First comprehensive analysis of Aedes aegypti bionomics during an arbovirus outbreak in west Africa: Dengue in Ouagadougou, Burkina Faso, 2016–2017

Background Dengue’s emergence in West Africa was typified by the Burkina Faso outbreaks in 2016 and 2017, the nation’s largest to date. In both years, we undertook three-month surveys of Aedes populations in or near the capital city Ouagadougou, where the outbreaks were centered. Methodology In 1200LG (urban), Tabtenga (peri-urban) and Goundry (rural) localities, we collected indoor and outdoor resting mosquito adults, characterized larval habitats and containers producing pupae and reared immature stages to adulthood in the laboratory for identification. All mosquito adults were identified morphologically. Host species (from which bloodmeals were taken) were identified by PCR. Generalized mixed models were used to investigate relationships between adult or larval densities and multiple explanatory variables. Results From samples in 1,780 houses, adult Ae. aegypti were significantly more abundant in the two urban localities (Tabtenga and 1200 LG) in both years than in the rural site (Goundry), where Anopheles spp. were far more common. Results from adult collections indicated a highly exophilic and anthropophilic (>90% bloodmeals of human origin) vector population, but with a relatively high proportion of bloodfed females caught inside houses. Habitats producing most pupae were waste tires (37% of total pupae), animal troughs (44%) and large water barrels (30%). While Stegomyia indices were not reliable indicators of adult mosquito abundance, shared influences on adult and immature stage densities included rainfall and container water level, collection month and container type/purpose. Spatial analysis showed autocorrelation of densities, with a partial overlap in adult and immature stage hotspots. Conclusion Results provide an evidence base for the selection of appropriate vector control methods to minimize the risk, frequency and magnitude of future outbreaks in Ouagadougou. An integrated strategy combining community-driven practices, waste disposal and insecticide-based interventions is proposed. The prospects for developing a regional approach to arbovirus control in West Africa or across Africa are discussed.

Background Dengue's emergence in West Africa was typified by the Burkina Faso outbreaks in 2016 and 2017, the nation's largest to date. In both years, we undertook three-month surveys of Aedes populations in or near the capital city Ouagadougou, where the outbreaks were centered. Methodology In 1200LG (urban), Tabtenga (peri-urban) and Goundry (rural) localities, we collected indoor and outdoor resting mosquito adults, characterized larval habitats and containers producing pupae and reared immature stages to adulthood in the laboratory for identification. All mosquito adults were identified morphologically. Host species (from which bloodmeals were taken) were identified by PCR. Generalized mixed models were used to investigate relationships between adult or larval densities and multiple explanatory variables. Results From samples in 1,780 houses, adult Ae. aegypti was significantly more abundant in the two urban localities (Tabtenga and 1200 LG) in both years than in the rural site (Goundry), where Anopheles spp. were far more common. Results from adult collections indicated a highly exophilic and anthropophilic (>90% bloodmeals of human origin) vector population, but with a relatively high proportion of bloodfed females caught inside houses. Habitats producing most pupae were waste tires (37% of total pupae), animal troughs (44%) and large water barrels (30%). While Stegomyia indices were not reliable indicators of adult mosquito abundance, shared influences on adult and immature stage densities included rainfall and container water level, collection month and container type/purpose. Spatial analysis showed autocorrelation of densities, with a partial overlap in adult and immature stage hotspots. Conclusion Results provide an evidence base for the selection of appropriate vector control methods to minimize the risk, frequency and magnitude of future outbreaks in Ouagadougou. An integrated strategy combining community-driven practices, waste disposal and insecticide-based interventions is proposed. The prospects for developing a regional approach to arbovirus control in West Africa or across Africa are discussed.   In 1200LG (urban), Tabtenga (peri-urban) and Goundry (rural) localities, we collected indoor and 33 outdoor resting mosquito adults, characterized larval habitats and containers producing pupae and 34 reared immature stages to adulthood in the laboratory for identification. All mosquito adults were 35 identified morphologically. Host species (from which bloodmeals were taken) were identified by 36 PCR. Generalized mixed models were used to investigate relationships between adult or larval 37 densities and multiple explanatory variables. 38

39
From samples in 1,780 houses, adult Ae. aegypti was significantly more abundant in the two urban 40 localities (Tabtenga and 1200 LG) in both years than in the rural site (Goundry), where Anopheles 41 spp. were far more common. Results from adult collections indicated a highly exophilic and 42 anthropophilic (>90% bloodmeals of human origin) vector population, but with a relatively high 43 proportion of bloodfed females caught inside houses. Habitats producing most pupae were waste 44 tires (37% of total pupae), animal troughs (44%) and large water barrels (30%). 45 While Stegomyia indices were not reliable indicators of adult mosquito abundance, shared 46 influences on adult and immature stage densities included rainfall and container water level, 47 collection month and container type/purpose. Spatial analysis showed autocorrelation of densities, 48 with a partial overlap in adult and immature stage hotspots. 49 Conclusion 50 1200 Logements: (1200 LG) (12°22'N; 1°29'W) is an urban setting of 1.2 km 2 in central 122 Ouagadougou, less than 1 km from the international airport. Roads are paved and the area is 123 connected to centralized water, waste and electricity systems. Houses are relatively modern single-124 or two-storey and typically comprise one living room and 2-3 bedrooms, often with air 125 conditioning. Vegetation, often within gardens, is common on both private and common land. 126 Tabtenga: (12°22'N; 1°27'W) is a 10 km 2 peri-urban setting located within Ouagadougou, 127 approximately 5 km east of 1200 LG. Roads are unpaved, there are no electricity or waste 128 management systems and the majority of households obtain water at communal pumps. Typical 129 households are single-storey structures with 1-4 rooms, within walled compounds. Vegetation is 130 sparse. 131 Goundry: (12°30'N, 1°20'W) is a small rural farming community village situated 30 km north-east 132 of Ouagadougou, with unpaved roads, low housing density, and is surrounded by fields and trees. 133 There is a dam in the center of the villages enabling people to practice gardening during the dry 134 season. Goundry has no electricity or waste management systems and households obtain water at 135 communal pumps. Livestock, mainly cattle and sheep, and dogs are common. 136

Study Design 137
Longitudinal surveys were carried out during the wet season from August to October in both 2016 138 and 2017. Prior to each survey, all prospective houses were visited to inform the population of the 139 proposed project, and the proposed sampling procedures. Each day, the first house was selected at 140 random and the second and subsequent houses chosen where the family were present and agreed to 141 participate, avoiding the houses nearest to the previous one. An average of ten houses were visited 142 per day. Houses were sampled at 06:00-09:00 or 16:00-19:00 by two teams of four persons, each 143 working as follows: one person collected written informed consent and household data (S1Table) 144 from participants, a second person searched for adult mosquitoes indoors and outdoors; two people 145 recorded breeding habitat characteristics and collected immature stages of mosquitoes. Areas of 146 public or communal land adjacent to sampled houses were mapped and inspected for containers, 147 from which mosquito immature stages were processed as described below. 148

Breeding site characterization and collection of immature mosquitoes 149
At each property, the team worked indoors and outdoors inspecting every container capable of 150 holding sufficient water for immature mosquitoes, recording its dimensions, water volume, water 151 level, material (natural, wood, metal, cement, etc.) and utility (whether the container was in use 152 (yes) or discarded waste (no)). Where possible, water from each container was poured into a 153 graduated beak and any immature mosquitoes collected using a sieve. The water volume of heavy 154 or immovable containers, including potable water, was measured by removal with buckets, again 155 using a sieve to collect all mosquitoes, before being returning to the container. All larvae and pupae 156 were transferred alive to containers labelled by house number and breeding site location, for 157 subsequent identification. 158 Breeding habitats were categorized using the WHO operational guide

Pupal indices 173
To identify the most productive dengue vector breeding sites, and to determine whether certain 174 habitats that were relatively more productive for pupae might be identified for future targeting, we 175 recorded pupae separately from larvae. Pupal mortality is typically low meaning that the number of 176 pupae is highly correlated with the number of adults [38]. To identify the most productive Ae. 177 aegypti immature stage habitats, the percentage contribution of each container type to the total 178 count of pupae is calculated as the total number of pupae per container type, divided by the total 179 number of pupae in all containers throughout the study area [28]. Outdoors, all walls, eaves, vegetation (flowers) and shaded areas within or behind containers, stored 185 materials, and car tires within the walled area marking the perimeter of each household were then 186 inspected for an additional 10 minutes. 187

Processing and identification of collected mosquitoes 188
All immature mosquito stages were sorted based on morphology then reared in cups with ground 189 Tetramin as food until adult emergence. Pupae were transferred to fresh cups and emerging adults 190 killed and preserved by freezing for identification. 191 Mosquitoes were identified by microscopy using morphological keys. Aedes aegypti [42][43][44] was

Meteorological data 216
Daily records of minimal, maximal, mean temperatures, relative humidity and daily rainfall were 217 obtained from the National Meteorological Agency records from Ouagadougou station for 2016 and 218 2017. Intermediate calculations were made for the cumulative rainfall for previous 4 days, one 219 week, 12 days or 14 days to be included in the models.

Statistical analyses 221
To investigate factors associated with Aedes sp. abundance we developed four Generalised Linear 222 Mixed Models (GLMMs) with a negative binomial link function using the R package "glmmTMB". 223 The first (adult model) and second (bloodfed model) were fitted to the number of adults and 224 bloodfed Aedes collected in each house, respectively, as a function of the abiotic and biotic 225 covariates: the year of collection, locality (1200 LG, Tabtenga, Goundry), house type (mixed, 226 modern or traditional), location (indoors/outdoors), number of children, ITNs presence, ITNs 227 number, animal presence, animal number, day of collection, month of collection, the number of 228 immatures collected in breeding sites located in that house, and also climate factors including 229 temperature and cumulative rainfall. The interaction terms 'collection location' and locality, and 230 locality and year were also included, and to account for variation arising from the sampling design 231 we included date of collection and house identifier as random effects. The third (larval model) and 232 fourth (pupal model) models were fitted to the number of larvae and pupae per container, 233 respectively, as a function of the abiotic and biotic covariates: locality, container type, water level, 234 water volume, container material, the 'location' (i.e. indoors/outdoors) of the containers utility, 235 rainfall and temperature, number of adults collected in the house. As in the adult mosquito model, 236 the variables location and locality, and year and locality were also included as interaction terms, and 237 date of collection and house identifier were added as random effects. 238 From these full models we selected the minimal model using a stepwise backward model selection 239 procedure based on the lowest AIC values by removing factors with highest p-value in the model. If 240 removing a variable resulted in a change of the AIC value of more than 2 and the resultant model 241 was still parsimonious. e.g. following residuals diagnostics in DHARMa [51], the simplified model 242 was kept. This procedure was repeated until removing variables no longer improved the model. 243 To ensure the models described above were appropriate, we also explored the use of alternative 244 families such as poisson and quasipoisson, but DHARMa residuals diagnostics indicated that these 245 failed to capture the dispersion in data. We tested for spatial autocorrelation both in the data and residuals of the models using the Moran's I test [52] and found no significant spatial autocorrelation 247 (for a p-value of 0.05), hence inclusion of a spatial term was not required. Finally, we estimated the 248 correlation among our potential covariates, if two variables were >50% correlated, one of them was 249 excluded from the final full model. These included temperature and relative humidity, immatures 250 total number and larvae number, number of residents and number of children. Since larvae and 251 immatures were correlated (53%), for the adult model we used the covariate 'immatures' (sum of 252 larvae and pupae). 253 To estimate the overlap between the spatial distributions of adult and immature stages, we used the 254 "nicheOverlap" function in the R package 'dismo' [53], which estimates an index of similarity 255 between rasterized density distributions based on [54] , and calculated the Pearson correlation 256 coefficient between these two distributions using the 'layerStats' function in the package 'raster' 257 [55]. 258 To analyse whether the location of blood-feeding (indoor vs outdoors) was in line with collection 259 densities, the expected numbers of indoor blood fed females were predicted from the total bloodfed 260 collections in each year and locality multiplied by the relative indoor density. A chi-square 261 goodness of fit test was used to determine whether observed values deviated from predictions. 262 Other comparisons between proportions used chi-square contingency table tests or Fisher exact tests 263 (depending on the expected values). Stegomyia index results were compared among localities using 264 non-overlapping confidence intervals as indicative of a significant difference. 265

Characteristics of sampled houses 267
A total of 1,163 houses were sampled in 2016 and 631 in 2017, plus an additional 24 public spaces 268 in 2017 (S1Table), including among others, places of worship, schools, market stalls and stores. All 269 houses in 1200 LG (urban) were cement block buildings of modern designs, whereas the houses in 270 houses. Average occupancy rates in each locality in 2016 were 5.3, 5.0 and 2.9 residents per house, 272 and insecticide-treated nets (ITNs) were seen in 75.9%, 91.5% and 75.9% of houses, with an 273 average of 0.44, 0.43 and 0.43 ITNs per person, respectively. In 2017, the average occupancy rates 274 in each locality were 5.2, 6.2 and 3.3, and ITNs were seen in 60.6%, 87.8% and 88.7% respectively, 275 with an average 0.36 ITNs per person in each localities. 276

Adult mosquito species abundance 277
A total of 47,255 adult mosquitoes were collected during both years in all localities by indoor and 278 outdoor resting catches ( and Anopheles ziemani) and the predatory culicine Lutzia tigripes were recorded much more

Resting location and diurnal activity of adult Aedes aegypti 305
Significantly more adult female Ae. aegypti were collected resting outdoors than indoors in all 306 localities in each year (S2Table), with a remarkably consistent proportion outdoors each year 307 (mean=0.73 in both 2016 and 2017), equivalent to an outdoor: indoor ratio of 2.7-fold (binomial 308 test P<0.001). As can be seen in Figure 1  However, the proportion of bloodfed females caught outdoor was only slightly greater than those 317 indoor (ratio = 1.05), which represents approximately twice the expected number predicted from 318 the total indoor: outdoor catch ratio (Table 2). This suggests that a preference for exophily may not 319 be sustained through the entire gonotrophic cycle. 320 Trends in morning vs. afternoon collections of Ae. aegypti were inconsistent across years (S3Table). 321 In 2016, there appeared to be a bias toward morning collections (59.6%; binomial test, P<0.001), 322 but in 2017 morning and afternoon collections were very similar (50.3%; binomial test, P=0.83). 323 Similarly, strong variation between localities in morning:afternoon collections was evident in 2016, 324 with relatively morning biased collections in the urban and peri-urban sites but afternoon-biased in 325 Goundry (χ 2 2=497, P<0.001), but was barely-evident in 2017 (χ 2 2=6.1, P=0.047).  Table 3. Consistent with the 339 analyses presented above, locality (urban, peri-urban, rural) was a strong determinant, as was the 340 collection location (indoors vs. outdoors), with variation in the indoor:outdoor ratio between 341 localities shown by the significant interaction term. Whilst collections in Tabtenga and Goundry 342 were quite consistent between years, the number of Ae. aegypti collected in 1200 LG was much 343 lower in 2017 than 2016 (evident in the significant year and year x locality interaction terms). 344 However, it should be noted that the proportion of Ae. aegypti in the total mosquito catch was 345 similar, and actually slightly higher, in 1200 LG in 2017 as a result of a much lower abundance of 346 Cx. quinquefasciatus (Table 1). Collection month and rainfall also exerted significant effects with 347 reduced abundance in October, and a positive relationship with elevated previous over the past 14 348 days rainfall. Importantly, adult abundance was also predicted by the abundance of immature stages 349 from containers in or around the same household, with larval and pupal collections pooled due to 350 low pupal numbers. House type was retained in the minimal model, but explained little variation, 351 whilst other factors did not add to predictive value and were excluded from the final model.

Immature mosquito species abundance 362
Aedes aegypti comprised over 85% of immature stages collected from habitats in the urban and 363 peri-urban localities, 1200 LG and Tabtenga, but only 46.4% in rural Goundry. In contrast, Aedes 364 vittatus, which comprised less than 0.5% of the immatures collected in the two urban localities, 365 amounted to 40.6% of the total in Goundry (S5Table). Aedes vittatus immatures were found mainly 366 in water residues in the animal drinking troughs that were more common in Goundry than in the 367 urban sites (S1Figure). Few immature Anopheles gambiae s.l. were collected in any site, as would 368 be expected for a species that typically breeds in ground water pools rather than the containers 369 sampled in this study. 370 Overall mosquito community composition was strongly influenced by collection location, which 371 was consistent across the two sampling years, with significant difference between rural Goundry 372 and the two urban sites (S1Figure). 373

Habitats of immature stage Aedes aegypti 374
Across all three localities, a total of 1,445 containers were inspected during the study of which 666 375 (46%) contained Aedes aegypti larvae or pupae (Fig.2). Immatures were found in all container types 376 inspected and although infestation rates of container types were significantly different across the 377 localities (χ 2 =170; P<<0.001; Figure 2), there were some consistencies. Tires were among the most 378 heavily infested in both urbanised localities, reaching rates of 31% and 32% in Tabtenga and 1200  379 LG respectively, but they were of minimal importance in Goundry, where infested domestic water 380 storage drums (40%) were the most heavily infested (Figure 2). 381 Not all containers found to contain larvae may support development to the pupal stage and be regarded as productive for breeding. However for pupae, in urban and peri-urban sites, tires were 387 highly productive habitats, containing 37% and 34% of all pupae in 1200 LG and Tabtenga 388 respectively (Table 4; Figure 2). 389 Other highly productive containers included large water storage containers (drums, jars or barrels) 390 from which, in Tabtenga, an area without piped water, 40% of pupae were found. In contrast, these 391 large containers produced only 13% of pupae in 1200 LG, an area with piped water, and where 392 instead, small containers including miscellaneous objects ( plastic and metallic boxes, terracotta 393 pots, plastic shoes,.. ) were responsible for nearly 40% of the vector population. 394 Large water drums or barrels were also important in the semi-rural locality Goundry (30%), where 395 the most productive habitats were animal water troughs (44%). showed an effect of year (2017 collections > 2016 collections) and significant differences between 405 localities, which were consistent across years, with both the urban and peri-urban localities having 406 much higher densities than the rural site, Goundry (S6Table). Container type influenced larval 407 density with highest larval densities found in tires. Higher water levels in containers were also 408 associated with higher larval density and highest larval densities occurred in September compared 409 to August and October, though all were similar. Factors such as cumulative rainfall totals measured 410 over 2 or 7 days, temperature, container purpose, number of residents, and numbers of adult 411 mosquitoes collected had no significant associations with larval density. 412 Pupal density did not differ significantly between years, and was higher only in Tabtenga than 413 Goundry, with urban 1200 LG not significantly different (S7Table). Container type influenced 414 pupal density, but in contrast to larvae, tires were not significantly more productive, with only 415 animal drinking troughs significantly higher than the reference category. Container utility was 416 important as pupal density was reduced by 43% in functional containers compared with non-417 functional/discarded containers Pupal density was also negatively associated with mean temperature 418 and was also positively associated with the number of adult mosquitoes collected in the same house. 419

Stegomyia and pupae/person indices 420
The Stegomyia indices are summarized in Figure 3 The objective of this study was to generate detailed baseline data on the biology and behaviour of 445 Ae. aegypti in Burkina Faso, contributing to the essential evidence-base for developing dengue 446 prevention and outbreak plans. The key findings indicate that the arbovirus vector Ae. aegvpti is 447 common throughout Ouagadougou, and most abundant in the highly populated central areas where 448 infestation rates reach 78% of dwellings. Adult females are predominantly anthropophagic and also 449 highly exophilic, though feeding more indoors than exophily rates would suggest. Females also 450 appear to oviposit in all container types, both in use or discarded. 451

Typology of breeding sites 452
The key container habitats, those harboring the highest numbers of pupae, and from which the 453 greatest numbers of adults emerge, were discarded car tires, large domestic water containers (drums 454 and barrels) and small containers (including discarded vessels). Aedes aegypti were caught in both 455 mornings and evenings, consistent with their expected pattern of diurnal activity , but it is unclear to 456 what extent they may be also nocturnally active. 457 The profile in the rural outskirts of the city appears to be quite different. Here there was a greater 458 diversity of mosquito species, the most common of which were Anopheles gambiae and Culex 459 quinquefasciatus, and Aedes vittatus. The most productive containers in Goundry for both Ae. Larviciding can also be used for larval reduction using temephos, an organophosphate and two 492 biological insecticides Bti and pyriproxyfen. Preliminary data shows organophosphates are effective 493 against Aedes aegypti larvae in Burkina [63], but data on efficacy of Bti and pyriproxyfen area 494 awaited. Although the identification of key containers for pupal productivity may reduce the 495 challenge, the diversity and the number of breeding containers will compromise larviciding as a 496 stand-alone method for dengue control in Burkina Faso. 497

Aedes aegypti resting and blood feeding behavior 498
Aedes aegypti adults can be prevented from entering buildings by screens fitted to windows, which 499 do not necessarily need to be insecticide-treated to be effective [64,65]. Indoor resting can be 500 controlled by targeted indoor residual spraying (TIRS), where only the lower half of the walls are 501 treated with insecticide or by using hand-held aerosol cans to spray known resting indoor sites [66][67][68]. Clearly, further work is required to fully elucidate resting preferences, a critical question for 503 planning control, and the efficacy of IRS against Ae. aegypti in Africa need to be evaluated 504 experimentally as a priority even if the exophilic behaviour may recommend additional control 505 methods. The high levels of exophily recorded in this study (Fig. 1)  Indoors Residual Spray) on Ae. aegypti were also performed in Mexico. In Burkina Faso, a 518 significantly higher than expected proportion of bloodfed females are found resting indoors, which 519 suggests that exophily may not dominate the entire adult stage and that adult females are likely 520 endophagic or come indoors at some stage after bloodfeeding. 521 The majority of bloodmeals were identified as human in origin, with the remainder from dogs and 522 only one sample from cattle (Table S4). Notably, only 6-7% of bloodmeals were non-human in 523 samples from the urban localities, but 28% of bloodmeals in the semi-rural site were from dogs. predominantly Aaa behaviour and a rural population that is still anthropophilic but with a far greater 532 likelihood of zoophagy [4]. Clarification of host preference is an important element of a mosquito 533 population's vectorial capacity but doing so will require a larger sample size and additional studies 534 to gain more insight. This should be prioritised together with the studies on indoor/outdoor feeding 535 and resting preferences. 536

Mosquito species diversity 537
Higher densities of adult Ae. aegypti were found in urban and peri-urban localities of 1200 LG and 538 Tabtenga, compared to Goundry, the rural locality. Adults rested mainly outdoors in all sites, 539 feeding preferentially on humans with rare canine or bovine bloodmeals and their densities were 540 affected by month and the year of collection, the locality, the indoors/outdoors location and at a 541 lesser extent by the immature stage abundance and the cumulative rain of 14 previous days. 542 Ae. aegypti was the main Aedes species collected at all developmental stages, in all localities, in 543 both years. Ae. aegypti adults and larvae densities followed a negative gradient from urban to rural 544 localities. Urbanisation has been identified as the main driver of Ae. aegypti proliferation in Africa 545 [4,78], and other environmental changes resulting from human activities promote higher abundance 546 and lower species diversity; lower abundance and higher species diversity are more typical of 547 natural environments and ecosystems [79]. In our study, the diversity of all culicines was greater in 548 the rural site, Goundry, than in the urban and peri-urban localities. Goundry is predominantly 549 agricultural land at the edge of the bush, with trees and scrubland beyond. Aedes vittatus was 550 common here only, preferring animal drinking troughs as larval habitats. Also found only at 551 Goundry, the predatory larvae of Lutzia tigripes shared some habitats with a prey species, Ae. 552 aegypti and may have contributed to the lower densities of Ae. aegypti in Goundry compared to 553 common here though not as abundant as at the other localities, and Anopheles gambiae s.l., which 555 was common. 556

Determinants of larval and adult mosquito densities. 557
Multiple containers types were found to contain larvae in urban and peri-urban sites, with tires the 558 most common. The typology of containers may vary according to the locality. Drums and barrels 559 that are used for water storage, are more abundant in the rural and peri-urban localities of Goundry Breeding site characteristics that affect immature stages abundance and adult life history traits 567 include among others, dissolved oxygen, water temperature, pH, conductivity and salinity [85]. 568 Characteristics such as dissolved solids, ammonia, nitrate, and organic matter vary significantly 569 between urban and rural containers, which might explain some urban-rural differences in breeding 570 of Ae. aegypti [86]. We examined a limited number of breeding sites characteristics and 571 environmental variables and found that container types and water levels within, can increase larval 572 density while containers that are in use, or classed as useful, decrease pupal density. Cumulative 573 rainfall of the previous 14 days and mean temperature affect adult and pupal densities respectively. 574 Investigations in Iquitos used a generalised additive model to highlight the important contribution to 575 Ae. aegypti adult density of weather-related covariates including temperature, rainfall and wind 576 [87]. Though our study did not consider other covariates related to breeding sites, the density of 577 immature stages was the strongest covariate in the model contributing to adult density. 578 Immature stages as well as adult densities were more affected by the month of collection with 579 immature stages contributed to adult density, only pupal density was affected by adult abundance. In this study the number of total adults collected was highly correlated with the number of fed 598 females collected and their density models shared the same explanatory variables. In contrast to 599 Stegomyia indices, which are based on immature stage numbers, estimating the total number of 600 adult Ae. aegypti can be more informative but remains challenging, with precision depending of the 601 sampling methods and sampling efforts [37]. A pattern of similarities between spatial distribution of 602 adults and immature stages are found, with hotspots overlapping in the urban sites, at least. More 603 consistent indices are needed, to take into account Ae. aegypti resting and blood feeding behaviour, 604 and the typology of breeding sites that are productive for pupae, yet potentially specific to the 605 locality. Associating dengue active case detection in the community with holistic collection of Ae.
aegypti bionomics collection could allow more consistent inferences to be made. 607

Perspective for Aedes aegypti-transmitting disease control 608
Without vaccines for three of the four arboviruses transmitted by Ae. aegypti, Burkina Faso,  This study has done little to alter the view that the Stegomyia indices have limited epidemiological 650 value or that they are likely to be more relevant in Africa than they have been elsewhere. 651 Identifying alternatives however remains elusive. While there are similarities between immature 652 stage and adult densities and similarities in spatial distribution, determining how these might be 653 applied or how they could be combined with additional epidemiological parameters to generate 654 more accurate indices reflecting the transmission potential of Ae. aegypti and disease risk remains a 655 challenge. 656 All of these topics fit well within a regional approach to arbovirus control. Networks mapping Ae. 657 aegypti key behaviors across the African continent, together with accurate indices of arbovirus risk, 658 insecticide susceptibility/resistance status and key epidemiological parameters would be an 659 important step towards an effective regional/ global control strategy. Such an initiative is suspected 660 to have broad support.    We are grateful to the reviewers for their comments and have revised the manuscript accordingly, and provided point-by-point responses (with the reviewer queries in red). We have also made very minor corrections for typo errors and to improve English. We hope that our amendments will make this version suitable for publication.

Reviewer#1
We thank the reviewer for their helpful corrections throughout the manuscript all of which have been made as suggested.

Title:
The title seems to be underselling the manuscript a little bit. One could emphasize that this is the first comprehensive analysis of a West African population The title is changed to "First comprehensive analysis of Aedes aegypti bionomics during an arbovirus outbreak in west Africa: 2016-2017 dengue in Ouagadougou, Burkina Faso".

Abstract:
(1) L32: punctuation missing after PCR. The full stop has been added after PCR and the sentence has been modified to include parentheses as suggested by reviewer 1.
(2) L36: Isn't Culex the most abundant mosquito sampled? See also my comments on table 1. Culex were the most common in urban sites, and we have made a change to better illustrate the urban vs rural differences, focusing on Aedes aegypti and Anopheles spp.

Methods
(3) L119 ff: How do the housing densities in the two urban areas compare?. Tabtenga was the locality with highest housing density compared to 1200LG, clarification is given in the text.

Results
(4) L296 ff: Table1 indicates that Culex is the most abundant in the adult collections, which is opposing the statement in the Abstract that Aedes aegypti is the most abundant one, or does that only refer to immature stages, as indicated in figure S1? If Culex is most dominant in the adult samples but Aedes in the immature ones, how are these differences explained?
The abstract has been corrected as described above; this refers to adult samples

Response to Reviewers
Also on table 1: I assume the fractions given for the Anopheles gambiae species indicate their percentage in the species complex and not in the overall sample. This should be indicated. But why do the numbers of Anopheles gambiae species do not add up? For Anopheles gambiae species, not all the collected mosquitoes have been subjected to PCR identification, explaining why the numbers don't sum to the An. gambiae sl. total. This is indicated in Table 1 legend by addition of 'The proportions of Anopheles gambiae complex members were calculated using the total number of An. gambiae s.l. identified to species (rather than total collected) as denominator.' In the meantime, we increased the sample size for identified Anopheles gambiae complex mosquitoes to species in Table 1. (5) L325 ff: Table 2: remove ' §' from headline. On right hand side some numbers seem to have shifted? Under % indoor the fraction of 1 rather than the percentage is given, which could be misunderstood at first glance. The symbol is removed from headline. The numbers are expressed as percentages now and we made sure that the numbers have not shifted.
(6) L434 ff: Is there any idea/explanation why the overlap of immature and adult distribution is notably lower in the rural area? Why are adults completely missing from some sites where larvae were sampled?. We have added the following possible explanation in the discussion 'The lower overlap in adult vs. larval densities in the rural compared to urban locations might be explained by a closer proximity of vegetation around households in the former, which offer potentially suitable, but difficult to sample, resting places for adults.
(7) L656 ff (Figure legends) The legends to figure 2 refers to figure 3 and vice versa, this needs to be corrected and the entire text should be checked if references to figures follow the right numbering. Also on figure legend 3 (referring to figure 2): The colour code is not given. In addition, the description 'Proportion of positive containers for larvae' does not fully fit what is shown. It seems to rather be a distribution of all larvae across the different container types, adding up to 100% in total The legends have been checked and now refer to the appropriate figures and the color codes are given in figure 3 and the title has been changed to correspond to the figure.
L667/ figure 4: I assume the dark black spots represent the houses that were sampled? This should be indicated in the figure legends This is correct and has been added to Fig 4 legend Discussion: (7) General: I would suggest the use of sub-headings in the discussion to improve structure and readability Sub-headings are now added to the discussion