Background
Soon after the start of the rains, populations of the malaria vector Anopheles gambiae sensu lato increase explosively in sub-Saharan Africa. The immature stages of this vector can then be found in numerous transient habitats created by the rains, such as hoof prints, car tracks, borrow pits and ditches 1. The relationship between rainfall on the one hand and mosquito population dynamics and malaria risk on the other is well established and has been modelled accordingly 234. In contrast, the effects of a period of drought on mosquito survival remain less clear.
Previous studies have suggested that adult members of the An. gambiae complex survive the dry season either as aestivating females 56 or as populations in hidden refugia 78. These dry season survival strategies facilitate rapid recolonisation of the area at the onset of the rainy season. During a short spell of drought, Charlwood et al. 9 found a complete cessation of recruitment of young adults to the population and a drastic decline in adult Anopheles arabiensis in Tanzania. This was ascribed to mass larval mortality due to the rapid drying up of the numerous breeding sites, normally present during the wet season. Once the main rains had started, the populations began to build up again.
Recently, Minakawa et al. 8 showed that An. gambiae s.s. oviposits on moist soil, if no better alternative, e.g. flooded soil, is available. This was also true for Anopheles melas, a salt-water breeding member of the An. gambiae complex 10. Since eggs of An. gambiae s.l. remain viable for 12–16 days under dry conditions 1112, egg dormancy has been proposed as a short-term survival mechanism of this species. Field observations, whereby soil from larval habitats that had dried up was investigated for the presence and viability of immature stages, support this hypothesis 81113. Similar observations were made with other malaria vectors, such as Anopheles albimanus 14 and Anopheles balabacensis 15.
During the rainy season, the immature stages of An. gambiae may experience desiccation, due to drying up of their habitats after a few days without rainfall. In addition, eggs may have been laid on damp soil as a result of the female's oviposition choice. The unpredictable rainfall pattern will select those females that make optimal habitat choices for their offspring. Investigating the trade-off between female oviposition choice and optimal larval survival is of key importance in understanding mosquito population dynamics 16.
In this study, we looked at the consequences of a female's choice by testing the ability of eggs to hatch on damp soil and the extent to which emerging first-instar larvae were able to reach a nearby artificial breeding site at varying distances from a simulated oviposition site. Second, we investigated the potential of the four larval development stages to survive on damp soil.
Methods
Mosquitoes
For the egg hatching experiment, we used eggs of An. gambiae s.s. (Kisumu strain), maintained at the Centre for Vector Biology and Control Research (CVBCR) of the Kenya Medical Research Institute (KEMRI) in Kisumu, Kenya. For the larval survival experiment, we used larvae of An. gambiae s.s. (Ifakara strain), maintained at the Mbita Point Research and Training Centre (MPRTC) of the International Centre of Insect Physiology and Ecology (ICIPE), Mbita, Kenya. Adults of both strains were maintained in 30 cm cubic cages with ad libitum access to a 6% glucose solution. The Kisumu strain was maintained on rabbits, while the Ifakara strain was fed on a human arm twice a week.
Egg hatching and larval movement
Plastic plates (18 cm ![]()
, 4 cm deep) or trays (40 × 30 × 11 cm) were filled to a depth of 4 cm with local, black cotton soil, which was first saturated with river water. The topsoil layer was smoothed and in the middle of the plates or trays, breeding sites were simulated by making circular depressions (5 cm ![]()
, 2 – 3 cm deep), which were filled with river water. Fifty eggs of An. gambiae s.s., that had been incubated for one day on wet filter paper, were placed on a small piece of filter paper (0.5 × 1.0 cm). This filter paper was placed directly in the site with water (control) or on the damp soil at 0, 2, 5 or 10 cm from the edge of the artificial breeding site (Figure 1). Each distance experiment was replicated 5 times, each replicate (plate or tray) receiving one batch of 50 eggs. The control, 0 and 2 cm distance experiments were carried out in the plates, while the 5 and 10 cm distance experiments were done in the trays. Plates and trays were placed in a laboratory of the CVBCR, without controlled environmental conditions. During the experiment, mean minimum indoor temperature was 19°C and mean maximum indoor temperature 30°C. Plates and trays were not exposed to direct sunlight.
<p>Figure 1</p>
Experimental set-up of the egg hatching experiment showing the positions of the filter paper relative to the simulated breeding site
Experimental set-up of the egg hatching experiment showing the positions of the filter paper relative to the simulated breeding site.
![]()
At 8 am and at 6 pm the water level in the artificial breeding sites was topped up. At the same times, the soil was dampened by spraying water with a plant sprayer from a height of approximately 1 meter. We made sure there was no water run-off or collection of water on the damp soil.
The number of larvae that was recovered in the artificial breeding site was counted and removed daily for 3 consecutive days. For each replicate, the number of larvae reaching the water surface was expressed as the proportion of the total number of eggs tested. After 3 days, the number of hatched and unhatched eggs was determined by counting the number of opened and closed eggs on the filter paper under a stereomicroscope. For each replicate, the number hatched was expressed as the proportion of the total number of eggs tested. The filter paper was then placed directly in the artificial breeding site, and the number of larvae that hatched overnight was scored the next day.
To test for differences between distances, the observed proportions of first-instar larvae in the artificial breeding site were arcsine square root transformed 17. Analysis of variance (ANOVA) was done on transformed data, followed by Tukey HSD post hoc tests for pairwise comparisons between treatments to determine significance levels. All calculations were carried out with SPSS 11.0 software.
For each replicate, the proportion of larvae expected to reach the site was calculated with the following two equations:
pexp = h * (2 * α /360) (1)
α = invsin ((r + d + 0.25) / r) (2)
whereby pexp is the proportion expected in breeding site, h the proportion hatched, r the radius of the breeding site (in cm), d the distance between the placement of eggs and the breeding site (0, 2, 5 or 10 cm) and 0.25 the distance between the centre and the edge of the filter paper (1.0 × 0.5 cm) (see Figure 1).
Calculations of pexp assumed that larvae move randomly in all directions upon hatching and that mortality during movement is negligible. To test these assumptions, we compared observed and expected proportions of larvae present in the breeding site. Proportions were arcsine square root transformed 17, after which paired t-tests were done. Calculations were carried out with SPSS 11.0 software.
Larval survival
Volcanic soil from within the compound of MPRTC was sun-dried, filtered through wire mesh (3 mm) and saturated with water from Lake Victoria (1 litre of soil : 0.38 litre of water). For each of the four instars tested, 12 metal trays (26 × 26 × 9 cm) were filled with the saturated soil and four depressions (4 cm ![]()
, 8 cm deep) were evenly distributed in the soil of each tray. At the beginning of the experiment, 20 first, second, third or fourth-instar An. gambiae s.s. larvae were placed on the damp soil at the bottom of each depression. Six hours after the beginning of the experiment, 4 depressions per instar were filled completely with water to simulate rainfall. Larvae that appeared alive at the water surface were counted and removed, and the depressions were re-examined 6 and 12 hours later. Twelve hours after the beginning of the experiment, another 4 depressions per instar were filled with water and larvae appearing at the water surface counted. This procedure was continued at 6 hour intervals up to 72 hours after the start of the experiment. All metal trays were placed inside a greenhouse in a large basin of water to prevent ants from disturbing the experiments. During the experiment, trays were exposed to ambient conditions and allowed to dry out. Outdoor temperatures during the experiment ranged from 17.9°C to 28.7°C and averaged 23.2°C.
For each replicate, the total number of larvae appearing at the water surface within 12 hours was expressed as the proportion of the total number of larvae tested (20 per replicate) and these proportions were used as a measure of survival. The proportions were arcsine square root transformed 17, after which a linear regression model was fitted through the data to evaluate survival through time for each instar. All calculations were done with the SPSS 11.0 software
Results
Egg hatching and larval movement
At all distances from the breeding site, eggs of An. gambiae s.s. hatched on filter paper placed on damp soil, but the proportion that hatched within 3 days at 2 cm distance of the artificial breeding site was significantly lower than the control, 0 and 5 cm treatments (Table 1). When, after the experiment, the unhatched eggs were placed directly in the water, the highest proportion additionally hatching originated from the 2 cm treatment, followed by the 10, 5 and 0 cm treatment (Table 1). Although ANOVA showed a significant difference between these proportions (P = 0.039), Tukey HSD post hoc tests could not detect significant differences when proportions were compared pairwise.
<p>Table 1</p>
Mean proportions (± s.e.) of An. gambiae s.s. eggs hatched on damp soil at different distances from a simulated breeding site and mean proportions (± s.e.) of eggs that hatched after 3 days after placement in water. Different letters within the second column indicate significant differences (P < 0.05) between the treatments as determined by Tukey HSD post hoc comparisons.
Treatment
Proportion hatched on damp soil within 3 days
Proportion additionally hatched in water on the 4th day
0 cm
0.79 ± 0.14 a
0.012 ± 0.027
2 cm
0.30 ± 0.12 b
0.120 ± 0.121
5 cm
0.67 ± 0.20 a
0.024 ± 0.026
10 cm
0.54 ± 0.18 ab
0.075 ± 0.034
Control (in water)
0.85 ± 0.05 a
The proportion of newly emerged larvae that reached the site differed significantly between distances (ANOVA, P < 0.001; Figure 2). Eighty-five percent of the eggs that were placed directly in the water, i.e. the control treatment, hatched within 3 days. A significantly lower proportion of larvae was recovered from the site, when eggs had been placed at the edge of the water surface (Tukey HSD; P < 0.001). At 2, 5 and 10 cm again a significantly lower proportion of larvae was recovered in the site, but no differences in proportions were observed between these three distances (Tukey HSD, P > 0.05).
<p>Figure 2</p>
Observed and expected proportions of larvae in the simulated breeding sites, placed during the egg stage at different distances from the site
Observed and expected proportions of larvae in the simulated breeding sites, placed during the egg stage at different distances from the site. Different letters indicate significant differences (P <0.05, Tukey HSD) between the observed proportions at the different distances. Significant differences (P <0.05, paired t-test) between observed and expected proportion of a distance are indicated by '*'; n.s. = not significant
![]()
The observed proportion of free-swimming larvae in the artificial breeding site, that were placed during the egg stage on the edge of the artificial breeding site (0 cm treatment), was significantly greater than expected (paired t-test, P < 0.05), while at 2 and 5 cm no significant differences were found between observed and expected proportions (paired t-test, P > 0.05; Figure 2). At 10 cm distance, the proportion reaching the site was significantly lower than expected (paired t-test, P < 0.05).
Larval survival
Figures 3 and 4 show the average proportion of larvae that appeared at the water surface, and thus survived, after exposure to different periods of drought for all four larval stages. The estimated regression parameters of the linear models that were fitted through the data after arcsine square root transformation are shown in Table 2. By comparing the 95% confidence intervals of regression coefficient a of the fourth-instar with that of the other three instars, we found that survival of fourth-instar larvae was significantly higher than of the other three instars (P < 0.05). Using the model, we predicted that 50% of first, second and third-instar larvae had died after 31, 29 and 33 hours, respectively, while 50% of the fourth-instar larvae had died after 53 hours (Table 2). We estimated that first, second and third-instar larvae could survive up to 64, 65 and 69 hours, respectively, while fourth-instar larvae survive up to 113 hours (4.7 days) under the experimental conditions.
<p>Figure 3</p>
Average proportion (± s.e.) of larvae of An. gambiae s.s. surviving on damp soil after exposure to different periods of drought
Average proportion (± s.e.) of larvae of An. gambiae s.s. surviving on damp soil after exposure to different periods of drought. The lines show the models after back-transformation of the linear model (arcsine(√ p) = a * time + b) obtained after arcsine square root transformation of the original data. Model descriptions are given in Table 2. A: first-instar larvae, B: second-instar larvae.
![]()
<p>Figure 4</p>
Average proportion (± s.e.) of larvae of An. gambiae s.s surviving on damp soil after exposure to different periods of drought.
Average proportion (± s.e.) of larvae of An. gambiae s.s surviving on damp soil after exposure to different periods of drought. The lines show the models after back-transformation of the linear model (arcsine(√ p) = a * time + b ) obtained after arcsine square root transformation of the original data. Model descriptions are given in Table 2. C: third-instar larvae (observed for 48 hours only), D: fourth-instar larvae
![]()
<p>Table 2</p>
Regression coefficients (with 95% confidence intervals (C.I.)) after arcsine square root transformation of survival proportions for the four different instars of An. gambiae s.s.. Model predictions of 50% and 0% survival are given in hours.
Model: arcsine (√ p) = a * time + b
Model predictions
instar
a
95 % C.I.
b
95 % C.I.
R2
50 % survival
(hours)
0 % survival
(hours)
L1
-0.024
-0.026 – -0.022
1.53
1.46 – 1.61
0.931
31
64
L2
-0.022
-0.024 – -0.019
1.43
1.34 – 1.52
0.895
29
65
L3
-0.022
-0.027 – -0.018
1.52
1.39 – 1.66
0.727
33
69
L4
-0.013
-0.015 – -0.011
1.47
1.38 – 1.56
0.748
53
113
Discussion
When natural larval habitats of An. gambiae s.l. dry up, their contribution towards the population dynamics of the adult stage is often neglected, since mass larval mortality is assumed to occur 9. We found that eggs of An. gambiae s.s. hatched and that emerging larvae showed limited capability of reaching a nearby breeding site, when placed on damp soil. With this experiment we simulated a natural situation whereby eggs remain on damp soil when the water level drops or eggs are oviposited deliberately on damp soil after desiccation of the habitat 810. In addition, we found that larvae of An. gambiae s.s. survive for several days in sites that are drying up, depending on the larval stage.
Larvae that hatch from eggs on damp soil were able to reach a breeding site within a distance of 10 cm, albeit in low numbers. We induced hatching of the eggs by keeping the soil of our experimental sites damp by daily spraying of water, but during spraying we made sure that no water ran off or that a film of water was created. Surprisingly, the proportion that hatched at 2 cm distance was significantly lower than at 0 and 5 cm distance. During the experiment we noticed that the plates with the 2 cm treatments dried out more quickly than the other plates. This may be explained because the 2 cm plates were placed closest to the door of the experimental room, and could have dried faster as a result of draught. The observation that the highest proportion of additional hatching occurred in the 2 cm treatment, supports the idea that these eggs may have been exposed to excessively dry conditions during their egg stage and may not have hatched during the first 3 days as a result.
Assuming random movement of newly hatched larvae, we found that more larvae than expected reached the artificial breeding site, when hatched directly on the edge of the site. We observed that a ring of water film (of ~1 cm) was present, directly around the edge of the artificial site, which may explain the higher than expected numbers in the water at this distance. At 2 and 5 cm, there was no difference between observed and expected, while at 10 cm fewer than expected larvae reached the site. This can be explained by assuming that at greater distances mortality during movement towards the site becomes important. Energy reserves may get depleted and larvae may die before reaching the water.
All different larval instars of An. gambiae s.s. survived on damp soil, but the duration of survival depended on the developmental stage of the larvae. From our data we estimated that first, second and third-instar larvae could survive 64–69 hours maximum (2.7 – 2.9 days), while fourth-instar larvae could survive markedly longer up to 113 hours (4.7 days). However, this latter estimate should be considered with caution, since the model was based on data up to 72 hours (3 days) only. Other published experimental tests with larval survival of An. gambiae in drying habitats are few. Muirhead-Thomson 13 found that young larvae could survive for two days on damp soil, while older instars 'were more susceptible to drying up'. Unfortunately, no precise maximum survival time was given. Our study, however, suggests the opposite for survival of fourth-instar larvae, since they survived significantly longer than larvae of the first 3 instars. Holstein 12 observed that larvae (instar undetermined) still developed into pupae after desiccation on damp clay for 4 days. Other evidence of survival on damp soil is anecdotal and comes from personal observations of larvae appearing after adding water to recently desiccated natural habitats.
Female An. gambiae mosquitoes are assumed to use semiochemicals to select suitable aquatic habitats for oviposition 1819. Interestingly, they take a high risk when they select damp soil sites for oviposition, as has been shown for this species 8. This behaviour is not uncommon since other malaria vector species, such as An. balabacensis 20, An. albimanus 21 and the closely related An. melas 10 express this behaviour as well. Probably, the high risk of laying eggs on damp soil pays off when the sites are flooded, since these sites will then not have been colonised by predators, parasites and/or pathogens 2223.
Although eggs of An. gambiae can survive for 12–16 days under dry conditions 1112 and larvae for a few days on humid soil (this study, 1213), we argue that these strategies may not be sufficient to survive throughout the dry season, which generally lasts 6–8 months in sub-Saharan Africa. Other mosquito species, such as Aedes spp. 24 or other Diptera, such as Pylopedilum vanderplanki (Chironomidae) 25, have long dormant immature stages of a few months and several years, respectively, which are adaptive for surviving long-term adverse conditions. Nevertheless, the short-term survival of eggs and larvae of An. gambiae in drying habitats may be adaptive for rapid exploitation of extremely dynamic habitats. Sites may dry up relatively quickly, even during the wet season, but a rain shower normally follows within a few days and development can then resume immediately. We did not investigate, however, to what extent the desiccation period affected the fitness of the adult stages, but this will be addressed in future studies. Long-term survival throughout the dry season is most likely restricted to the adult stages, either as aestivating females 56 or as populations in hidden refugia 78.
Although the World Health Organization considers larval control to have limited applicability in Africa 26, strategies directed towards the aquatic stages of An. gambiae, such as biological control with Bacillus thuringiensis var. israelensis (Bti) or Bacillus sphaericus 272829 and drainage of larval habitats, may provide important tools for the Roll Back Malaria Initiative in Africa 3031. The data resulting from this study suggest that, if larviciding is applied, recently desiccated habitats should be treated as well. If not, larvae in these desiccated habitats may survive and the habitats may become an important source of rapid re-infestation. As a consequence, control attempts may be frustrated.
Apart from having direct implications for larval control operations, the results will also affect environmental management practices, in particular intermittent drainage of irrigated rice fields. Although rice irrigation is associated with an increased malaria risk in some areas and a decreased risk in others (reviewed in 32), there is general concensus that proper water management is required to obtain high rice yields with low water consumption and no additional malaria risk (reviewed in 33). Therefore, intermittent irrigation seems to be the best choice 3435, but if the dry field period is too short, the larval stages will survive, as shown in this study. Within three dry days, most of the young instar larvae will have been killed, but only after five dry days will the older instars have been killed. Iterative drainage for a period of several days seems feasible for rice irrigation schemes, but only where climate conditions favour rapid drying 33.