Main limitations of the study
This study represents a snapshot of a highly dynamic system and presents the first results of a larger malaria control program. More intensive longitudinal studies are currently being undertaken to complement these results.
The study was implemented during an exceptionally dry year and, although the study took place during the rainy season, it actually rained very little . As a result, it is unclear whether the breeding site structure found during the three months of our sampling period was representative of the structure found in years with normal rainfall. Similarly, the measured ecological parameters might be different during more normal years, as the selectivity of mosquitoes for oviposition sites can be greatly diminished during dry periods because of the limited availability of aquatic habitats . In years with normal rainfall, it is very likely that many habitats that contained larvae during the study period would be unproductive due to a flushing effect, while other small productive sites might appear. However, for the purpose of reducing mosquito density by larvae abatement, the time when the mosquito population is most vulnerable is the dry period. A. gambiae s.s., A. arabiensis and A. funestus survive during the dry period in discreet habitats, making them an easier target for control interventions . Dry season larval control is for example the rule in South Africa .
Very small breeding sites could not be detected by aerial pictures and were, therefore, largely excluded from the study. This could result in the omission of factors potentially important for mosquito control interventions. Further, mosquitoes do not necessarily lay eggs every day in each potential breeding site, thus the reason why a site did not contain larvae could simply be because no eggs were deposited within the last week.
One of the key objectives of the study was to detect ecological factors determining Anopheles sp. larvae productivity. But within a city, the influence of humans can never be neglected. The pollution from waste such as oil, soap or industrial by-products is important in potential breeding sites in Dar es Salaam. Larvae may have been absent due to pollution by elements that could not be detected in the frame of this study.
Anopheles sp. mosquitoes were not classified down to species level. The goal of the study was to characterize important breeding sites of Anopheles sp. mosquitoes and, consequently, potential foci of malaria transmission, regardless of the species. This is because in the context of sustainable operations in a routine mosquito abatement programme, municipal staff cannot be expected to identify all Anopheles larvae samples to species level without rendering sampling procedures prohibitively laborious and expensive. To achieve a satisfactory impact, exhaustive targeting of all potential vector species is necessary anyway. Furthermore, community acceptance of vector control programmes in Dar es Salaam has been shown to require suppression of all mosquito species, rather than only malaria vectors .
Ecological factors influencing mosquito larvae density
The abundance of Anopheles sp. and Culicine larvae seems to be influenced differentially by ecological parameters, as none of them had a significant effect in all four regression models. Only one factor influenced the presence or absence of both Anopheles sp. and Culicine larvae. In turbid breeding sites Culicine larvae were much more likely to be present, whereas Anopheles sp. larvae were much more likely to be absent. Bates  supports this finding of Anopheles sp. breeding in rather clear water, but Gimnig et al.  found increasing A. gambiae s.l. larvae densities with increasing turbidity. Robert et al.  found a clear water preference by A. ambiensis breeding in wells in urban Dakar. A study by Ye-Ebiyo et al.  found that the production of A. arabiensis was favoured in moderately turbid water, while excessive turbidity limited the production of larvae. The proximity to flowering maize with pollen as food source compensated for the development failure induced by excessive turbidity. Clearly, the simple definition of "turbidity" might not be precise enough. Water which is turbid from particles not edible for Anopheles sp. larvae could disfavour the production of larvae, while water turbid from food particles represents a very suitable habitat. The preference of Culicine mosquitoes for turbid water is coherent with their known breeding site preferences, as they breed successfully in rather polluted environments such as blocked drains and septic tanks [36, 37].
Key factors influencing the density of Anopheles sp. larvae
In small breeding sites with diameters less than 1 m, Anopheles sp. were more likely to be present than in larger habitats. This is consistent with the general description of breeding sites for A. gambiae s.l., but not with those for A. funestus [27, 36]. However, this finding could be biased by the fact that rain was lacking for several weeks, evaporating the bigger pools and concentrating the larvae, hence making them easier to detect. Further, larval density of small breeding sites might be increased due to a higher sampling intensity per unit area.
Anopheles larvae were less likely to be present in swamps, and when so, they were found in low densities. This finding matches the known preference of A. gambiae s.l. for temporary sites [36, 38]. Even though swamps were less likely to harbour Anopheles sp., the importance of these habitats should not be underestimated because of their great size and their role in supplying water for irrigation ditches, rice fields and various agricultural activities.
"Puddle" as a type of breeding site was not favoured by A. gambiae s.l. larvae, and when Anopheles sp. larvae were present there, they were likely to be present in low densities. This finding is not consistent with established habitat descriptions for Anopheles sp. larvae [25, 39].
Large drains were likely to have low Anopheles sp. densities. This result agrees with findings by Yamagata . Large drains were often organically polluted with waste water and, therefore, less suitable for Anopheles sp. breeding [21, 28, 38].
Key factors only influencing the density of Culicine larvae
A pH value of less than 7.3 favoured high Culicine densities and a value of less than 7.6 favoured their presence. These results showed clearly that Culicine larvae favoured a pH-neutral environment.
Anopheles sp. breeding sites: unexpected findings
The findings of this study revealed that A. gambiae s.l. was found in a sewage pond (Figure 3 and 4) and in one swamp extremely polluted with organic matter. These findings from Dar es Salaam, together with other studies, could indicate a change of Anopheles sp. breeding requirements in urban settings. In Lahore, Pakistan, Anopheles sp. mosquitoes were found in the waste water system . In Accra, Ghana, data collections since 1911 indicate that A. gambiae s.l. adapted to breeding in organically polluted water habitats . As most studies of Anopheles sp. larval ecology have been conducted in rural settings, it is likely that unpolluted breeding sites are found and described far more often [7, 22], giving a biased impression. One of the main problems is the term "polluted habitat", as it has never been clearly defined. The findings of this study show that Anopheles sp. larvae can breed in nearly every kind of water accumulation. For every ecological factor identified as enhancing or reducing Anopheles sp. larvae productivity, at least one breeding site was found that contradicted these findings. Hence, all water bodies in an urban environment should be considered as potential breeding places and a target for larval control.
Defined socio- and geo-ecological environments with high mosquito breeding sites density
Slopes to riverbeds, riverbeds, borders of swamps, stagnant drains and rivers, areas with restricted access and areas along railway lines represented environments where most breeding sites were found in Dar es Salaam. These sites where mostly associated with agriculture activities. Afrane et al.  stated that areas of high mosquito density tended to follow valleys, where breeding sites were most persistent. Also, in Brazaville the valleys with vegetable gardens and crops were identified as some of the most suitable places for Anopheles sp. breeding . In Dakar, one big marshy area was responsible for the production of nearly all adult Anopheles sp. mosquitoes within a distance of one kilometer . Similar results were found by Staedke et al. . Ponds close to the embankments of a railway line are presented as potential dry-season refugia for mosquito by Charlwood et al. .
The increasing urbanization of Dar es Salaam will probably help to reduce risk areas in the long-term, as open spaces get rare and swampy areas are filled to regain building land, but unplanned city growth along the city edges could increase the number of breeding sites . In Brazaville, areas with the lowest malaria transmission intensity corresponded to the oldest and most densely populated districts . Furthermore, high density housing has been shown to reduce breeding sites more than medium-density housing in Kisumu and Malindi, Kenya .