The study was implemented in Lupiro village (8.385°S, 36.670°E) (Figure 1), in the Kilombero valley, south-eastern Tanzania. Households were selected from four sub-villages namely: a) Ndoro; b) Libaratula; c) Mabatini and d) Lupiro Kati. Most residents here were peasants, cultivating rice, maize and other crops. Houses have brick or mud walls, and metal (corrugated iron sheets) or grass-thatched roofs. Annual rainfall is 1200-1600 mm, and temperatures range between 20.0°C and 32.6° [39, 40]. Principal malaria vectors in this area are Anopheles funestus and Anopheles arabiensis with the former contributing over 80% of transmission . Both An. arabiensis and An. funestus populations in the area have been shown to be resistant to multiple public health insecticides including pyrethroids, carbamates and organochlorides [41-43]. LLINs are the main malaria prevention method, most of which are distributed by the government .
Characterization of the peri-domestic spaces
Two hundred (200) households were surveyed, including 50 from each sub-village (Figure 1), selected via stratified random sampling. Data were collected using electronic tablets using KoboCollect™, an open access software programmed using Open Data Kit (ODK) . Trained research teams were assigned to each sub-village. Written informed consent was obtained from each of the 200 households. For each household, the peri-domestic spaces were observed directly to characterize them physically based on use, physical site and , as well as whether they were built-up or not. Digital pictures were taken of the different peri-domestic environments. The research team also administered survey questions to the household heads to capture: a) identification information such as age, b) education level, c) socio-economic data including source of income, possession of radio, television, cell phone among others, d) information on peri-domestic spaces such as presence of other peridomestic spaces apart from veranda, and e) their usage, presence of peri-domestic spaces if the house had no veranda and their usage.
The peri-domestic spaces were classified as either: a) built-up spaces attached to the main houses, i.e. veranda extensions; b) built-up spaces not attached to the main houses, e.g. separate kitchens, and c) non-built-up or other peri-domestic spaces commonly used for various outdoor activities. The outdoor built up structures were also characterized based on the roofing and wall types.
Transfluthrin-treated chairs and hessian ribbons
For the dry season experiment, six identical chairs made of wood and metal frame were constructed by a local carpenter while for the wet season experiment 15 chairs were made (Figures 2a and 2b). The chairs were fitted underneath with four standardized hessian fabric mats: two measuring 42cm × 43cm and fitted underneath the right and left sides of the chair and other two measuring 20cm × 33cm, which were fitted underneath the middle part of the chair (Figure 2c). These mats were made by a local seamstress at the Ifakara Health Institute fabrication facility (the MozzieHouse). The hessian mats had been treated in emulsified solutions containing 2% transfluthrin (Bayer AG, Germany), prepared as previously described [31, 33].
Similarly, the hessian ribbons were prepared as previously described by Mmbando et al., . Each ribbon had 15cm width and 10m length, and were also made locally at the MozzieHouse. More detailed descriptions of the hessian ribbons have previously been published by Ogoma et al  and Mmbando et al . The ribbons were also treated in a 2% emulsified solution of transfluthrin as previously described .
Assessing protective efficacies of transfluthrin-treated chairs and ribbons
This assessment was conducted in two seasons: dry and wet seasons, between September to October 2019 and between January to February 2020 as dry and wet seasons respectively. Following the characterization of the peri-domestic spaces as described above, eight households with outdoor kitchens were selected for a small-scale assessment of protective efficacies of the two candidate interventions in the dry season. The houses were paired and assigned as follows: a) a control arm, where neither transfluthrin-treated chairs nor transfluthrin-treated ribbons were used, b) a treatment arm where one transfluthrin-treated chair was used, c) a second treatment arm where two transfluthrin-treated chairs were used, and d) a third treatment arm where transfluthrin-treated hessian ribbons were used around the outdoor kitchens. In each arm, two houses were enrolled.
One consenting adult male volunteer was assigned to each household, to sit inside the exposure-free miniaturized double nets trap (DN-Mini)  from 1900hrs to 2300hrs. The volunteer spent 45 min each hour retrieving all host-seeking mosquitoes caught in the DN-Mini while attempting to bite him. For the households with transfluthrin-emanating chairs, the DN-Mini was installed 0.5m from the chairs (Figures 2d and 2e). For households with transfluthrin-treated hessian ribbons, the ribbon was fitted 1.3m above ground (Figures 2f) onto the outdoor kitchens. CDC light traps  were suspended inside these makeshift kitchens to collect host-seeking mosquitoes nightly, while DN-Mini traps were set beside the kitchens to assess biting risk in the general peri-domestic space (Figure 2f).
Each treatment arm was initially located in two houses per experimental night, but was rotated between the houses using a 4 × 4 Latin square design over 32 experimental nights, so that each treatment or control arm was tested at each of the eight houses four times. The primary outcome was number of mosquitoes of different species caught in the DN-Mini or the CDC light traps per house per night. All treated materials were carefully shifted between the houses to avoid any contamination during the rotations. As the experiments were conducted outdoors with enough airflow, there was no need to break for wash out. Instead, a control set up was used to monitor mortality of mosquitoes as described in the sub-section below. Each morning the collected mosquitoes were sorted and identified using morphological keys . In the wet season, 20 households were enrolled making five households in each arm for other 32 nights. The same procedure was adopted as described in dry season.
Assessing mortality effects of the transfluthrin-treated chairs on mosquitoes
This assay was done using three different groups of mosquitoes, as follows: a) field-collected An. arabiensis and An. funestus of unknown age, which are known to be pyrethroid resistant in this setting [41-43], b) laboratory-reared An. arabiensis from a pyrethroid-susceptible colony of local origin, and c) laboratory-reared Aedes aegypti from a pyrethroid-susceptible colony of local origin .
The wild-caught An. arabiensis females were collected using a separate set of eight DN-Mini traps  set outdoors at households without any transfluthrin treatments. Eight consenting adult male volunteers were involved in these collections each night from 1900hrs to 0100hrs. As population densities of An. funestus in this study area were very low, CDC light traps were used to collect adult females of this species from another village (Tulizamoyo (-8.3669, 36.7336)) approximately 30km away.
Each morning captured mosquitoes were sorted and An. arabiensis and An. funestus females separated in two cages containing 100 mosquitoes per species (four cages in total). Since the Anopheles gambiae sensu lato in this area are known to consist exclusively of An. arabiensis , no molecular identification was required. Similarly, since indoor collections of An. funestus sensu lato have consistently been found to be >90% An. funestus sensus stricto , it was assumed that these were the dominant species in the collections. The separated mosquitoes were kept at a field insectary (average temperature: 26.75 ± 0.09°C; relative humidity: 73.26 ± 0.46%) for acclimatization for at least 20 hours before testing the next evening.
For the tests, two chairs were placed within open verandas of two separate houses. One of the chairs was fitted underneath with transfluthrin-treated hessian mats, while the other was fitted with an untreated hessian mat (control). The caged mosquitoes were placed underneath each chair overnight (1900hrs to 0700hrs). A simple water moat was used to prevent ants from eating the mosquitoes. Each morning, the cages were returned to the field insectary and monitored for further 12 hours, totaling 24 hours of observation since start of exposure. This procedure was repeated 10 times (totaling 1140 mosquitoes) for field-collected An. arabiensis and five times (totaling 490 mosquitoes) for field-collected An. funestus tested in control and treated arms.
Similar tests were conducted using cages containing 100 laboratory-reared An. arabiensis or 100 laboratory-reared Ae. aegypti. Since Ae. aegypti mosquitoes are active during the day, they were exposed from 0800hrs to 1900hrs each day, as opposed to the Anopheles mosquitoes, which were exposed at nightPercentage mortality of mosquitoes was calculated for each species separately as a proportion of total exposed.
Testing susceptibility of local malaria vector populations to common public health pesticides
In order to determine phenotypic resistance status of local mosquito populations to common pesticides, standard discriminatory tests were performed using standard WHO susceptibility bioassays . Since transfluthrin is a pyrethroid, the tests also provided indication of how the transfluthrin-based interventions evaluated here (transfluthrin-treated chairs and transfluthrin-treated hessian ribbons) evaluated here would perform against wild pyrethroid-resistant mosquito populations. The susceptibility tests were done for: a) 0.1% bendiocarb, a carbamate; b) 4.0% dichlorodiphenyltrichloroethane (DDT), an organochloride; c) 0.25% pirimiphos methyl, an organophosphate, iv) 0.75% permethrin, a type I pyrethroid; and v) 0.05% deltamethrin, a type II pyrethroid.
Female An. arabiensis mosquitoes were collected from nearby rice fields as larvae, and reared to emergence at Ifakara Health Institute vector biology laboratory, the VectorSphere. The susceptibility tests were done using 3-day old adult females, using at least 100 mosquitoes per test (25 per replicate), with at least 4 replicates as described in the recent WHO guidelines .
The survey data was summarized in ODK analysis module  to generate descriptive statistics of peri-domestic spaces and their usage. Data on efficacy of the transfluthrin-treated chairs and ribbons was analyzed using R open-source statistical software , primarily using generalized linear mixed-effects models , each time modeling the numbers of mosquitoes of a given species caught as a function of the treatments (fixed factors), and fitting the data onto Poisson distributions. Volunteer ID, day and house ID were included as random factors in the models.