Temporal distribution and spatial pattern of abundance of the Rift Valley fever and West Nile fever vectors in Barkedji, Senegal

Authors


ABSTRACT:

The temporal distribution and spatial pattern of abundance of mosquito vectors of Rift Valley fever (RVf) and West Nile fever (WNf) were studied during the 2005 and 2006 rainy seasons at Barkedji, Senegal. Mosquitoes were collected every two weeks with CDC light traps with dry ice at 79 sites including temporary ponds, barren, shrubby savannah, wooded savannah, steppes, and villages at different distances (between 0 and 600 m) from the nearest pond. The temporal distributions of these vectors varied between 2005 and 2006 and were positively correlated with rainfall for Aedes (Aedimorphus) vexans Patton, with rainfall after a lag time of one month for Culex (Culex) poicilipes (Theobald) and Culex (Culex) neavei Theobald. All the vectors had their highest abundances and parity rates between September and November. The highest vector abundances were observed in the barren and temporary ponds. The distance of trap location to the nearest ponds was negatively correlated to the abundance of the vectors. Taking into account the linear regression equations, it was predicted that mosquitoes would not disperse and be collected by the light trap, up to 1,500 m to the nearest ponds. The implications of these findings in the epidemiology and control of RVF and WNF at Barkedji are discussed.

INTRODUCTION

Rift Valley fever (RVf) and West Nile fever (WNf) are two emerging mosquito-borne diseases of medical and veterinary importance. RVf virus (RVfv) is a phlebovirus of the family Bunyaviridae (Murphy et al. 1995). It causes an anthropo-zoonosis that affects mainly cattle, sheep, and goats leading to abortions among pregnant females and high mortality in newborn kids and lambs (Easterday 1965). Humans are generally infected by contact with blood, body fluids, or organs of infected animals and by mosquito bites. In humans, RVf manifestations are usually asymptomatic, but severe cases with complications such as encephalitis, hepatitis, and hemorrhagic syndromes associated with high mortality rates have been reported (Laughlin et al. 1979, Jouan et al. 1988). WN virus (WNv) is a flavivirus of the family Flaviviridae. WNv is maintained in nature in enzootic cycles involving birds and ornithophilic mosquitoes (Hayes et al. 2005). WNf may affect humans and horses under certain environmental conditions (Bunning et al. 2001). Infections in humans and horses are usually asymptomatic or mild, but cases of serious meningitis and encephalitis with high mortality rates were observed (O'Leary et al. 2002). RVf and WNf viruses are both transmitted by several species of mosquitoes, mainly of Aedes and Culex genera (Traore-Lamizana et al. 2001, Molaei et al. 2006).

In the last decades, several animal and human outbreaks of RVF occurred in Africa and the Middle East (Hoogstraal et al. 1979, Faye et al. 2007, Woods et al. 2002). WNF outbreaks were also reported in Africa, Europe, the U.S.A., and Asia (Murgue et al. 2002, Mostashari et al. 2001). In Barkedji, many mosquito species (Fontenille et al.1995, Diallo et al. 2000), as well as humans and livestock (Wilson et al. 1994, Thonnon et al. 1999), have been found to be associated with RVfv and WNfv. The first documented RVf epizootic and epidemic in Senegal occurred in the Senegal River basin in 1987 with an identification of 200 human cases and an estimation of more than 1,000 human cases (Jouan et al. 1988). After that, several serologic surveys in human and animals indicated that RVf virus continuously circulates in Senegal (Saluzzo et al. 1987, Thonnon et al. 1999). WNv has been isolated from two humans and several mosquito species in Senegal and seroprevalence of 5.5%, 78.3% and 80% has been recorded in birds, horses, and humans, respectively (Traoré-Lamizana et al. 2001, Chevalier et al. 2006, 2009).

These active viral circulations emphasize the need for more efficient surveillance and control measures. In this process of developing better surveillance tools, Vignolles et al. (2009) produced an early warning system by combining specific GIS, remote-sensing, entomological and livestock data. This tool may serve to produce a dynamic risk map but needs to be validated and extended to other vectors like Cx. poicilipes and Cx. neavei. A better understanding of the bio-ecology of the vectors could assist in the implementation of these measures.

Mosquito species with high population densities in Barkedji and which were found most often infected with RVFV and WNV in Africa are Culex (Culex) poicilipes (Theobald), Aedes (Aedimorphus) vexans Patton, and Culex (Culex) neavei Theobald. All these vectors breed in temporary ponds that are flooded after the first rainfalls (Fontenille et al. 1998, Traore-Lamizana et al. 1994, 2001, Ba et al. 2005, 2006). The maximum flight distances of Ae. vexans and Cx. poicilipes from these ponds was 620 and 550 m respectively (Ba et al. 2005). But the flight range of Culex (Culex) neavei is still unknown. In Senegal, although the associations among these viruses and mosquitoes, humans, and animals are frequent, the mechanisms by which vertebrates become infected (i.e., vector-borne or as aerosols) remain unclear. The assumption that the enzootic and epizootic/epidemic cycles of the diseases are due to the movement of infected mosquitoes from ponds to villages has been examined in a field study (Ba et al. 2005), but in this case, the movement of mosquitoes occurred over relatively short distances and the study was limited to non-infected populations of Cx. poicilipes and Ae. vexans. Moreover, it is unknown if infected and non-infected vectors have similar flight behaviors. Further information about the ecological environments associated with each vector species, the flight range of each species, and the radius of risk-of-infection of a given population of vectors is crucial for planning vector control campaigns to be certain that treatment is implemented over adequate areas for each environment.

Moreover, since 1990, most studies of the bio-ecology of mosquito vectors in Barkedji, a village in northern Senegal, have focused on temporary ponds and to a lesser extent on villages and did not explore other areas inhabited by known vertebrate hosts of the viruses. These studies were also carried out on a monthly basis, which may have underestimated the temporal fluctuations of mosquito populations, and were limited mainly to Cx. poicilipes and Ae. vexans (Fontenille et al. 1998, Traore-Lamizana et al. 2001).

The aims of this work were to analyze the temporal distribution and the spatial pattern of abundance of RVF and WNF vectors in relation to environmental factors like rainfall, habitat characteristics, and distance to larval breeding sites. Our specific goal was to identify the period and area at risk of transmission of these diseases.

MATERIALS AND METHODS

Study area

The study was carried out within a radius of 13 km centered on the Barkedji village (14° 47′-14° 53'W, 15° 13′-15° 20'N) during the 2005 to 2007 rainy seasons. This area (Figure 1) belongs to the Sahelian biogeographic domain characterized by a shrubby vegetation, hot dry climate, short rainy season (from June to October), and long dry season (November to May), with annual rainfall ranging from 300 to 500 mm. Many temporary ponds flood at the beginning of the rainy season and are the main source of water for herders and their livestock in this period. People grow mainly millet during the rainy season and herd sheep, goats, and cattle year-round. Family groups and their herds may become seasonal nomads, temporarily relocating for a few months during which they use grass huts for shelter.

Figure 1.

Location of the study site.

Mosquito sampling and processing

Different landscapes classes were defined by remote sensing and geospatial analyses from a SPOT 5 satellite image. The radiometric and geometric corrections were made by SPOT Image (http://www.spotimage.com.cn/spot5/ensavoirplus/eng/plus_niveau.html). The description of the vegetation classes is based on a combination of the FAO (1997) and Anon (1956). Thus the following landscape classes were selected for the study: 1) temporary ponds, 2) villages, and 3) vegetation classes of steppes, shrubby savannah, wooded savannah and barren.

Adult mosquitoes were collected using CDC light traps with dry-ice in 79 sites selected to include all the landscape classes. All the sampling sites were geo-referenced with a hand-held GPS receiver and the Euclidean distance (in meters) from each sampling site to the nearest pond was estimated based on the spatial map of the study area. Mosquitoes were collected for one night in each site on a bi-weekly basis from June to December, 2005, July to November, 2006. For each sampling night, one trap was set in each site at about 50 cm above ground level. Traps were set in the late afternoon and were retrieved the following morning. Collections were transported to the field lab where mosquitoes were killed in the freezer and transferred to a chill table. Mosquitoes were identified morphologically using the identification keys of Edwards (1941) for the culicines and Diagne et al. (1994) for the anophelines. Every two weeks, more than 50 specimens of each vector species were dissected to evaluate parity based on the presence of ovarian dilations, as described by Detinova (1962).

Data analysis

The temporal distribution, expressed as mean number of females per trap per night (F/T/N), was calculated for each vector every two weeks from July to December, 2005 and from June to November, 2006. The associations between the temporal distribution of the vectors and rainfall were evaluated by the Pearson correlation.

The mean numbers of mosquitoes collected annually in the different landscapes classes, expressed as mean number of females per trap per year (F/T/Y), were calculated and compared for each vector. Means were compared by ANOVA followed by Fisher PLSD tests. Statistical analyses were done after a log transformation log10 (n+1) of the data to normalize the distribution and minimize the standard error.

To study the possible concentration of females near ponds (larval breeding sites), the Pearson correlation between abundance and distance to the nearest pond was investigated by grouping all the sites in seven distance intervals (0, 100, 200, 300, 400, 500, and 600 m) from trap site to the nearest breeding site. A simple linear regression analysis was also conducted to evaluate the impact of distance to the nearest ponds on the mosquito's vector abundance.

Temporal mean parity rates (number of parous females divided by number of dissected females in percentage) of the vectors were determined. Data were analyzed using StatView 5.0®.

RESULTS

Species composition and relative abundance

A list of the mosquito species collected and their relative abundances during the study are presented in Table 1. A total of 14,908 mosquitoes, belonging to eight genera and 39 species were collected in 790 trap-nights in 2005. In 2006, 14,630 mosquitoes, belonging to seven genera and 38 species were collected in 636 trap-nights. Among females, Culex poicilipes (31.6%) was the most abundant species in 2005, followed by Aedes vexans (12.9%), Cx. neavei (11.4%), Ae. ochraceus (10.2%), and Mansonia uniformis (7.9%). In 2006, Ae. vexans (29.7%) was the most abundant mosquito, followed by Cx. poicilipes (25.1%), Cx. neavei (10.9%), Ae. ochraceus (10.7%), and Ma. uniformis (6.9%). The highest female to male ratio was that of Cx. poicilipes (148.9), followed by Ae. vexans (37.7) and Cx. neavei (9.4) in 2005. These ratios were 433.5 and 265.3 for Ae. vexans and Cx. neavei respectively in 2006. No Cx. poicilipes males were collected these years.

Table 1.  Female mosquitoes collected during the 2005 and 2006 rainy seasons at Barkedji, Senegal.
Mosquito species20052006
NRA (%)NRA (%)
  1. N=number of specimen collected for the year; RA=relative abundance as percentage of collected mosquitoes (%) for the year.

Aedes argenteopunctatus (Theobald)110.08820.56
Aedes dalzieli (Theobald)2061.411120.77
Aedes fowleri (Charmoy)690.47300.21
Aedes minutus (Theobald)290.20150.10
Aedes ochraceus (Theobald)1,48310.161,56210.71
Aedes vexans Patton1,88412.914,33529.72
Aedes furcifer (Edwards)10.0100
Aedes sudanensis (Karsch)5183.553862.65
Aedes mcintoshi (Huang)50.03250.17
Aedes aegypti (Linnaeus)80.0550.03
Aedes luteocephalus Newstead30.0200
Aedes metallicus (Edwards)80.0590.06
Aedes unilineatus (Theobald)590.40590.40
Total Aedes4,28429.346,62045.38
Aedomyia africana Neveu-Lemaire0010.01
Anopheles ziemanni Grünberg8545.853942.70
Anopheles funestus Giles s.1.0010.01
Anopheles gambiae Giles s.l.1641.12160.11
Anopheles pharoensis Theobald420.291060.73
Anopheles rufipes (Gough)510.35280.19
Anopheles squamosus Theobald30.02230.16
Anopheles wellcomei Theobald0020.01
TotalAnopheles1,1147.635703.91
Culex annulioris Theobald1450.99620.43
Culex antennatus (Becker)5854.012311.58
Culex aurantapex Edwards0010.01
Culex bitaeniorhynchus Giles2061.41310.21
Culex decens Theobald10.0100
Culex duttoni Theobald20.0100
Culex ethiopicus Edwards700.481851.27
Culex neavei Theobald1,66911.441,59210.91
Culex perfuscus Edwards3752.57790.54
Culex poicilipes (Theobald)4,61731.633,66025.09
Culex quinquefasciatus Say540.37100.07
Culex tritaeniorhynchus Giles2001.374172.86
Culex nebulosus Theobald100.0740.03
Culex tigripes De Grandpre & De Charmoy10.0120.01
Culex. sp.0050.03
Total Culex7,93554.376,27943.04
Ficalbia uniformis Theobald10.0110.01
Mansonia africana (Theobald)690.471090.75
Mansonia uniformis (Theobald)1,1517.891,0056.89
Total Mansonia1,2208.361,1147.64
Mimomyia mimomyiaformis (Newstead)10.0110.01
Mimomyia splendens Theobald310.2120.01
Total Mimomyia320.2230.02
Uranotaenia balfouri Theobald70.0500
Ur. sp.20.0100
Total Uranotaenia90.0700
Total females14,59510014,588100
Aedes argenteopunctatus (Theobald)0024.76
Aedes dalzieli (Theobald)10.3200
Aedes ochraceus (Theobald)51.624.76
Aedes vexans Patton5015.971023.81
Aedes furcifer (Edwards)30.9624.76
Aedes unilineatus (Theobald)51.6614.29
Aedomyia africana Neveu-Lemaire20.6400
Anopheles ziemanni Grünberg30.9600
Anopheles gambiae Giles s.l.113.5100
Anopheles pharoensis Theobald30.9612.38
Culex bitaeniorhynchus Giles41.2800
Culex neavei Theobald17756.55614.29
Culex perfuscus Edwards103.1900
Culex poicilipes (Theobald)319.900
Culex tritaeniorhynchus Giles0012.38
Mansonia uniformis (Theobald)72.241228.57
Mimomyia splendens Theobald10.3200
Total males31310042100
Total mosquitoes14,90810014,630100

Temporal patterns in species abundance

The rainfall and temporal patterns in mosquito populations in 2005 and 2006 are shown in Figure 2. The rainfall patterns differed between the two years. Culex poicilipes appeared at the beginning of the rainy season in July, 2005 (1.2 F/T/N), with two peaks at 18.3 and 9.6 F/T/N in early September and late October, respectively. In 2006, Cx. poicilipes also appeared at the beginning of the rainy season in June (0.01 F/T/N) and had an important peak in early October (23.7 F/T/N). Culex neavei was captured at the end of July, with 0.4 F/T/N and 0.6 F/T/N in 2005 and 2006, respectively. This species had a single peak, with 9.9 F/T/N in late October in 2005. In 2006 it had two peaks in late September (12.2 F/T/N) and in late October (5.2 F/T/N). In 2005, the highest Ae. vexans peak was observed at the beginning of the rainy season in July, with 24.3 F/T/N. This species reached a second peak at the beginning of October, with 8.1 F/T/N. In 2006 the highest peak of Ae. vexans populations was observed later in late September (51.2 F/T/N). The first peak was observed in late July, with 5.2 F/T/N.

Figure 2.

Temporal distribution of RVFV and WNV vectors during the 2005 and 2006 rainy seasons at Barkedji, Senegal.

The Pearson correlation analysis has shown a positive and significant correlation between temporal abundance and rainfall for Ae. vexans for both years (r= 0.64, P= 0.03 in 2005 and r= 0.62, P= 0.02 in 2006). This analysis showed a positive and significant correlation between temporal abundance and rainfall for Cx. poicilipes and Cx. neavei after one month, except for Cx. neavei in 2005 (r= 0.04, P= 0.9).

Spatial patterns of abundance

Mean vector abundances in the different landscape classes are shown in Table 2. The Anova showed a statistically significant difference among the different classes for Cx. poicilipes in (F= 7.25; df = 5, 137; P < 0.0001), Cx. neavei (F= 10.7; df = 5, 137; P < 0.0001), and Ae. vexans (F= 3.54; df = 5, 137; P= 0.005).

Table 2.  Mean numbers of RVf and WNf vectors collected in different landscape classes during the 2005 and 2006 rainy seasons in Barkedji, Senegal. For each species, means with different letters are significantly different (P≤ 0.05).
 Abundance (females /trap /year) ± SE
 Cx. poicilipesCx. neaveiAe. vexans
Barren145 ± 63.9 a46.8 ± 14.4 a172.8 ± 90.5 a
Temporary ponds124.3 ± 38.5 a41.9 ± 8.8 a68.2 ± 24.6 b
Shrubby savannah59.6 ± 24.9 b18.5 ± 3.8 b34.6 ± 9.1 b
Wooded savannah17.8 ± 4.1 b31.4 ± 9.3 a40.8 ± 21.6 b
Villages17.8 ± 4.1 b6.9 ± 0.9 b21.9 ± 4.8 c
Steppes9.5 ± 3.2 c12.7 ± 3.4 b22.6 ± 4.0 b

The highest Cx. poicilipes abundances were observed in the barren and temporary ponds, which were statistically comparable, followed by shrubby savannah, wooded savannah and steppes. Culex neavei abundances were statistically higher in barren, temporary ponds and wooded savannah compared to the other classes. Barren had the highest abundance of Ae. vexans followed by temporary ponds, shrubby, and wooded savannah. Villages and steppes had the lowest abundance of this species.

The average vector abundance according to the trap distance from nearest pond are presented in Figure 3. The Pearson correlation analysis has shown a strong negative and significant correlation between average abundance and trap distance from nearest ponds for Cx. poicilipes (r=– 0.95, P= 0.0004) and Cx. neavei (r=– 0.97, P < 0.0001) but not for Ae. vexans (r=– 0.55, P= 0.22),

Figure 3.

Relationship between vector abundance and distance to the nearest pond during the rainy seasons of 2005 to 2007 at Barkedji, Senegal.

Regression analysis of the abundance of the mosquito vector in relation to distance from the nearest ponds produced the equation: log (abundance + 1) = 1.394 – 0.001× distance for Ae. vexans, log (abundance + 1) = 1.356 – 0.001 × distance for Cx. neavei, and log (abundance + 1) = 1.686 – 0.002 × distance for Cx. poicilipes. Taking into account these equations, the predicted distances from which no mosquitoes would be collected by the light trap will be 1,394 m, 1,356 m and 843 m for Ae. vexans, Cx. neavei and Cx. poicilipes, respectively.

Parity determination

The temporal variations in the vectors’ parity rates were statistically different in 2005 and 2006 (Figure 4). Culex poicilipes parity rates presented a peak in early August and a plateau between late September and late November before becoming zero at the end of December in 2005. In 2006, all Cx. poicilipes females were nulliparous between late June and early August. Thereafter, this species showed two parity peaks in late August (60.7%) and October (79.8%). Culex neavei females showed three parity peaks in 2005, with 80%, 83.7%, and 100% of the dissected females being parous in late August, early November, and late December, respectively. There were two parity peaks in late August and late October 2006, with 100% and 93.5% of parous females, respectively. Aedes vexans also showed three parity peaks in 2005, with 65.3%, 69.2%, and 100% of the dissected females being parous in late July, late September, and late October, respectively. The single specimen of Ae. vexans collected in June 2006 was parous. This vector showed two parity peaks at 100% in late August and early October.

Figure 4.

Temporal dynamics of the RVFV and WNFV parity rates during 2005 and 2006 rainy seasons at Barkedji, Senegal.

DISCUSSION

Although the species diversity of the Barkedji mosquito populations was reasonably high in our study, we found fewer species than previous studies in this area (Wilson et al. 1994, Traore-Lamizana et al. 2001). These differences may be explained by the sampling tools used but also by potential changes in the mosquito fauna caused by general environmental change.

Our study has shown, as have previous studies in the area (Fontenille et al. 1998, Ba et al. 2005), a dominance of Cx. poicilipes and Ae. vexans in the mosquito fauna. Culex poicilipes dominated the fauna in 2005, while in 2006 it was Ae. vexans. This switch-over of domination between Cx. poicilipes and Ae. vexans at Barkedji has been observed on a regular basis between 1990 and 1995 (Traore-Lamizana et al. 2001). The reason for this phenomenon, as well as its impact on the epidemiology of RVf and WNf, remains unknown and needs further analysis. It is worth noting that, taking into account the data presented here and previous data (Traore-Lamizana et al. 2001, Ba et al. 2005), all RVfv strains were isolated at Barkedji during the years when Ae. vexans dominated the mosquito fauna.

The major peaks of Ae. vexans abundance were observed later in the season, in late September in 2006 compared to 2005. This could be explained by the difference in rainfall patterns between 2005 and 2006. Culex spp. dominated the end of the season during the two years. This yearly switch-over between Aedes and Culex has already been described for Barkedji (Traore-Lamizana et al. 2001) and for South Africa (McIntosh et al. 1980). It could be explained by the biology of these two genera. Indeed, the Aedes spp. spend the unfavorable season as resistant desiccated eggs that hatch as soon as the pools are flooded. Therefore, the Aedes populations reach their peak at the beginning of the rainy season. The Culex spend the dry season as nulliparous mated females that lay eggs at the beginning of the rainy season. The population grew gradually to reach its peak later. All the vectors had their highest parity rates between September and October, which is the period when outbreaks of RVf and WNf occurred at Barkedji.

The spatial dynamics of mosquitoes is important for understanding disease epidemiology as well as for formulating the most appropriate control strategy because it allows knowing where and estimating how far an outbreak could spread if a cohort of that vector became infected. Although RVf and WNf vectors were collected in all the study areas, our study has shown that they preferred barren and temporary ponds rather than villages and the others classes. This is in accordance with the result of Ba et al. (2005) who found, in a study of the dispersal of Cx. poicilipes and Ae. vexans that they were rare within villages. This fact implies that these vectors feed on the vertebrates hosts available near temporary ponds. Indeed, there are many rodents, small reptiles, amphibians, birds, and other wild animals living around the ponds. Additionally, nomadic herders migrating through the region and their livestock use barren for shelter. At night, equine also divagate around the ponds after their daily chores. In a recent study on the feeding pattern of these vectors, Ba et al. (2006) have found that Ae. vexans feed mainly on equine.

Our study also showed that abundances of RVf and WNf vectors are related to trap proximity to the ponds which are their larval breeding habitats. Indeed, the mosquito abundance decreased with increasing distance from the nearest pond. There is ample evidence for a relationship between the abundance of a specific vector species in a site and the distance of the site to that vector's breeding places. Indeed, several studies have shown a negative correlation between mosquito abundance and distance to larval habitats (Leonardo et al. 2005). In a dispersal study on Ae. vexans and Cx. poicilipes, in the same area, Ba et al. (2005) found that the number of mosquitoes recaptured decreased with increasing distance from the ponds. The maximum distances from this release point that marked females were recaptured were 620 and 550 m for Ae. vexans and Cx. poicilipes, respectively. This difference in the flight range of the two vectors was also observed in our study. Culex poicilipes have the shorter flying range compared to the other vectors. This fact is very important taking into account that Cx. poicilipes is the mosquito in which the highest numbers of RVfv and WNv have been isolated in Senegal (Diallo et al. 2000, 2005, Ba et al. 2005, Traore-Lamizana et al. 1994, 2001). It is also the only species in which RVfv strains have been isolated during outbreaks in West Africa (Diallo et al. 2005, Faye et al. 2007). Thus, vector control during outbreaks will target a broader area of activity of this species.

Mosquitoes must disperse to locate resting sites, mates, nectar sources, blood sources and oviposition sites. If these are close, then little dispersal is needed. It is probably for this reason that the vectors are aggregated around the ponds. The spatial aggregation of the mosquitoes near the ponds has three principal epidemiological implications. The first implication is that the hosts living close to the ponds would be more likely to be bitten by these vectors and are probably at more risk of disease transmission. There were probably sufficient numbers of hosts that can serve as blood source near the ponds, so the mosquitoes did not have to disperse long distances from their larval habitats. Thus, the distance of a host from the ponds is an important factor that could affect its risk of being infected by viruses. This assertion is supported by the fact that a higher than expected number of cases of Ross River virus have been reported in Australia from areas near major larval habitats (Whelan et al. 1997, Muhar et al. 2000) and for malaria transmission in Mozambique (Thompson et al. 1997). The second implication of the spatial aggregation of the vectors near the ponds is that the potential vertebrate reservoir hosts of the viruses probably live in these environments and should be examined there. Third, our data also support the idea that people rather than mosquitoes rapidly move the arboviruses within and among communities.

One of the direct implications of our study is that vertebrate hosts might be protected against vector borne disease by living at distance far away from mosquito larval habitats (Barrera et al. 1999, Carter et al. 2000). As shown in this study in the Barkedji area, people and their livestock could be protected by living more than 1,500 m from the nearest ponds. On the other hand, the heterogeneous spatial pattern of vector abundance suggests the need to target vector control operations rather than more expensive blanket coverage (Ribeiro et al. 1996). Indeed, the identification of geographic areas with consistently high or low numbers of vectors may allow control activities to be focused in areas with the greatest risk of transmission (Ryan et al. 2004). The potential control of RVf and WNf vectors in the Barkedji area should target the ponds and the 1,500 m area around them. However, in the context of an outbreak, killing infected mosquitoes around ponds alone is not likely to prevent the spread of infections in other areas because infected people are more likely to spread them.

Surveillance is an essential component of disease prevention and control. Indeed, it allows early detection of the virus before human and domestic vertebrates are involved. It is done mainly by entomological and serological surveys (Thonnon et al. 1999, Deegan et al. 2005). Among other factors, the ability of a sentinel surveillance program to detect the early presence of an arbovirus depends on the presentation of the sentinel flock in a microhabitat favoring mosquito attack (Reisen et al. 1995). For a more efficient surveillance of RVf and WNf virus activity at Barkedji, light traps and sentinel bait should be installed at the ponds and the area at their vicinity.

Because we hypothesize, as do other authors (Ryan et al. 2004), that areas with high vector abundance will be associated with a higher risk of infection, our results suggest that at Barkedji, and the area close to the ponds, are at higher risk for RVf and WNf transmission. Thus, the livestock of the nomadic herders and the equines are at higher risk to be infected with RVf and WNf than the local population and their domestic ungulates. Indeed, it's the livestock of the nomadic herders and the equines that live at the ponds and around them at night when the local population and their domestic ungulates shelter in villages. However, most vector-borne disease transmission cycles are complex and the density of vector mosquitoes is not always correlated with pathogen transmission intensity (Beier et al. 1999).

In conclusion, this study showed that the spatial pattern of RVf and WNf vectors is heterogeneous in the area of Barkedji. These vectors remain close to ponds and this area is probably at a higher risk of RVf and WNf transmission. Our results could be helpful in the attempts to map areas at risk of RVf and WNf transmission. Additional environmental and biological parameters and further research will be needed to better define this map, the first step in the planning and execution of cost-effective vector control operations.

Acknowledgments

We thank Sadio Sow and Bidiel Fall for their technical assistance. The authors also acknowledge Dr. Jacques-André Ndione from the Centre de Suivi Ecologique de Dakar, Sénégal for his help with remote sensing and geospatial analyses. This work was partially funded by EU grant GOCE-2003–010284 EDEN (The paper is catalogued by the EDEN steering comittee as EDEN0265). The contents of this publication are the responsibility of the authors only and do not necessarily reflect the views of the European Commission.

Ancillary