Simple strategies such as locally cultivated cover crops and plants to provide shade over mosquito breeding habitats as well as the use of predatory fish in combination with Bti are feasible options for the control of immature mosquitoes, including malaria vector species. All habitats provided with shade from Azolla, sweet potatoes, arrow root, Napier grass, rice and papyrus reeds supported significantly fewer anopheline larvae than the controls.
Habitats provided with shade did not have An. gambiae s.l. except for the unweeded rice, while An. coustani, a less important malaria vector, was recorded in all the habitats except in unweeded rice habitats. These effects exceeded our expectations, because An. gambiae s.l. was never recorded from treated sites except from unweeded rice, where it was found once. By contrast, An. gambiae was frequently present in unshaded control sites. Anopheles funestus, a malaria vector species of secondary local importance, was more sampled in canal habitats covered with couch grass. Longitudinal studies in Nyalenda established that An. arabiensis was the main species of the An. gambiae complex found in this area . Anopheles arabiensis thrives best in open, sunlit transient habitats [29, 31] hence shading potential breeding habitats might have affected its abundance.
There was an overall reduction in larval populations within habitats provided with cover crops or plants. As the crops grew taller, increase in height was directly proportional to shade over the mosquito habitat before the crop reached maturity and started to senesce. The findings are in agreement with other studies that found heavy shade to be negatively correlated with larval abundance of anophelines in breeding sites [2, 32–35]. Previous studies carried out in Uganda showed that An. gambiae s.l. did not breed in the interior of papyrus swamps in their natural, undisturbed state . In our study papyrus reed seedlings were transplanted into man-made habitats, hence they were not in their natural state and a reasonable abundance of both anopheline and culicine larvae were recorded within the habitats. The main malaria vector in western Kenya, prefers open, sunlit pools of water, however such habitats become unsuitable for ovipositing females when shade increases . This is probably caused by the action of shade, which lowers the water temperature and reduces algal growth. Gravid female mosquitoes select to oviposit in sun-exposed sites .
The numbers of An. gambiae s.l. mosquito larvae recorded 24 h and 48 h after exposure to treatment was significantly influenced by treatment type. However, man-made habitats provided with both Bti and fish resulted in greater reductions of anopheline larval population densities when compared to habitats where only G. affinis was introduced. These results are comparable to the outcome of experiments conducted by Blaustein  where G. affinis alone failed to control mosquitoes in experimental rice habitats. These results indicate that the predatory effectiveness of mosquito fish on anopheline mosquito larvae diminished when introduced into the man-made larval habitats. The contrast in the findings could be attributed to other factors that we did not investigate/foresee, such as fish preying on other aquatic organisms, external food or invertebrate sources and physical factors such as turbidity. Homski et al.  found that higher turbidity in man-made habitats may have favored a higher abundance of invertebrates and reduced visibility of anopheline larvae for mosquito fish than in sites covered with emergent vegetation. In addition, under natural circumstances other fish species may be better predators on anopheline larvae . The findings on larval control options suggest that G. affinis and Bti, when used together in the right quantities complement each other and are more effective in reducing mosquitoes in man-made habitats.
In this study, Bti was applied in habitats once in a fortnight, which matched with larviciding studies, carried out in Eritrea . With a two weeks interval, our results show a low impact of Bti only on larval abundance. However studies by Fillinger and Lindsay  and Majambere et al. , report microbial larvicides such as Bti to have greater efficacy (95%) when applied to anopheline larval habitats in optimum quantities on a weekly basis. If weekly application of Bti would have been followed, then habitats provided with Bti only may have been as effective as those provided with Bti and fish on larval abundance. In addition, the persistence of Bti endotoxins in our study may have reduced rapidly under field conditions, hence showing no apparent effect on anopheline larval abundance. As previous studies clearly showed that Bti is non toxic to non-target organisms [16, 30], we used this property of Bti to serve as a basis for integrating this product with G. affinis for increased efficacy of larval control.
The trials in this study were done under field conditions in man-made habitats that were naturally colonized by mosquito larvae. Under these conditions, external factors were not controlled and could have played an important role in the colonization and growth of mosquito larvae in the respective habitats. The variations in water level and occasional flooding of habitats could not be avoided, as the sites were exposed to ambient conditions. Factors such as water turbidity, nutrient content in water, cannibalism, predation of immature stages, parasitism, pathogens, competition, water temperature and plant odours that could have either repelled or attracted female mosquitoes during oviposition [41–47] were not controlled and hence could have played a role in the results obtained. All larval stages of culicine mosquitoes increased as vegetation cover increased progressively from man-made ponds, small water canals, rice paddies to swamps. Habitats with few anopheline larvae recorded more culicine larvae, while those that recorded more anophelines had fewer culicine larvae. This suggests selective oviposition behaviour among these mosquito families [37, 48, 49]. In Nyalenda, the water present in breeding habitats was often polluted with debris and human waste, which might have favored proliferation of culicine mosquitoes (data not shown) and at the same time water quality may have had a negative impact on the efficacy of the treatments provided. Competition and differences in the physical-chemical characteristics of the water may have played a role in structuring larval populations, although these factors were not investigated in this study. The standard dipping method was used to estimate mosquito larval densities, which may have underestimated larval abundance [28, 50, 51] and consequently, may have influenced the amounts of Bti and numbers of G. affinis used, leading to contrasting results.
In western Kenya, areas that were previously natural swamps and forests have been transformed into agricultural fields mainly due to human population pressure . These agricultural developments have an impact on the ecological characteristics of the local mosquito vector in terms of density, local microclimate and malaria incidence [2, 33, 52, 53]. Communities in western Kenya are willing to take part in malaria control  and to effect this, simple control strategies suitable for the local of mosquito vectors need to be available as a way of adapting to the changes in land use. Results from this study indicate that locally available leafy plants could be used for mosquito control especially in areas under traditional agriculture. Use of edible fish  and mosquito fish are other options that can easily be put into practice, especially in areas where water is always present.
The effectiveness of biological larvicides for the control of African anophelines has already been demonstrated by several studies [16, 17, 54] and in areas where locally available solutions are not feasible and where water cannot be drained, then application of microbial larvicides could be the best option. Although our results are spatially and temporally limited, the option of using shade from locally available crops and predatory fish seems an easily applicable alternative for the control of mosquito larvae. More importantly, as the level of morbidity resulting from the specific problem of malaria is a net result of a balance between livelihood and ecosystem factors  an ecohealth approach to malaria control is bound to produce discernable and long lasting effects.
This study was part of an ongoing project in different agro-ecological settings in two highland villages (Lunyerere and Fort Ternan) and one peri-urban area (Nyalenda), where most larval habitats were man-made . The field studies reported in this paper were done in the peri-urban area of Kisumu town to assess the best options of controlling immature mosquitoes. For Nyalenda, larviciding and use of predatory fish seem promising and can be supplemented with the existing adulticiding options.