (March–May) season and a larger second peak occurs in autumn (September–December). No other region outside the Indian subcontinent shows a similar bi-annual cholera incidence pattern.

Figure 1.

(a) Monthly climatology of recorded cholera incidences in four different regions of the world. (b) Cholera incidence (% infected among all patients), recorded and detrended, during 1980–2000 at ICDDR, Bangladesh. (c) Autocorrelation function of monthly cholera incidence.

[3] The seventh cholera pandemic, which started in 1961 in Indonesia, has been reported in over 50 countries. The global pattern and magnitude of the pandemic suggests that cholera outbreaks primarily originate in coastal environments and then spread inland through secondary means [Colwell, 1996; Mouriño-Pérez, 1998]. The cholera-coastal connection is usually explained by the fact that cholera is caused by two particular pathogenic strains of the bacterium Vibrio cholerae, which are found mainly in marine plankton [Epstein, 1993; Colwell and Huq, 2001]. Phytoplankton and zooplankton, by serving as the primary sources of nutrients and also physical carriers of the bacteria, play an important role in facilitating the survival, multiplication, and transmission of Vibrio cholerae in the natural aquatic environment [Lipp et al., 2002; Mouriño-Pérez, 1998].

[4] Cholera epidemics in Bangladesh have been historically linked to a range of environmental and climate variables including precipitation [Pascual et al., 2002; Hashizume et al., 2008], floods [Koelle et al., 2005], peak river level [Schwartz et al., 2006], sea surface temperature (SST) [Lobitz et al., 2000; de Magny et al., 2008], sea surface height [Lobitz et al., 2000], coastal salinity [Miller et al., 1982], and fecal contamination [Islam et al., 2006]. None of these studies, however, successfully quantified the role of the seasonal hydroclimatological processes with bi-annual cholera incidences in Bangladesh.

[5] A key purpose of this paper is to show that the hydroclimatology of the GBM basin explains the bi-annual peaks of cholera incidence in Bangladesh. Our analyses provide a plausible explanation of the hydroclimatological processes causing the first cholera outbreaks in a given year, when low flow volumes in the GBM Rivers during spring can help seawater intrusion thus enabling transport of coastal plankton into the estuarine southern districts of Bangladesh (Figure 2 inset). Higher flow volumes from these rivers and the extent of inundated floodplains, on the other hand, cause a bigger second cholera peak in the autumn season.

Figure 2.

Monthly climatology of streamflow volumes (m3/s) of the Ganges, the Brahmaputra, and the rivers combined for 1958–2007. (Inset map: CIA World Factbook)

[6] The prevalence of Vibrio cholerae bacterium in brackish estuarine waters and the initial outbreaks of cholera near coastal areas link the initiation and transmission of this disease with coastal ecosystems [Colwell, 1996; Collins, 2003; Worden et al., 2006]. If low and high flows were the two possible drivers for dual peak cholera incidences in Bangladesh, then, one would expect to see a spatial signature of such a process with the first cholera outbreaks occurring in coastal areas and the second peak in a wider geographical area. In the case of GBM basin region, Bouma and Pascual [2001] provide strong evidence of such a coastal link to cholera outbreaks in Bengal, showing endemic outbreaks in spring in coastal districts and epidemic outbreaks during autumn in regions situated further inland, suggesting roles of large physical events such as floods. Sack et al. [2003], focusing on the evolution of cholera outbreaks in four locations in Bangladesh, found that spring outbreaks are more common in the two locations closer to the coast, while the other two remote locations are affected mostly by the autumn outbreak. Similar spatial progression of cholera outbreaks, diffusing through coastal rivers and water networks in South Africa, has been shown by Bertuzzo et al. [2008].

[7] We hypothesize that the pre and post monsoon cholera peaks are governed by two distinctly different hydroclimatological drivers. To make this point, we disaggregate cholera incidence data from Bangladesh into seasonal components and analyze those with corresponding seasonal GBM streamflow and Bay of Bengal (BoB) SST. More specifically, the paper attempts to answer the following two questions: What is the role of spring low flow volumes in the GBM Rivers on the first cholera outbreaks in the region? How high monsoon streamflow volumes and subsequent flooding impact the post-monsoon outbreaks?

2. Data and Background

[8] Daily river discharge data for the GBM Rivers were obtained from the Bangladesh University of Engineering and Technology. The Reynolds 1°×1° SST database [Reynolds and Smith, 1994] was used to extract SST information for the coastal zone of the GBM Rivers in BoB (areas north of 21°N) based on available bathymetry and active production zone information [Chamarthi et al., 2008]. The Flood Affected Area (FAA) time series, showing the annual extent of flooded land area, was obtained from Bangladesh Water Development Board.

2.1. Hydroclimatology of the Ganges-Brahmaputra-Meghna Basin

[9] The GBM river system, one of the largest freshwater flow regimes in the world, shows a strong seasonal pattern (Figure 2). The GBM system is formed by two of the largest Himalayan Rivers, the Ganges and the Brahmaputra, which are joined by the Meghna in Bangladesh (Figure 2 inset). Most of the annual precipitation in this basin occurs only during four monsoon months (Jun–Sep). The region thus has a contrastingly dry low flow season compared to its typically wet rainy reason. The lowest flows in the Brahmaputra and the Ganges are recorded during January–April and are typically one-tenth and one-twentieth of the average peak flows in respective rivers. Such drastic reductions cause a drop in the hydraulic head of the GBM system and its tributaries, which can accelerate saltwater intrusion from the coast towards the inland freshwater resources [Rahman et al., 2000]. Recent studies on coastal ecosystems of Bangladesh have pointed out significant salinity increases during spring [Islam and Gnauck, 2008; Wahid et al., 2007]. Miller et al. [1982] suggested strong links between coastal salinity and cholera outbreaks in Kolkata and London. Louis et al. [2003] and Vital et al. [2007] have found increased Vibrio cholerae population in water with brackish salinity conditions.

[10] The rivers begin to rise rapidly due to monsoon rainfall starting from June and reach peak flow levels during the months of August and September. The entire country is situated on alluvial floodplains and thus prone to inundation by the overflowing of the GBM. On an average, about 20 percent land area of Bangladesh is inundated every year, with as much as 60 percent in high flood years, such as in 1988 and 1998 [Chowdhury and Ward, 2007]. As open mixing of water between sewers, exposed drains, reservoirs, and rivers is very common during floods, the submerged areas quickly become contaminated with Vibrio cholerae through other infected water sources. Islam et al. [2006] found an astounding 62.5% water samples carrying the cholera bacteria in the suburban reservoirs around Dhaka during the 2004 floods. Although river levels fall rapidly from September through November, water levels on adjoining flood plains fall more slowly because of low gradients, congested drainage, and substantial depression areas. Some areas stay submerged until December–January [Mirza et al., 2001], which can serve as ideal habitats for Vibrio cholerae even after the flood recedes and act as conduits of transmission within the surrounding population [Islam et al., 2006].

2.2. Seasonality of Cholera Incidences in Bangladesh

[11] The Bengal Delta with its extensive estuary formed by the Ganges and the Brahmaputra rivers has been considered the native homeland of cholera since the early 19th century [Bouma and Pascual, 2001]. The cholera surveillance program at the International Center for Diarrhoeal Disease Research, Bangladesh (ICDDR,B) provides some of the longest and largest records available in the world. The on-going program carries out a systematic sub-sampling of all patients visiting the hospital, which serves as the main treatment center for the most concentrated population center in Bangladesh.

[12] The cholera incidence climatology, constructed by averaging the monthly cholera records of the 1980–2000 time series, exhibits significant seasonal and inter-annual variability (Figures 1a and 1b). A closer look at the above reveals that the beginning of the spring cholera outbreaks coincides with the low flow season of the GBM Rivers [Akanda et al., 2007]. Monthly cholera numbers decrease in high monsoon and peak streamflow months (June through September). However, cholera infection starts increasing again in later monsoon, and a larger second peak is observed during the early winter months of November and December. Our analysis focuses on understanding the roles of the regional hydroclimatology behind these two seasonal peaks observed in cholera incidences, and the dominant processes behind each.

3. Methodology and Results

3.1. Seasonal Flow Versus Cholera Analysis

[13] We separated out the mean cholera incidences of the two seasons, spring (MAM: Mar–Apr–May) and autumn (OND: Oct–Nov–Dec) and created two new time series to examine interannual variability of cholera. These seasonal outbreaks are analyzed with the Ganges and Brahmaputra combined flow and Bay of Bengal SST. Mean streamflow values for the two lowest flow months, February and March, are combined to develop a low flow time series, and the two highest flow months (August and September) to develop the high flow time series.

[14] We have performed autocorrelation analysis for both the seasonal and the original monthly time series to validate the presence of two peaks in a year. Figure 1c on monthly autocorrelation function values shows the presence of two peaks in a year. The observed two peaks in the monthly climatology occur approximately in the months of May and November. The lag-1 autocorrelation value for the spring (MAM) and autumn (OND) time series are 0.55 and 0.22, respectively. In addition, the low flow values and pre-spring (DJF: Dec–Jan–Feb average) SST, show similar autocorrelation values (∼0.50) as spring (MAM) cholera incidences, and high flow values and (JJA: Jun–Jul–Aug average) SST show similar autocorrelation values (∼0.20) as autumn (OND) cholera incidences. Taken together, these values suggest the role of two separate hydroclimatological processes behind the two seasonal cholera peaks, with different temporal autocorrelation structures.

[15] Seasonal cross-correlation values are calculated between low flow values and spring cholera incidences; and for high flow values and autumn cholera incidences for the period 1980–2000 (Table 1). We also use the seasonal mean (e.g., DJF, JFM, FMA) coastal BoB SST information for a similar seasonal cross-correlation analysis with spring and autumn cholera incidences, respectively. In addition to the calculated Pearson correlation coefficients, we perform non-parametric Kendall Tau significance tests on and report the corresponding Kendall Tau-a correlation and statistical significance information in the paper.

Table 1. Correlation Matrix for Low and High Seasonal Streamflow, Flood Affected Area (FAA), Bay of Bengal (BoB) SST, and Seasonal (MAM and OND) Cholera Incidencesa
 Cholera MAMCholera ONDLow FlowHigh FlowFAADJF SSTJJA SST
  • a

    DJF, Dec–Jan–Feb; MAM, Mar–Apr–May; JJA, Jun–Jul–Aug; OND, Oct–Nov–Dec; Low (Feb–Mar) and High (Aug–Sep) Flow: Ganges & Brahmaputra combined. Values are given as Pearson and (Kendall Tau in parenthesis) coefficients.

  • b

    (highly significant) for p < 0.01.

  • c

    (significant) for 0.01 < p < 0.05.

Cholera MAM-      
Cholera OND0.68 (0.50)b-     
Low Flow−0.65 (−0.44)b−0.34 (−0.23)-    
High Flow0.15 (0.08)0.55 (0.34)c−0.43 (−0.29)-   
FAA0.67 (0.40)c0.97 (0.86)b−0.53 (−0.36)c0.86 (0.46)b-  
DJF SST−0.69 (−0.32)−0.21 (−0.01)0.48 (0.22)−0.19 (−0.01)−0.23 (−0.01)- 
JJA SST0.11 (0.01)0.72 (0.37)−0.20 (−0.09)0.53 (0.27)0.61 (0.40)c−0.14 (−0.04)-

[16] We find strong negative cross-correlation (Table 1, r = −0.65 and −0.44, p < 0.01) between seasonal combined low flow volumes for the Ganges and the Brahmaputra and average spring (MAM: March–May) cholera incidences. An implication of these results is that more spring cholera outbreaks are likely in a drought year than in an average flow year. Peak streamflow volumes cause substantial inundation along the major riverbanks of Bangladesh during monsoon, and usually lead to large-scale contamination of water systems such as rivers, canals, and ponds [Schwartz et al., 2006]. Autumn cholera incidence (OND: Oct–Dec) values are found to be strongly correlated to high flow volumes (r = 0.55 and 0.34, p < 0.05), and also to flood inundation extent (r = 0.97 and 0.86, p < 0.01), as seen in Table 1. In summary, excess availability or the extreme lack of flow, both impact cholera dynamics in Bangladesh.

[17] Our results show complementary evidence of two separate hydroclimatological processes within the context of coastal SST variability in the BoB and seasonal cholera incidences in Bangladesh (Table 1). Winter DJF (Dec–Feb) SST along the coast is negatively correlated to spring cholera incidences; the correlation coefficient between DJF SST and MAM cholera incidences is found to be −0.69 and −0.32 (p = 0.11). On the other hand, summer JJA (June–August) SST in BoB shows high positive correlation with autumn cholera outbreaks. The correlation values of JJA SST values with OND cholera incidences and flood-affected area (FAA) in Bangladesh are both found to be very strong (Table 1, r = 0.70 and 0.37 (p = 0.08) and r = 0.61 and 0.40 (p < 0.05), respectively). Results from Salahuddin et al. [2006], which show high positive correlation between monsoon rainfall and summer SST in BoB, further reinforce the high correlation between SST and FAA reported in Table 1.

3.2. High Year Versus Low Year Analysis

[18] If the correlations linking low flow with spring cholera, and high flow with autumn cholera are to be physically consistent, one would expect to see the manifestations of these processes in the high and low cholera incidence years. In other words, the relationships should hold true for the entire probability distribution of flow values, including the extreme years. To make this point, cholera incidence exceedance probability plots are constructed based on seasonal low and high flow values for the analysis period.

[19] Ten extreme drought years, five strongest and five weakest, are identified out of the twenty years based on streamflow volumes of the low flow months (average monthly cholera incidence during spring for these years are shown in Figure 3a). Figure 3b shows the probability of exceedance values for the entire period, the strongest drought years, and the weakest drought years. Cholera incidence information has been standardized by subtracting mean from seasonal incidence and dividing by standard deviation. The clear demarcation of the three categories shows the strong relationship between severity of low flow and spring cholera incidence. The probability of a spring cholera outbreak to exceed a certain threshold of incidence rates is always significantly higher in a water scarce year than in a water abundant year. Similarly, ten extreme flood years, five highest and five lowest, are identified within the same period, 1980–2000, based on combined monthly peak flow volumes. The autumn cholera incidence rates (Figure 3c) and associated probability of exceedance values (Figure 3d) are calculated for the entire time series and the lowest and highest flood year sets. The probability of a large cholera outbreak during autumn is found distinctively higher for high flood years, however, the probability in a lower than average flood year is indistinguishable from that of an average year.

Figure 3.

(a) Mean spring cholera incidence during weak and strong drought years. (b) Exceedance probability of spring cholera in low flow. (c) Mean autumn cholera incidence during low and high flood years. (d) Exceedance probability of autumn cholera in high flow.

4. Discussion

[20] The combination of seasonal hydroclimatology, high population density, floodplain geography and coastal ecology has made the Bengal Delta region especially vulnerable to periodic cholera outbreaks. In this study, we provide a working hypothesis on how the first outbreaks of cholera may be related to low flow discharge of the GBM Rivers and subsequent plankton intrusion during spring. Cholera incidence values in this season are inversely related to streamflow, i.e., bigger spring cholera peaks are seen in strong drought years. On the other hand, autumn cholera outbreaks are positively correlated to peak streamflow volumes, i.e., bigger autumn peaks are seen in high flood years. Evidence points to the role of fecal contamination of open water and inundation extent, which often impact a large portion of the population, gathered in few remaining dry areas.

[21] This is perhaps the first study that attempts to link the seasonal dual cholera peaks in Bangladesh directly with river discharge and points at two separate, pre and post monsoon, hydroclimatological drivers. These results provide the rationale and motivation to identify complementary physical evidences, e.g., coastal phytoplankton measurements, salinity patterns in coastal Bangladesh, fecal contamination sampling during autumn, and flood inundation patterns. These measurable environmental signatures will provide a way to strengthen our hypothesis and estimate the risk of cholera outbreaks with a reasonable lead-time to develop targeted intervention strategies and preempt future cholera outbreaks.

Acknowledgments

[22] This research was supported, in part, by a grant from the National Science Foundation (EAR-0510429), and a National Institutes of Health Fellowship (DK070117-05) administered by the Water: Systems, Science and Society (WSSS) program at Tufts University, Medford, MA.

Ancillary