Malaria transmission risk variations derived from different agricultural practices in an irrigated area of northern Tanzania


  • J. N. Ijumba,

    Corresponding author
    1. Tropical Pesticides Research Institute, Arusha, Tanzania,
    2. Danish Bilharziasis Laboratory, Charlottenlund, Denmark and
      Dr Jasper Ijumba, Centre for Enhancement of Effective Malaria Interventions, PO Box 9653, Dar es Salaam, Tanzania. E-mail: or
    Search for more papers by this author
  • F. W. Mosha,

    1. Tropical Pesticides Research Institute, Arusha, Tanzania,
    Search for more papers by this author
  • S. W. Lindsay

    1. Danish Bilharziasis Laboratory, Charlottenlund, Denmark and
    2. School of Biological and Biomedical Sciences, University of Durham, U.K.
    Search for more papers by this author

  • 1

    Some samples of the An. funestus sensu lato showed high degrees of exophily, exophagy and zoophily, suggesting that they were probably not An. funestus Giles sensu stricto, which is very endophilic, endophagic and anthropophilic, whereas other members of the An. funestus group are not (c.f. Gillies & DeMeillon, 1968; Gillies & Coetzee, 1987). In Kisangasangeni and perhaps Mvuleni, some samples of An. funestus s.l. were sufficiently endophilic and anthropophilic to be regarded as An. funestus s.s., but we lacked the technical capability to confirm this by PCR. Therefore, we assessed results for all females of the An. funestus group collectively, without distinction between the sibling species.

Dr Jasper Ijumba, Centre for Enhancement of Effective Malaria Interventions, PO Box 9653, Dar es Salaam, Tanzania. E-mail: or


Abstract Malaria vector Anopheles and other mosquitoes (Diptera: Culicidae) were monitored for 12 months during 1994–95 in villages of Lower Moshi irrigation area (37°20′ E, 3°21′ S; ∼700 m a.s.l.) south of Mount Kilimanjaro in northern Tanzania. Adult mosquito populations were sampled fortnightly by five methods: human bait collection indoors (18.00–06.00 hours) and outdoors (18.00–24.00 hours); from daytime resting-sites indoors and outdoors; by CDC light-traps over sleepers. Anopheles densities and rates of survival, anthropophily and malaria infection were compared between three villages representing different agro-ecosystems: irrigated sugarcane plantation; smallholder rice irrigation scheme, and savannah with subsistence crops. Respective study villages were Mvuleni (population 2200), Chekereni (population 3200) and Kisangasangeni (population ?1000), at least 7 km apart.

Anopheles arabiensis Patton was found to be the principal malaria vector throughout the study area, with An. funestus Giles sensu lato of secondary importance in the sugarcane and savannah villages. Irrigated sugarcane cultivation resulted in water pooling, but this did not produce more vectors. Anopheles arabiensis densities averaged four-fold higher in the ricefield village, although their human blood-index was significantly less (48%) than in the sugarcane (68%) or savannah (66%) villages, despite similar proportions of humans and cows (ratio 1 : 1.1–1.4) as the main hosts at all sites. Parous rates, duration of the gonotrophic cycle and survival rates of An. arabiensis were similar in villages of all three agro-ecosystems.

The potential risk of malaria, based on measurements of vectorial capacity of An. arabiensis and An. funestus combined, was four-fold higher in the ricefield village than in the sugarcane or savannah villages nearby. However, the more realistic estimate of malaria risk, based on entomological inoculation rates, indicated that exposure to infective vectors was 61–68% less for people in the ricefield village, due to the much lower sporozoite rate in An. arabiensis (ricefield 0.01%, sugarcane 0.1%, savannah 0.12%). This contrast was attributed to better socio-economic conditions of rice farmers, facilitating relatively more use of antimalarials and bednets for their families. Our findings show that, for a combination of reasons, the malaria challenge is lower for villagers associated with an irrigated rice-growing scheme (despite greater malaria vector potential), than for adjacent communities with other agro-ecosystems bringing less socio-economic benefits to health. This encourages the development of agro-irrigation schemes in African savannahs, provided that residents have ready access to antimalaria materials (i.e. effective antimalaria drugs and insecticidal bednets) that they may better afford for protection against the greater vectorial capacity of An. arabiensis from the ricefield agro-ecosystem.


Human population growth rate is greater in Africa than other continents (WRI/UNEP/UNDP, 1995), rising by nearly 20 million people annually (USBC, 2000) and projected to reach 1.58 billion by the year 2005. Already the population has outstripped food production in all but three of the 41 sub-Saharan countries (the exceptions being Cameroon, Central African Republic and Côte d'Ivoire), so African countries seek better ways of expanding agricultural production. Unfortunately, almost half the potential arable land in Africa is too dry for rain-fed agriculture: precipitation fluctuates greatly from year to year and is generally insufficient for reliable cultivation (WRI/IIED, 1986). Irrigation may be the most effective way to increase crop production through increased yield, acreage, number of cropping cycles per year, and by reducing the risk of crop failure (Oomen et al., 1994). At present, only 4% of Africa's arable land is irrigated and 70% of this is in four countries: Egypt, Madagascar, Nigeria and Sudan (FAO, 1987, 1996). Although irrigation can facilitate improved production of food and cash crops, not all arable land in Africa is suitable for irrigation, because of poor soils or unsuitable locations (WRI/IIED, 1986). Moreover, there are grounds for concern that irrigation may aggravate the health risks of local communities, by enhancing conditions favouring transmission of particular communicable diseases and an upsurge of specific vector mosquitoes, particularly when irrigated rice is cultivated (Bradley, 1988; Gratz, 1988; Hunter et  al., 1993).

Rice production is becoming increasingly important in Africa, as a cash crop as well as for local food (Grist, 1986; Juliano, 1993). The introduction of high yield varieties (HYV) in Africa in the mid-1960s resulted in increased rice cultivation, and promotion of its dietary importance has helped replace the traditional coarse cereals, such as millet and sorghum. Most African countries already have some kind of rice improvement activity, although varietal information is limited, and most countries lack statistics on areas devoted to individual varieties. HYVs are widely distributed and used to a greater degree than generally recognized (Dalrymple, 1986).

Tanzania is the second largest producer of rice in eastern Africa; by the early 1980s the harvested area was c. 280 000 ha, mostly rain-fed and only intermittently mosquitogenic. Unfortunately, irrigated rice cultivation provides ideal breeding sites for members of the Anopheles gambiae Giles complex, the principal vectors of malaria in Africa (Holstein, 1954; Gillies & De Meillon, 1968; Surtees, 1970; Surtees et al., 1970; White, 1974; Snow, 1983; Coluzzi, 1984; Ijumba & Lindsay, 2001) and by 1984 irrigated rice covered c. 20 000 ha in Tanzania (Dalrymple, 1986). Numerous studies in Africa have shown that irrigated rice cultivation can generate prodigious numbers of mosquitoes, although malaria sporozoite rates in Anopheles are generally lower in irrigated areas than elsewhere (Ijumba & Lindsay, 2001). Exposure to malaria risk in rice-growing areas may therefore be less than in neighbouring areas with various other agro-ecosystems.

Sugarcane also requires irrigation for optimal growth, although it is very susceptible to water-logging and therefore needs efficient drainage (Arnon, 1972). Whereas sugarcane is usually irrigated and cultivated continuously, there is little published data on the influence of sugarcane irrigation on malaria transmission in Africa. Ross (1908) demonstrated how drainage of sugarcane effectively reduced malaria vectors (An. arabiensis) in Mauritius. Normally the shade from mature sugarcane would be inimical to breeding of malaria vectors, but malaria resurgence in Swaziland was encouraged by neglected sugarcane irrigation (Packard, 1986).

As part of a comprehensive investigation on irrigation and malaria (Ijumba, 1997), we report here the results of an entomological study conducted in adjacent villages of the Lower Moshi irrigation area in north-east Tanzania, to assess effects of irrigated sugarcane and a smallholder rice irrigation scheme on the transmission of malaria, compared with traditional farming practices in this savannah area. This study design compares three villages chosen to represent different agro-ecosystems. For logistical reasons it was not possible to replicate each type of village for routine sampling, yet we regard these as typical examples based on our malariological experience of many such communities in East Africa.

Materials and Methods

Study area and villages

The Lower Moshi area (Moshi Chini) is hyperendemic for falciparum malaria at c. 700 m above sea level, between the Masai savannah and foothills of Mount Kilimanjaro. Rainfall is seasonally concentrated in March–May, accounting for about 60% of the annual total ∼800 mm precipitation at Moshi town (10–15 km north of our study villages), while the remainder falls during October–December. Between these two rainy seasons are a hot dry season during January–February and a cool dry season during June–September. Figure 1 shows locations of three study villages associated with contrasted agro-ecosystems in Lower Moshi. Distances of 7–14 km between these villages would not preclude mosquitoes flying between them, but each village has sufficient hosts and nearby breeding sites to discourage mixing of the vector populations.

Figure 1.

Sketch map of Lower Moshi irrigation scheme area, showing location of study villages. Inset shows position of study site * in Tanzania. Moshi Town is located at 3°21′ S 37°20′ E.

Chekereni village is situated near a rice irrigation scheme. Of 2300 ha available for cultivation, 1100 ha was irrigated and used for growing HYV rice and the remainder used for non-irrigated crops. Although these ricefields are collectively large-scale, they belong to smallholder farmers, each owning a few plots. Water used for irrigation comes from the rivers Njoro and Rau. For the main growing season, rice was sown in mid-June and transplantation was carried out one month later. Irrigation extended from June to October. A second growing season involves sporadic cultivation during September–February. Most of the 3330 adult male inhabitants of Chekereni are farmers, the remainder work either at the local rice mill or in Moshi town as civil servants or at Kilimanjaro Agricultural Training Centre (KATC) where rice farmers are trained. Livestock at Chekereni are predominantly cattle, goats, sheep and poultry.

Sugarcane cultivation in the Lower Moshi irrigation area covers 6313 ha belonging to the Tanganyika Planting Company (TPC). Water for continuous irrigation comes from the River Weruweru through a system of canals, reservoirs and pumping stations supporting sprinkler and surface irrigation. Fields are irrigated until the ground is saturated for planting. Staggered planting and harvesting are practised, requiring irrigation continuously in some parts of TPC plantations. Mvuleni village (population 2200) is situated beside the sugarcane irrigation scheme, 5–6 km west of the rice irrigation scheme (Fig. 1). Most of the adult male residents of Mvuleni are labourers on the sugarcane plantations or work at the TPC sugarcane-processing factory.

Kisangasangeni village is situated near a small dam on River Miwaleni (Fig. 1), flanked by undeveloped savannah. Among c. 1000 inhabitants the majority of men are subsistence farmers, growing mainly maize and bananas.

Environmental studies

In each study village, large domestic livestock were counted and the ratio of humans to cattle was determined. Proportions of inhabitants regularly sleeping under bednets for personal protection against mosquitoes was estimated by household inspections. Meteorological observations were made at the office of Kilimanjaro Agricultural Development Project (KADP) in Chekereni, at the TPC campus for Mvuleni and in Kisangasangeni. At each village the temperature and relative humidity were recorded using a thermohygrograph in a Stevenson screen and rainfall was measured using a Snowdon rain gauge.


Adult mosquitoes were sampled fortnightly by five methods in each village, from June 1994 to June 1995. To assess the biting rates, human bait collections were made by pairs of local collectors with their limbs exposed, sitting on chairs and using glass tubes and torch-light, trying to catch mosquitoes before they could bite. Female mosquitoes landing on the human baits were captured in two houses between 18.00 and 06.00 hours and outdoors between 18.00 and 24.00 hours at each village on different nights. To standardize catching efficiency, some collectors rotated between villages. On the same night as human-landing catches in each village, CDC light traps (J. W. Hock Ltd, Gainesville, FL, U.S.A.) were run from 18.00 to 06.00 hours in two houses, each with a single sleeper. For operation, each light trap was suspended with its base c. 45 cm above the head of a person sleeping inside a conical bednet (Lines et al., 1991). To assess the indoor-resting mosquito density in each village, four small houses (each having one bedroom and a single sleeper) were sprayed with 0.3% pyrethrum in kerosene at 06.30–07.30 hours and knocked-down mosquitoes were collected from white sheets spread on the floor of the whole house (WHO, 1975). Hand-catches were made of all mosquitoes from artificial pit shelters (Service, 1976) in each village using an aspirator for 15 min between 06.30 and 07.30 hours.

Mosquitoes were kept cool until dissection. Specimens were sorted to species, categorized by abdominal condition and counted. Anopheles were identified morphologically (Gillies & De Meillon, 1968; Gillies & Coetzee, 1987) and graded according to abdominal condition. Some semi-gravid females of the An. arabiensis from each village were preserved in Carnoy's fixative for cytotaxonomic identification of sibling species in the An. gambiae Giles complex (Coluzzi, 1968; Green, 1972). Ovaries of An. arabiensis collected by human-landing catches and light traps were removed, covered with a glass slip and dried before parity was scored by Detinova's method (WHO, 1975). The gut of each specimen was removed and the stomach wall examined for oocysts; salivary glands were examined for sporozoites (WHO, 1975). Other anophelines were bisected dorso-ventrally between the second and third pairs of legs, and the anterior portion preserved in a desiccator for subsequent detection of Plasmodium falciparum sporozoites by enzyme-linked immunosorbent assay of circumsporozoite protein, CSP-ELISA (Wirtz et al., 1987). Bloodmeals from anopheline females collected by pyrethrum spray-catch and from pit shelters were squashed on filter paper (Whatman no. 1) and preserved in a desiccator for identification of host source using a modification of the ELISA technique described by Burkot et al. (1981).

To measure duration of the gonotrophic cycle, freshly blood-fed females of An. arabiensis were collected indoors by aspirator at 06.00–06.30 hours. Specimens were kept individually in paper cups at ambient temperature with moist cotton wool covered by filter paper as substrate for oviposition. Time until oviposition was recorded for each specimen and the gonotrophic cycle period was presumed to have begun at midnight before collection.

Statistical analyses

Mosquito counts were log-transformed [ln (x + 1)] to stabilize the variance and statistical analysis used Epi-Info® and SPSS® software. Entomological inoculation rates (EIR) were determined by multiplying the geometric mean number of bites by the sporozoite rate and by 365 days for a year. Vectorial capacity (sensuGarrett-Jones & Shidrawi, 1969) was estimated from the formula:


where the vectorial capacity C represents the power of the vector population to multiply malaria, unity being theoretical equilibrium; ma is the human-biting density; a is the human-biting habit = human-blood index (HBI) × 1/x days, where HBI is the proportion of bloodmeals taken from humans; p is the average daily survival of the female mosquito (p1/x); x is the duration of gonotrophic cycle (days), p is the parous rate and n is the extrinsic period of development of the parasite, assumed to be 12 days at the mean temperature of 25°C in the study area (Garrett-Jones & Shidrawi, 1969).


Meteorology and environmental factors

Mean temperature and relative humidity values are tabulated in Fig. 2 for the 12 months observation period. Total precipitation ranged from 540 to 659 mm in the three study villages. The rainfall pattern was typically bimodal, concentrated in February–May and October–December (Fig. 2).

Figure 2.

Summary of main meterological factors in three study villages and monthly rainfall pattern, June 1994–May 1995.

The ratio of humans to cattle was similar in all villages, 1 : 1.4 in the rice irrigation scheme, 1 : 1.2 in the sugarcane irrigation area and 1 : 1.1 in the savannah village. Proportions of people regularly sleeping under bednets were estimated to be ≈ 7% in the ricefield village, ≈ 6% in the sugarcane village, but only ≈ 1% in the savannah village (Ijumba, 1997). None of the bednets were said to be insecticide-treated and there appeared to be little use of household insecticides or repellents. Agricultural and irrigation activities were as follows.

Ricefield village Due to shortage of water only <60% of the normally cultivated area around Chekereni was planted with rice during the main growing season of 1994 and a much smaller area (< 20%) cultivated during the September 1994–February 1995 season. Maize was cultivated in small plots near the village and beyond the irrigated fields, but these were never flooded.

Sugarcane village Plantations near Mvuleni were often irrigated, and persistent pools formed from canal leakages, although sugarcane fields were not flooded continuously.

Savannah village Most of the land around Kisangasangeni remained uncultivated and bare, with sparse Acacia bushes. Patches of bananas and maize were not irrigated and relied on rainfall. The harvest was so low in 1993 that Kisangasangeni received donations of food from the government and church organizations.

Vector bionomics

Tables 1 and 2 show the relative abundance of mosquitoes collected in each village by different trapping methods. The overwhelmingly predominant anopheline species in all samples was Anopheles arabiensis (total 19 029) and no other members of the An. gambiae complex were found among 42 specimens identified cytotaxonomically. Anopheles funestus sensu lato was relatively scarce (total 221), mostly collected from pits or biting outdoors at the sugarcane irrigation scheme and the savannah village, with very few An. funestus s.l. in the ricefield village. Other anophelines collected comprised 192 An. pharoensis Theobald, 93 An. coustani Laveran and 91 An. rufipes Gough, plus 65 other specimens, mostly indeterminate. Culicines were most abundant in the sugarcane village and least abundant in the savannah village, with Culex quinquefasciatus Say and Mansonia (Mansonioides) spp. predominating (Tables 1 and 2).

Table 1.   Relative abundance of mosquito species in villages of three contrasted agro-ecosystems at Lower Moshi, sampled by human-biting catches and light traps. Overall percentages of each species from fortnightly collections during 12 months 1994–95, are shown in parentheses.
SpeciesSampling method
Human-biting catch (indoors)Human-biting catch (outdoors)Light-trap collections (indoors)
An. arabiensis2917(73.2)2593(40.7)1232(89.1)1596(44.3)893(21.7)522(34.1)2471(82.1)712(29.8)370(70.7)
An. funestus0(0)12(0.2)3(0.2)3(0.1)38(0.9)26(1.7)1(0.1)4(0.2)4(0.7)
An. pharoensis12(0.3)21(0.3)3(0.2)80(2.2)51(1.2)11(0.7)7(0)7(0.3)0(0)
An. coustani2(0.1)8(0.2)2(0.1)11(0.3)35(0.9)33(2.2)0(0)2(0.1)0(0)
An. rufipes0(0)6(0.1)0(0)0(0)9(0.2)0(0)0(0)0(0)0(0)
Other anophelines4(0.1)7(0.1)0(0)1(0)24(0.6)0(0)0(0)4(0.2)0(0)
Cx. quinquefasciatus906(22.7)3139(49.3)85(6.2)1065(29.5)1806(43.9)377(24.6)412(13.7)1439(60.2)89(17)
Mn. uniformis19(0.5)206(3.2)10(0.7)157(4.4)442(10.8)55(3.6)39(1.3)72(3)10(1.9)
Mn. africana17(0.4)190(3)8(0.6)166(4.6)200(4.9)54(3.5)13(0.4)37(1.5)5(0.9)
Other culicines39(2.7)189(2.9)40(2.9)526(14.6)612(14.9)453(29.6)65(2.2)114(4.8)45(8.6)
Table 2.   Indoor collections of adult mosquitoes from daytime resting-sites in study villages of three agro-ecosystems at Lower Moshi. Overall percentages of each species from 12 months sampling 1994–95, are given in parentheses.
SpeciesPyrethrum-spray collections (indoors)Pit collections (outdoors)
An. arabiensis2124(77.7)456(40.2)531(87.6)1884(93.8)468(66.3)260(72.2)
An. funestus4(0.1)4(0.4)38(6.3)5(0.3)28(4)51(14.2)
An. pharoensis1(0.1)1(0.1)0(0)1(0.1)2(0.3)0(0)
An. coustani0(0)1(0.1)0(0)0(0)0(0)0(0)
An. rufipes0(0)0(0)0(0)0(0)0(0)0(0)
Other anophelines0(0)0(0)4(0.7)0(0)11(1.6)10(2.8)
Cx. quinquefasciatus567(20.8)618(54.5)25(4.1)93(4.6)135(19.1)30(8.3)
Mn. uniformis12(0.4)3(0.3)0(0)0(0)7(1)0(0)
Mn. africana0(0)0(0)0(0)0(0)1(0.1)0(0)
Other culicines24(0.9)51(4.4)8(1.3)25(1.2)53(7.6)9(2.5)

Anopheles arabiensis reached higher densities seasonally in the ricefield agro-ecosystem than in the other two villages (Fig. 3). Following flooding and planting of the paddy in June, the peak population density of An. arabiensis occurred in the ricefield village during July–August 1994, with slight increases in the other two villages during August that might be attributed to dispersal from the ricefield agro-ecosystem. Despite sporadic rice cultivation during September 1994–February 1995, the population density of An. arabiensis was highest in the sugarcane village until the main rains of February–May 1995, when the upsurge was greatest in the ricefield village and least in the savannah village. No increase of An. arabiensis was associated with the short rains of October–December 1994 (Figs 2 and 3).

Figure 3.

Seasonal variation in man-biting rate of Anopheles arabiensis at three study villages in relation to rice irrigation.

Considerably more An. arabiensis were collected in the ricefield village, by all sampling methods, than in either of the other study villages (Table 3). Comparing the sugarcane and savannah villages, similar densities of An. arabiensis were obtained by three sampling methods, whereas human-baits caught 2.7-fold more in the Mvuleni than in Kisangasangeni. Due to this inconsistency, which may reflect human bias, we have based our estimates of biting rates on light trap collections instead of relying on landing catch data for computation of vectorial capacity and malaria risk (Table 6). This is justified by the positive correlation between yields from light-traps and human-biting collections in all three agro-ecosystems, viz: ricefield r2 = 0.42, n= 21, P < 0.001; sugarcane r2 = 0.20, n= 21, P < 0.025; savannah r2 = 0.81, n= 21, P < 0.001; pooled data for three villages: r2 = 0.48, n= 63, P < 0.001.

Table 3.   Geometric mean number of Anopheles arabiensis females collected using different sampling methods at villages representative of three agro-ecosystems.
Sampling methodVillage type
Human-biting catches34.123.38.7
Light trap catches33.38.67.3
Spray catches21.64.13.3
Pit traps19.52.14.4
Table 6.   Comparative values of vectorial capacity (C) and its components as observed in villages of three agro-ecosystems at Lower Moshi, where ma = man-biting rate, HBI = human-blood index, a = man-biting habit, g= gonotrophic cycle (days), p= parity, P= daily survival rate, n= extrinsic cycle (days), s= sporozoite rate and EIR = entomological inoculation rate.
  1. * ma based on light-trap collections (interpreted as per Lines et al., 1991).

Anopheles arabiensis
Anopheles funestus
Pooled An. arabiensis plus An. funestus

The gonotrophic cycle of An. arabiensis was completed within 3 days at ambient temperature throughout the year (Table 6). Daytime samples of females therefore comprised four categories based on abdominal condition: unfed, freshly blood-fed, semi-gravid and almost fully gravid, i.e. ready for oviposition. Excluding the few unfed specimens, the other three categories should occur in approximately 1 : 1 : 1 ratio (disallowing for interim mortality), which may be simplified as the ‘fed : gravid ratio’ expected to approximate 1 : 2 for a 3-day gonotrophic cycle. The fed : gravid ratio for samples of anopheline females from indoor and outdoor resting sites can be used as an indicator of relative endophily vs. exophily during the gonotrophic period. Table 4 summarizes fed : gravid ratios of our An. arabiensis samples, showing mostly higher ratios than expected (see Discussion), with no indication of postprandial exophily.

Table 4. Anopheles arabiensis fed : gravid ratios among samples of females from indoors (spray-catch) and outdoors (pit shelters) at villages representing three agro-ecosystems. Those grouped as gravid comprise all abdominal conditions later than freshly blood-fed, i.e. semigravid up to fully gravid.
Agro-ecosystemRatio fed:gravid females

Of those collected on human bait (Table 1), the proportion of An. arabiensis attempting to bite outdoors was 35.4% in the ricefield village, significantly higher than 25.6% and 29.8% in the sugarcane and savannah villages, respectively (χ2 = 86.7 and 17.48, P < 0.001).

Among indoor-resting females of An. arabiensis, the human-blood index (HBI) of 47.9% in the ricefield village was significantly lower than 68.4% in the sugarcane or 66.4% in the savannah village (χ2 = 25.3 and 24.5, P < 0.001). Among outdoor-resting An. arabiensis, the HBI of 25.8% was significantly higher than 4% in both other villages (Table 5). In contrast, the ratio of humans to cattle as main hosts was similar in all three agro-ecosystems. Bloodmeals identified from the An. funestus group were only three bovine-fed specimens from pits at the ricefield village; HBI = 29% (5/17) from pits at the sugarcane village; and HBI = 36% (11/31) from the savannah village, of which 10 were from indoors.

Table 5.   Summary of data from mosquito dissections and assays for bloodmeal identification and malaria parasite detection.
SpeciesVillage type
Measurement RiceSugarcaneSavannah
  • *

    Human-Blood Index, percentage human fed/total blood-fed.

An. arabiensisHBI*Indoors381/795 (47.9%)132/193 (68.4%)160/241 (66.4%)
Outdoors21/501 (4.2%)45/174 (25.8%)5/121 (4.1%)
Parous rate 567/1099 (51.6%)498/950 (52.4%)261/262 (50.0%)
Oocyst rate 2/595 (0.3%)6/765 (0.8%)2/419 (0.5%)
Sporozoite rate dissection 0/610 (0%)0/777 (0%)0/420 (0%)
ELISA 1/9941 (0.01%)4/4357 (0.10%)3/2481 (0.12%)
An. funestusELISA 0/13 (0%)1/84 (1.2%)1/121 (0.82%)

Although parous rates of An. arabiensis averaged 50–52% in all three villages (Table 5), the malaria sporozoite rate of 0.01% in this vector at the rice irrigation scheme was 10-fold less than at the sugarcane village (χ2 = 3.69, P < 0.05) and 12-fold less than at the savannah village (χ2 = 4.53, P < 0.05). Sporozoite-positive An. arabiensis were all detected by CSP-ELISA (8/16 779 = 0.048%), none by dissection (0/1807) (Table 5).

Among the relatively few An. funestus s.l. collected and tested (Table 5), sporozoite-positives were detected at the sugarcane and savannah villages (1/84 and 1/121, respectively), so this vector was included in our estimates of EIR and found to contribute about 14% of the malaria risk (Table 6).

Transmission potential, assessed from vectorial capacity of An. arabiensis, was three- or four-fold higher in the ricefield village than in the other agro-ecosystem villages (Table 6). Conversely, the risk of malaria transmission, based on EIRs of An. arabiensis plus An. funestus, was apparently about three-fold lower in the ricefield irrigation scheme than in sugarcane and savannah villages.


In a highly malarious part of East Africa with limited rainfall, vectorial capacity and malaria transmission risks were compared between communities associated with irrigated rice and sugarcane vs. traditional savannah with subsistence agriculture. Anopheles arabiensis was found to be the predominant vector in all three agro-ecosystems, with An. funestus1s.l. of much less importance. None of the other anophelines obtained (An. coustani, An. pharoensis, An. rufipes) is regarded as a malaria vector of any importance in this area (Gillies & De Meillon, 1968).

Seasonal irrigation of ricefields in June leading to peak abundance of An. arabiensis during July–August shows typical population dynamics of the An. gambiae complex. These pioneer species rapidly colonize recently flooded land and flourish for a few weeks before declining in abundance as the vegetation grows and shades the water surface (Gillies & De Meillon, 1968; Snow, 1983; Lindsay et al., 1991). Productivity of An. arabiensis instead of An. gambiae s.s. from irrigated ricefields has been reported elsewhere in East Africa (e.g. White, 1972; Githeko et al., 1993; Duchemin et al. 2001). There was no discernible upsurge of An. arabiensis when rice cultivation was not synchronized during October–December, when rains are unpredictable. Anopheles funestus s.l. was a secondary vector of malaria in the irrigated sugarcane and savannah villages, but apparently not in the rice irrigation scheme. This vector was associated with poorly maintained water systems in the sugarcane village, as reported in sugar plantations elsewhere (Ijumba & Lindsay, 2002).

The human-biting rate (ma) of An. arabiensis in the rice irrigation scheme was about four-fold greater than in villages of both other agro-ecosystems at Lower Moshi. This study therefore confirms the importance of ricefields as major sources of the An. gambiae complex (Ijumba & Lindsay, 2001) and illustrates that irrigated sugarcane cultivation is relatively unproductive of malaria vectors, unless leakages are allowed to form swampy pools providing breeding sites for An. funestus.

Among indoor-resting An. arabiensis females containing a fresh bloodmeal, proportionately fewer had bitten people (i.e. HBI was significantly lower) in the ricefield village of Chekereni compared with both other villages, despite the availability of similar host ratios of humans to cattle (1 : 1.1–1.4) in all three villages. Although bednets were apparently used by similar proportions of inhabitants of the ricefield and sugarcane villages, the condition of the nets may have been better for the more prosperous ricefield farmers and their families, giving more protection for this community. Considering possible density-dependent effects, whereby people may implement personal protective measures in proportion to the biting density of mosquitoes, reducing the HBI and hence the proportion of potentially malaria-infected bloodmeals (Lindsay et al., 1992, 1993), our findings were ambivalent because biting densities of anophelines were greater in the ricefield village, whereas biting densities of culicines were greater in the sugarcane village. However, as both duration of the gonotrophic cycle and daily survival rate of An. arabiensis were similar in all three agro-ecosystems, we infer that the much lower malaria sporozoite rate prevailing in the ricefield village resulted from the reduced HBI due to lower success rate of biting humans vs. other hosts.

Our two estimates of exposure appear to be contradictory at first sight. We found that vectorial capacity as an indicator of potential risk was greatest in the ricefield village and three- to four-fold lower in the other villages, mainly due to the different biting rates. In contrast, exposure estimated in terms of the EIR was lowest in the ricefield community, mainly due to the lower sporozoite rate, but about three-fold greater in the sugarcane and savannah villages. Any bias due to our use of light-trap data for measurement of the man-biting rate would have affected both estimates of exposure equally, whereas other factors (i.e. An. arabiensis parous rates being so low, fed:gravid ratio being so high, gonotrophic cycle being so short), could have contributed to the unrealistically low values determined for vectorial capacity. In this situation, however, the EIR serves as a more realistic reflection of exposure by taking into account the contrasted sporozoite rates. Neither measure is entirely appropriate because they consider only the risk of exposure to unprotected individuals. Actually in both irrigation communities the majority of people slept under an untreated bednet, likely to provide a substantial reduction in mosquito bites (Lindsay et al., 1989). Therefore, residents of the study villages would be less exposed than we estimated from entomological data.

Neither estimate of risk fits the pattern of malaria seen in these villages. From initial parasitological surveys, the prevalence of malaria parasitaemia in 1–4-year-old children was 15.8% in the ricefield village, 19.8% in the sugarcane village and 41% in the savannah village (Ijumba et al., 2001) and these prevalences remained relatively consistent through four subsequent surveys during the year (Ijumba, 1997). The difference between our entomological and parasitological findings evidently reflects the difficulty of estimating exposure by means of conventional entomological techniques (WHO, 1975), which may miss any influences and effects of human behaviour on transmission. Thus, we suspect that the explanation for least malaria being in the ricefield village comes from the greatest wealth of this community and their more effective use of antimalarial drugs and better bednets. Quite how wealth moderates risk remains unclear, except that in the savannah village relatively few people slept under bednets and there was poor access to efficient healthcare (Ijumba, 1997). The sugarcane village was intermediate socio-economically, with the access to bednets resembling the ricefield community and malaria risk resembling the savannah.

Results of this study confirm other findings in Africa, where the cultivation of irrigated rice gives rise to large numbers of potential vectors but low inoculation rates. For example, in rice-growing areas of Rusizi valley, Burundi, the biting density of An. gambiae s.l. was about 10-fold higher than in adjacent cotton-growing areas, but the sporozoite rate was three-fold lower in the rice irrigation scheme (Coosemans et al., 1985). Likewise, the sporozoite rate of An. gambiae s.l. in Ahero rice irrigation scheme in Kenya was five-fold less than in the nearby sugar-belt (Githeko et al., 1993). In the rice-growing areas of Bobo-Dioulasso, Burkina Faso, the man-biting density of An. gambiae s. l. was 10-fold higher than in the nearby savannah areas, but the sporozoite rate was 10-fold lower in the rice-growing area (Robert et al., 1985). These anomalies can no doubt be explained from specific responses of the different malaria vectors in combination with and human behavioural and socio-economic factors of malariological relevance.

Development of rice irrigation in Africa has seldom resulted in increased malaria transmission. Perhaps the only exception to this was in Burundi, an area of low transmission, where an influx of migrant workers may have brought parasites into an area where indigenous people had little prior exposure or immunity (Coosemans et al., 1985). Generally, in places where large-scale irrigated rice is cultivated for commercial purposes, local communities may benefit from sales and accumulate enough disposable income that they invest in countermeasures such as construction of screened houses, the use of insecticide-treated bednets and other personal protection against anthropophilic mosquitoes, as well as the timely use of effective antimalaria drugs. Therefore, development projects involving rice irrigation should be encouraged, providing that they are supported by effective healthcare.


Approval for this project was given by the Ethical Clearance Committees of DBL and TPRI, by the Tanzanian Ministry of Health, local authorities and village committees. We thank the people of Chekereni, Mvuleni and Kisangasangeni, particularly the village leadership for their co-operation and assistance. For specific services we thank Phillip Kayuni, Lucy Lyaruu, Magdalena Lucas, Charles Masenga, William Shafuri, Michael Shiyyo and the local the mosquito collectors. We appreciate the help of Dr Henry Madsen with preparing the manuscript and thanks to Drs Jacob Koella and Jo Lines for their helpful comments on the draft. Our work was funded by the Danish Bilharziasis Laboratory.

Accepted 8 September 2001