Article Figures and data Abstract Introduction Materials and methods Results Discussion Appendix 1 Data availability References Decision letter Author response Article and author information Metrics Abstract Background: Monitoring malaria transmission is a critical component of efforts to achieve targets for elimination and eradication. Two commonly monitored metrics of transmission intensity are parasite prevalence (PR) and the entomological inoculation rate (EIR). Comparing the spatial and temporal variations in the PR and EIR of a given geographical region and modelling the relationship between the two metrics may provide a fuller picture of the malaria epidemiology of the region to inform control activities. Methods: Using geostatistical methods, we compare the spatial and temporal patterns of Plasmodium falciparum EIR and PR using data collected over 38 months in a rural area of Malawi. We then quantify the relationship between EIR and PR by using empirical and mechanistic statistical models. Results: Hotspots identified through the EIR and PR partly overlapped during high transmission seasons but not during low transmission seasons. The estimated relationship showed a 1-month delayed effect of EIR on PR such that at lower levels of EIR, increases in EIR are associated with rapid rise in PR, whereas at higher levels of EIR, changes in EIR do not translate into notable changes in PR. Conclusions: Our study emphasises the need for integrated malaria control strategies that combine vector and human host managements monitored by both entomological and parasitaemia indices. Funding: This work was supported by Stichting Dioraphte grant number 13050800. Introduction National malaria control programmes, working in collaboration with global stakeholders, have achieved extensive intervention coverage over the last two decades, leading to significant reductions in morbidity and mortality due to malaria (Bhatt et al., 2015a). However, malaria is still a leading global health problem. The previous successes and current challenges have motivated ambitious, yet feasible, global and national targets towards malaria elimination. A key component of efforts to achieve these targets is surveillance and monitoring, which is critical for continued assessment of intervention effectiveness, identification of areas or groups at the highest risk, and guiding the development and implementation of new intervention strategies (World Health Organization, 2015). A wide range of metrics exists for monitoring malaria parasite transmission. The strengths and limitations of each metric are related, in part, to the step of the parasite transmission cycle it measures (Tusting et al., 2014). These strengths and weaknesses, including the sensitivity of each metric, vary across epidemiological settings and as parasite transmission declines within a given setting (The malERA Refresh Consultative Panel on Characterising the Reservoir and Measuring Transmission, 2017). Two of the most commonly monitored metrics are the prevalence of Plasmodium parasites and the entomological inoculation rate (EIR), especially in moderate to high transmission settings. The prevalence of Plasmodium parasites in the human population at a given time point (i.e. the parasite rate; PR) approximates the reservoir of hosts potentially available to transmit the parasite from humans to mosquitoes. While only the gametocyte stage of the parasite contributes to transmission, it remains relatively expensive to detect this stage of the parasite. Conversely, rapid diagnostic tests (RDTs) primarily detect asexual-stage antigens, yet they are inexpensive and easily deployed in large-scale community-based surveys (Poti et al., 2020). Still, the limit of detection (50–200 parasites/µl) for RDTs is higher than that of expert microscopy or PCR (Chiodini, 2014), so that RDT-based estimates of PR are biased by excluding low-density infections. Despite these limitations, RDT-based cross-sectional surveys to measure PR capture both symptomatic and asymptomatic infections, which is important because both are likely to contribute to transmission (Bousema et al., 2014; Slater et al., 2019), and changes in PR over time can indicate changes in transmission. EIR provides an estimate of the intensity of parasite transmission from mosquitoes to humans, expressed as the number of infectious bites received per person per unit time. EIR is calculated by multiplying the number of malaria vector bites per person per unit time, also known as the human biting rate (HBR), by the proportion of vectors carrying the infectious sporozoite stage of malaria parasites, referred to as the sporozoite rate (SR) (Onori and Grab, 1980). The accuracy and precision of EIR estimates, therefore, depends on the accuracy and precision with which HBR and SR can be measured (Tusting et al., 2014). Two common methods for measuring HBR are the human landing catch and the Centers for Disease Control and Prevention Light Trap, but inter-individual variation in attractiveness to mosquitoes restricts standardisation across sampling points for both of these methods (Knols et al., 1995; Qiu et al., 2006). Alternative methods for estimating HBR include the Suna trap, which uses a synthetic blend of volatiles found on human skin and carbon dioxide to attract host-seeking Anopheles mosquitoes (Mukabana et al., 2012; Menger et al., 2014; Hiscox et al., 2014). The standardised odour blend allows for reliable comparisons among trapping locations (Mburu et al., 2019). Regardless of the method used to estimate HBR, the precision of SR decreases as the number of mosquitoes collected decreases. Despite these limitations, EIR is a vital metric of malaria parasite transmission because it directly describes human exposure to malaria parasites before post-inoculation factors such as immunity, nutrition, and access to health care (Killeen et al., 2000). Moreover, EIR provides information about the relative contributions of different vector species to transmission, which can impact malaria intervention effectiveness based on interspecies differences in biting behaviours related to time and location, non-human blood-meal hosts, larval ecology, and insecticide resistance profiles (Ferguson et al., 2010). Malaria parasite transmission is heterogeneous in space and time at fine resolution due to several factors, including the availability of larval mosquito habitat, socioeconomics, human behaviour and genetics, and malaria intervention coverage (Carter et al., 2000; Bousema et al., 2012; McCann et al., 2017a). Repeated cross-sectional surveys continuously carried out in communities can reveal this fine-resolution heterogeneity (Roca-Feltrer et al., 2012), providing timely estimates of malaria control progress at the sub-district level and potentially identifying hotspots of malaria parasite transmission for targeted intervention (Kabaghe et al., 2017; Bousema et al., 2016). However, understanding this heterogeneity and identifying hotspots in a way that is meaningful for control programmes remains challenging (Stresman et al., 2019), in part because hotspot location and size can depend on which metric is used (Stresman et al., 2017). Given that PR and EIR are indicative of components of the parasite transmission cycle that are separated by multiple complex steps, each metric provides partial but useful information about the underlying risk of transmission. Therefore, measuring and mapping both metrics can provide a fuller picture of parasite transmission (Cohen et al., 2017). Additionally, modelling the functional relationship between EIR and PR can provide further insights into the underlying malaria epidemiology. Previous studies have demonstrated that this relationship is non-linear, such that small changes in EIR are associated with large changes in PR when EIR is low, but PR saturates rather than changing at a constant rate when EIR is high (Beier et al., 1999; Smith et al., 2005). These previous studies were meta-analyses using paired estimates of EIR and PR, with one estimate of each outcome per site, from sites representing a wide range of EIR and PR in Africa. Their findings had a number of important implications, which included providing estimated ranges for the change in PR that may be expected for a given change in EIR. However, these estimates implicitly assumed that the relationship is constant across space on a continental scale, such that differences in EIR and PR between sites would be indicative of differences over time within a site. Yet no previous study has explicitly examined this relationship over time within a single geographical region. In the current study, we use a series of repeated cross-sectional surveys conducted over 38 months in one region of southern Malawi to map the fine-scale spatiotemporal dynamics of P. falciparum entomological inoculation rate (PfEIR) and P. falciparum parasite prevalence (PfPR). The joint monitoring of these two outcomes in space and time allows us to identify and compare the spatial heterogeneities and temporal patterns of PfEIR and PfPR in a region with moderately intense, seasonally variable malaria parasite transmission. We then investigate the PfEIR-PfPR relationship based on changes in these outcomes observed at both annual and subannual scales within our study site using several statistical models, which can be distinguished as follows: mechanistic models that are based on different epidemiological assumptions and empirical models where the data inform the PfEIR-PfPR relationship. These approaches allows us to address the following questions. (1) How do spatiotemporal patterns of EIR and PR compare? (2) Do EIR and PR lead to the identification of the same malaria hotspots? (3) As EIR changes over time, how do those changes in EIR affect PR? (4) Does EIR have a lagged effect on PR? (5) Does the EIR-PR relationship vary between women of reproductive age and children between 6 and 59 months of age? Materials and methods Study site Request a detailed protocol This study was part of the Majete Malaria Project (MMP), an integrated malaria control project in Chikwawa District, Malawi. The catchment area of MMP consisted of three distinct geographical regions, referred to as Focal Areas A, B, and C (Figure 1), with a total population of about 25,000 people living in 6600 households in 65 villages. Figure 1 Download asset Open asset Map of study site. Map of Malawi (insert) highlighting the Majete Wildlife Reserve and the borders of 19 community-based organisations (CBOs) surrounding the Majete perimeter. Three focal areas (red patches), labelled as A, B, and C, show the households (black points) selected for the parasitaemia and entomological surveys by the Majete Malaria Project (MMP). The base map was obtained from Google Maps. Chikwawa experiences highly variable rainfall during its single rainy season, which spans November/December to April/May. Temperatures are generally high, with daily maximum temperatures in December averaging 37.6°C, and in July averaging 27.6°C (Joshua et al., 2016). A wide range of both permanent and temporary water bodies create suitable larval habitats in the region for Anopheles funestus s.s., Anopheles arabiensis, and Anopheles gambiae s.s., including dams, swamps, ponds, borehole runoffs and drainage channels (Gowelo et al., 2020). Malaria control in the district is implemented through the Chikwawa District Health Office. During the study period, interventions applied throughout the study area included the continuous provision of insecticide-treated nets (ITNs) to pregnant women and children under five years old, mass distribution campaigns of ITNs targeting universal coverage, intermittent preventative therapy for pregnant women, and malaria case diagnosis and treatment with artemisinin-based combination therapy. The only mass distribution of ITNs in the district during the study period occurred in April 2016. As part of the MMP, a randomised trial was conducted to assess the effectiveness of additional, community-implemented malaria interventions between May 2016 and May 2018 (McCann et al., 2017b). The trial interventions were implemented at the village level, with villages assigned to one of four groups: (a) no additional interventions; (b) larval source management; (c) house improvement; and (d) both larval source management and house improvement (McCann et al., 2017b; van den Berg et al., 2018). Data To quantify PfPR and PfEIR over the course of the study, a rolling malaria indicator survey (rMIS) (Roca-Feltrer et al., 2012) was conducted in conjunction with mosquito sampling, forming a series of repeated cross sectional surveys. Sampling was carried out over 17 rounds, with each round spanning a period of 2 or 3 months. In the first two rounds of data collection (baseline, from April through August 2015), an inhibitory geostatistical sampling design (IGSD) was used to select 300 and 270 households, respectively, for the rMIS from an enumeration database of all households in the catchment area (Chipeta et al., 2017). The IGSD helped to ensure that randomly sampled households are relatively uniformly spaced over the study region by requiring each pair of sampled households to be separated by a distance of at least 0.1 km, which increases the efficiency of hotspot detection (Kabaghe et al., 2017). In the three subsequent rounds of data collection during the baseline, an adaptive geostatistical sampling design (AGSD) was used to select 270 households per round (Chipeta et al., 2016). With AGSD, new households for the current round of rMIS were chosen from regions with high standard errors of estimated prevalence, based on data from all previous rounds. In the baseline period, previously sampled households were not eligible for sampling in subsequent rounds. For the trial period (starting May 2016), IGSD was again used to select households from the enumeration database of all households. All households were eligible for selection in each round of the trial period regardless of whether they were selected in a previous round. At each round of rMIS data collection in the baseline and trial phases, respectively, 75% and 72% of the households chosen at each round of the rMIS were then randomly selected for mosquito sampling. In each sampled household, children under five (0.5–5 y/o) and women of reproductive age (15–49 y/o) were tested for P. falciparum using an RDT (SD BIOLINE Malaria Ag P.f. HRP-II, Standard Diagnostics, Yongin-si, Republic of Korea). Mosquitoes were sampled from 5pm to 7am using Suna traps (Biogents AG, Regensburg, Germany) with MB5 blend plus CO2 to mimic human odour (Hiscox et al., 2014; Mburu et al., 2019). For a selected household in a surveillance round, the trap was set for one night indoors and one night outdoors, with the order of indoor/outdoor determined randomly. For households where the residents were sleeping in more than one building, a trap was set at each building. Trapped female anophelines were preserved using a desiccant and identified using standard morphological and molecular techniques (Gillies and Coetzee, 1987; Koekemoer et al., 2002; Scott et al., 1993). Female anophelines were further tested for the presence of P. falciparum in their head and thorax, after removing the abdomen, using quantitative polymerase chain reaction (qPCR ) (Bass et al., 2008; Perandin et al., 2004). Specimens with a Ct value below 37.0 were considered positive for P. falciparum. Environmental and climatic factors Request a detailed protocol Environmental and climatic factors affect the abundance and suitability of water bodies that support the development of immature mosquitoes (Madder et al., 1983; Loetti et al., 2011), the duration of mosquito development (Ciota et al., 2014; Loetti et al., 2011; Craig et al., 1999), mosquito host-seeking and biting behaviour, and the development rate of malaria parasites in mosquitoes (Rumisha et al., 2014; Amek et al., 2011). Using hourly measurements of temperature and relative humidity (RH) from a weather station in each focal area, we computed the average temperature and RH for different ranges of days before the day of data collection (Appendix 1 – Procedure for building the HBR, PfSR and PfPR models). Spectral indices, namely normalised difference vegetation index (NDVI) and enhanced vegetation index (EVI), were computed using remotely sensed multi-spectral imagery from the Landsat 8 satellite. These data are freely available from the United States Geological Survey (USGS) Earth Explorer (earthexplorer.usgs.gov) as raster files at a spatial resolution of 30 × 30 m for every 16 days. For our analysis, we averaged each spectral index over 5 years, from April 2013 to April 2018, while omitting scenes that were dominated by cloud artefacts. We extracted raster data of surface elevation from the global digital elevation model (DEM) generated using measurements from the Advanced Space-borne Thermal Emission and Reflection Radiometer (ASTER) (Tachikawa et al., 2011). These data are also freely available for download from the USGS Earth Explorer. Using a flow accumulation map derived from the DEM, a river network map was generated and used to calculate and store as raster images the distance to small rivers and large rivers (henceforth, DSR and DLR, respectively). Geostatistical analysis The number of mosquitoes trapped by Suna traps can be used to estimate HBR, as these traps primarily target host-seeking mosquitoes. Hence, we first estimated HBR and the P. falciparum sporozoite rate (PfSR). We then estimated PfEIR as the product of these two quantities. We carried out separate analyses for A. arabiensis and A. funestus s.s., using explanatory variables and random effects structures that we found to be suitable for each species. Details of the variable selection process and the final sets of explanatory variables for each of the models later described in this section are given in Appendix 1 – Procedure for building the HBR, PfSR, and PfPR models. The correlation structures adopted for the geostatistical models were informed by the variogram-based algorithm described in Giorgi et al., 2018. The geostatistical models for the HBR and PfPR data described below were fitted using PrevMap (Giorgi and Diggle, 2017), freely available from the Comprehensive R Archive Network (CRAN, www.r-project.org). The PfSR models were fitted using the glmm package, also available on CRAN. Human biting rate Request a detailed protocol Let Y(xi,ti),i=1,…,M, where M=2432 is the total number of houses, denote counts of mosquitoes trapped at location xi in month ti∈{1,…,38}, where ti=1 denotes April 2015. We modelled the Y(xi,ti) using Poisson mixed models expressed by the following linear predictor (1) log{HBR(xi,ti)}=d(xi,ti)⊤β+f(ti;α)+S(xi)+Zi, where: d(xi,ti) is a vector of spatiotemporal explanatory variables with associated regression coefficients β; the f(ti;α) is a linear combination of several functions of time, including sines, cosines and splines, with an associated vector of regression parameters α, accounting for trends and seasonal patterns; the Zi are independent and identically distributed Gaussian random variables with variance τ2; S(x) is a zero-mean stationary and isotropic Gaussian process with variance σ2 and exponential correlation function ρ(u)=exp(-u/ϕ), where ϕ regulates the pace at which the spatial correlation decays for increasing distance u between any two locations. We allow the explanatory variables d(xi,ti) and f(ti;α) to differ between different mosquito species since different species may respond differently to environmental changes. We point out that the stationarity of the process S(x) implies that all of its properties, including the variance (σ2) and scale of the spatial correlation (ϕ), are constant over space. The estimation of the model parameters is then carried out using Monte Carlo Maximum Likelihood (Christensen, 2004). Plasmodium falciparum sporozoite rate Request a detailed protocol Let Y*(xi,ti) be the number of mosquitoes that tested positive for the presence of P. falciparum sporozoites. We assumed that the Y*(xi,ti) follow a Binomial mixed model with number of trials N*(xi,ti), that is the total number of successfully tested mosquitoes, and probability of testing positive PfSR(xi,ti). We model the latter as a logit-linear regression given by (2) log{PfSR(xi,ti)1-PfSR(xi,ti)}=d(xi,ti)⊤β*+f*(ti;α*)+Zi*, where each term in (2) has an analogous interpretation to those of (1). A spatial process S(x) was not included in the sporozoite rate model because we found no evidence of residual spatial correlation in the sporozoite rate data (Appendix 1—figure 1). Estimating the Plasmodium falciparum entomological inoculation rate Request a detailed protocol Let PfEIRf(x,t) and PfEIRa(x,t) denote the PfEIR for A. funestus s.s. and A. arabiensis at a given location x and month t. We estimated each of these two as PfEIRf(x,t)=HBRf(x,t)PfSRf(x,t)l(t)PfEIRa(x,t)=HBRa(x,t)PfSRa(x,t)l(t), where l(t) is the number of days in month t. Finally, we estimated the overall PfEIR as (3) PfEIR(x,t)=PfEIRf(x,t)+PfEIRa(x,t). We then mapped PfEIR as in (3) over a 30 × 30 m regular grid covering the whole of the study area for each month across 38 months. To map PfEIR for each month, we first simulate at each prediction location (i.e. the centroid of each grid cell) 10,000 samples from the conditional distribution of the random effects (corresponding to S(x)+Z in the case of the PfHBR and Z in the case of PfSR) given the data. We then transform these to obtain 10,000 predicted surfaces for PfHBR and PfSR, and by applying (3) we obtain 10,000 predictive samples for PfEIR. The predicted PfEIR at each prediction location is taken to be the median of the 10,000 samples at that location. The associated 95% predictive interval is the 2.5th to 97.5th percentile of the 10,000 predictive samples. In this procedure, all the parameters corresponding to the regression coefficients, the scale and variance of the spatial process, and variance of Gaussian noise are fixed at their MCML estimates. Plasmodium falciparum prevalence Request a detailed protocol We mapped PfPR in women and in children by fitting a geostatistical model to each group. More specifically, let I(xi,ti) denote the number of RDT positives out of Nit sampled individuals at location xi in month ti. We then assumed that the I(xi,ti) follow a Binomial mixed model with probability of a positive RDT result p(xi,ti), such that (4) log{p(xi,ti)1-p(xi,ti)}=d(xi,ti)⊤φ+g(ti;ϱ)+T(xi)+Ui, where T(xi) is a stationary and isotropic Gaussian process with exponential correlation function and Ui are Gaussian noise, g(ti,ϱ) is a linear combination of splines, and sine and cosine functions of time accounting for trends and seasonality, and φ and ϱ are vectors of regression parameters to be estimated. Hotspot detection using PfEIR and PfPR Request a detailed protocol We demarcated hotspots for PfEIR and PfPR using an exceedance probability approach. Using the resulting 10,000 predictive samples for PfEIR and PfPR, as described above, we then obtained the exceedance probability for each outcome at each space-time location by computing the proportion of the 10,000 predictive samples that exceeded the respective, predefined thresholds, which were set at 0.1 ib/person/month for EIR, 31% for PfPR in children, and 17% for PfPR in women. Finally, we mapped these exceedance probabilities and demarcated hotspots as areas where these probabilities were at least 0.9. The PfPR thresholds were defined to correspond to the PfEIR threshold based on the best of six functional relationships between PfEIR and PfPR as described in the next section. Modelling the relationship between PfEIR and PfPR Because PfEIR may have a delayed effect on PfPR, possibly due to the time taken for P. falciparum to develop in the human host, we considered that current PfPR may depend on PfEIR l months prior. In particular, we considered l=0,1,2. We then assumed that the number of RDT positive individuals, I(xi,ti), follow independent Binomial distributions such that (5) PfPR(xi,ti)=h{PfEIR^(xi,ti-l)}, where h(⋅) is a function depending on a vector of parameters θ that governs the relationship between PfPR and PfEIR, and PfEIR^(xi,ti-l) is the estimated PfEIR as in Equation (3). We considered six models, each of which provided a different specification for h(⋅). We now describe the six models for h(⋅). Models 1 to 4 make explicit assumptions on the underlying mechanism of transmission, whereas models 5 and 6 describe the functional relationship between PfEIR and PfPR through regression methods. Model 1: The susceptible-infected-susceptible (SIS) model Request a detailed protocol Let b be the probability that an infectious mosquito bite results in an infection, referred to as the transmission efficiency. Then, infections at (xi,ti-l) are assessed to occur at a rate of b×PfEIR(xi,ti-l). We assumed that each infection cleared independently over a duration 1/r so that the ratio γ=b/r is the time taken to clear infection per infectious bite. We assumed that the relationship between PfEIR and PfPR holds throughout the study region. If PfEIR(x,t-l) is constant, the relationship between PfEIR(x,t-l) and PfPR(x,t) is described by Ross, 1911 (6) ∂PfPR(x,t)∂t=b×PfEIR(x,t-l)(1-PfPR(x,t))-r×PfPR(x,t). We obtained our first model as the non-zero equilibrium solution of (6), given by (7) PfPR(x,t)=γPfEIR(x,t-l)γPfEIR(x,t-l)+1. Model 2: The SIS model with different infection/recovery rates (D.I/R) Request a detailed protocol Model 1 assumes that women and children get infected and recover at the same rate. However, the transmission and recovery rates in children may differ from those in women. We, therefore, modified Model 1 by allowing different values of b and r for each category of people. Let ξ1,it and ξ2,it respectively be the proportion of children and women sampled at (xi,ti) and γk=bk/rk, where k=1 denotes children and k=2 denotes women. The resulting Model 2 is (8) PfPR(x,t)=∑k=12ξk,itγkPfEIR(x,t-l)γkPfEIR(x,t-l)+1. Model 3: The SIS model with superinfection (S.I.) Request a detailed protocol If individuals are super-infected with P. falciparum, then the rate at which infections clear depends on the infection rate, with clearance being faster when infection rate is low, and slower when infection rate is high. To capture this feature, we modelled infection clearance rate as g(ϑ,r)=ϑ/(eϑ/r-1), where ϑ=b×PfEIR(Smith et al., 2005; Walton, 1947; Dietz et al., 1974; Aron and May, 1982). The resulting model for PfPR(x,t) is (9) PfPR(x,t)=1-exp{-γPfEIR(x,t-l)} Model 4: The SIS model with S.I and D.I/R Request a detailed protocol Combining the assumptions of heterogeneous infection/recovery rates, as in Model 2 and superinfection, as in Model 3, we obtain Model 4, (10) PfPR(x,t)=∑k=12ξk,it(1-exp{-γkPfEIR(x,t-l)}). Model 5: The Beier model Request a detailed protocol Beier et al., 1999 assumed that the log of PfEIR is linearly related to PfPR, and fitted the regression model (11) PfPR(x,t)=a+blog(PfEIR(x,t-l)), the so called ‘log-linear model’. Model 6: The logit-linear model Request a detailed protocol The Beier model has the limitation that PfPR approaches -∞ as PfEIR goes to 0 and approaches ∞ as PfEIR goes to ∞. To constrain PfPR to lie between 0 and 1, we applied the logit-link function to PfPR to give Model 6, (12) log(PfPR(x,t)1-PfPR(x,t))=a+blog(PfEIR(x,t-l)). Parameter estimation of the PfEIR-PfPR relationship models Request a detailed protocol We estimated the parameters of each of the six models by maximising the log-likelihood function (13) ∑ti∑xiI(xi,ti)log(PfPR(xi,ti))+(Nit-I(xi,ti))log(1-PfPR(xi,ti)). To fit each model, we first obtained 10,000 bootstrapped data sets of predicted PfEIR as in (3) at the set of all space-time locations sampled for the rMIS. We did this for two reasons: to obtain PfEIR data at locations (xi,ti) that were sampled for rMIS but not for the entomological surveillance; and to account for the uncertainty in PfEIR. The predicted PfEIR values were then paired with respective empirical PfPR values at (xi,ti). By fitting each model to each of the 10,000 datasets, we then obtained 10,000 bootstrapped samples {θ^1,…,θ^10000} for the vector of parameter estimates θ^ of each the six candidate models. We then summarised these samples by their mean and central 95% probability interval. We repeated this process for l=0,1,2. We compared the fit of the six models based on their predictive ability as measured by the bias and root-mean-square error when each model is used to predict prevalence at all the observed space-time locations. Results rMIS and mosquito sampling From April 2015 to May 2018, a total of 6870 traps (3439 indoors; 3431 outdoors) were placed at 2432 houses over 17 rounds of sampling (Figure 2), resulting in the collection of 657 female Anopheles mosquitoes (Table 1). Following PCR of the 478 A. gambiae s.l. collected, 92% were identified as A. arabiensis, 2% as A. gambiae s.s., 1% as A. quadriannulatus, and 5% could not be identified further. From the 179 A. funestus s.l. collected, 95% were identified as A. funestus s.s. by PCR, while the remaining 5% could not be identified further. The observed vector composition is therefore 71%, 27%, and 2% for A. arabiensis, A. funestus s.s., and A. gambiae s.s., respectively. Figure 2
Abstract. The strongly anthropophilic behaviour of Anopheles gambiae Giles sensu stricto (Diptera: Culicidae), the most important malaria vector in Africa, has been demonstrated by field and laboratory studies. Other members of the An. gambiae complex express varied degrees of anthropophily. Anopheles quadriannulatus (Theobald) species A and B are more zoophilic members of the complex and hence are considered to be of no medical importance. Olfactometer experiments with An. quadriannulatus species A have demonstrated attraction to both human and cow odour. To extend these olfactometer observations a choice experiment was conducted in an outdoor cage with a human and a calf as baits, using laboratory‐reared mosquitoes. Anopheles gambiae s.s. (from Liberia) and two strains of An. quadriannulatus species A (SKUQUA from South Africa, SANGQUA from Zimbabwe), marked with different coloured fluorescent powders for identification purposes, were released simultaneously and given an equal opportunity to feed on either host. The experiment was repeated six times. Bloodmeals were identified using the precipitin technique. Anopheles gambiae s.s. showed highly anthropophagic behaviour, taking 88% of bloodmeals from the human host. In contrast, both strains of An. quadriannulatus fed with equal frequency on the human or the calf; the response to either host was not significantly different. These results confirm the olfactometer findings and demonstrate anthropophagic behaviour not previously recorded in this species. This finding has implications for prospective manipulation of host preference for genetic control purposes.
~~ ~ Summarythe field in Zimbabwe, the behaviour of tsetse (Diptern: Glossinidae) released 10 m downwind of an odour source was studied with a video camern and with electric nets. Video studies showed that in the absence of odour, 46% of the released Glos.rina pallidipes Austen turn downwind and 32 % turned upwind. Tsetse left the box at a constant rate. When an artificial odour mixture containing carbon dioxide, acetone, octenol and phenols was used significantly fewer tsetse, 35 %, turned downwind and more tsetse, 37 %, turned upwind. In the presence of odour, tsetse left the TRB later and not at a constant rate. When the TRB was placed in a complete ring of electric nets, the release of natural ox odour changed the distribution of tsetse to the downwind electric nets compared to the no odour treatment. Anificial odour, with and without carbon dioxide, had no effect on the distribution of tsetse over the electric nets. The difference between the video study and die electric net study is attributed to the 50% efficiency of electric nets. We infer from the results that 10% of the tsetse departing from the TRB reacts to the presence of odour immediately.
As the ecology of mosquito larvae can be complex there is need to develop a rational framework for undertaking larval ecological studies. Local environmental characteristics, such as altitude, climate and land use, can significantly impact on phenology and population dynamics of mosquito larvae, and indirectly affect the dynamics of mosquito-borne diseases. The aim of this study was to assess the feasibility of implementing an integrated approach to larval source management under the distinct ecological settings. The study was conducted in two highland villages and one village, at a lower altitude, in the Lake Victoria basin, where malaria is endemic and transmitted by the same Anopheles mosquito species. In each village the stability of mosquito larval habitats was classified as either temporary or permanent. The productivity of these habitat types was quantified by carrying out weekly larval sampling using a standard dipping method for a period of two years. During sampling the physical characteristic of the larval habitat, including the vegetation cover were noted. Ambient temperature, rainfall and relative humidity were recorded on a 21 × Micro-datalogger in each study site. Anopheles gambiae sensu lato larvae were found in all study sites. Anopheles arabiensis was more abundant (93%) in Nyalenda (Lake Victoria basin) and Fort Ternan (highland area; 71%). In Lunyerere (highland area), An. gambiae sensu stricto comprised 93% of the total An. gambiae s.l. larvae. Larvae of An. gambiae s.l. mosquitoes were present in both temporary and permanent habitats with monthly variations dependent on rainfall intensity and location. Anopheles larvae were more likely to be found in man-made as opposed to natural habitats. Grassy habitats were preferred and were, therefore, more productive of Anopheles larvae compared to other habitat types. Weekly rainfall intensity led to an increase or decrease in mosquito larval abundance depending on the location. The majority of mosquito breeding habitats were man made in all sites. Both temporary and permanent habitats were suitable for An. gambiae breeding. In Fort Ternan temporary sites were favoured for mosquito breeding above permanent sites. Significant differences in larval abundance were found depending on weekly rainfall intensity. Larval source management programmes should target permanent and temporary habitats equally and work closely with land and home owners as a majority of the breeding habitats are man made.
'%3e%3cpath fill='%23012169' d='M0 0v30h60V0z'/%3e%3cpath stroke='%23fff' stroke-width='6' d='m0 0 60 30m0-30L0 30'/%3e%3cpath stroke='%23C8102E' stroke-width='4' d='m0 0 60 30m0-30L0 30' clip-path='url(%23b)'/%3e%3cpath stroke='%23fff' stroke-width='10' d='M30 0v30M0 15h60'/%3e%3cpath stroke='%23C8102E' stroke-width='6' d='M30 0v30M0 15h60'/%3e%3c/g%3e%3c/svg%3e)
'/%3e%3c/defs%3e%3cpath fill='%23fff' d='M-120-80h240V80h-240z'/%3e%3cpath d='M-120-80h240v48h-240z'/%3e%3cpath fill='%23060' d='M-120 32h240v48h-240z'/%3e%3cg id='b'%3e%3cuse xlink:href='%23a' stroke='%23000'/%3e%3cuse xlink:href='%23a' fill='%23fff'/%3e%3c/g%3e%3cuse xlink:href='%23b' transform='scale(-1 1)'/%3e%3cpath fill='%23b00' d='M-120-24v48h101c3 8 13 24 19 24s16-16 19-24h101v-48H19C16-32 6-48 0-48s-16 16-19 24z'/%3e%3cpath id='c' d='M19 24c3-8 5-16 5-24s-2-16-5-24c-3 8-5 16-5 24s2 16 5 24'/%3e%3cuse xlink:href='%23c' transform='scale(-1 1)'/%3e%3cg fill='%23fff'%3e%3cellipse rx='4' ry='6'/%3e%3cpath id='d' d='M1 5.85s4 8 4 21-4 21-4 21z'/%3e%3cuse xlink:href='%23d' transform='scale(-1)'/%3e%3cuse xlink:href='%23d' transform='scale(-1 1)'/%3e%3cuse xlink:href='%23d' transform='scale(1 -1)'/%3e%3c/g%3e%3c/svg%3e)
