How does an Ethiopian dam increase malaria? Entomological determinants around the Koka reservoir

Authors


Corresponding Author Solomon Kibret, International Water Management Institute, PO BOX 14001, Addis Ababa, Ethiopia. E-mail: s.kibret@gmail.com

Abstract

Objectives  To identify entomological determinants of increased malaria transmission in the vicinity of the Koka reservoir in Central Ethiopia.

Methods  Larval and adult mosquitoes were collected between August 2006 and December 2007 in villages close to (<1 km) and farther away from (>6 km) the Koka reservoir. Adult mosquitoes were tested for the source of blood meal and sporozoites.

Results  In reservoir villages, shoreline puddles and seepage at the base of the dam were the most productive Anopheles-breeding habitats. In villages farther from the dam (control villages), rain pools were important breeding habitats. About five times more mature anopheline larvae and six times more adult anophelines were found in the villages near the reservoir. Anopheles arabiensis and Anopheles pharoensis were the most abundant species in the reservoir villages throughout the study period. The majority of adult and larval anophelines were collected during the peak malaria transmission season (September–October). Blood meal tests suggested that Aarabiensis fed on humans more commonly (74.6%) than Apharoensis (62.3%). Plasmodium falciparum-infected Aarabiensis (0.97–1.32%) and Apharoensis (0.47–0.70%) were present in the reservoir villages. No P. falciparum-infected anophelines were present in the control villages.

Conclusions  The Koka reservoir contributes to increased numbers of productive Anopheles-breeding sites. This is the likely the cause for the greater abundance of malaria vectors and higher number of malaria cases evidenced in the reservoir villages. Complementing current malaria control strategies with source reduction interventions should be considered to reduce malaria in the vicinity of the reservoir.

Abstract

Objectifs:  Identifier les déterminants entomologiques de la transmission accrue du paludisme dans les environs du réservoir de Koka, dans le centre de l’Ethiopie.

Méthodes:  Les larves et adultes de moustiques ont été recueillis entre août 2006 et décembre 2007 dans les villages proches de (<1 km) et éloignés (> 6 km) du réservoir de Koka. Les moustiques adultes ont été testés pour la source de sang sucé et pour les sporozoïtes.

Résultats:  Dans les villages riverains du réservoir, les flaques d’eau et infiltrations à la base du barrage étaient les habitats les plus productifs pour la reproduction des anophèles. Dans les villages plus éloignés du barrage (villages témoins), les eaux de pluie stagnantes étaient des habitats de reproduction importants. Environ cinq fois plus de larves matures d’anophèles et six fois plus d’anophèles adultes ont été trouvés dans les villages autour du lac. Anopheles arabiensis et An. pharoensisétaient les espèces les plus abondantes dans les villages autour du réservoir tout au long de la période étudiée. La majorité des anophèles adultes et des larves ont été collectées durant la saison de pic de transmission du paludisme (septembre-octobre). Les tests sur les sources de sang sucé ont suggéré qu’An. arabiensis s’alimente plus souvent sur l’homme (74,6%) qu’An. pharoensis (62,3%). 0,97%à 1,32% d’An. arabiensis et 0,47%à 0,70% d’An. pharoensis infectés par Plasmodium falciparumétaient présents dans les villages du réservoir. Aucun anophèle infecté par Plasmodium falciparum n’était présent dans les villages témoins.

Conclusions:  Le réservoir de Koka contribue à une augmentation du nombre de sites productifs de reproduction d’Anopheles. Ceci est la cause probable de la plus grande abondance des vecteurs du paludisme et du nombre plus élevé des cas de paludisme mis en évidence dans les villages du réservoir. La complémentation des stratégies actuelles de lutte contre le paludisme avec des interventions de réduction des sources devraient être envisagée pour réduire le paludisme dans les environs du réservoir.

Abstract

Objetivos:  Identificar los determinantes entomológicos del aumento en la transmisión de malaria en el vecindario de la reserva de Koka en Etiopía Central.

Métodos:  Se recogieron larvas y mosquitos adultos entre Agosto del 2006 y Diciembre del 2007 en poblados muy cercanos a (<1 km) y más lejanos (>6km) a la reserva de Koka. Los mosquitos adultos se examinaron en busca de la fuente de la comida de sangre y de esporozoitos.

Resultados:  En los poblados de la reserva, los hábitats más productivos para la reproducción de Anopheles eran los charcos cercanos a la orilla y por filtración en la base de la presa. En los poblados más alejados de la presa (poblados controles), los charcos producidos por la lluvia eran los hábitats de reproducción importantes. Se encontraron cerca de cinco veces más larvas maduras de Anopheles y seis veces más adultos anofelinos en los poblados cercanos a la reserva. Anopheles arabiensis y An. pharoensis fueron las especies más abundantes en los poblados de la reserva a lo largo del periodo de estudio. La mayoría de anofelinos adultos y larvas se recogieron durante el pico de estación de transmisión de malaria (Septiembre-Octubre). Las pruebas realizadas a las comidas de sangre sugerían que An. arabiensis se alimentaba de humanos más a menudo (74.6%) que An. pharoensis (62.3%). Se hallaron An. arabiensis (0.97% - 1.32%) y An. pharoensis (0.47% - 0.70%) infectados con Plasmodium falciparum en los poblados de la reserva. No había anofelinos infectados con Plasmodium falciparum en los poblados control.

Conclusiones:  La reserva de Koka contribuye a aumentar el número de lugares activo de reproducción de Anopheles. Probablemente esta sea la causa de la mayor abundancia de vectores de malaria y el mayor número de casos de malaria en los poblados de la reserva. Deberían complementarse las estrategias de control de malaria actualmente disponibles con el fin de reducir la malaria en el vecindario de la reserva.

Introduction

Water reservoirs have long been recognised to be a risk factor for malaria transmission (Hunter et al. 1993; Jobin 1999). In tropical Africa, where the vast majority of the malaria disease and death burden is concentrated (WHO 2011), there is ample evidence that dams have amplified transmission of malaria (e.g. Jobin 1999; Keiser et al. 2005). In Ethiopia, there is evidence that proximity to the Gilgel-Gibe dam (Yewhalaw et al. 2009), the Koka dam (Lautze et al. 2007; Kibret et al. 2009) and several microdams in Tigray (Ghebreyesus et al. 1999) are associated with elevated rates of malaria transmission. Around the microdams in Tigray, increased abundance of malaria vectors, owing to greater quantities of standing water, was the cause for greater malaria transmission (Ghebreyesus et al. 1999; Yohannes et al. 2005). However, research related to the entomological and ecological determinants of the rising malaria burden in the vicinity of large dams is limited.

Retrospective analyses of clinical malaria data from villages around the Koka reservoir in the Rift Valley region of Ethiopia found proximity to the reservoir to be closely linked with elevated malaria transmission (Lautze et al. 2007; Kibret et al. 2009). Malaria case-rates decreased with distance from the reservoir, ranging from 90 per 1000 in villages within 1 km, to 5 per 1000 in villages 5–9 km from the reservoir (Kibret et al. 2009). Villages close to the reservoir were at increased malaria risk mainly during seasons of greatest transmission intensity, chiefly from Plasmodium falciparum (Kibret et al. 2009).1

It is likely that factors such as an increase in water-related activities, including irrigation, an expansion of breeding sites along the reservoir shore, and changes in vector composition and abundance, all contribute to intensified malaria transmission. This paper reports the findings of an entomological study conducted to identify the factors that explain the elevated malaria around Koka.

Materials and methods

Study sites

Our study focused on the area around the Koka dam and its reservoir in Ethiopia (Figure 1). The Koka dam is located in a rural setting 100 km to the south-east of Addis Ababa at 1590 m above sea level. Originally constructed to generate electricity, its purpose has since expanded to include flood mitigation and downstream irrigation, especially for the 6000 ha Wonji sugarcane irrigation scheme.

Figure 1.

 Map of the study site in Ethiopia.

In the Ethiopian Rift Valley, malaria is seasonally demarcated. Peak transmission occurs from mid-September to mid-November, after the wet season. Because malaria transmission is seasonal, there is little build up of immunity within local populations. Local malaria control authorities seek to reduce malaria risk by case management and indoor residual spraying (IRS).

Four villages were selected for the study (Figure 1). They were chosen because of their similar socio-economic status and the absence of known confounding factors such as major differences in bed net use, livestock ownership or other factors including those driving treatment-seeking behaviour (Deressa et al. 2007). No IRS was applied during the period of the study. Two study villages are located near the reservoir: Ejersa and Siree-Robe, which are, respectively, 0.4 and 0.8 km from the shore at average reservoir capacity. The other two villages (control villages) are situated farther from the reservoir: Kuma and Gudedo, respectively, 6.5 and 9.8 km from the shore at average reservoir capacity (Kibret et al. 2009). The angle of the slope from the edge of the reservoir to the adjacent villages varies. The average slope between Siree-Robe and the reservoir shore (7.9%) is steeper than the average slope between Ejersa and the shore (2.3%).

People generally live in traditional mud huts that mostly comprise just one room. Most households possessed some form of livestock: cattle, horses, donkeys and chickens were present in the villages. Studies in the nearby area of Ziway have shown no association between livestock ownership and malaria prevalence (Kibret et al. 2010), different from evidence from Kenya (Mutero et al. 2004).

Mosquito sampling

We employed standard entomological methods to identify mosquito-breeding sites and assess vector abundance (Amerasinghe et al. 1997). Between August 2006 and December 2007, mosquitoes (both larval and adult) were collected every 2 weeks from the study villages. Larval surveys were undertaken in all water bodies within 1 km of each village. The type of larval site (‘agricultural field puddles’, ‘manmade pools’ such as construction pits, ‘rain pools’ in the villages and on roads, ‘seepage pools below the dam’, reservoir shoreline puddles’) was recorded along with the site’s approximate surface area and water depth in the middle. In each survey, larval sampling was conducted using a standard 350-ml dipper at the rate of 6 dips per square metre. The number and developmental stage of anopheles larvae were recorded in positive breeding sites. Third and fourth instars anopheline larvae were collected from each type of breeding habitat. Direct pipetting was used to transfer larvae to separate vials, where larvae were killed and preserved in 70% alcohol. The samples were transported to Addis Ababa University where species were identified from morphological characteristics (Verrone 1962b).

Adult mosquitoes were sampled every 2 weeks from each of the four villages over the same time period. Six Centre for Disease Control and Prevention (CDC) light traps (Model 512; JW Hock Co., Atlanta, GA, USA) were deployed in four houses and two outdoor locations in each village. Houses were selected to ensure a widespread distribution over each village and where bed nets had not been used before the study. Untreated bed nets were provided to protect the people sleeping in the room with the light trap that was placed near the bed, 1.5 m above the ground. The outdoor traps were placed just outside occupied houses. All traps were operated overnight between 18.00 and 07.00. Collected mosquitoes were transported to Addis Ababa University. Adult Anopheles were sorted by species and sex using the morphological descriptions of Verrone (1962a). Based on the information available from previous studies, it was assumed that only Anopheles arabiensis was present among the Anopheles gambiae s.l. Giles sibling species (Abose et al. 1998). The human blood index and sporozoite rates were determined using direct and sandwich ELISA techniques, respectively (Burkot et al. 1981; Wirtz et al. 1987).

Data analysis

For statistical analysis, daily larval and adult mosquito collections were tested for normality and then log-transformed (log10 [n + 1]). Larval density was determined by dividing the number of anopheline larvae by square metre of surface area. Adult density was determined as the mean number of mosquitoes collected in each light trap each night. Variations in larval and adult mosquito densities were determined using the nonparametric Mann–Whitney U-test. Confidence intervals were defined at the 95% range. The sporozoite rate was multiplied by man-biting rate to determine entomological inoculation rates (EIR); Man-biting rates were derived from light trap catches (i.e. density divided by a conversion factor 1.5; Yohannes et al. 2005). All analyses were carried out using Microsoft Excel 97 and SPSS version 13.

Results

Larval abundance

Of 298 potential mosquito-breeding sites that were surveyed in the reservoir villages, 160 (53.6%) contained anopheline larvae (Table 1). In comparison, in the control villages, anopheline larvae were found in only 46 of 157 (29.3%) potential mosquito-breeding sites. Among the potential larval-breeding habitats encountered, shoreline puddles and seepage below the dam were the most important, together comprising 75% of the total positive larval sites in the reservoir villages. Temporary rain pools and agricultural pools were the most important breeding sites in control villages (Table 1).

Table 1.   Larval breeding sites of anopheline mosquitoes in villages near Koka reservoir and farther away in 34 sampling rounds from August 2006 until December 2007
Type of larval habitatReservoir villagesControl villages
Potential breeding sites
N (%)
Positive breeding sites
N (%)
Potential breeding sites
N (%)
Positive breeding sites
N (%)
  1. *The difference among larval habitats is significant (< 0.001), Mann–Whitney U-test.

Agricultural field puddles20 (6.7)9 (5.6)56 (35.7)21 (45.6)
Manmade pools22 (7.4)10 (6.3)14 (8.9)1 (2.2)
Rain pools41 (13.8)20 (12.5)87 (55.4)*24 (52.2)*
Seepage pools below the dam55 (18.5)25 (15.6)
Reservoir shoreline puddles160 (53.7)*96 (60)*
Total298 (100)160 (100)157(100)46 (100)

Evidence on breeding site abundance was largely paralleled by evidence on larval abundance. Anopheles arabiensis and Anopheles pharoensis were the most abundant species in the reservoir villages, constituting 85% of the mature anopheline larvae collections (Table 2). The main breeding sites for these species appeared to be reservoir shoreline puddles and seepage under the dam, together comprising 69.4% (1757 of 2531) of the total mature larvae in the reservoir villages. In the control villages, A. arabiensis and Anopheles coustani were the most common species, comprising 60.7% and 20.4% of the total mature larval collection, respectively. These species were predominantly abundant in temporary rain pools and man-made pools (Table 2).

Table 2.   Species of anopheline larvae (third and fourth instar) in various breeding sites around Koka reservoir
  N positive sitesType of larval habitat
Anopheles arabiensis
n
(%)
Anopheles pharoensis
n
(%)
Anopheles coustani
n
(%)
Anopheles funestus
n
(%)
Total
n
Reservoir villages
 Agricultural field puddles920 (1.5)16 (1.9)16 (5.1)0 (0.0)52
 Manmade pools10139 (10.7)57 (6.7)41 (13.0)0 (0.0)237
 Rain pools20292 (22.5)130 (15.3)63 (19.9)0 (0.0)485
 Seepage pools at dam base25319 (24.5)280 (32.9)67 (21.2)19 (29.2)685
 Reservoir shoreline puddles96530 (40.8)367 (43.2)129 (40.8)46 (70.8)1072
 Total1601300 (100)850 (100)316 (100)65 (100)2531
Control villages
 Agricultural field puddles2156 (17.1)9 (8.8)8 (7.3)0 (0.0)73
 Manmade pools155 (16.8)24 (23.5)32 (29.1)0 (0.0)111
 Rain pools24216 (66.1)69 (67.7)70 (63.6)0 (0.0)355
 Total46327 (100)102 (100)110 (100)0 (100)539

The abundance of positive larval sites rose after the beginning of the main rainy season in June and peaked between August and October, at the end of the rainy season in all study villages (Figure 2). Between November and May, there were a very small number of positive sites in the reservoir villages and no positive sites in the control villages as most habitat had dried-up. The two malaria vectors Aarabiensis and Apharoensis were predominantly found in reservoir-associated pools (i.e. seepage below the dam and reservoir shoreline puddles), with greatest larval densities during (July–December) (Figure 3). During the dry season, reservoir-associated pools were the only available breeding sites in the reservoir villages.

Figure 2.

 Positive anopheline larval sites each month from August 2006 to December 2007, stratified by village (the bar graph shows total monthly rainfall).

Figure 3.

Anopheline larval density (mean larvae per m2) in each season, stratified by village. The bars indicate the confidence interval.

Adult abundance

Over the study period, a total of 2952 adult anophelines were collected from the study villages (Table 3). 2514 (85.2%) of the overall total were from the two reservoir villages. Anopheles arabiensis and A. pharoensis constituted 55.3% (1632 of 2952) and 29.8% (880 of 2952) of the total adult collections, respectively. Anopheles arabiensis did not show significant difference in its density indoors and outdoors, signifying that during the night the species is equally indoor and outdoor feeding. Other Anopheles species were collected mainly outside in all study villages, which reveals exophagic feeding behaviour (Table 3).

Table 3.   Total numbers and densities (mean anophelines per light trap per night) of adult anopheline mosquitoes collected indoor and outdoor in villages near to and farther away from Koka reservoir, on 34 nights with 6 traps per village in four villages, between August 2006 and December 2007
 VillagesTotal
EjersaSiree-RobeGudedoKuma
  1. N = 816 (sample size, total number of trap nights times the number of traps employed per night).

  2. *The difference between indoor and outdoor adult densities (mean number of mosquitoes in each trap each night) was significant (< 0.001).

Anopheles arabiensis
 N (%)824 (49.3)532 (63.0)161 (64.9)117 (61.6)1634 (55.3)
 Indoor3.47*2.390.620.43*1.73
 Outdoor2.612.010.530.291.36
Anopheles pharoensis
 N (%)574 (34.4)213 (25.2)42 (16.9)50 (26.3)879 (29.8)
 Indoor1.690.590.130.160.64
 Outdoor2.52*1.91*0.230.34*1.25*
Anopheles coustani
 N (%)252 (15.1)90 (10.7)45 (18.2)23 (12.1)410 (13.9)
 Indoor0.410.360.140.080.25
 Outdoor1.38*1.60*0.200.16*0.84*
Anopheles funestus
 N (%)20 (1.2)9 (1.1)0 (0.0)0 (0.0)29 (1.0)
 Indoor0.030.00000.01
 Outdoor0.080.02000.03

Throughout the study period, the density of anopheline mosquitoes was significantly greater (< 0.001) in the reservoir villages (4.80 Anopheles/trap/night, 95% CI = 2.41–6.69) than in the control villages (0.79 Anopheles/trap/night, 95% CI = 0.08–1.50). Density of adult anophelines was greatest during the main rainy season in all villages (Figure 4).

Figure 4.

 Density of Anopheles arabiensis and Anopheles pharoensis (mean anophelines per trap each night) in the reservoir and control villages around Koka in 34 nights with 24 traps between August 2006 and December 2007.

Among female Aarabiensis and Apharoensis with blood in their gut, 71.0% (323 of 455) and 61.9% (266 of 430) had fed on humans, respectively, and 24.2% (110 of 455) and 29.3% (126 of 430) on cattle, respectively (Table 4). Anopheles arabiensis was therefore most important in terms of feeding on human blood, followed by Apharoensis. In contrast, both Acoustani and Anopheles funestus were mainly zoophagic (i.e. animal feeding). No significant difference was evident in Anopheles feeding tendencies between reservoir and control villages.

Table 4.   Sources of blood in anopheline mosquitoes in four villages around the Koka reservoir, collected in 34 samples between August 2006 and December 2007
  N Positive for human blood (%)Positive for bovine blood (%)*Unidentified (%)†
  1. *Samples that were positive for both human blood and bovine blood were included in both categories. Hence, the total in each row could be more than 100%.

  2. †Unidentified blood was neither from humans nor from cattle.

  3. ‡No A. funestus were collected in the control villages.

Ejersa
 Anopheles arabiensis208149 (71.6)38 (18.3)21 (10.1)
 Anopheles pharoensis231138 (59.7)69 (29.9)40 (17.3)
 Anopheles coustani7031 (44.3)43 (61.4)11 (15.7)
 Anopheles funestus62 (33.3)3 (50)2 (33.3)
Siree-Robe
 A. arabiensis11189 (80.2)29 (26.1)10 (9.0)
 A. pharoensis154102 (66.2)42 (27.3)32 (20.8)
 A. coustani3313 (39.4)20 (60.6)5 (15.1)
 A. funestus20 (0.0)1 (50.0)1 (50.0)
Gudedo
 A. arabiensis8957 (64.0)26 (29.2)18 (20.2)
 A. pharoensis2818 (64.3)9 (32.1)3 (10.7)
 A. coustani154 (26.7)10 (66.7)2 (13.3)
 A. funestus00 (0.0)0 (0.0)0 (0.0)
Kuma
 A. arabiensis4728 (59.6)15 (31.9)11 (23.4)
 A. pharoensis178 (47.1)6 (35.3)5 (29.4)
 A. coustani81 (13.5)3 (37.5)4 (50.0)
 A. funestus00 (0.0)0 (0.0)0 (0.0)

Anopheles arabiensis and APharoensis infected with P. falciparum were found in only the reservoir villages; no anophelines in the control villages were infected (Table 5). Greater P. falciparum sporozoite rates were found in Aarabiensis (0.97–1.32%) than in Apharoensis (0.47–0.70%). All Plasmodium-infected mosquitoes were collected between August and October. Fewer mosquitoes were collected in control villages, and no P. falciparum sporozoites were found in A. coustani and A. funestus mosquitoes from control villages.

Table 5.   Sporozoite rates (Plasmodium falciparum) in four species of Anopheles in villages around Koka, from 34 collections obtained between August 2006 and December 2007
 Reservoir villagesControl villages
EjersaSiree-RobeGudedoKumo
Anopheles arabiensis
 N824532161117
 Positive (%)  9 (0.97)  7 (1.32)  0 (0.00)  0 (0.00)
Anopheles pharoensis
 N574213 42 50
 Positive (%)  4 (0.70)  1 (0.47)  0 (0.00)  0 (0.00)
Anopheles coustani
 N252 90 45 23
 Positive (%)  0 (0.00)  0 (0.00)  0 (0.00)  0 (0.00)
Anopheles funestus
 N 20  9  0  0
 Positive (%)  0 (0.00)  0 (0.00)  0 (0.00)  0 (0.00)

The EIR calculated for reservoir villages showed that in Ejersa inhabitants received 0.30 and 0.19 P. falciparum-infective bites each night by Aarabiensis and Apharoensis, respectively. In Siree-Robe, people received 0.12 and 0.03 P. falciparum-infective bites each night by Aarabiensis and Apharoensis. Inhabitants of Ejersa receive on average 45 infective bites during the transmission season (August–October), inhabitants of Siree-Robe receive an average of 14.

Discussion

The study results indicate that the elevated malaria transmission in villages near the Koka reservoir is likely to be a consequence of increased anopheline abundance, mainly due to breeding sites directly related to the reservoir. The two most abundant species breeding in those sites, A. Arabiensis and A. pharoensis, are confirmed major malaria vectors, feeding predominantly on humans, indoors and outdoors. People living close to the reservoir may receive on average 45 bites from anophelines-infected with P. falciparum during the transmission season, whereas people living in the villages farther away are likely to receive far fewer infective bites. These findings are consistent with evidence from the Tigray region of northern Ethiopia, where the abundance of A. arabiensis was reported to be about seven times greater in a village close to microdams (Yohannes et al. 2005).

Our observations on the seasonality of adult vector abundance are consistent with epidemiological evidence (Lautze et al. 2007; Kibret et al. 2009) that the reservoir’s primary impact is an amplification rather than extension of the transmission season. However, these findings differ from those of others who found that water reservoirs create perennial transmission in areas of historically seasonal transmission (Keiser et al. 2005). A possible explanation for the different findings may relate to the size of impoundments from which observations were derived. Around small dams, distance between the shore and nearby communities may not fluctuate substantially through the year. The same was true for small-scale irrigation near Ziway, 60 km south of Koka in the Rift valley (Kibret et al. 2010). In contrast, around the Koka reservoir the distance from the shore to certain nearby villages – particular those with relatively flat terrain – may change by as much as a kilometre as a result of fluctuations in the water level (S. Kibret & J. Lautze, personal observations). High water levels, which are associated with the creation of shoreline puddles much closer to villages, coincide with the months of peak transmission.

Disparity in geomorphic conditions may explain the difference in abundance of malaria vectors between the two reservoir villages. In Ejersa, where the land is rather flat (slope of 2.3 m vertical per 100 m horizontal), anophelines mainly originated from shoreline sites. However, in Siree-Robe, the steeper slope of 7.9 m per 100 m is less conducive to the formation of puddles in which mosquitoes can develop, and Anopheles were associated mostly with the seepage pools below the dam. These sites are washed out when the reservoir spills and not all larvae will develop into adults. As a result, consistent with other evidence on the impact of topography on malaria transmission (Cohen et al. 2008), abundance of adult anophelines was much greater in Ejersa than Siree-Robe.

A primary reason for undertaking this investigation was to derive recommendations for strengthening malaria control efforts around Koka and other reservoirs. The study’s principal finding that the elevated malaria transmission around Koka can be attributed to increased vector populations, resulting largely from breeding sites directly linked with the reservoir, highlights an important potential role for source reduction measures to disrupt vector development in shoreline sites. In addition to strengthening conventional malaria control efforts, a recommendation is to draw on the sparse, yet important, body of literature outlining water management options for malaria reduction around the reservoirs (e.g. TVA 1947; Gartrell et al. 1972).

There indeed may be limitations to control strategies relying on single measures such as IRS or insecticide treated net (ITN) distribution. Evidence exists from the region that anophelines may modify their behaviour to feed mainly outdoors in response to repeated indoor spraying with insecticides (Ameneshewa & Service 1996). In addition, because some malaria vectors feed early in the evening (Abose et al. 1998; Kibret et al. 2010; Yohannes & Boelee 2012), the potential contributions from ITNs may be partially undermined.

As such, complementing current malaria control strategies (ITN, IRS) with source reduction interventions should be considered to reduce malaria in the vicinity of the reservoir. Recent studies (e.g. Lautze & Kirshen 2007; Kibret et al. 2009; Reis et al. 2011) have focused on adapting water manipulations once employed around water reservoirs in the American South (TVA 1947; Gartrell et al. 1972) to control malaria in sub-Saharan Africa. Around the Koka reservoir in particular, evidence (Kibret et al. 2009; Reis et al. 2011) suggests that a more rapid water level drawdown could be employed to desiccate Anopheles shoreline breeding sites and that implementation of this measure would impose minimal conflict with other reservoir uses such as hydropower production.

In conclusion, the Koka reservoir has substantially increased the abundance of malaria vectors in adjacent villages, mainly due to breeding in reservoir-associated sites, notably reservoir shoreline puddles. These puddles are both abundant and productive in the flat reservoir area, leading to high numbers of vector mosquitoes near the reservoir. This is the most likely explanation why villages close to the reservoir are at increased malaria risk.

Footnotes

  • 1

    These findings held in the context of a multiple variable analysis that included variables such as rainfall and temperature.

Acknowledgements

The studies in this paper were funded through support of the CGIAR Challenge Program on Water and Food (CPWF) to project number 36 entitled ‘Improved planning of large dam operation: using decision support systems to optimise livelihood benefits, safeguard health and protect the environment’. Jonathan Lautze had a training grant from the National Institutes of Health. The authors acknowledge the kind cooperation of the inhabitants of Ejersa, Siree-Robe, Gudedo and Kuma during the mosquito collection. We thank Paul Kirshen for his advice at the beginning of the study, Gayathri Jayasinghe for critical feedback, Dereje Olana for facilitating field data collection and Richard Pollack for intellectual guidance in the development of this document. We are also grateful to the Oromia Malaria Control Authorities and in particular to Tesfaye Abebe and Lemma Regassa.

Ancillary