Effect of Irrigation and Large Dams on the Burden of Malaria on a Global and Regional Scale

EFFECT OF IRRIGATION AND LARGE DAMS ON THE BURDEN OF MALARIA ON A GLOBAL AND REGIONAL SCALE

JENNIFER KEISER, MARCIA CALDAS DE CASTRO, MICHAEL F. MALTESE, ROBERT BOS, MARCEL TANNER, BURTON H. SINGER, AND JÜRG UTZINGER Swiss Tropical Institute, Basel, Switzerland; Department of Geography, University of South Carolina, Columbia, South Carolina; Saint Antony’s College, Oxford University, Oxford, United Kingdom; Water, Sanitation and Health, World Health Organization, Geneva, Switzerland; Office of Population Research, Princeton University, Princeton, New Jersey

Abstract. Human-made ecologic transformations have occurred at an unprecedented rate over the past 50 years. Prominent among them are water resource development projects. An estimated 40,000 large dams and 800,000 small dams have been built, and 272 million hectares of land are currently under irrigation worldwide. The establishment and operation of water projects has had a history of facilitating a change in the frequency and transmission dynamics of malaria, but analyses of these environmental risk factors are sparse. Here, we present a comprehensive review of studies that assessed the impact of irrigation and dam building on malaria prevalence or incidence, stratified by the World Health Organization’s (WHO) sub-regions of the world, and link these studies with the latest statistics on disability adjusted life years, irrigated agriculture, and large dams. We also present estimates of the population at risk due to proximity to irrigation schemes and large dam reservoirs. In WHO sub-regions 1 and 2, which have 87.9% of the current global malaria burden, only 9.4 million people are estimated to live near large dams and irrigation schemes. In contrast, the remaining sub-regions concentrate an estimated 15.3 million people near large dams and up to 845 million near irrigation sites, while here only 12.1% of the global malaria burden is concentrated. Whether an individual water project triggers an increase in malaria transmission depends on the contextual determinants of malaria, including the epidemiologic setting, socioeconomic factors, vector management, and health seeking behavior. We conclude that in unstable malaria endemic areas, integrated malaria control measures, coupled with sound water management, are mandatory to mitigate the current burden of malaria in locations near irrigation or dam sites.

INTRODUCTION water projects must also be juxtaposed with the positive effect that dams and irrigation schemes contribute substantially to Currently, more than two billion people live at risk of con- renewable energy production, food security, and social and tracting malaria, and the estimated global annual incidence of economic development. This, in turn, can provide rural clinical malaria is greater than 300 million cases. More than households with greater capacity to purchase essential com- one million people die every year from the direct causes of modities, including drugs and insecticide-treated nets (ITNs), malaria, with children less than five years of age living in as well as improved access to health care services and educa- sub-Saharan Africa at highest risk.1 The disease accounts for tion. an estimated loss of 46.5 million disability adjusted life years Reliable analyses of environmental risks to health are fun- (DALYs) with almost 90% currently concentrated in sub- damental for the prevention and control of diseases, for evi- Saharan Africa.2 Approximately 90% of this burden is related dence-based guidance of health policy and planning, and for to environmental factors.3 The establishment and operation the promotion of intersectoral action for the reduction of of water resource development projects represents an impor- transmission. However, to our knowledge, an in-depth analy- tant aspect of these factors, since dams and irrigation schemes sis of the malaria burden attributable to the development and transform ecosystems and can substantially change the nature operation of water projects has not been carried out. of malaria risk proximal to their location. There is a substan- In this report, we present the outcomes of a systematic tial body of literature documenting the facilitation of in- review of the literature spanning the past 25 years by linking creases in malaria incidence and prevalence as a consequence malaria prevalence and incidence data in relation to major of such projects.4 water projects, with an emphasis on irrigation and large dams. In 2001, the total area under irrigation worldwide was es- The global database on the effect of small dams and flood timated at 272 million hectares (ha) compared with 139 mil- control is inadequate to support generic conclusions from a systematic review. Our primary objectives are 1) to estimate סlion ha in 1961 (http://apps.fao.org/page/collections?subset agriculture). Concurrently, it is estimated that at least 40,000 the size of the populations at risk of malaria due to their large dams (i.e., defined as impoundments more than proximity to irrigation schemes and large dams, and 2) to 15 meters high or storing more than 3 million m3 of water) assess the impact of irrigation and large dams on the burden and 800,000 small dams have been built worldwide. The ma- of malaria at global and regional scale. We use the 14 sub- jority of the large dams serve irrigation purposes. Most of the regions articulated in the statistical analyses of the annual large dams were constructed after 1950, during the post-war World Health Report of the World Health Organization development era, when large-scale infrastructures were re- (WHO).2 garded as symbols of patriotic pride and technological ad- In the next section, we describe our data sources and meth- vance. More than 400,000 km2 have been inundated by res- odology for producing estimates of the sizes of at-risk popu- ervoirs worldwide.5 These ecologic transformations go hand- lations and the impact of large dams and irrigation schemes in-hand with the creation of new mosquito breeding sites. on the burden of malaria. Detailed illustrations of our calcu- Water resources development is usually also coupled with lations are given in Appendix 1. After presenting our results demographic changes, and thus alters human-vector-parasite in the subsequent section, we conclude with a discussion of a contact patterns. The potential for negative health impacts of myriad of unresolved issues that need to be addressed if the 392 IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 393 impact of major water projects on the burden of malaria is to high and moderate malaria transmission and excluded coun- be estimated with greater precision than is currently feasible. tries with sporadic malaria risk (e.g., Kazakhstan). The coun- The requirement for such measurement is directly connected tries included in our review are located in 10 of the 14 sub- to ongoing policy debates about the pressing need for defen- regions in the WHO classification and are listed in Table 1. sible health impact assessments associated with development Irrigated areas and affected population. For each country projects quite generally. we compiled data on the total, agricultural, irrigated, and rice-harvested areas, and the potential area for irrigation using the latest Food and Agricultural Organization (FAO) da- MATERIALS AND METHODS tabases (http://www.fao.org). We calculated the sum of the areas for each individual sub-region using data for the year Systematic literature review. We systematically reviewed 2000. Data on DALYs and the total population were ob- the literature with an emphasis on research findings published tained from the World Health Report.2 over the past 25 years on any form of water resource devel- We gathered statistics on population assigned to mixed ir- opment and management and its effect on the frequency and rigation schemes (areas that combine cropping with livestock transmission dynamics of malaria. Publications were searched with at least 10% of the area irrigated) from a global data set through Medline (National Institutes of Health, Bethesda, of irrigated areas.6 To have a second range (since the irriga- MD), the Environmental Sciences and Pollution Manage- tion population provided by Thornton and others6 might be ment Database (Cambridge Scientific Abstracts, Cambridge, overestimated by a factor as high as 10), we based our calcu- MA) and the website of the World Commission on Dams lations on the irrigated area of each country and a hypotheti- (http://www.dams.org/). Pertinent dissertation abstracts, book cal average population density of 200 people/km2 in the irri- chapters, and unpublished documents (“gray literature”) gated areas. The later figure is justified as follows. Although were also consulted. We only included those studies that as- rural population densities vary from province to province and sessed malaria prevalence or incidence before and after the country to country, in general irrigation schemes are well- construction of a water project, or compared two or more developed and highly attractive areas, and the villages might settings that primarily differed with regard to a water re- be even overcrowded. For example, in the Bura and Mwea source development project. irrigation schemes in Kenya, population densities of 223 Malaria-endemic countries according to WHO sub- people/km2 and 320 people/km2 have been reported, whereas regions. We used the recent WHO classification of countries the overall population density in Kenya is several-fold lower, into 14 epidemiologic sub-regions, which is based on a com- namely 54 people/km2 as of 2002.7,8 bination of WHO regions, and child and adult mortality rates, To determine the population living in proximity to irriga- as described in the annexes of the annual World Health Re- tion schemes in malaria-endemic areas, we retrieved data for port.2 From this list we included only those countries with each country on the percentage of the population living in

TABLE 1 Countries included in our analysis based on World Health Organization (WHO) epidemiologic sub-regions and propensity for malaria transmission*

Africa Percentages of the population in malaria endemic areas available at http://www.rbm.who.int/amd2003/amr2003/table6.htm WHO sub-region 1: Angola, Benin, Burkina Faso, Cameroon, Chad, Comoros, Equatorial Guinea, Gabon, Gambia, Ghana, Guinea, Guinea- Bissau, Liberia, Madagascar, Mali, Mauritania, Niger, Nigeria, Sao Tome and Principe, Senegal, Sierra Leone, Togo WHO sub-region 2: Botswana, Burundi, Central African Republic, Congo, Côte d’lvoire, Democratic Republic of the Congo, Eritrea, Ethiopia, Kenya, Malawi, Mozambique, Namibia, Rwanda, Swaziland, Uganda, (South Africa), United Republic of Tanzania, Zambia, Zimbabwe The Americas Percentages of the population in malaria endemic areas available at http://165.158.1.110/english/hcp/hctmalaria.htm WHO sub-region 4: (Argentina), Belize, Brazil, Colombia, Costa Rica, Dominican Republic, French Guiana, Guyana, Honduras, Mexico, (Panama), Paraguay, (Suriname), (Venezuela) WHO sub-region 5: Bolivia, (Ecuador), Guatemala, Haiti, Nicaragua, Peru Eastern Mediterranean Percentages of the population in malaria endemic areas available at http://www.emro.who.int WHO sub-region 6: Saudi Arabia WHO sub-region 7: Afghanistan, Djibouti, (Islamic Republic of Iran), (Iraq), (Somalia), Sudan, Pakistan, Yemen Europe Percentages of the population in malaria endemic areas available at http://www.emro.who.int/rbm/meetings/muscat02/Presentations/Sunday/ Malaria%20Situation%20in%20EUR%20(M.Ejov).ppt WHO sub-region 9: Georgia, Tajikistan, Turkey, (Uzbekistan) Southeast Asia Percentages of the population in malaria endemic areas available at http://whqlibdoc.who.int/searo/2002/SEA_MAL_229.pdf WHO sub-region 11: Indonesia, Sri Lanka, (Thailand) WHO sub-region 12: Bangladesh, (Bhutan), Democratic People’s Republic of Korea, India, Myanmar, Nepal Western Pacific Percentages of the population in malaria endemic areas available at http://www.wpro.who.int/themes_focuses/theme1/focus2/t1f2country.asp WHO sub-region 14: Cambodia, (China), Lao People’s Democratic Republic, (Malaysia), Papua New Guinea, (Philippines), Solomon Islands, Viet Nam * Countries in parentheses have < 20% of the population at risk of malaria. 394 KEISER AND OTHERS malaria risk areas. The sources of these data are given in tion) have been, or will be, displaced due to large dam build- Table 1. We then determined for each country the population ing. When we standardized the populations from the year of at risk by multiplying the sizes of the irrigation populations by construction of the individual dams to the year 2000, we found the fraction of the population living in malaria-endemic areas. that the densities are as low as 1.2 people/km2 and as high as Populations at risk of malaria due to their proximity to 2,478 people/km2. The median population density calculated reservoirs of large dams. Components of dam sites include for each relevant WHO sub-region ranges from 25.8 people/ the reservoir, upper catchment area, irrigation schemes, and km2 in WHO sub-region 2 to 764 people/km2 in WHO sub- flood plains. To estimate the size of the at-risk populations of region 7. malaria, we focus on the environment immediately surround- Irrigated areas, large dam sites, malaria burden, and people ing the reservoir. at risk in endemic WHO sub-regions. Sub-Saharan Africa In a first step, we got an estimate of the population density (WHO sub-regions 1 and 2). Table 3 summarizes estimated near dam sites stratified by WHO sub-regions that are en- DALYs lost due to malaria, total surface area, agricultural demic for malaria by collecting information on displacement area, irrigated area, rice-harvested area, as well as total popu- and resettlement of population in relation to the size of the lation, irrigation population, and irrigation population in ma- reservoir for many dams for 8 of the 10 relevant sub-regions laria-endemic areas (“population at risk”). At present, irri- (Table 2).9,10 For each individual dam we standardized the gated agriculture or rice harvested areas are marginal in calculated population density according to the year 2000, us- WHO sub-regions 1 and 2 because they represent only ing the average annual rate of change of the rural popula- 0.2−0.5% of the total surface area. While some countries have tion.11 We then calculated the median for each relevant WHO virtually no areas under irrigation (e.g., Central African Re- irrigation is more pronounced in others ,(0.02% ס sub-region. public However, as irrigation provides .(1.5% ס In a second step, we collected data on the area of the res- (e.g., South Africa ervoirs and the length of the dam for WHO sub-regions 1 and an opportunity for agriculture in arid areas and stabilizes 2 by consultation of the World Register of Dams. For the yields in regions with unpredictable rainfall (e.g., Sahel), ir- South African dams, we used the geo-referenced database on rigated areas continue to grow in sub-Saharan Africa: the African dams (FAO; http://www.fao.org) and the malaria risk predicted irrigation potential of WHO sub-regions 1 and 2 is map generated by the “Mapping Malaria Risk in Africa” 39.3 million ha (Table 3). This represents a 10-fold increase of (MARA; http://www.arma.org.za) to examine, which of the the current irrigated area. dams are located in the malaria-endemic area. In sub-Saharan Africa 1,039 large dams have been con- Figure 1 shows how we estimated the area of risk near dam structed, more than half of them are located in South Africa. reservoirs. A detailed description of our calculations of the Employing the geo-referenced database provided by FAO, at-risk populations of malaria and 2 examples are given in we found that of the 539 South African dams only 25 are Appendix 1. located in the malaria-endemic parts of the country. For 287 large dams, information on the size of the reservoir and the 12 RESULTS length of the dams is given in the World Register of Dams. These dams have a total reservoir size of 24,792 km2. The Causal web. The various levels of causality between ma- calculated mosquito risk area comprises 45,594 km2 from the laria and different types of water projects are shown in Figure borders of the reservoirs at full water level in endemic areas 2. As detailed earlier, the current review focuses on irrigation (Table 4 and Appendix 1). schemes and large dam sites. In principle, proximity to a dam, Our literature review found 11 studies that have been car- including command areas and/or an irrigation scheme, implies ried out in areas of stable malaria transmission and 2 in un- proximity to new bodies of standing water that can serve as stable malaria transmission areas in WHO sub-regions 1 and Anopheles larval development sites. Whether this general ex- 2 that compared malaria incidence or prevalence rates among pectation is realized largely depends on the ecology of the people living close to an irrigation project with those ob- local vectors. In particular, it requires that the new bodies of served in distant villages (Table 5). The cross-village compari- standing water have pH, sunlight or shade, surrounding veg- sons assume that prior to the water project, the two sets of etation, turbidity, etc., compatible with the larval habitats for sites were approximate ecologic replicates in terms of factors at least one local vector species. Consequently, the creation of influencing malaria transmission. None of the 11 studies in new breeding sites might have an effect on the development stable transmission areas found a higher malaria prevalence of vector species and survival rates, and tradeoffs among them in the irrigated villages compared with non-irrigated villages. in terms of their role in local transmission. In addition, dams For example, in the Kou valley in Burkina Faso, malaria and irrigation schemes operate in the presence of diverse prevalence rates ranged from 16% to 58% in an irrigated combinations of preventive and curative interventions against village, compared with 35−83% in a non-irrigated village.13 In malaria. The details of these intervention packages vary sub- two study villages in Senegal, children were found to have a stantially from one location to another in the malarious re- malaria prevalence of 8.7% in the irrigated village and 16.5% gions of the world. Interventions and also the social and eco- in the non-irrigated village.14 Furthermore, in Mali, a two- nomic changes that occur will affect the pool of parasites in fold reduction in the annual malaria incidence was observed humans and the human-vector-parasite contact patterns. after the implementation of irrigation, although rice cultiva- Population density near large dam sites. We present data tion changed transmission from seasonal to perennial.15 A on the number of displaced people in relation to the size of lower malaria incidence or prevalence in the irrigated villages the reservoir for 71 large dams in Table 2. Between 1,000 when compared with non-irrigated villages (the so-called (Epupa dam, Namibia) and 1.2 million people (Three Gorges “paddies paradox”16) has been explained by improved socio- dam, People’s Republic of China; currently under construc- economic status, effective vector control programs, or IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 395

TABLE 2 Population density estimates near dam sites standardized for the year 2000*

Population density (people/km2)

Estimated number of Year of dam WHO sub-region Name of dam (country) Year of construction people displaced Area of reservoir (km2) construction Year 2000 1 Akosomba (Ghana) 1965 80,000 848 94.3 173.6 Kpong (Ghana) 1981 6,000 35 171.4 279.4 Manantali (Mali) 1993 12,000 477 25.1 28.3 Selingue (Mali) 1982 12,000 430 27.9 38.7 Dadin Kowa (Nigeria) 1988 26,000 300 86.7 102.3 Kiri (Nigeria) 1982 19,000 110 172.7 305.7 Kainjii (Nigeria) 1986 50,000 1,260 39.7 48.0 Nangbeto (Togo) 1987 11,000 180 61.1 80.1 91.2 ס Median 2 Kossou (Côte d’Ivoire) 1972 75,000 1,780 42.1 90.3 Kiambere (Kenya) 1988 6,000 25 240.0 291.0 Epupa (Namibia) Proposed 1,000 350 2.9 2.9 Cahora Bassa (Mozambique) 1974 25,000 2,660 9.4 12.2 Kariba (Zambie, Zimbabwe) 1959 57,000 5,100 11.2 25.8 25.8 ס Median 4 Yacyreta (Argentinia) 1994 50,000 1,720 29.0 29.0 Aqua Vermelha (Argentina) 1979 4,345 643 6.8 6.1 Piedra del Aguilai (Argentina) 1991 9,000 292 30.8 30.0 Salto Grande (Argentina) 1979 8,000 783 10.2 9.1 Tucurui (Brazil) 1984 30,000 2,430 12.3 9.9 Itaparica (Brazil) 1986 49,500 800 61.8 45.7 Itaipu (Brazil) 1983 59,000 1,350 43.7 35.0 Sobradinho (Brazil) 1979 65,000 4,150 15.6 12.3 Ita (Brazil) 1998 11,500 138 83.3 83.0 Foz de Areira (Brazil) 1980 8,400 167 50.3 39.7 Marimbondo (Brazil) 1975 5,500 438 12.6 9.7 Guatape (Colombia) NK 5,000 63 78.8 78.8 Guavio (Colombia) 1989 5,500 14 381.9 381.0 Upia (Colombia) NK 5,000 396 12.6 12.6 Cerron Grande (El Salvador) 1977 13,339 139 96.0 95.9 El Cajon (Honduras) 1985 4,694 94 49.9 58.1 La Angostura (Mexico) 1974 5,500 644 8.5 9.8 30.0 ס Median 5 No record 6 No record 7 Tarbela (Pakistan) 1976 96,000 242 396.6 723.8 Mangla (Pakistan) 1967 90,000 253 357.0 804.3 764 ס Median 9 Aslantas (Turkey) 1984 5,000 49 102.0 97.0 Ataturk (Turkey) 1991 40,000 817 49.0 50.3 Sir (Turkey) 1991 7,000 47.5 147.4 153.1 Keban (Turkey) 1974 30,000 675 44.4 43.3 Karakaya (Turkey) 1987 20,000 298 67.1 67.0 67.0 ס Median 11 Kedung Ombo (Indonesia) 1989 25,000 46 534.5 527.6 Cirata (Indonesia) 1988 56,000 62 903.2 892.5 Kota Panjang (Indonesia) 1997 24,930 124 201.0 199.0 Saguling (Indonesia) 1986 60,000 49 1,234.6 1,256.0 Victoria (Sri Lanka) 1984 45,000 23 1,982.3 2,342.7 Kothmale (Sri Lanka) 1988 13,000 9.5 1,368.4 1,525.7 Pak Mun (Thailand) 1990 8,000 60 133.3 150.5 Ubol Ratna (Thailand) 1986 30,000 410 73.2 88.7 Srinagarind (Thailand) 1981 6,400 419 15.3 20.0 Khao Laem (Thailand) 1986 10,800 388 27.8 33.7 363.3 ס Median 12 Kaptai (Bangladesh) 1962 100,000 777 128.7 269.5 Bargi (India) 1990 113,600 809.0 140.4 162.8 Sardar Sarovar (India) 2001 320,000 376 851.3 851.3 Kabini (India) 1974 15,000 61 245.9 367.5 Nagarjunasagar (India) 1974 28,000 285 98.3 147.3 Pong (India) 1974 150,000 290 517.2 773.6 Tehri (India) 1997 100,000 42 2,380.9 2,478.0 Bedhi (India) NK 5,100 124 41.1 41.1 Godavari (India) NK 38,100 1,008 37.8 37.8 Sapta Kosi (Nepal) Proposed 75,000 195 384.6 384.6 Khali Kola (Nepal) Proposed 40,000 108 370.4 370.4 367.5 ס Median 396 KEISER AND OTHERS

TABLE 2 Continued

Population density (people/km2)

Estimated number of Year of dam WHO sub-region Name of dam (country) Year of construction people displaced Area of reservoir (km2) construction Year 2000 14 Sambor (Cambodia) Proposed 5,120 880 5.7 5.7 Strung Treng (Cambodia) Proposed 9,160 640 14.3 14.3 Three Gorges (China) 1994–2008 1,200,000 1,084 1,107.0 1,107.0 Shuikoi (China) 1993 67,000 94 712.8 695.0 Yantan (China) 1995 40,000 121 330.6 322.1 Dongiian (China) 1989 53,000 160 331.2 323.1 Sanmenxia (China) 1960 319,000 799 399.2 591.8 Wuqiangxi (China) 1996 306,000 170 1,800.0 1,746.0 Longtan (China) 2005 73,392 370 198.3 198.3 Nam Ngum (Lao PDR) 1994 4,400 58 75.9 83.2 Nam Theun 2 (Lao PDR) 2007 (planned) 5,700 450 12.7 12.7 Nam Ou 2 (Lao PDR) 2007 (planned) 26,200 107 244.8 244.8 Hao Bin (Viet Nam) 1994 58,000 52,700 1.1 1.2 244.8 ס Median .People’s Democratic Republic ס not known; PDR ס World Health Organization; NK ס WHO * changes in health-seeking behavior in the irrigated villages.17 Rosso, Richard Toll, and Podor in the Senegal River Basin.20 In addition, as described earlier, the epidemiologic setting, in Irrigated villages in the Rusizi Valley of Burundi, an area of particular the entomologic parameters, are key contextual de- unstable malaria transmission, had higher malaria preva- terminants whether an irrigation project causes an increase in lences and a 150-fold higher vectorial capacity of An. arabi- the malaria incidence or prevalence. A water resource devel- ensis compared with a neighboring non-irrigated village.21 In opment project may also cause a change toward less endo- Madagascar, since 1878 several malaria epidemics have oc- philic and anthropophilic malaria vectors, thereby resulting in curred on the plateau where rice is grown in monoculture. A a lower vectorial capacity, as for example the replacement of huge increase in the level of malaria transmission in these Anopheles funestus by An. arabiensis.16 In addition, a greater rice-irrigated settings was found to be essentially related to larval competition in irrigated areas might result in reduced the proliferation of An. funestus, a much more anthropoph- adult longevity. Furthermore, in irrigated villages in sub- agic and endophagic vector than An. arabiensis. The parity Saharan Africa, high An. gambiae and An. arabiensis densi- rate of An. funestus was greater than 75% throughout the ties were correlated with low anthropophily, a decrease in the whole year.18 parity rate, low sporozoite indices, and low mosquito survival We found only three studies assessing the impact of large rates.18 Again, significant use of ITNs or antimalarial drugs dams in WHO sub-regions 1 and 2 (Table 6). No malaria due to a greater wealth in irrigated villages, or a higher com- transmission was observed in a village near the Gleita dam in pliance to ITN use, often driven by the nuisance caused by a Mauritania in 1984 in the fifth month of the dry season, al- higher mosquito density, might play an important part in though the malaria situation in the region is unstable.22 In these findings.19 Another explanation for lower malaria trans- Cameroon, a malaria prevalence of 36% was observed near mission in irrigated villages might also be differing presence the Bamendjin dam compared with a malaria prevalence of of cattle in the villages. Domestic animals are often kept close 25% in a village located 14 km away from the dam.23 In to the house and ITNs might divert mosquitoes away to the addition, year round malaria transmission was observed in unprotected animals.17,19 villages near the Manantali dam reservoir in Mali, which were In areas of unstable malaria transmission, the introduction previously characterized by seasonal transmission.24 of irrigation was found to place the non-immune population In addition, one study examined the effect of small dams: in at a high risk of acquiring the disease. It may alter malaria the unstable malaria transmission Tigray region of northern transmission from seasonal to perennial, and malaria ende- Ethiopia at altitudes above 1,800 meters, numerous small micity from mesoendemic to hyperendemic, as observed in dams and irrigation systems were put in place with the broad aim of reducing dependence on rain-fed agriculture, and thus improving food production. Comparative appraisal of a series of cross-sectional malaria surveys among children carried out in villages in close proximity to these newly constructed small dams and in villages farther away showed a seven-fold increase in malaria risk for those residing near dams.25 We estimate that of the 637.3 million people living in WHO sub-regions 1 and 2, approximately 9 million people (1.4%) live close to irrigation schemes (Table 3). Approximately two-thirds of these people live in malaria-endemic areas, and thus are at a risk of the disease. In addition, 3.1 million people are living near large dam sites in malaria endemic areas. In FIGURE 1. Estimation of the area at risk of malaria near dam Figure 3, we depict these key numbers in relation to the ma- reservoirs. laria burden. IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 397

FIGURE 2. Causal web (relationship between malaria and different types of water projects).

As the review of studies has shown and as depicted in Fig- belt and coastal belt of India, characterized by large areas ure 2, the impact of irrigation schemes or a large dam site on under irrigation, the risk of acquiring malaria is very small.27 malaria depends on several contextual determinants since An estimated 6.0% of the estimated global malaria burden personal protective measures, access to effective treatment, rests in WHO sub-regions 11 and 12.2 Whether irrigation and and acquired immunity factors strongly counterbalance nega- dam sites present a risk factor for malaria in these sub-regions tive effects. Although studies on large dam sites in sub- again depends on contextual determinants, which make the Saharan Africa are rare, we can conclude that irrigation pro- attribution of the fraction to these potential risk factors pres- jects and dam sites in general might not present a risk to ently impossible. First, there is a great diversity of vectors in inhabitants of stable malaria areas, in particular when control WHO sub-regions 11 and 12 and several of these (e.g., siblings programs have been launched simultaneously. It has recently of An. culicifacies or An. stephensi) have limited breeding in been estimated that approximately 10% of the African popu- irrigated rice fields.27 Conversely, a shift in vector species lation lives in epidemic, unstable, malaria risk areas.26 We composition may occur. In addition, the review of studies in therefore assume that currently approximately 0.9 million these two WHO sub-regions has shown that the local setting, people live near irrigation and large dam sites in unstable malaria endemicity, the deployment of control programs, and malaria transmission areas. Without the implementation of knowledge on the disease were key determining factors. malaria control programs, inhabitants, particularly young We retrieved studies that assessed the impact of surface children, in these settings are at high risk of disease- irrigation projects in India28–31 or Sri Lanka.32–34 We are not associated morbidity and mortality. Since the irrigated area in aware of studies assessing the impact of irrigation on malaria sub-Saharan Africa is anticipated to increase strongly, in par- prevalence or incidence in Bangladesh, Bhutan, Indonesia, ticular in arid and semi-arid environments, this number is Myanmar, Nepal, the People’s Democratic Republic Korea, likely to increase significantly. or Thailand. Southeast Asia (WHO sub-regions 11 and 12). In contrast Sharma and others have analyzed data over a 21-year pe- to WHO sub-regions 1 and 2, irrigated agriculture plays a riod, commencing in 1963 in 25 states of India, representing much greater role in southeast Asia: 10.6% of the total sur- state-wide annual parasite incidence and the area under rice face area is currently irrigated, mainly for rice production. irrigation. Significant positive associations were only found in The irrigated area is expected to further grow significantly, the two states of Punjab and Nagaland.27 However, paddy potentially up to 22.4% of the total area (Table 3). A total of cultivation did not cover a huge area and the relationship, 4,431 large dams have been built in the selected countries of which was generally poor, was only found when both sets of WHO sub-regions 11 and 12; the large majority of them in data were pooled at the state level.27,31 Studies focusing on Table 4). At maximum capacity the res- individual irrigation projects have demonstrated the impact) (4,010 ס India (n ervoirs constitute a total area of 53,265 km2. Between 145.1 irrigation has on malaria: after the implementation of the and 771 million people have been assigned to irrigation Mahi-Kadana project in India, the annual parasite index in- schemes, and 122.9–659.6 million (7.7–41.5% of the total creased from 0.01 in 1961 to 37.9 in 1976. As a consequence, population) live in malaria-endemic areas. We furthermore a malaria control program was stepped up. Two years later, estimate that 10.9 million people are at risk of malaria due to the annual parasite index in the Mahi-Kadana irrigation proj- large dam sites (Figure 3). However, this number of people at ect had decreased to 11.4.29 In Meerut and Gurgaon, the risk may be overestimated because dams and irrigation incidence in canal irrigated villages increased up to nine- schemes are not distributed homogenously between malaria- fold.28 Of particular concern are reports of malaria outbreaks endemic and non-endemic areas. For example, in the eastern due to irrigation schemes from areas that have been only 398 KEISER AND OTHERS

TABLE 4 Number of large dams, estimated area of mosquito flight range, and population at risk for the different malaria-endemic World Health Organization (WHO) sub-regions )§ 6 –8.5%) (2.8–3.1%) (4.4–4.7%) (9.1–46.3%) (0.4–3.1%) (14.5–34.7%)** (4.4–5.8%) (2.3–20.4%) Number Estimated area of mosquito flight range WHO of large for all dams in malaria-endemic area Population at risk 2 risk (×10 Population at 2.3 (0.6%) 4.0 (1.4%) sub-region dams* at full water level (km ) at full water level 1 179 28,932 2,638,598 2† 346 16,662 429,887 4 1,389 12,789 383,670 5 60 2,823 98,837‡ 7 156 2,509 1,916,876 9 656 8,053 539,577 11 346 967 351,211 12 4,085 28,846 10,600,537 )‡

6 14 1,974 5,626 1,377,245 sub-regions Total 9,191 107,207 18,336,438 (×10 4.9 (1.4%) 4.0 (1.4%) * Source: World Register of large dams.12 † Only the 25 South African dams located in malaria-endemic areas have been included.

Estimated population in ‡ The population density of WHO sub-region 4 has been applied. irrigated agricultural area were included. in irrigated areas. 2 mildly prone to malaria, e.g., the Thar desert in the Rajastan )* 6 State of India. As many as 13 epidemic outbreaks have been 73.8 4.9–5.4 (6.6–7.3%) 3.3–3.5 23.5 3.2–3.5 (13.6–14.9%) 1.8–2.0 (7.6 359.1 278.2 327.8 68.3–139.9 (20.8–42.6%) 47.6–114.0 414.8 26.3–28.8 (6.3–6.9%) 12.0–12.9 107.4 19.7–19.9 (18.3–18.5%) 4.8–6.2 298.2 20.9–132.3 (7.0–44.3%) 7.0–61.1 reported in this area up to 2002 because extensive irriga- 1,291.5 127.2–639.5 (9.8–49.5%) 117.6–598.5 1,510.7 119.5–928.4 (7.9–61.4%) 6.4–46.8 tion has altered the physiography and malaria transmis- in 2002 (×10 Total population sion parameters. An. culicifacies, which was previously un- known in the desert, has taken over from the original vectors, causing a high percentage of the Plasmodium falciparum ma- is at risk. )† 2 laria.31 km

3 In Sri Lanka, a five-fold higher malaria incidence was re- (×10 for irrigation ported following the introduction of the Mahaweli Systems H Potential area Djibouti, and the Islamic Republic of Iran and B.33 Another study comparing the malaria prevalences in 3 four villages, two relatively new villages and two ancient vil-

ABLE lages, of which two were irrigated and two non-irrigated, T )† and an estimated population density of 200 people/km 2 showed a prevalence of 4.8% in the irrigated compared with km

3 2.5% in the non-irrigated villages. However, the new villages, in 2000 (×10 risk stratified according to World Health Organization (WHO) in irrigated but also non-irrigated areas, had much higher Irrigated area malaria prevalences compared with the old villages, which was explained by changing livelihoods, less knowledge on ma- )† 2 laria, and fewer personal protection measures in the new vil- —the Global Picture”

km 32 3 area; no data for Afghanistan, Yemen, cultivable area. lages. In a more recent study, irrigated rice cultivation in the that approximately 20% of the irrigation population Uda Walawe region was found to have a lower malaria risk Area rice paddies in 9.5 (0.58%) 24.8 (1.50%) 253.3 (15.6%)¶ 2.0 (0.14%) 99.6 (6.9%) 106.1 (7.3%) 34 No record 16.2 (0.75%) No record 50.1 (0.30%) 131.7 (0.80%) 591.5 (3.6%) 18.2 (0.20%) 24.7 (0.27%) 183.5 (2.0%) 50.9 (0.52%) 20.0 (0.21%) 209.2 (2.1%) 29.7 (0.41%) 341.5 (4.7%) 105.9 (1.5%)# than non-irrigated areas. As in the African cases discussed 2000 (×10 453.3 (4.1%) 597.6 (5.4%) 745.7 (6.5%) 639.4 (14.5%) 636.0 (14.4%) 1,248.6 (28.2%) 223.8 (9.0%) 104.7 (4.2%) 275.2 (11.4%) before, these claims also presume that the two groups of com- munities were approximate ecologic replicates prior to the Livestock and Poverty )†

2 introduction of irrigation. Several studies have assessed the

km impact on dam building in southeast Asia (Table 6). For ex- 3 ample, the Bargi dam in India has been studied in consider- Agricultural (×10 area in 2000 able detail: after the construction of the Bargi dam, a 2.4-fold increase in malaria cases and a more than four-fold increase )† 2 in annual parasite incidence among children were recorded in km at risk was available for Iraq; we estimated living in malaria-endemic areas. 3 villages closer to the dam (head end) compared with more was estimated from data on the total potential

was estimated from the total potential cultivable distant villages (tail end). In addition, there was a strong in- Malaria burden, irrigated areas, and population at

Total surface crease in the prevalence rates in partially submerged villages, . area (×10 2 as seen from routinely collected malaria data in the nearby hospital.35,36 ) 3 Again, integrated vector management or other control in-

“Mapping People, Livestock Production Systems, terventions were found to have a strong influence on the

malaria in malaria transmission parameters. For example, a study car- 2002* (×10 DALYs due to ried out in Uttaranchal, India comparing the parasitologic indices in a dam area with those in forest or plain areas showed a prevalence and annual parasite incidence of 0 in the 5 25 (0.05%) 1,650 889 (53.8%) 4 86 (0.2%) 16,522 6,397 (38.7%) 2 20,785 (44.7%) 9,184 4,092 (44.5%) 1 20,070 (43.2%) 9,702 3,853 (39.6%) 6 92 (0.2%) 2,150 1,737 (80.0%) 79 2,158 (4.6%) 20 (0.05%) 7,230 1,435 3,318 (45.8%) 742 (51.7%) The potential irrigation area for Nicaragua Source: http://www.fao.org. Range obtained from Multiplication with the fraction of population 14 441 (1.0%) 11,469 5,843 (51.2%) 12 2,275 (4.9%) 4,422 2,085 (47.1%) 11 502 (1.1%) 2,483 659 (26.5%) dam area. An elevated economic status, indoor residual * Source: World Health Report, 2004 † ‡ ** No data on the percentage of the population # The potential irrigation area for Pakistan § ¶ WHO

sub-region spraying, and more awareness of malaria risk were reported IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 399

TABLE 5 Effect of irrigation on malaria prevalence and/or incidence in Africa (World Health Organization [WHO] sub-regions 1, 2, and 7)*

Overall malaria prevalence and/or incidence Study site, Irrigation scheme or period, reference Population sample construction Irrigated village Non-irrigated village Comment WHO sub-region 1 Kou valley, Children (ages 0–14 Rice fields Prevalence: 16% Prevalence: 35.4% High consumption of Burkina Faso, years) from 31 (May); 58% (May); 82.5% chloroquine due to 1985–198613 families (October) (October) better socioeconomic status Kou valley, 2,362 individuals Rice fields Prevalence: 44.5% Prevalence: 60.5% Epidemiologic paradox: Burkina Faso57 (January); 33.9% (January); 58.5% bed nets and shift in (October) (October) biting behavior? SEMRY rice 924 individuals (all age SENRY I, II: Prevalence: 3.0–7.6% Prevalence: 3.1% development, groups) 35,000-ha lake (April) (3 villages (April) (50 km Cameroon zone for rice close to lake and from lake) 198158 development; rice irrigation) 5,300-ha rice SEMRY II rice 4,611 children (ages SENRY I, II: Before irrigation: scheme 2–9 years) 35,000-ha lake Prevalence: 13.8% Mayo-Danai zone for rice (March 1979) Cameroon development; After irrigation: 1979–198559 5,300 ha rice Prevalence: 30.1% (November 1979), 11.5% (March 1981), 7.1% (November 1981), 12.9% (April 1985) The Gambia, 1,465 children (ages River Gambia Prevalence: 34.2% Prevalence: Low prevalence rates 199160 1–4 years) and rice 28.7–71.2% close to productive swamps breeding sites (could be explained by use of bed nets) Niono, Mali 3,669 children (ages Rice irrigation Prevalence: 31.3% Prevalence: 46.5% 199515 0–14 years) Annual incidence: 0.3 Annual incidence: per 1,000 0.7 per 1,000 Senegal river Children (ages 0–9 Rice irrigation Prevalence: 8.7% Prevalence: 16.5% No increase in central valley, years) (July); 8.3% (July); 7.1% transmission Senegal14 (November) (November) Ferlo’s water 1,548 children (ages Prevalence: 37.6% Prevalence: 34.3% launching, 0–14 years) (September, (September, Senegal, 199661 October) October) Moyamba district, 1,106 individuals Rice swamps Prevalence 38.8–58% Prevalence: 57.1% Sierra Leone, (several swamp 199162 sites) WHO sub-region 2 Burundi, 198163 Not given Rice fields Prevalence: 24.4% Prevalence: 4.5% (February); 69.2% (January); 29.6% (June) (July) Coˆ te d’lvoire64 Rice production Single cropping: No irrigation: Different immunity systems annual malaria annual malaria acquisition within the incidence: 0.6 per incidence 0.9 per 3 rice agro-eco- 1,000 1,000 systems, double Double cropping: cropping system annual malaria extends malaria risk incidence 0.8 per in the beginning of 1,000 the dry season Lower Moshi 2,951 children (ages Rice fields and Prevalence: 12.5% Prevalence: 29.4% Use of bed nets due to area Tanzania 1–4 years) sugar cane (rice field); 16.9% elevated socio- 1994–199517 fields (sugar cane) economic status WHO sub-region 7 Gezira, Sudan40 Population of Gezira: Rice, groundnut, Before irrigation: estimated at 2 and vegetables malaria was not an million important health problem in the area After introduction of irrigation: malaria prevalence up to 20% .hectare סha* 400 KEISER AND OTHERS to be the main factors accounting for the lack of malaria where houses were located closely to fields and irrigation transmission at the dam site.37 In addition, in Thailand, no canals compared with villages in the dry areas.41 The second increase of malaria incidence was observed near the Nong study was carried out in the Lao People’s Democratic Repub- Wai dam and the Ubol Ratana dam. However, this is prob- lic. The malaria infection rate was higher in villages sur- ably because all walls inside of houses were sprayed with rounded by rice fields compared with non-irrigated villages.42 DDT compared with the Srinagarind dam, where an increase Finally, in Turkey, the implementation of a network of irri- in malaria prevalence was reported, but where there was no gation channels and a subsequent domestic migration from mention of any vector control measures.38,39 malaria-endemic regions to the area caused a serious epi- Eastern Mediterranean (WHO sub-regions 6 and 7). Irri- demic outbreak.43 gated areas range from 0.04% of the total surface area in The health impacts of three large Brazilian dams, namely 1) Djibouti, 0.45% in Somalia, 0.94% in Yemen, 1.5% in Sudan, the Balbina power plant, 2) the Itaipú dam, and 3) the Tucu- 4.6% in the Islamic Republic of Iran, 6.2% in Afghanistan, ruí Hydropower dam have been studied in detail. We sum- and 8% in Iraq to 22% in Pakistan. These percentages cor- marize data on the malaria incidence before and after their respond to an estimated irrigation population ranging from construction in Table 6. An increase of malaria cases was 0.09% in Yemen to 73% in Pakistan. We allocated between reported at all three sites.44–47 Overall, these studies show 71.5 and 143.4 million people (with the great majority in Pak- that despite a limited malaria burden and a small population istan) to irrigation (Table 3); 49.4–116 million of these live in at risk, irrigation and large dam sites might have a strong malaria-endemic areas. influence on disease parameters in WHO sub-region 4. There are 156 large dams located in these countries, which report malaria as a health problem and are part of WHO sub-region 7. The majority of these large dams are located in DISCUSSION .(66 ס and the Islamic Republic of Iran (n (71 ס Pakistan (n We estimate that in these regions 1.9 million people live We have presented a systematic review of the literature to within the estimated mosquito flight range of 2,509 km2, and provide numbers of the current population living at risk of thus might be at risk of acquiring malaria (Table 4). We could malaria due to proximity to irrigated agro-ecosystems and not retrieve data on the size of the reservoirs for the 38 large large dams, and to assess the impact of these types of water dams in Saudi Arabia (WHO sub-region 6). resource development on the burden of malaria at regional The Eastern Mediterranean (WHO sub-regions 6 and 7) and global scale. We have highlighted that the estimated total have 4.8% of the current estimated global malaria burden.2 population living in proximity to the reservoirs of large dams New water resource development projects were reported to in malaria-endemic areas is small, namely 18.3 million people, increase malaria transmission in Afghanistan and in the with the majority of them living in India. Conversely, as many Gezira scheme in Sudan.27,40 However, this data is insuffi- as 851.3 million people live in or close to irrigation systems in cient to determine the attributable fraction of irrigation and malaria-endemic areas. In WHO sub-regions 1 and 2, where large dam sites to the malaria burden; thus, further studies are 87.9% of the currently estimated global malaria burden is warranted. concentrated, only 9.4 million people are living near large The Americas, Europe, and the Western Pacific sub-regions dam reservoirs and irrigation sites. In contrast, the remaining (WHO sub-regions 4, 5, 9, and 14). Only 1.3% of the global WHO sub-regions of the world, where malaria is also en- malaria burden is currently estimated to occur in WHO sub- demic, have a maximum of 860.3 million people near large regions 4, 5, 9, and 14.2 Irrigated areas account for less than dam and irrigation sites, but here, only 12.1% of the global 1% (WHO sub-region 4) and up to 6.9% (WHO sub-region 9) malaria burden rests. of the total surface area, as shown in Table 3. A large popu- It was not possible to quantify the attributable fraction of lation (170.4–982.5 million people) can be associated with ir- the malaria burden due to dam building and irrigation for the rigation. However, the majority of these individuals live in individual WHO sub-regions (e.g., using the methodology of parts of the countries where no malaria transmission occurs comparative risk assessment)48 due to many confounding fac- (e.g., in the non-malarious parts of China); only 1.2–3.3% of tors and the scarcity of the currently available global data- the irrigation population (26.5–69.4 million people) is esti- base. Sadly, even the extensive report authored by the World mated to live in malaria-endemic areas. Commission on Dams, derived from 17 exhaustive reviews on A total of 4,079 large dams have been constructed in the dams, allocates a mere 2 pages to health.49 At a given loca- countries of WHO sub-regions 4, 5, 9, and 14 and are included tion, if malaria incidence or prevalence data are available in our review. The countries with the highest number of large both before and after the introduction of a dam and/or an -irrigation scheme, we can ascertain the impact of the envi ס Turkey (n ,(1,905 ס dams in these regions are China (n The reservoir ronmental transformation. However, most extant studies .(536 ס and Mexico (n ,(594 ס Brazil (n ,(625 areas range from 385 km2 (large dams of WHO sub-region 5) based their results on the comparison of two villages. Care is to 58,480 km2 (large dams of WHO sub-region 12). We esti- needed in the interpretation of these results because many mate that a total of 2.3 million people are living close enough studies comparing malaria rates in villages proximal to a wa- to reservoirs in endemic areas; thus, they are at risk of malaria ter resource development project with villages that are rela- transmission (Table 4 and Figure 3). tively distant do not give a clear picture of the extent to which We retrieved only three studies assessing the impact of nearby villages were approximate ecologic matches/replicates irrigation on the malaria incidence and prevalence in the se- of the distant villages prior to the introduction of the water lected countries of these WHO sub-regions. The first is a resource project. There might be subtle differences in eco- recent study conducted in a dry coastal area of Peru, where logic, epidemiologic, and socioeconomic features; thus, result- malaria incidence was found to be five-fold higher in villages ing in different transmission characteristics, even in neighbor- IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 401 After construction Paraguayan side contributed tointroduction the of cases onside. the Brazilian side: 43 in 1986,Paraguayan 1,084 side: in 1,707 1989. in1989 1986, 4,883 in outbreak was registered inDDT 1989; spraying thus, was reintroduced compared with distant villagelake 14 (25%) km from depending on village). In 1968–1969: the lake, suggesting year-round transmission, compared to 27.3%29.6% and downstream of therespectively dam, cases in all districts,than however, number no of greater casesdisease in respiratory increased annually by 21% 1982. 1980, 7.5 in 1982, and 4.4 in 1989 0.6% (0.2–1.1%, depending onLow village). prevalence due to DDT spraying construction, peaking to more10,000 than cases in 1984. sub-regions Precarious control measures on the Number of autochthonous cases: Brazilian Malaria incidence increased in 1988, and an Prevalence: village close to lake (36%) Prevalence: 0% (fifth month of dry season) Prevalence in 1967–1968: 1.2% (0.4–2.1%, Prevalence: in July 1994 up to 47% around Great upward trend in number of malaria 1981–1984: 837 cases; number of cases Positivity index was 6.8 in 1977 andAPI 0.8 was in 192.7 in 1977, 131.1 in 1978, 0 in Number of cases increased after API was 60.4 in 1980 and 26.7 in 1996. Malaria prevalence/incidence Before construction ´ episodes after flooding events. 1.15%, although it includedmunicipalities some endemic for malaria that were only indirectlyby affected Itaipu during rainy season July–September) transmission from January to July health center and the positivity index was 0.13. 1975. Malaria was epidemic with seasonal 1975–1976: the positivity index was Unstable malaria transmission (only Seasonal transmission, little August 1981: 143 cases in Riakanau In 1972 the population was small 106 positive cases in 1962 and 251API in was 29.6 in 1970. 6 ABLE T or incidence in different World Health Organization (WHO) 176 meters, ס 1.5 km); artificial lake Characteristics of dam ס dam for hydropower. Phase I dam (height ´ completed in 1983 1974 to regulate Edeapower hydroelectric 1980, comprises retention lakelong 10 and km 15 kmgravity wide irrigation. to Displacement facilitate of 9,000 people completed 1966–1967 for irrigation purposes length provide hydropower and irrigation irrigation along the riverprevent and saline to water frominland flowing 1981) initiated in 1977; reservoir construction commenced in 1981, project begun operation inAfter 1989. registering an annualindex parasite (API) of 192.7workers when came the to first themeasures area; were control initiated construction was started inwas 1975 completed and in 1984.began Phase in II 1998. Manatali dam completed in 1987 to Diama dam completed 1986 for Tana River lakes (e.g., Masinga dam Itaipu Construction of the power plant Tucuruí Dam with retention lake completed in Dam across Gorgol River completed in Nong Wai dam, Ubol Ratana dam Population sample Effect of large dam construction on malaria prevalence 567 individuals 525 individuals 8,931 individuals 44 ´ State, 45,46 47 ´ 23 38 municipal 65 Hydropower, Study site, 22 dam, Parana 24 66 ´ period, reference Presidente Figueiredo municipality, Brazil Cameroon 1984 area State, Brazil Senegal area Thailand Tucuruí Brazil, 1975 district, Para Balbina power plant, Bamendjjin dam area, Gleita dam, Mauretania Mali Manatali dam St. Louis Diama dam, Tucuruí Kenya Tana river lake Itaipu Khon Kaen Province, WHO sub-region 1 WHO sub-region 2 WHO sub-region 4 WHO sub-region 11 402 KEISER AND OTHERS

ing villages. In addition, the possible presence of multiple malaria control interventions in the two sets of localities makes clear interpretation of claims about impact of dams and irrigation schemes on at-risk population difficult to inter- pret, since most studies do not give sufficient attention to this issue. Our calculations depended on a number of assumptions and they are therefore inevitably subject to a level of uncer- tainty. The possibility that we have overestimated the risk cannot be ruled out. First, we assumed that the whole population assigned to irrigation in malaria-endemic areas is at risk of the disease. However, not all forms of irrigation actually present a risk for the local population. There are three com- villages: 49.4%; in submerged71.4%. villages: district: 4,279 in 1996. Number of cases infrom head dam): end 2.4-fold (44–50 highertail km compared end to villages (75–78 km) 48.5% in forest villagearea and 1.8% in plain state (population has increased by 2.5%) Indira Ghandi canal; thus,malaria recent outbreak) focal mon classes of irrigation systems, namely 1) pressurized dis- Hospital prevalence in partially submerged Health post records for Naranyanga Prevalence: 38% (mass survey of children) Prevalence: 0% in dam area compared to Prevalence in 1976: 25% (on dam site) In 1994: 229,772 positive cases in Rajasthan Prevalence: 85% (in two villages near tribution as in sprinkler or trickle systems, 2) gravity flow distribution as in surface irrigation, and 3) subsurface irrigation. If they are well maintained, sprinkler irrigation, drip irrigation, and subsurface irrigation provide irrigation water without creating suitable breeding sites for Anopheles vectors.50 Second, we did not include annual fluctuations of the water level of the reservoir, which in turn has important implica- tions for the estimation of the population at risk from large

in 1972: 16% dam sites. At the end of the low water period, the area of the reservoir, and thus the mosquito flight range area is, in general, considerably reduced. For example, the reservoir area of 2 39%. district: 184 cases in 1979 (preliminary survey) Rajasthan state the Manantali dam in Mali decreases from 477 km to 275 Hospital prevalence in dry villages: Health post records for Naranyanga In 1961: 8,494 positive cases in km2 at the minimum operating level of the dam.24 Further- more, not every dam reservoir might actually be a good breeding site for malaria vectors. Each Anopheles species is characterized by specific habitat preferences, including expo- sure to sunlight, turbidity of the water, presence of vegetation, pH, and nitrate and phosphate concentrations of the water.51 These environmental factors are specific for each dam and its shoreline. In addition, settlement around the reservoirs might not be possible at certain locations due to topography and other reasons. Third, we have assumed that the population densities around dams are similar to the ones of the resettled commu- nities and that the population near the dam is subjected to the and irrigation projects completed1988 in marshy conditions) systems (Gang, Sirhind feeder,Ghandi) Indira for irrigation, hydropower, water supply same population growth as the rural areas of the respective Bargi dam multipurpose hydropower Nanak Matta dam (created swampy, Canalization project with 3 major canal Srinagarind dam completed in 1978 Prevalence countries. New villages might have been constructed further away than the 2 km (estimated mosquito flight range, Appen- dix 1) from the dam sites and, consequently, the population would not be at risk attributable to the dam. Conversely, dam sites are characterized by marked demographic impacts, in particular during the construction and early operational phases. These sites attract visitors, fishermen, and farmers who often have low immunities to malaria. Thus, during construction, the population might be larger than before. As the dam ages, however, temporary workers leave and the popu- febrile cases. febrile cases. survey. villages patients in dam area;forest 272 area in and 849area in plain lation density consequently decreases. 602 individuals 2,016 blood smears from 1,714 blood smears from 379 children for mass blood Active case detection in 10 56 blood smears from febrile Fourth, since a geo-referenced database exists only for the African dams, it is difficult to determine the exact population at risk from dams in countries that are only partially endemic 37 for malaria on the remaining continents. Without knowledge of the geographic coordinates, we would have presumed that 68,69 Ն 20% or more than 100 of the 539 South African large dams 39 and their reservoirs are located in malaria-endemic areas. In

35,36,67 reality, however, only 25 (< 5%) of these are located in areas where malaria transmission occurs. Similarly, calculation of Thailand Uttaranchal, India Thar desert Jabalpur, Bargi dam Kanchanari Province, WHO sub-region 12 the irrigation population at risk in partially malaria-endemic IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 403

FIGURE 3. Malaria burden and estimated at-risk populations due to proximity to irrigation and large dam sites in the different World Health .disability adjusted life years ס Organization (WHO) sub-regions where malaria is endemic. DALYs countries is based on the assumption that the total population Furthermore, an estimated 15,000 small dams have been con- and the population living near irrigation schemes are equally structed in Zimbabwe, and more than 50,000 small dams were at risk of the disease. built in Kenya within three years during the late 1950s.52 It is also conceivable that we might have underestimated Finally, studies investigating the consequences of the con- the actual population at risk of malaria from water resource struction of flood control, water projects for recreational pur- development, which is justified on the following grounds. poses, or pumps and drains for water supply and sanitation on First, we could not include the impact of a large dam on malaria have, to our knowledge, not been conducted. It fol- malaria downstream of the project site. However, the change lows that no estimates of their impact on malaria could be of the water regimen can stretch for many kilometers and presented in this review. strongly influence larval breeding. When irrigation schemes and dams are proximal to areas of Second, it is unfortunate that no systematic inventory of unstable transmission, integrated multiple-intervention ma- small dams and only very few studies assessing their cumula- laria control holds promise for mitigation. In several of the tive impact on malaria are currently available. Their impact studies, which we have reviewed here, malaria control pro- on the frequency and transmission dynamics of malaria could grams, consisting mainly of early diagnosis and treatment, be significant because their total shoreline is much greater residual spraying, or distribution of ITNs, have been success- when compared with large dams. For example, 1,110 Nigerian fully conducted. It is important to note that environmental small dams were described to have an area of 400,000 ha management presents an additional option for malaria con- compared with a surface area of 116,000 ha of 34 large dams. trol in such settings. For example, vector control by means of 404 KEISER AND OTHERS water management has been carried out with success for sev- 5. Gujja B, Perrin M, 1999. A Place for Dams in the 21st Century. eral decades, particularly in areas where malaria is unstable. Discussion Paper. Washington, DC: World Wildlife Fund. The first studies on intermittently irrigated rice fields, which 6. Thornton PK, Kruska RL, Henninger N, Kristjanson PM, Reid RS, Atieno F, Odero AF, Ndegwa T, 2002. Mapping Poverty led to greatly reduced Anopheles densities and often in- and Livestock in the Developing World. London: Department creased rice yields, were carried out more than 70 years ago.53 for International Development. At the same time, elimination of mosquito breeding sites has 7. Mwadime RK, Omwega AM, Kielmann N, Korte R, 1996. Pre- been achieved in rivers and streams of Sri Lanka and Malay- dictors of nutritional status among participants in a rice irrigation scheme in Kenya. Ecol Food Nutr 35: 263–274. sia by means of different types of siphons and small dams.54 8. Waiyaki PG, 1987. The history of irrigation development in Significant reductions of Anopheles breeding sites has been Kenya and the associated spread of schistosomiasis. Effects of achieved in the reservoirs of the Tennessee River Valley by Agricultural Development on Vector-Borne Diseases. Seventh implementation of several types of environmental and water annual meeting of the Joint WHO/FAO/UNEP Panel of Ex- management measures. Among them was an integrated op- perts on Environmental Management for Vector Control. Rome: FAO, 23−26. erating rule consisting of a fluctuation cycle with an amplitude 9. Cernea M, 1997. Hydropower Dams and Social Impacts. A So- 24,55 of 0.3 meters over 7−10-day periods. ciological Perspective: Washington, DC: The World Bank. We conclude that future water resource development proj- 10. Cernea M, 1997. African Involuntary Population Resettlement in ects should include in-depth assessment of potential health a Global Context: Washington, DC: The World Bank. effects, positive or negative, including malaria, where this dis- 11. United Nations, 2002. World Urbanization Prospects: The 2001 Revisions. New York: Population Division Department of ease is endemic. Indeed, institutionalization of health impact Economics and Social Affair of the United Nations. assessments for development projects quite generally, analo- 12. International Commission on Large Dams, 1998. World Register gous to environmental impact assessments, would lead to in- of Dams (computerized version). Paris: International Commis- formation requirements that could fill many of the data gaps sion on Large Dams. described in this report.56 Introduction of sound monitoring 13. Boudin C, Robert V, Carnevale P, Ambroise-Thomas P, 1992. Epidemiology of Plasmodium falciparum in a rice field and a and surveillance systems proximal to such water projects savanna area in Burkina Faso. Comparative study on the ac- would facilitate systematic evaluation of the impact of these quired immunoprotection in native populations. Acta Trop 51: ecosystem interventions over time. This, in turn, would 103–111. greatly improve our understanding of the role of dams and 14. Faye O, Fontenille D, Herve JP, Diack PA, Diallo S, Mouchet J, 1993. Malaria in the Saharan region of Senegal. 1. Entomo- irrigation systems in either promoting or reducing malaria logical transmission findings. Ann Soc Belg Med Trop 73: 21– transmission. In addition, mitigation strategies to alleviate po- 30. tential negative health effects, of which malaria might be only 15. Sissoko MS, Dicko A, Briet OJ, Sissoko M, Sagara I, Keita HD, one component, are mandatory to reduce the current burden Sogoba M, Rogier C, Touré YT, Doumbo OK, 2004. Malaria of malaria in settings near irrigation or dam projects, particu- incidence in relation to rice cultivation in the irrigated Sahel of Mali. Acta Trop 89: 161–170. larly in areas where malaria transmission is unstable. 16. Ijumba JN, Lindsay SW, 2001. Impact of irrigation on malaria in Africa: paddies paradox. Med Vet Entomol 15: 1–11. Received April 9, 2004. Accepted for publication August 13, 2004. 17. Ijumba JN, Shenton FC, Clarke SE, Mosha FW, Lindsay SW, 2002. Irrigated crop production is associated with less malaria Financial support: This work was part of the project Burden of Wa- than traditional agricultural practices in Tanzania. Trans R Soc ter-Related Vector-Borne Diseases: An Analysis of the Fraction At- Trop Med Hyg 96: 476–480. tributable to Components of Water Resources Development and 18. Marrama L, Jambou R, Rakotoarivony I, Leong Pock Tsi JM, Management, which was kindly funded by the World Health Orga- Duchemin JB, Laventure S, Mouchet J, Roux J, 2004. Malaria nization. Marcia Caldas de Castro is grateful to the Center for Health transmission in southern Madagascar: influence of the envi- and Wellbeing at Princeton University and Jennifer Keiser and Jürg ronment and hydro-agricultural works in sub-arid and humid Utzinger to the Swiss National Science Foundation (Project no. regions. Part 1. Entomological investigations. Acta Trop 89: PMPDB-106212 and PPOOB–102883) for financial support. 193–203. Authors’ addresses: Jennifer Keiser, Marcel Tanner, and Jürg Utz- 19. Dolo G, Briet OJT, Dao A, Traoré SF, Bouaré M, Sogoba N, inger: Swiss Tropical Institute, PO Box, CH-4002 Basel, Switzerland. Niare O, Bagayogo M, Sangare D, Teuscher T, Toure´ YT, Marcia Caldas de Castro Department of Geography, University of 2004. Malaria transmission in relation to rice cultivation in the South Carolina, Callcott Hall-125, Columbia, SC 29208. Michael F. irrigated Sahel of Mali. Acta Trop 89: 147–159. Maltese, St. Antony’s College, Oxford University, Oxford OX2 6JF, 20. USAID, 1994. Senegal River Basin Health Master Plan Study. United Kingdom. Robert Bos, Water, Sanitation and Health (WSH/ Washington, DC: United States Agency for International De- PHE), World Health Organization, Avenue Appia 20, CH-1211 velopment. Geneva 27, Switzerland. Burton H. Singer, Office of Population Re- 21. Coosemans M, Mouchet J, 1990. Consequences of rural develop- search, Wallace Hall, Princeton University, Princeton, NJ 08544. ment on vectors and their control. Ann Soc Belg Med Trop 70: 5–23. Reprint requests: Jennifer Keiser, Swiss Tropical Institute, PO Box, 22. Baudon D, Robert V, Darriet F, Huerre M, 1986. Impact of CH-4002 Basel, Switzerland, Telephone: 41-61-225-2666, Fax: 41-61- building a dam on the transmission of malaria. Malaria survey 225-2678, E-mail: [email protected]. conducted in southeast Mauritania. Bull Soc Pathol Exot Fil- iales 79: 123–129. REFERENCES 23. Atangana S, Foumbi J, Charlois M, Ambroise-Thomas P, Ripert C, 1979. Epidemiological study of onchocerciasis and malaria 1. Breman JG, 2001. The ears of the hippopotamus: manifestations, in Bamendjin dam area (Cameroon). Malacologic fauna and determinants, and estimates of the malaria burden. Am J Trop risks of schistosomian introduction. Med Trop (Mars) 39: 537– Med Hyg 64 (Suppl): 1–11. 543. 2. WHO, 2004. The World Health Report 2004 - Changing History. 24. Jobin WR, 1999. Dams and Disease: Ecological Design and Geneva: World Health Organization. Health Impacts of Large Dams, Canals, and Irrigation Systems. 3. WHO, 1997. Health and Environment in Sustainable Develop- London: Routledge. ment. Geneva: World Health Organization. 25. Ghebreyesus TA, Haile M, Witten KH, Getachew A, Yohannes 4. Service MW, 1991. Agricultural development and arthropod- AM, Yohannes M, Teklehaimanot HD, Lindsay SW, Byass P, borne diseases: a review. Rev Saude Publica 25: 165–178. 1999. Incidence of malaria among children living near dams in IRRIGATION AND LARGE DAMS AND MALARIA BURDEN 405

northern Ethiopia: community based incidence survey. BMJ ment: The Need for Intersectoral Negotiation. Geneva: World 319: 663–666. Health Organization. 26. Snow RW, Craig M, Deichmann U, Marsh K, 1999. Estimating 47. World Commission on Dams, 2000. Tucurui Hydropower Com- mortality, morbidity and disability due to malaria among Af- plex Brazil. Cape Town, South Africa: World Commission on rica’s non-pregnant population. Bull World Health Organ 77: Dams. 624–640. 48. Ezzati M, Lopez AD, Rodgers A, Vander Hoorn S, Murray CJM, 27. Sharma VP, Srivastava A, Nagpal BN, 1994. A study of the re- Comparative Risk Assessment Collaborating Group, 2002. Se- lationship of rice cultivation and annual parasite incidence of lected major risk factors and global and regional burden of malaria in India. Soc Sci Med 38: 165–178. disease. Lancet 360: 1347–1360. 28. Sharma VP, Uprethy HC, 1982. Preliminary studies on irrigation 49. Sleigh AC, Jackson S, 2001. Dams, development, and health: a malaria. Indian J Malariol 19: 139–142. missed opportunity. Lancet 357: 570–571. 29. Jayaraman TK, 1982. Malaria impact of surface irrigation pro- 50. Amerasinghe FP, 1987. Changes in irrigation techniques as a jects: a case study from Gujarat, India. Agric Environ 7: 23–34. means to control disease vector production. Effects of Agricul- 30. Mistry JF, Purohit MU, 1990. Environmental effects of the Mahi- tural Development on Vector-Borne Diseases. Seventh Annual Kadana and Dharoi projects at Gujarat, India. Water Re- Meeting of the Joint WHO/FAO/UNEP Panel of Experts on sources J 164: 93–97. Environmental Management for Vector Control. (Rome: 31. Tyagi BK, 2004. A review of the emergence of Plasmodium fal- FAO), 82−86. ciparum-dominated malaria in irrigated areas of the Thar Des- 51. van der Hoek W, Amerasinghe FP, Konradsen F, Amerasinghe ert, India. Acta Trop 89: 227–239. PH, 1998. Characteristics of malaria vector breeding habitats 32. Dixon RA, Pinikhana JP, 1994. Malaria and proximity to irriga- in Sri Lanka: relevance for environmental management. South- tion projects: a parasitaemia prevalence study from Sri Lanka. east Asian J Trop Med Public Health 29: 168–172. Mosq Borne Dis Bull 11: 116–121. 52. Jewsbury JM, Imevbore AMA, 1988. Small dam health studies. 33. International Irrigation Management Institute, 1986. Proceedings Parasitol Today 4: 57–59. of the Workshop on Irrigation and Vector-Borne Disease 53. Keiser J, Utzinger J, Singer BH, 2002. The potential of intermit- Transmission. Colombo, Sri Lanka: International Irrigation tent irrigation for increasing rice yields, lowering water con- Management Institute. sumption, reducing methane emissions, and controlling ma- 34. Klinkenberg E, van der Hoek W, Amerasinghe FP, 2004. A malaria in African rice fields. J Am Mosq Control Assoc 18: 329– laria risk analysis in an irrigated area in Sri Lanka. Acta Trop 340. 89: 215–225. 54. Konradsen F, van der Hoek W, Amerasinghe FP, Mutero C, Boelee E, 2004. Engineering and malaria control: learning 35. Singh N, Mehra RK, Sharma VP, 1999. Malaria and the Nar- from the past 100 years. Acta Trop 89: 99–108. mada-river development in India: a case study of the Bargi dam. Ann Trop Med Parasitol 93: 477–488. 55. Kitron U, Spielman A, 1989. Suppression of transmission of malaria through source reduction: antianopheline measures ap- 36. Singh N, Mishra AK, 2000. Anopheline ecology and malaria plied in Israel, the United States, and Italy. Rev Infect Dis 11: transmission at a new irrigation project area (Bargi Dam) in 391–406. Jabalpur (central India). J Am Mosq Control Assoc 16: 279– 56. Kemm J, Parry J, Palmer J, 2004. Health Impact Assessment: 287. Concepts, Theory, Techniques and Applications. Oxford, 37. Shukla RP, Sharma SN, Kohli VK, Nanda N, Sharma VP, Sub- United Kingdom: Oxford University Press. barao SK, 2001. Dynamics of malaria transmission under 57. Carnevale P, Robert V, 1987. Introduction of irrigation in changing ecological scenario in and around Nanak Matta Dam, Burkina Faso and its effect on malaria transmission. Effects of Uttaranchal, India. Indian J Malariol 38: 91–98. Agricultural Development on Vector-Borne Diseases. Seventh 38. Harinasuta C, Jetanasen S, Impand P, Maegraith BG, 1970. Annual Meeting of the Joint WHO/FAO/UNEP Panel of Ex- Health problems and socio-economic development in Thai- perts on Environmental Management for Vector Control. land. Southeast Asian J Trop Med Public Health 1: 530–552. (Rome: FAO), 57−67. 39. Bunnag T, Sornmani S, Pinithpongse S, Harinasuta C, 1979. Sur- 58. Couprie B, Claudot Y, Same-Ekobo A, Issoufa H, Leger- veillance of water-borne parasitic infections and studies on the Debruyne M, Tribouley J, Ripert C, 1985. Epidemiologic study impact of ecological changes on vector mosquitoes of malaria of malaria in the rice-growing regions of Yagoua and Maga after dam construction. Southeast Asian J Trop Med Public (North Cameroon). Bull Soc Pathol Exot Filiales 78: 191–204. Health 10: 656–660. 59. Audibert M, Josseran R, Josse R, Adjidji A, 1990. Irrigation, 40. el Gaddal AA, Haridi AAM, Hassan FT, Hussein H, 1985. Ma- schistosomiasis, and malaria in the Logone Valley, Cameroon. laria control in the Gezira-Managil irrigated scheme of the Am J Trop Med Hyg 42: 550–560. Sudan. J Trop Med Hyg 88: 153–159. 60. Thomson MC, D’Alessandro U, Bennett S, Connor SJ, 41. Guthmann JP, Llanos-Cuentas A, Palacios A, Hall AJ, 2002. Langerock P, Jawara M, Todd J, Greenwood BM, 1994. Ma- Environmental factors as determinants of malaria risk. A de- laria prevalence is inversely related to vector density in The scriptive study on the northern coast of Peru. Trop Med Int Gambia, West Africa. Trans R Soc Trop Med Hyg 88: 638–643. Health 7: 518–525. 61. Faye O, N’Dir B, Correa J, N’Dir O, Gaye O, Bah IB, Dieng T, 42. Kobayashi J, Somboon P, Keomanila H, Inthavongsa S, Nam- Dieng Y, Diallo S, 1998. Evaluation of parasitic risks related to banya S, Inthakone S, Sato Y, Miyagi I, 2000. Malaria preva- the revitalization of the Ferlo fossil valley (Senegal). Dakar lence and a brief entomological survey in a village surrounded Med 43: 183–187. by rice fields in Khammouan province, Lao PDR. Trop Med 62. Gbakima AA, 1994. Inland valley swamp rice development: ma- Int Health 5: 17–21. laria, schistosomiasis, onchocerciasis in south central Sierra 43. Gratz NG, 1987. The effect of water development programmes Leone. Public Health 108: 149–157. on malaria and malaria vectors in Turkey. Effects of Agricul- 63. Coosemans MH, 1985. Comparison of malarial endemicity in a tural Development on Vector-Borne Diseases. Seventh Annual rice-growing area and a cotton-growing area of the Rusizi Meeting of the Joint WHO/FAO/UNEP Panel of Experts on Plain, Burundi. Ann Soc Belg Med Trop 65: 187–200. Environmental Management for Vector Control. Rome: FAO, 64. Henry MC, Rogier C, Nzeyimana I, Assi SB, Dossou-Yovo J, 27−28. Audibert M, Mathonnat J, Keundjian A, Akodo E, Teuscher 44. Chagas JA, Barroso MAB, Amorim RD, Robles CRQ, 1982. T, Carnevale P, 2003. Inland valley rice production systems Controle da malária em projeto hidrelétrico no Estado do and malaria infection and disease in the savannah of Côte Amazonas. Rev Bras Malariol Doencas Trop 34: 68–81. d’Ivoire. Trop Med Int Health 8: 449–458. 45. Consolim J, Luz E, Pellegrini NJ, Torres PB, 1991. The Anoph- 65. Sow S, de Vlas SJ, Engels D, Gryseels B, 2002. Water-related eles (Nyssorhynchus) darlingi root, 1926 and malaria in the disease patterns before and after the construction of the Itaipú lake, Paraná, Brazil: a revision. Arq Biol Tecnol 34: Diama dam in northern Senegal. Ann Trop Med Parasitol 96: 263–286. 575–586. 46. Hunter JM, 1993. Parasitic Diseases in Water Resources Develop- 66. Roggeri H, 1985. African Dams: Impacts in the Environment: The 406 KEISER AND OTHERS

Social and Environmental Impact of Dams at the Local Level: We found that those dams with a reservoir size <1 km2 were Those dams .(2.5 ס A Case Study of Five Man-Made Lakes in Eastern Africa. Nai- close to square shapes (median ratio b/l robi, Kenya: Environment Liaison Centre. 2 67. Singh N, Shukla M, Chand S, Sharma V, 1997. Outbreak of fal- with reservoir sizes >100 km had a much longer base than .(0637.4 ס ciparum malaria in submerged villages of Narayanganj PHC, length (median b/l district Mandla due to Narmada irrigation project, central In- We assumed a mosquito flight range of 2 km, which is dia (Madhya Pradesh). Curr Sci 73: 686–691. justified on the following grounds. First, some individual mos- 68. Tyagi BK, Chaudhary RC, 1997. Outbreak of falciparum malaria in the Thar Desert (India), with particular emphasis on physi- quito species might have a very long flight range, up to 12 km ographic changes brought about by extensive canalization and have been reported for Anopheles sinensis; however, the great their impact on vector density and dissemination. J Arid En- majority (66.5%) of An. sinensis were recaptured 1–3km viron 36: 541–555. from their release points.70 Second, An. darlingi and An. al- 69. Tyagi BK, Yadav SP, Sachdev R, Dam PK, 2001. Malaria outbreak in the Indira Gandhi Nahar Pariyojna command area in bimanus were abundant in houses <1 km from the river and 71 Jaisalmer district, Thar Desert, India. J Commun Dis 33: 88– not present in houses further away. Third, WHO has sug- 95. gested to locate villages 1.5–2 km from the edge of the res- 70. Cho SH, Lee HW, Shin EH, Lee HI, Lee WG, Kim CH, Kim JT, ervoir, which has proven successful in reducing malaria inci- Lee JS, Lee WJ, Jung GG, Kim TS, 2002. A mark-release- 72 recapture experiment with Anopheles sinensis in the northern dence. Thus, we calculated the area 2 km around the hypo- part of Gyeonggi-do, Korea. Korean J Parasitol 40: 139–148. thetically rectangular reservoirs for all four groups applying ×b × 2)+2) × 2 ס Roberts DR, Manguin S, Rejmankova E, Andre R, Harbach RE, the following formula: Area at risk .71 Vanzie E, Hakre S, Polanco J, 2002. Spatial distribution of (l × 2)+22␲ (Figure 1). adult Anopheles darlingi and Anopheles albimanus in relation to riparian habitats in Belize, Central America. J Vector Ecol For the 226 large dams in WHO sub-region 2, we deter- 2 51 ס mined a total flight range area of 11,578 km (mean .30–21 :27 72. WHO, 1982. Manual on Environmental Management for Mos- km2/reservoir). Since we have no data on the reservoirs of the quito Control. Geneva: World Health Organization. remaining 120 dams in this WHO sub-region, we assume that they have a similar average flight range, namely 51 km2/ APPENDIX 1 reservoir. Consequently, the estimated total area for all 346 ESTIMATION OF AT-RISK POPULATION DUE TO 2 registered large dams in WHO sub-region 2 is 17,726 km . PROXIMITY OF LARGE DAM SITES (TWO Using the percentage of population in malaria-endemic re- EXAMPLES FOR WORLD HEALTH gions for each country, we obtained an estimate of the mos- ORGANIZATION (WHO) SUB-REGIONS 2 AND 4) quito flight range around large dams in endemic areas. In our An estimated 346 large dams are located in the malaria- example of WHO sub-region 2, we assume that 94% of the endemic countries of WHO sub-region 2. Only 25 of the 539 dams’ surface areas are located in endemic areas. South African dams are included in this number because only Multiplication of the mosquito flight range in malaria- these are located in areas where malaria is endemic. We clas- endemic areas with the obtained population density for WHO sified 226 of the dams (where detailed information on the sub-region 2 (25.8 persons/km2; Table 2) gave an estimate of area of the reservoir is available12) according to their reser- the at-risk population of 429,887. voir surface into four groups as follows: 1) <1 km2,2)1–10 For WHO sub-regions 4–14 we classified all large dams in km2,3)10–100 km2, and 4) >100 km2 (see table below). As- these countries into only two categories (since there are sev- suming a reservoir with a hypothetical rectangular shape (Fig- eral thousand dams), namely 1) area of the reservoirs Յ100 ure 1), we calculated for each dam the base b of the reservoir km2, and 2) area of the reservoirs >100 km2. For each group A/l, where A represents the area of the we calculated the area of a 2-km mosquito flight range with ס according to b ס reservoir and l the length of the dam (both parameters have the aid of two hypothetical rectangles of A1 15 l × l (for the ס Յ 2 12 been obtained from the World Register of Dams ). To get an small reservoirs 100 km ) and A2 500 l × l (for the large idea of the shape of the reservoirs (whether the rectangle has reservoirs >100 km2) (see example for WHO sub-region 4 a long or short perimeter), we then calculated for each group below). The calculation of the people at risk has been con- the median of the ratio b/l [A/(l × l)] of the dams’ reservoirs. ducted as described above for WHO sub-region 2.

Area flight Area flight range Area flight range for all for dams located in At-risk Total size of Median range at full large dams at malaria-endemic areas population Reservoir sizes Number reservoirs base/length water level full water level at fullwater level at full (×103 m2) of dams* (×103 m2) of dam (×103 m2) (×103 m2)† (×103 m2) water level 25–960 100 47,396 2.5 73,414 1,010–8,700 95 279,681 7.1 215,808 10,000–91,050 23 706,151 36.1 668,796 120,000–5,100,000 8 11,005,700 637.4 10,620,121 226 12,038,928 11,578,139 17,725,823 16,662,273 429,887 * Only the 25 South African dams located in malaria-endemic areas have been included. † No data were available on the area of the reservoir of 120 dams.

Total size of Estimated median Area flight range Area flight range Area flight range Reservoir sizes Number reservoirs base/length (×103 m2) for all dams for endemic dams At risk population (×103 m2) of dams (×103 m2) of dam (full water level) (×103 m2) (full water level)* (×103 m2) (full water level) (full water level) 25–100,000 516 4,643,105 15 1,138,457 100,000– 53 23,772,133 500 13,830,507 569 28,415,238 14,968,964 36,541,108 12,789,387 383,670 * No data were available on the area of the reservoir of 820 dams.