Mosquitoes are important vectors of viruses and parasites of human, veterinary and wildlife health concern [1]. Major mosquito-borne infectious diseases of humans include malaria, dengue, filariasis, chikungunya fever, Rift Valley fever and the arboviral encephalitides, including Japanese encephalitis and West Nile neuroinvasive viral disease. Infections can cause high morbidity and mortality and are of great public health concern in endemic areas. Globally, malaria parallels HIV/AIDS and is slightly behind diarrhoeal disease in terms of disability-adjusted life years lost [2]. The current estimated global burden of dengue is higher (390 million infections per year) than previously appreciated [3] and will likely increase as the range of Aedes expands [4, 5] and as humans are increasingly concentrated in high risk urban areas [6]. Lymphatic filariasis often leads to permanent disability and its global burden ranks high amongst other neglected tropical diseases [2]. West Nile virus is the most widespread mosquito-borne neuropathogen and has dramatically expanded its geographic range across the Western Hemisphere after its introduction into North America in 1999 [7].

The vectorial capacity [8] is the ideal theoretical metric for describing the potential intensity of mosquito-borne pathogen transmission and responses to control. There is a substantial body of research, including studies utilizing mosquito mark-release-recapture (MMRR) as a method, which have measured the intensity of transmission and other mosquito bionomic parameters that affect pathogen transmission [9, 10]. Vectorial capacity contains the entomological components of a pathogen’s basic reproductive number (R₀) and offers clarification about the relative importance of different aspects of mosquito behavior and ecology for transmission [11], i.e., mosquito population density relative to the pathogen’s host population density, daily mosquito survival, the blood index describing the proportion of bites taken from a pathogen’s host, the blood feeding rate describing the duration of the gonotrophic cycle or the interval between two successive blood meals, and the duration of the extrinsic incubation period. Vectorial capacity posits that transmission will be reduced linearly by reductions in mosquito density, quadratically by slowing the rate at which mosquitoes bite amplifying hosts, and approximately cubically with increases in mosquito mortality [12]. Recent theoretical developments refined the connection between mosquito ecology and transmission further [13, 14] by, for example, identifying that the scale of mosquito movement determines the spatial scale of transmission and thereby the number of hosts that can potentially become infected by a single mosquito [15]. Studies of transmission have focused on human biting rates and the entomological inoculation rates, depending on the pathogen [16], but there are persistent questions about the precision and accuracy of such estimates [10]. MMRR provide a method for cross-validation and unique insight into how mosquito movement patterns result in their dispersal through the environment [17]. Quantitative estimates of these parameters have been used to inform models of malaria [18] and of other mosquito-transmitted pathogens [9, 19].

Although models have identified those aspects of mosquito ecology that are most important for transmission and that should be the most promising targets for control, field measurements of those parameters are required to inform model-based evaluations of different control measures in different ecological contexts [20, 21]. An aspirational goal is to develop an evidence-base that could establish average bionomic parameters for various vector species and to explain and predict variation in mosquito bionomic parameters in various ecological settings across a species’ geographical range. These parameters would be useful for implementing control as well as for modeling the pathogens. An important common feature in the measurement of entomological parameters describing transmission is that historically they involved marking and releasing mosquitoes and recapturing them after some period of time. These MMRR experiments have been performed hundreds of times over decades on many different species, in a variety of locations and for many different purposes [22].

Despite their abundance and known importance for informing transmission models of mosquito-borne pathogens, data from the numerous MMRR studies have not been compiled into a single repository so that their usefulness in this context can be critically assessed. Global databases are increasingly assembled to support vector-borne disease entomological and epidemiological research [23, 24]. A notable example is the recent effort in malaria research to provide data repositories on disease prevalence [24] and vector occurrence [25–27]. These large data sets have been used as the basis of geo-spatial models of disease endemicity [28–30] and of the distribution of mosquito vectors [31] at national, regional and global scales. They have also provided the scientific community with unique opportunities to explore intrinsic relationships between entomological and epidemiological variables and to advance the theory behind practical, disease prevention questions [32]. Here, we report on an extensive global database of MMRR studies conducted over the last century with the ultimate aim to help improve mosquito-borne pathogen transmission models. This paper is the first on this topic and as such is limited to presenting a description of the database and preliminary analyses that give a general overview of its contents, strengths and limitations to inform modeling approaches.

MMRR experiments are a powerful tool for the study of different aspects of mosquito behavior and ecology, many of which are essential for understanding mosquito-borne pathogen transmission. The database assembled describes 774 MMRR experiments on Aedes, Anopheles and Culex species documenting the release of 3,341,116 adult female mosquitoes between 1913 and 2010, of which 90,524 (2.7%) were recaptured. Despite the considerable resources invested in creating this wealth of data, we found that the methods used were highly heterogeneous and the representation of species and ecological contexts was ad hoc; i.e. not intended to be adapted for purposes other than planned by the investigations. Further careful study of this database will be essential to identify detailed patterns that may be taxa specific, critique methodology, and guide design of future studies so that they can be better aggregated to more reliably inform public health. Preliminary biologically and epidemiologically relevant insights were, nevertheless, gained from a crude description of the assembled MMRR data and summary analyses.

Recapture success varied significantly across studies largely due to biological differences amongst mosquito genera/species. Typically, recapture success was significantly lower for highly vagile outdoor Culex species, particularly when compared with the generally less dispersive indoor Aedes species, reflecting important inherent behavioural dissimilarities and the propinquity of species-specific life resources such as blood meal hosts, resting sites and oviposition sites. Poor recapture success also may reflect a decrease in collection effort as a function of distance from the point of release and losses by emigration from the study area. Within-genus, and even within-species, variation in recapture success was also informative, particularly for Aedes species. There were two important considerations to account for these differences.

The first concerns the ecological context in which adult mosquitoes were released and recaptured. Local climate variables such as temperature affect daily mosquito survival [33, 39], whereas excessive wind may limit dispersive flight activity [35]. Additionally, local weather conditions, land use and landscape features, such as natural or artificial barriers, can affect dispersal [40, 41]. It is important, therefore, to consider site-specific MMRR data [42]. By using latitude as a crude proxy of climatic and ecological conditions of the study areas, we determined that recapture success varied significantly according to geographic location even when considering only studies of Ae. aegypti (Figure 9). For this species, MMRR experiments had significantly greater recapture success in study sites closer to the equator, suggesting a strong influence of climate, socioeconomic factors, mosquito population structure and other environmental factors. Public health applications of MMRR studies would benefit from designs that attempt to understand how bionomic patterns vary by ecological context.

Second, variation in experimental outcomes can be explained by the diversity of methodologies used in marking and recapturing mosquitoes [33], some of which clearly influenced recapture success. Different marking techniques, for example, can affect mosquito survivorship and behavior in different ways. If marking reduced longevity, this would limit the number of mosquitoes that were subsequently recaptured and affect estimates of dispersal and survival [33], even though common marking techniques applied correctly would not be expected to affect mosquito survival [43]. Similarly, laboratory sourced adult mosquitoes may exhibit atypical dispersal, activity rhythms and survival in MMRR experiments because of different behaviour, nutritional status or body size compared to their wild counterparts [33, 44]. The distribution, density and frequency of operation of sampling assets critically influenced recapture success. Typically, the more recapture attempts that were made near the release site, the greater the recapture success. This approach could, however, compromise information on daily or net dispersal.

Although it is fair to conclude that the MMRR studies we examined have contributed to an improved, basic understanding of mosquito biology and behavior that are important for pathogen transmission, it remains unclear how effectively these studies can be used to inform transmission models. Our preliminary results highlight intriguing patterns, but how well those patterns can be generalized across species or different ecological contexts is still uncertain and will be addressed in subsequent work on the MMRR database. Aspirational goals for modeling pathogen transmission include development of functions that could assign baseline parameter values to mosquito species in different ecological contexts. An evidential basis for doing so would require understanding the ecological factors that account for variability in mosquito feeding habits, feeding frequency, survival, dispersal and population density. Our database provides a basis for examining existing methodologies and their potential to address these issues, but a cursory examination of the data suggests that the ad hoc approach will not provide the information required to develop ecological theory for medical entomology. Some thought must be given to how studies should be designed and where studies must be conducted to describe a mechanistic basis for phenomenological descriptions of mosquito behavior. Next steps will focus on critical questions about ideal methodologies and study designs. It is our hope that this assembly of data will motivate such future studies and increased efforts of standardization.

It should be noted that the database has a global scope and its development was based on comprehensive searches and relaxed inclusion criteria for MMRR studies. Despite an extensive effort, the database revealed a general paucity of MMRR data and limited spatial coverage. The MMRR data were available from only a handful of geographically clustered study sites across the world (Figures 1, 2 and 3), due, in part, to the specific objectives of associated research programs. Lack of spatial coverage is of obvious concern, particularly in light of the wide geographic distribution of mosquito vectors and the pathogens they transmit, including malaria, dengue, lymphatic filariasis, and some of the widely spread arboviral encephalitides [3, 31, 45, 46]. For example, our database showed that even though Ae. aegypti, the primary vector of dengue viruses globally, has been investigated more than any other vector mosquito, MMRR data for this species originated from only 22 different study areas. Data for An. gambiae s.l. are even scarcer and are limited to eleven study areas scattered across six sub-Saharan African countries. This is a disappointing finding given that this species complex is responsible for most of the transmission and burden of Plasmodium falciparum malaria in the region [26]. Although our results suggest that many aspects of mosquito ecology that are important for transmission differ by ecological context, once the MMRR studies are broken down by geographical location and species, the amount of data available to examine geographic variation for the most important vectors of human infectious diseases remains limited.

For the past hundred years MMRR experiments have been used mainly to investigate adult mosquito dispersal and survivorship, two indisputably critical components of mosquito-borne pathogen transmission dynamics [42, 47–49]. Adult mosquito dispersal can be estimated with a variety of MMRR approaches [33]. Investigators in most studies released adults at a central point and then recaptured them as they dispersed over time at traps placed in a variety of configurations and at different distances and sampling densities from the release point [50, 51]. However, few studies adjusted recapture success by the reduction in sampling effort as a function of distance from the release point, which generally decreased as the square of the radius. The daily survival rate of adult individuals in the field, on the other hand, is commonly estimated through fitting single release and multiple recapture data to exponential models that assume age-independent survival rates [52], although the validity of this assumption has been questioned [42, 53, 54]. Our database was developed with the spatial resolution of MMRR data derived from dispersal experiments and the temporal resolution provided by survival experiments in mind. This means that despite the overall relative paucity of MMRR experiments, the database contains substantial amounts of raw, disaggregated release-recapture data that merit exploration for estimating dispersal and survival parameters, something that is of particular interest for mathematical and simulation models of mosquito-borne pathogen transmission.

The aggregated MMRR database presented here contains a wealth of information describing mosquito populations, their movements and demography. Consequently, it constitutes the basis for critically analyzing one of the standard methods used by medical entomologists to measure the components of vectorial capacity, or the entomological potential for pathogen transmission. Importantly, the breadth of the assembled data can be used to reveal important phenomena and differences among taxa that are not apparent with any one MMRR study. The data assembled are directly applicable to an improved understanding of pathogen transmission by mosquitoes and the public health impact of vector control. More careful analyses that consider how different methods might affect the reported results remain to be done.

The database is available to anyone as *.csv files (Additional file 3). We hope that the methodologies used and results will be critiqued with the goal of stimulating improvements within the limitations of good ethical practices and biosafety, which ultimately constrain the design of MMRR studies, especially within areas supporting pathogen transmission. To that end, we developed a checklist with recommended guidelines for those who are designing new studies and who may want to augment this database.