Relationships between water quality and mosquito presence and abundance: a systematic review and meta-analysis

Abstract

Mosquito-borne diseases (MBDs) are emerging in response to climate and land use changes. As mosquito (Diptera: Culicidae) habitat selection is often contingent on water availability for egg and larval development, studies have recognized water quality also influences larval habitats. However, underlying species-, genera-, and mosquito level preferences for water quality conditions are varied. This systematic review and meta-analysis aimed to identify, characterize, appraise, and synthesize available global data on the relationships between water quality and mosquito presence and abundance (MPA); with the goal to further our understanding of the geographic expansion of MBD risks. A systematic review was conducted to identify studies investigating the relationships between water quality properties and MPA. Where appropriate, random-effects meta-analyses were conducted to provide pooled estimates for the association between the most reported water quality properties and MPA. The most reported water quality parameters were pH (87%), nitrogen concentrations (56%), turbidity (56%), electrical conductivity (54%), dissolved oxygen (43%), phosphorus concentrations (30%), and alkalinity (10%). Overall, pH (P = 0.05), turbidity (P < 0.0001), electrical conductivity (P = 0.005), dissolved oxygen (P < 0.0001), nitrogen (P < 0.0001), and phosphorus (P < 0.0001) showed significantly positive pooled correlations with MPA, while alkalinity showed a nonsignificant null pooled correlation (P = 0.85). We observed high heterogeneity in most meta-analyses, and climate zonation was shown to influence the pooled estimates. Linkages between MPA and water quality properties will enhance our capacity to predict MBD risks under changing environmental and land use changes.

Graphical Abstract

mosquito, climate, water quality, systematic review, meta-analysis

Introduction

Mosquito-borne diseases (MBDs), such as malaria, dengue, chikungunya, and yellow fever, are the leading cause of approximately 1 million deaths and over 700 million infections attributed to vector-borne diseases annually (World Health Organisation 2020). High MBD prevalence is most common in tropical and arid regions; however, climate change is broadening the geographic distribution of MBD globally (Reiter 2001, Parkinson et al. 2014, Villeneuve et al. 2021). In addition to climate factors, land use is a major contributor to the dispersal of mosquito arboviruses (Norris 2004). Importantly, poorly managed urban and agricultural water management systems can increase mosquito (Diptera: Culicidae) presence and abundance (MPA) by providing prolonged access to stagnant water, which is a primary condition for the oviposition of most female mosquitoes (Clements 1992, Leisnham et al. 2004, Dale et al. 2007, Gardner et al. 2013). Some species such as Aedes albopictus and Aedes aegypti, have adapted to ovipositing and immature maturation (OIM) in surface waters found in urban environments (Ferraguti et al. 2016, Loaiza et al. 2019, Wilke et al. 2019, Perrin et al. 2022); consequentially increasing the risk of MBD transmission in more densely populated areas (Gardner et al. 2014, Madzokere et al. 2020). Thus, the identification of (OIM) sites conducive to mosquito propagation remains pivotal in predicting mosquito expansion and associated disease risks.

While the availability of OIM habitats have a direct link to MBD transmission, the water quality of these habitats is an important determinant of mosquito development and survival (Clements 1992, Gardner et al. 2014, Neff and Dharmarajan 2021). For example, mosquito larvae and pupae are most successful around a neutral pH (Clark et al. 2004, Okogun et al. 2005, Afolabi et al. 2019, Medeiros-Sousa et al. 2020), with lower survivorship outside pH levels of 6–8 (Ukubuiwe et al. 2020). Furthermore, turbidity has been positively associated with MPA (Muturi et al. 2008, Mbuya et al. 2014, Alkhayat et al. 2020, Villarreal-Treviño et al. 2020), although water clarity necessities are more pronounced in artificial OIM sources (Juliano et al. 2004). In addition, ion content has been shown to influence the rate of larvae and pupae growth (Ukubuiwe et al. 2020, Mamai et al. 2021). Likewise, dissolved oxygen has been categorized as vital for MPA in eutrophic OIM environments (Yamada et al. 2020), while larvae survival remains negatively impacted by reduced dissolved oxygen regardless of atmospheric oxygen availability (Reiter 1978, Silberbush et al. 2015). Eutrophication from excess nutrients has not deterred productive mosquito OIM, as nutrient-rich habitats have been shown to favor mosquito development (Leisnham et al. 2004, Darriet and Corbel 2008, Nikookar et al. 2017, Carvajal-Lago et al. 2021). Other water quality determinants, such as metal and sulfate concentrations, have also been associated with increased MPA (Rao et al. 2011, Nikookar et al. 2017, Djamouko-Djonkam et al. 2019, Neff and Dharmarajan 2021).

The magnitude and direction of relationships between water quality properties and MPA vary across studies (e.g., Ranjeeta et al. 2008, Soumendranath et al. 2015, Alam et al. 2018, Aklilu et al. 2020), suggesting MPA is influenced by other context dependent determinants (Leisnham et al. 2005, Mukhtar et al. 2006, Nikookar et al. 2017). Water quality-mosquito relationships may be species-dependent, adding to the complexities of predicting MPA broadly (Mercer et al. 2005, Burroni et al. 2013, Gardner et al. 2013, Abai et al. 2016, Cepeda-Palacios et al. 2017). The impact of water quality on MBD persistence has been acknowledged in recent years (Yee et al. 2019, Nagy et al. 2021, Neff and Dharmarajan 2021, Fazeli-Dinan et al. 2022, Kinga et al. 2022); however, the extent of variation in the magnitude and direction of its effects on mosquitoes at the species-, genus-, and family-level is less clear. Therefore, there is a need to investigate water quality determinants for MPA to increase the fidelity of MBD risk assessments.

Here, we conducted a systematic review and meta-analysis to appraise, identify, characterize, and quantitatively synthesize currently available studies across the globe, that investigate relationships between water quality and MPA; with the aim to further our understanding of the geographic expansion of MBD risks. Our primary objectives were to determine the impact of water quality on MPA and the degree to which it influences MPA. Results from this review will help to classify water quality properties that are linked to MPA, identify research gaps, improve modeling approaches for MBD risk assessments, and contribute to information that can be conveyed to the public on OIM hotspots in urban and rural environments.

Materials and Methods

Questions and Expectations

The following 2 questions have driven this systematic review and meta-analysis: (i) What is the evidence linking water quality properties to MPA? (ii) To what extent do water quality properties influence MPA?

We expected that pH, alkalinity, turbidity, electrical conductivity, dissolved oxygen, nitrogen, and phosphorus would be the most reported properties in studies investigating linkages between water quality and MPA. These expectations were based on previous reports establishing the influences of these water quality properties on the life history traits of mosquitoes, compared to other properties that have been less recognized to impact MPA. Therefore, we anticipated that these 7 water quality properties would be sufficiently reported on for purposes of meta-analyses. Given the ecological plasticity of various mosquito species, we expected that the magnitude and direction of the effects would be dependent on species and the type of available habitat. We hypothesized that pooled effects of most water quality properties would be influenced by the regional climate of the study settings, which is in line with studies investigating the adaptations of vector species in anthropized landscapes that have shown poor water quality regimes in urban and agricultural settings play a role in the spread of MBDs (Reiter 2001, Norris 2004, Brugueras et al. 2020, Perrin et al. 2022).

Review Approach and Search Strategy

This research was conducted using standard systematic review and meta-analysis methodologies (Young et al. 2014, Higgins et al. 2019), and applying the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement (Moher 2009). A comprehensive search strategy was developed and tested through an iterative process with the assistance of a trained librarian. A list of search terms was used for the following categories: mosquito genera (e.g., Aedes, Anopheles, Culex), exposure (e.g., water quality, turbidity, dissolved oxygen), and outcome (e.g., abundance, presence, larval density). This was applied to search for relevant articles in the following databases on 2 August 2022: Zoological Records, Scopus, CAB Direct, and Biological Abstracts. Search results were verified by screening references from 15 handpicked articles to ensure that relevant articles were successfully captured by the search strategy and to include potentially missed studies. Databases were searched without date or language restrictions. Search documentation and full search algorithms used for each database are included in Supp. Dataset S1. Duplicates were identified and removed using EndNote (version 20; Clarivate, Philadelphia, United States). Reviewing was conducted using the online systematic review management software DistillerSR (Evidence Partners, Ottawa, Canada). All steps of the systematic review were conducted using pre-tested forms by 2 independent reviewers. The relevance screening form was pre-tested on 50 abstracts until a kappa agreement of ≥0.80 was reached, while the other forms were each pre-tested on 3–5 articles to ensure both reviewers interpreted items clearly and consistently.

Eligibility Criteria

Studies selected for inclusion in the review were primary peer-reviewed quantitative research studies investigating MPA and water quality conditions of the reported habitats published in English, French, or Spanish in nonpredatory journals as per Beall’s List of potential predatory journals and publishers. All mosquito species and water quality properties were considered for inclusion. Outcomes of interest were the abundance and presence of mosquitoes in any stage of their life cycle (i.e., eggs, early/late instar larval stage, pupal stage, and adult stage). Any primary studies not directly reporting on water quality and MPA were excluded.

Relevance Screening and Risk of Bias Assessment

The titles and abstracts of citations identified during the search were assessed for their relevance using a structured screening form shown in Supplementary Text S2. Full texts of relevant references were obtained and then confirmed for relevance (Supplementary Text S2). Relevant studies that met all eligibility criteria underwent a risk of bias assessment to assess the internal validity of the studies and evaluate various biases (e.g., selection bias, detection bias, reporting bias, and confounding bias). An overall risk of bias of low, unclear, or high was determined for each outcome using a structured form shown in Supplementary Text S2. The risk of bias tool was developed and modified using previously established risk of bias tools for observational studies (Sterne et al. 2016, Higgins et al. 2019). The form contained 7 criteria that evaluated various biases (e.g., selection bias, confounding bias) and allowed multiple outcome assessments per study. In this review, overall unclear or high risk of bias was attributed to studies with unclear, unexplained, or insufficient information for the following elements: (i) quality of outcome and exposure measurement methodology, (ii) whether sites were selected in a way that makes them comparable across groups and/or unlikely to influence the outcome, (iii) possible influences of confounding elements on outcomes measured, (iv) complete assessment of all intended outcomes by authors, or (v) reported exclusions from final analyses. All relevance and risk of bias assessments were done by 2 independent reviewers, and conflicts were resolved by discussing until a consensus was met.

Data Extraction and Effect Size Conversion

A data characterization form (Supplementary Text S2) was used to extract the study characteristics from relevant articles including publication year, language, study design, location, climate, land cover, mosquito species, water quality properties, data collection methods (e.g., instruments used, sampling frequency), natural habitat drivers, and anthropogenic drivers. Studies sufficiently reporting outcome measurements for meta-analysis underwent an additional data extraction to collect association measures between water quality and MPA (Supplementary Text S2). Relevant data included correlation outcomes, continuous outcomes (e.g., mean differences), dichotomous outcomes (e.g., odds ratios), and contingency tables. Data were descriptively summarized using the following elements: mosquito species, stage of the life cycle, and water quality property for which the outcomes were measured.

Since the effect size metrics used in different studies were not consistent, it was not feasible to compare them directly. Therefore, odds ratios and continuous data were converted to standardized mean differences (Cohen’s d), where odds ratios were first collapsed into 2 categories while ensuring the same direction of effect was used across effect sizes. We then utilized conversion methods from Borenstein et al. (2009) to derive a common metric of effect size, the Pearson coefficient r. Studies that reported Spearman and Kendall correlation coefficients were converted to Pearson r coefficients using previously established methods (Gilpin 1993). To satisfy the conditions of meta-analytical tests, such as the normality of effect size distribution, we applied Fisher’s z transformation (Cooper et al. 2019). After conducting the analyses, the Fisher z estimate means were back-transformed to r means for interpretation purposes.

Meta-analyses and Potential Publication Bias

Although all water quality properties were eligible for inclusion in our review, we only synthesized those reporting sufficient data for meta-analyses. To draw meaningful conclusions and reduce the likelihood of spurious pooled estimates, we set a conservative minimum of >25 data points to be eligible for meta-analysis. This also ensured that analyses were performed with sufficient mosquito species to elucidate relationships at the mosquito level. We conducted a random-effects meta-analysis to estimate the overall mean of the distribution of effect sizes of the relationships between each of the most reported water quality properties and MPA. Estimates of heterogeneity in effect sizes between studies were measured by I², which measures the portion of the variance in effect sizes that is unrelated to sampling error, and the impact of heterogeneity was estimated using 95% prediction intervals (95% PI). As variability in effect size was expected, thus the random-effects meta-analysis incorporates variability between studies by assuming the true effect size is a distribution. The variance of each effect size was calculated using 1/(n − 3) (Borenstein et al. 2009), where n was the determined sample size of each effect size based on the number of mosquito sampling sites and/or the duration of sampling. To determine if overall mean effect sizes varied between climates, we further analyzed effect sizes of each of the most reported water quality properties by stratifying the effect sizes per climatic subgroups based on its Köppen classification (i.e., arid, tropical, temperate, and continental zones) when >1 effect sizes were present within each subgroup (Sterne et al. 2011). For each analysis, we reported the following: pooled correlation (r), 95% confidence intervals (95% CI), P value, number of effect sizes (n), median sample size (Mdn), I², and 95% PI.

To evaluate the possibility of publication bias, we conducted an Egger’s regression to test the symmetry of the funnel plots for each of the meta-analyses (Egger et al. 1997, Rothstein et al. 2005, ). Publication bias can occur when the decision to publish a study is influenced by its results, which can lead to overestimation or underestimation of the true effect size (Rothstein et al. 2005). By examining the results of Egger’s regressions for each meta-analysis, we determined if there was an asymmetry that indicated potential publication bias. In line with the suggestion of Nakagawa and Santos (Nakagawa and Santos 2012) for biological meta-analyses, we also performed a trim-and-fill analysis (Duval and Tweedie 2000) to address the potential nonindependence issue that can arise when multiple effect sizes are derived from the same study or species.

All analyses were performed using the metafor (Viechtbauer 2010) packages available in R software (version 4.2.1; R Core Team, 2022).

Results

Study Characteristics and Risk of Bias Assessment

Our comprehensive search strategy identified 4058 unique references; from these, 79 relevant articles met the inclusion criteria (Fig. 1). Most were cross-sectional studies (n = 44), followed by longitudinal (n = 37), controlled before-and-after experimental designs (n = 2), and a case-control study (n = 1). All germane articles were found to be published in English between 1986 and 2022 (Fig. 2).

Fig. 1.

PRISMA flow (Moher et al. 2009) of references through the systematic review process of studies captured by the keyword strategy (see Supplementary Dataset S1).

Fig. 2.

Distribution of publications, including (counts and study designs), of the 79 included primary research publications relevant to relationships between water quality and MPA.

Thirty-four of the included articles (n = 34, 43%) were based in Africa; followed by studies set in tropical and arid regions of Asia (n = 30, 38%); and a small proportion was set in warm, mild, and continental regions of North America (n = 7, 9%). The rest were set in South America (n = 5, 6%), Australia (n = 2, 3%), and Europe (n = 1, 1%). The most reported mosquito genera were Anopheles (n = 52, 66%), Culex (n = 33, 42%), and Aedes (n = 28, 35%) (Table 1). Mosquito abundance was reported in most studies (n = 67, 85%) compared to mosquito presence/absence, which was investigated less frequently (n = 27, 34%). Most studies reported only on outcomes regarding the immature life stage of mosquitoes (n = 77, 97%), while only 2 studies investigated adults (n = 2, 3%). Some studies made multiple measurements of the same outcome due to periodic sampling. The most reported water quality properties were pH (n = 69, 87%), nitrogen (n = 52, 56%), turbidity (n = 52, 56%), electrical conductivity (n = 43, 54%), dissolved oxygen (n = 34, 43%), phosphorus concentrations (n = 24, 30%), and alkalinity (n = 8, 10%). These 7 properties were the only ones sufficiently reported for meta-analyses (>25 effect sizes). Reported nitrogen concentrations include ammonium (NH₄), ammonia (NH₃), nitrate (NO₃), nitrite (NO₂), and total nitrogen (TN), while reported phosphorus concentrations include total phosphorus (TP), dissolved phosphorus (DP), phosphate (PO₄^3-) and organophosphate. Other water quality properties were less frequently reported, such as biochemical oxygen demand (BOD), chemical oxygen demand (COD), total organic carbon (TOC), oxidation-reduction potential (ORP), sulfate, and metal concentrations (e.g., copper, iron, magnesium, and calcium). Some of their reported relationships with MPA are summarized in Table 2. The most common generalized habitat environments in the included articles were rural regions (n = 50, 63%), followed by urban areas (n = 32, 40%), and natural landscapes (n = 28, 28%). The 2 experimental studies did not report biotopes as the study areas were artificially designed. As most studies were conducted in Africa and Asia, most were equatorial (n = 34, 43%) and arid (n = 27, 34%), followed by temperate climate zonations (n = 14, 18%). Many studies reported the specific land covers of their study areas, and most frequently reported, as defined by the Food and Agriculture Organization of the United Nations (Food and Agriculture Organization of the United Nations 2000), artificial surfaces (n = 41, 52%), inland water bodies (n = 33, 42%), and vegetated, aquatic, and regularly flooded areas (n = 18, 23%). Numerous natural and anthropogenic drivers were reported across studies. Specifically, drivers included consideration of air/water temperature (n = 58, 73%); qualitatively typified water containment systems housing mosquito communities (n = 47, 59%); vegetation diversity/density (n = 44, 56%); water depth (n = 39, 49%); variables related to shade/sunlight exposure of habitats and potential habitats (n = 38, 48%); and amount of precipitation during the study period (n = 30, 38%) (Table 1). Anthropogenic effects were less frequently reported (vs. natural drivers) with most not discussing any anthropogenic driver (n = 49, 62%). However, some did consider the distance of OIM sites to the closest building in the area (n = 23, 29%). Other drivers, although rarely reported (n = 10, 13%), included point source pollution, sediment runoff, livestock, farm waste, human-made floods, and chemical treatments. A full list of study characteristics is available in Supplementary Dataset S1.

Table 1.

General characteristics of the 79 included primary research publications

Category	No. of studies (n)	%
Continent
Africa	34	43
Asia	30	38
North America	7	9
South America	5	6
Oceania	2	3
Europe	1	1
Mosquito generaa
Anopheles	52	66
Culex	33	42
Aedes	28	35
Ochlerotatus	5	6
Culiseta	4	5
Otherb	19	24
Outcomesa
Mosquito abundance	67	85
Mosquito presence/absence	27	34
Mosquito life stage
Immature	77	97
Adult	2	3
Water quality propertiesa
pH	69	87
Nitrogenc	52	56
Turbidity	52	56
Conductivity	43	54
Dissolved oxygen (DO)	34	43
Phosphorusd	24	30
Othere	39	49
Type of mosquito samplinga
Larval dips	57	72
Pipetting	7	9
Centre for Disease Control (CDC) Miniature Light trap	2	3
Otherf	16	20
Not reported	8	10
Type of water quality testinga
In situ sonde measurement	53	67
Chemical testing	48	60
Physical testing	43	54
Biotopea
Rural	50	63
Urban	32	40
Natural	28	35
Not reported/ Not applicable	2	3
Land coverag
Artificial surfaces (including urban and associated areas)	41	52
Inland water bodies	33	42
Shrub/herbaceous vegetation, aquatic or regularly flooded area	18	23
Tree-covered areas	10	13
Herbaceous crops	9	11
Multiple/layered crops	6	8
Coastal water bodies and intertidal areas	6	8
Sparsely natural vegetated areas	5	6
Grassland	4	5
Terrestrial barren land	3	4
Shrub-covered areas	3	4
Mangroves	2	3
Woody crops	1	1
Not reported/ Not applicable	6	8
Climate (Köppen classification)a
Equatorial (tropical) zone	34	43
Arid (dry) zone	27	34
Warm/mild temperate zone	14	18
Continental zone	3	4
Not reported/ Not applicable	1	1
Natural drivers (confounding factors)a
Air/water temperature	58	73
Water-holding containment types	47	59
Vegetation (diversity and/or density)	44	56
Water depth	39	49
Sunlight/shade exposure	38	48
Precipitation	30	38
Size of water surface	25	32
Dissolved solids	25	32
Algae presence	23	29
Salinity	21	27
Water source elevation	18	23
Water velocity	19	24
Detritus content	16	20
Humidity	16	20
Substrate type	13	16
Bacterial abundance	9	11
Ion content	7	9
Otherh	27	34
Not reported/ Not applicable	0	0
Anthropogenic drivers (confounding factors)a
Distance to buildings	23	29
Point source pollution	3	4
Sediment runoff	1	1
Otheri	6	8
Not reported/ Not applicable	49	62

^aTotal numbers may exceed 79 when more than 1 option has been selected within a category.

^bOther lesser reported mosquito genera including Mansonia, Armigeres, Lutzia, and Toxorhynchites.

^cIncludes all forms of nitrogen such as ammonium, ammonia, nitrate, nitrite, and total nitrogen.

^dIncludes all forms of phosphorous such as total and dissolved phosphorus, phosphate, and organophosphate.

^eIncludes all other properties such as alkalinity, BOD, COD, ORP, sulfate, metal concentration, etc.

^fIncludes all other forms of sampling such as area samples, nets, ladles, and hand collections.

^gLand cover categories defined by the Food and Agriculture Organization of the United Nations (FAO).

^hIncludes all other natural drivers such as altitude, wind speed, habitat permanence, and suspended solids.

ⁱIncludes all other anthropogenic drivers such as livestock, farm waste, man-made floods, and chemical treatments.

Table 2.

Summary of outcomes of the 79 included primary research publications relevant to relationships between water quality and mosquito presence/abundance

Reference	Country (Climate)a	Sampling year(s)b	Water source typec	Biotoped	Outcome	Water quality propertye	Main findings
Abai et al. (2016)	Iran (Arid/Dry, Temperate)	2008 - 2009	Ponds Streams	Natural	Abundance	Conductivity Turbidity Nitrate Nitrite Phosphate Metals Other	Water quality parameters did not show any significant differences among different mosquito species. There was no significant correlation between the abundance of larvae and any of the water quality parameters.
Abdel-Meguid (2022)	Egypt (Arid/Dry)	2020	Artificial Ponds	Urban Rural	Abundance	pH Conductivity Turbidity Nitrate Phosphate Other	The density of Cx. pipiens s.l. L. was significantly negatively correlated with pH, turbidity, phosphates, sulfates, and nitrates, while there was no significant correlation between larval density and conductivity.
Aklilu et al. (2020)	Ethiopia (Arid/Dry)	2012–2013	Ponds Streams Puddles	Rural	Abundance	pH Turbidity	An. pretoriensis Theobald larval density was significantly negatively associated with turbid habitats.
Alam et al. (2018)	Bangladesh (Tropical)	2011–2012	Artificial Ponds Streams	Rural Natural	Abundance	pH Turbidity	The abundance of An. peditaeniatus Leicester decreased significantly with increased pH. The abundance of An. vagus increased with a rise in pH.
Alkhayat et al. (2020)	Quatar (Arid/Dry)	2015–2016	Artificial Ditches	Urban Rural Natural	Abundance Presence/absence	pH Turbidity DO	Cx. pipiens s.l. was positively associated with turbidity and pH. Cx. quinquefasciatus Say was negatively associated with dissolved oxygen.
Bashar et al. (2016)	Bangladesh (Tropical)	2012	Artificial Ponds	Urban Natural	Abundance	DO pH Turbidity Ammonia Other	Dissolved oxygen is found to be one of the main predictors for the abundance of all species’ larvae (except Ae. aegypti L.). Some associations were found between the abundance of Culex spp. and chemical oxygen demand.
Burroni et al. (2013)	Argentina (Temperate)	2002	Ponds	Natural	Abundance	Turbidity	There was a significant negative relationship between the abundance of Oc. albifasciatus Macquart and turbidity.
Carver et al. (2011)	Australia (Temperate)	2009	Saltmarshes	Urban Natural	Abundance	DO pH Conductivity Turbidity Nitrate Nitrite Total phosphorus Total nitrogen Metals Other	During dry season, Ae. camptorhynchus Thomson mosquito abundance was significantly negatively associated with pH, turbidity and dissolved magnesium content. No other water quality parameter showed significant association with the specie’s abundance.
Cepeda-Palacios et al. (2017)	Mexico (Arid/Dry)	2015–2016	Artificial	Rural Urban Rural	Abundance Presence/absence	DO pH Turbidity	Water samples from troughs (peridomestic water containers) where Ae. aegypti larvae were present had significantly greater turbidity and DO compared to samples without the presence of Ae. aegypti larvae.
Chaiphongpachara et al. (2018)	Thailand (Tropical)	2016	Ponds Streams Rivers	Rural Natural	Abundance	DO pH Turbidity	Abundance of 2 malaria vector species An. subpictus Grassi and An. barbirostris s.l. van der Wulp was found to be significantly positively associated to DO and significantly negatively associated with pH. Turbidity had no significant associations.
Chirebvu and Chimbari (2015)	Botswana (Arid/Dry)	2013	Ponds Streams	Rural	Abundance	DO pH Conductivity Turbidity	Abundance of larvae was significantly negatively correlated to conductivity, and positively associated with turbidity. There was no statistically significant correlation between abundance and DO/pH.
David et al. (2021)	Brazil (Tropical)	2010	Artificial	Urban	Abundance	pH Conductivity DOC Total phosphorus Total nitrogen	Higher levels of dissolved organic carbon was the best predictor for the abundance of Ae. albopictus Skuse.
Djamouko-Djonkam et al. (2019)	Cameroon (Tropical)	2017–2018	Artificial	Urban	Presence/absence	pH Conductivity Turbidity Organophosphate Metals Other	Conductivity, turbidity, and organophosphate were found to be at significantly higher levels in Anopheles Meigen positive samples compared to negative samples.
El-Naggar et al. (2013)	Egypt (Arid/Dry)	2012	Artificial	Urban	Abundance	pH Turbidity	Cx. pipiens s.l., Cx. perexiguus Theobald, and Cx. antennatus Becker abundances were significantly positively correlated with turbidity. Cx. perexiguus, Cx. antennatus, Cx. pusillus Macquart, and Oc. detritus Haliday abundances were significantly positively associated with pH.
Emidi et al. (2017)	Tanzania (Tropical)	2015–2016	Artificial Ponds Streams Puddles	Rural	Abundance Presence/absence	pH Conductivity	Upper percentiles of conductivity (OR) was significantly associated with the presence and the observed increase of abundance of Anopheles mosquitoes.
Fillinger et al. (2009)	Gambia (Arid/Dry)	2005	Ponds Rice paddies Puddles Rivers	Natural	Abundance Presence/absence	DO pH Conductivity Turbidity	Case-control study: An. gambiae s.l. Giles larval density was found to be higher in areas where conductivity level was over 2,000 μS/cm. (statistically significant)
Gadiaga et al. (2011)	Senegal (Arid/Dry)	2007–2010	Artificial Puddles Ponds Rivers Marshes Lake	Urban Rural	Abundance Presence/absence	pH Conductivity Turbidity	Water pH >= 8.0 was significantly positively associated with Anopheles larvae presence and abundance.
Gadzama et al. (2018)	Nigeria (Arid/Dry)	Not reported	Artificial	Urban	Abundance Presence/absence	DO pH Conductivity Turbidity Nitrate Nitrite Ammonium Phosphate Metals	Turbidity, pH, conductivity, sulfate, calcium content, and magnesium content were all significantly positively associated to An. gambiae s.l. presence.
Gardner et al. (2013)	U.S.A. (Continental)	2009	Artificial	Urban	Abundance	pH Ammonia Nitrate Phosphate	Ammonia and nitrate were significantly positively associated to larval abundance, whereas pH was negatively associated to larval abundance.
Getachew et al. (2020)	Ethiopia (Arid/Dry)	2014–2016	Puddles Streams Swamps	Rural	Abundance	pH Turbidity	No statistically significant associations between Anopheles larval abundance and WQP.
Ghosh et al. (2020)	India (Tropical)	2017–2018	Artificial Puddles Swamps	Urban Rural	Abundance	DO pH Conductivity, Ammonia Nitrate Other	Positive correlations were found between all WQP and the density of Cx. vishnui Theobald and Ae. albopictus, except for chemical oxygen demand and alkalinity. Hardness showed a positive correlation with An. stephensi Liston and Cx. vishnui but showed a negative correlation with Ae. albopictus density.
Gouagna et al. (2012)	Reunion Island (Republic of France) (Tropical)	2010–2011	Artificial Puddles Ponds Streams	Natural	Abundance Presence/absence	pH Conductivity Turbidity	Within all aquatic habitat sampled, turbidity was significantly positively associated with An. arabiensis Patton larval presence. Conductivity was associated with larval abundance.
Gowelo et al. (2020)	Malawi (Tropical)	2017–2018	Artificial Streams Puddles	Rural	Abundance	pH Turbidity	Turbidity was significantly negatively associated with larval abundance.
Hafeez et al. (2022)	Pakistan (Arid/Dry)	Not reported	Artificial Puddles Ponds Streams	Urban Rural Natural	Presence/absence	DO pH Conductivity Turbidity	The most remote sites had highest Ae. albopictus presence in conductivity conditions of ≤1,000 µS/m and DO conditions of 2–5 mg/liter.
Hawaria et al. (2020)	Ethiopia (Temperate)	2017–2018	Artificial	Rural	Abundance Presence/absence	Turbidity	Turbidity was significantly positively associated with larval abundance.
Imai and Panjaitan (1990)	Indonesia (Tropical)	1982	Saltmarshes Lagoons Ponds	Rural	Abundance	pH Turbidity	Within the Anopheles group, turbidity was responsible for the variation of habitat preferences for each species (at a statistically significant level).
Kindu et al. (2018)	Ethiopia (Arid/Dry)	2011–2012	Artificial Puddles Ponds Streams	Rural	Abundance Presence/absence	pH Turbidity	Using multiple regression analysis, An. gambiae s.l. abundance was significantly positively associated to low turbidity, and significantly negatively associated to pH.
Keno et al. (2022)	Ethiopia (Tropical)	2020	Artificial Puddles Swamps	Rural	Abundance	DO Conductivity Turbidity	Anopheles abundance was significantly negatively correlated to conductivity, but no significance was attributed to DO.
Laboudi et al. (2012)	Morocco (Arid/Dry)	2009	Swamps Rivers Rice paddies	Rural	Abundance	DO pH Conductivity Turbidity	An. labranchiae Falleroni abundance was significantly negatively associated to pH and turbidity. No significant findings were found of the relationships of abundance and DO/conductivity.
Leisnham et al. (2004)	New Zealand (Temperate)	2003	Artificial Swamps	Urban Rural Natural	Abundance	pH Conductivity DOC Ammonia Nitrate Nitrite Phosphate Other	Study of relationships between environmental variables in various landuses and mosquito abundance. Dissolved organic carbon was the only WQP that was significantly positively associated with mosquito abundance.
Liu et al. (2012)	China (Temperate)	2010	Artificial	Rural	Abundance Presence/absence	pH Ammonia Other	The majority of An. sinensis Wiedemann larvae were found in chemical oxgyen demand conditions of < 2 mg/liter, ammonia of < 0.4 mg/liter, and sulfate < 150 mg/liter.
Low et al. (2016)	Ethiopia (Arid/Dry)	Not specified	Artificial Ponds Streams	Natural	Presence/absence	pH	There were some significant positive associations between pH and mosquito larvae abundance.
Ma et al. (2016)	China (Temperate)	2013	Rivers	Urban	Abundance	DO pH Ammonium Nitrate Nitrite Total phosphorus Dissolved phosphorus Total nitrogen Other	In urban rivers, larval density was significantly positively correlated to ammonium, TP, and DP, whereas larval density was significantly negatively correlated to DO, pH, and nitrate.
Mala and Irungu (2011)	Kenya (Arid/Dry)	2005	Streams	Rural	Abundance	pH Conductivity Turbidity Total phosphorus Total nitrogen	An. gambiae s.l. larvae density was significantly positively correlated with turbidity and TN, whereas there was a significant negative correlation with pH.
Mala et al. (2011)	Kenya (Arid/Dry)	2008–2010	Artificial Ponds Streams	Rural	Abundance	pH Conductivity Turbidity	There was a weak significant positive correlation between mosquito abundance and conductivity.
Mbuya et al. (2014)	Kenya (Arid/Dry)	2006	Natural swamps vs. cow-dung treated swamps	N/A	Abundance	DO pH Conductivity Turbidity Ammonium Nitrate Phosphate	Controlled before-and after study: Both anopheline and culicine larvae abundance was significantly positively correlated to DO, conductivity, and turbidity, whereas there was a significant negative correlation with pH.
Mercer et al. (2005)	U.S.A. (Continental)	2002–2003	Wetland	Natural	Abundance	DO pH Conductivity Turbidity Ammonia Nitrate Nitrite Phosphate	In 2002, mosquito abundance was significantly positively correlated with nitrate and phosphate. In 2003, mosquito abundance was significantly positively correlated with nitrate, phosphate, and turbidity.
Mukhtar et al. (2006)	Pakistan (Arid/Dry)	2001–2002	Artificial	Rural	Abundance	pH Conductivity Turbidity Ammonium Total phosphorus Other	Anopheles and Culex abundance was significantly negatively associated to conductivity at levels over 1.5 dS/m, and also significantly positively associated to TP.
Muturi et al. (2009)	Kenya (Arid/Dry)	2006	Rice paddies	Rural	Abundance	Turbidity	Cx. quinquefasciatus significantly positively associated to turbidity.
Muturi et al. (2008)	Kenya (Arid/Dry)	2006	Rice paddies	Rural Natural	Abundance	DO pH Conductivity Turbidity	An. arabiensis and Cx. quinquefasciatus were significantly positively correlated (weakly and strongly, respectively) to DO.
Mwangangi et al. (2007)	Kenya (Arid/Dry)	1999	Rice paddies	Rural	Abundance	pH Conductivity Turbidity	No significant associations were found between mosquito abundance and WQP.
Mwangangi et al. (2010)	Kenya (Arid/Dry)	2004–2005	Artificial Ponds Swamps	Rural	Abundance	Turbidity	Anopheles abundance was significantly negatively associated to turbidity.
Nabar et al. (2011)	India (Arid/Dry)	Not specified	Artificial Rice paddies	Urban	Abundance	pH Other	Anopheles, Aedes Meigen, and Culex abundances were positively correlated to pH.
Nambunga et al. (2020)	Tanzania (Arid/Dry)	2018–2019	Ponds Streams	Rural	Presence/absence	pH Conductivity Turbidity Nitrate	An. funestus Giles larvae presence had no significant association with WQP
Ndenga et al. (2012)	Kenya (Arid/Dry)	2008–2009	Artificial Swamps Ponds Rivers	Natural	Abundance Presence/absence	pH Ammonium Nitrate Nitrite Metals	Anopheles late instar larvae abundance was significantly negatively associated to iron content. Mean nitrate content was higher in mosquito present habitats (statistically significant).
Nihad et al. (2022)	India (Tropical)	2019–2020	Artificial Ponds	Urban Rural	Presence/absence	pH	Ae. albopictus presence was negatively associated to pH.
Nikookar et al. (2017)	Iran (Temperate)	2014	Artificial Ponds Streams	Urban Rural Natural	Abundance	pH Conductivity Turbidity Nitrate Phosphate	Cx. pipiens s.l. had a significant positive correlation with conductivity. No other significant correlations were found between all species sampled and WQP.
Noori et al. (2015)	U.S.A. (N/A)	N/A	Artificial Streams	N/A	Abundance	Ammonium Nitrate Phosphate	Controlled before-and after study: Higher levels of nitrate and phosphate was significantly positively associated to Culex spp. survival rate in artificial breeding sites.
Obi et al. (2019)	Nigeria (Tropical)	2013	Rock pools	Natural	Abundance	DO pH Conductivity Turbidity Nitrate Phosphate Other	Mosquito abundance was significantly positively correlated to conductivity, whereas DO was weakly negatively associated to abundance.
Okanga et al. (2013)	South Africa (Arid/Dry, Temperate)	Not specified	Artificial Wetlands	Natural	Abundance	DO pH	Abundance of malaria prevalent mosquitoes was significantly positively correlated to DO.
Onchuru et al. (2016)	Kenya (Arid/Dry)	2012	Ponds	Natural	Abundance	DO pH Conductivity ORP Ammonium Nitrate Phosphate Metals Other	Ae. aegypti and some other Aedes species larval abundances were significantly positively correlated to conductivity, ammonium, and phosphate. Culex species larval abundance was significantly positively correlated to ORP and free copper content. Total Anopheles larval abundance was significantly positively associated to DO.
Oussad et al. (2021)	Algeria (temperate)	2018–2019	Artificial Ponds	Urban Rural	Abundance	DO Ph Conductivity	DO negatively correlated with Cx. hortensis Ficalbi, conductivity positively correlated with Cs. longiareolata Macquart and Cx. pipiens s.l.,Cx. perexiguus negatively correlated with pH, An. labranchiae strongly negatively correlated with pH
Overgaard et al. (2017)	Columbia (Tropical)	2011	Artificial	Urban Rural	Abundance Presence/absence	DO pH Conductivity Ammonia Nitrate Phosphate	Ae. aegypti presence was significantly positively associated to DO, and negatively associated to pH. Ae. aegypti abundance was significantly negatively associated to DO.
Pinault and Hunter (2012)	Ecuador (Tropical)	2008–2010	Artificial Streams Swamps	Rural Natural	Presence/absence	DO pH Conductivity	An. punctimacula Dyar & Knab larval presence was significantly positively associated to DO.
Piyaratne et al. (2005)	Sri Lanka (Tropical)	1997–1998	Stream	Rural Natural	Abundance	DO pH Conductivity Turbidity Ammonia Nitrate Phosphate Metals Other	Only 1 significant association was found for both Anopheles species investigated, where An. varuna Lyengar was positively correlated to calcium content.
Rajavel (1992)	India (Tropical)	1998	Artificial	Urban Rural	Abundance	DO pH Conductivity Ammonia Nitrate Other	Ar. subalbatus Coquillett larval abundance was significantly positively correlated only to ammonia.
Ranathunge et al. (2020)	Sri Lanka (Tropical)	2013–2015	Artificial Ponds Swamps Streams	Urban Rural	Abundance	DO pH Conductivity Turbidity	The abundance of Anopheles spp. larvae showed a significant positive correlation with DO and turbidity.
Ranjeeta et al. (2008)	India (Tropical)	2005–2006	Artificial	Urban	Abundance	pH Conductivity Metals Other	Conductivity, pH, and calcium content were significantly positively associated to Anopheles and Culex larvae abundances.
Rao et al. (2011)	India (Tropical)	Not reported	Artificial	Urban	Abundance	pH Conductivity Turbidity Nitrate Phosphate Metals Other	Ae. albopictus larval density showed significant positive moderate to strong correlations with all WQP except for pH which showed a significant moderate correlation with larval abundance.
Ratnasari et al. (2020)	Indonesia (Tropical)	2019	Artificial Ponds Streams	Rural Natural	Abundance	pH	Ae. albopictus and Ae. aegypti larvae abundance were significantly positively correlated to pH.
Reiskind and Hopperstad (2017)	U.S.A. (Temperate)	2016	Artificial	Urban	Presence/absence	Turbidity	Aedes albopictus presence was significantly negatively associated to turbidity.
Reji et al. (2013)	India (Tropical)	2010–2011	Artificial	Urban Rural	Abundance	DO pH Conductivity Turbidity	Larval abundance was significantly negatively associated to conductivity. No other significant associations were found.
Rejmankova et al. (1993)	Belize (Tropical)	1990–1991	Artificial Puddles Ponds Swamps Streams	Rural Natural	Presence/absence	DO pH Conductivity Ammonia Nitrate Phosphate Metals	During the dry season, conductivity was significantly positively associated to An. albimanus Wiedemann presence, and negatively associated to An. pseudopunctipennis and An. argyritarsis Robineau-Devoidy presence. During the wet season, pH was significantly positively associated to An. albimanus presence, and negatively associated to An. crucians Wiedemann presence.
Rosmanida et al. (2020)	Indonesia (Tropical)	2017	Artificial Puddles Ponds Swamps Streams	Urban	Abundance	DO pH Turbidity Ammonia Nitrate Other	Only DO showed a significant correlation (weak positive) with total larval abundance.
Sasikumar et al. (1986)	India (Tropical)	1984–1985	Artificial	Rural	Abundance	pH Metals	Larval abundance was significantly positively associated to pH.
Seal et al. (2019)	India (Tropical)	2015–2016	Artificial Ponds Swamps	Rural	Abundance Presence/absence	DO pH Turbidity	Larval abundance was significantly positively associated to DO. No other significant associations were found.
Sérandour et al. (2010)	France (Tropical)	2003–2007	Wetlands	Natural	Presence/absence	DO pH Conductivity Nitrate Nitrite	Coquillettidia sp. Dyar presence/absence was significantly positively associated to conductivity, and negatively associated to nitrate.
Soares Gil et al. (2021)	Brazil (Tropical)	2016–2018	Rivers	Rural Natural	Abundance	DO pH Conductivity	Data was collected from aquatic plant species and associated water. Associations between Mansonia Blanchard species and WQP varied for each plant species.
Soleimani-Ahmadi et al. (2014)	Iran (Arid/Dry)	2009–2010	Ponds Rivers	Rural Natural	Abundance Presence/absence	pH Conductivity Turbidity Metals Other	Anopheles abundance was significantly negatively associated to turbidity, and was positively correlated to pH, conductivity, and sulfate.
Soumendranath et al. (2015)	India (Tropical)	2012–2014	Artificial	Urban Rural	Abundance	pH Conductivity Metals Other	Aedes abundance was significantly positively associated to conductivity.
Suryadi et al. (2019)	Indonesia (Tropical)	Not specified	Artificial	Rural	Abundance	pH Turbidity	No WQP had any significant associations with mosquito abundance.
Tarekegn et al. (2022)	Ethiopia (Tropical)	2018–2019	Artificial Puddles	Rural	Abundance Presence/absence	pH Conductivity Turbidity	Mildly turbid habitats were associated to the presence of Anopheles larvae.
Tedjou et al. (2020)	Cameroon (Tropical)	2018	Artificial	Urban	Abundance Presence/absence	Turbidity	Ae. aegypti and Ae. albopictus abundance was positively associated to turbid waters.
Thomas et al. (2016)	India (Tropical)	2013–2014	Artificial	Urban	Abundance	DO pH Turbidity Nitrate Nitrite Phosphate Other	General larval abundance was significantly positively associated to conductivity, sulfate, fluoride, and total hardness. When Anopheles abundance was investigated alone, it was also significantly positively associated to nitrate.
Villarreal-Treviño et al. (2020)	Mexico (Arid/Dry)	2012–2016	Artificial Puddles Streams	Urban Rural Natural	Abundance Presence/absence	Turbidity	An. pseudopunctipennis Theobald larval abundance and presence was significantly positively associated to turbidity, whereas An. albimanus was significantly negatively associated to turbidity.
Vong et al. (2021)	Thailand (Tropical)	2016–2017	Pitcher plants	Natural	Abundance	pH Conductivity	Total mosquito larvae abundance was significantly positively correlated to pH.
Wang et al. (2020)	China (Temperate)	2018	Artificial Puddles Rice paddies	Urban Rural	Abundance	DO pH Ammonia	An. sinensis larval abundance was significantly positively associated to DO, whereas Cx. p. pallens was significantly positively associated to ammonia.
Wang et al. (2021)	China (Continental)	2019	Artificial Rice paddies	Urban Rural	Abundance	DO pH Conductivity Turbidity Ammonia Other	Six different WQP were investigated, and their association to 6 species were assessed. The directions and strength of associations varied across species.
Zogo et al. (2019)	Côte d’Ivoire (Tropical)	2016–2017	Artificial Streams Rivers Rice paddies	Rural	Abundance Presence/absence	Turbidity	Anopheles abundance and presence was significantly positively associated to turbidity.

WQP, Water quality property; Ae, Aedes; An, Anopheles; Ar, Armigeres; Coq, Coquillettidia; Cx, Culex; Cs, Culiseta; Mn, Mansonia; Oc, Ochlerotatus; Tr, Tripteroides Giles;Tx, Toxorhynchites Theobald.

^aSome studies have reported more than 1 climate, as sampling has occurred in different climatic regions of their respective countries.

^bTime period of the study’s sampling. The years shown are the intervals in which mosquito and water sampling has occurred.

^c“Artificial” water type includes throughs, human-made containers, canals, drains, animal hoof-prints, potholes, water reservoirs, catch basins, dams, irrigation canals, drainage ditch, water-treatment ponds, septic tanks.

^dUrban – any region in city bounds; Rural – any region outside of urban areas, usually where agriculture and farming occur; Natural – any landform with no man-made infrastructure (e.g., fluvial, aeolian, coastal landforms).

^eDO – Dissolved Oxygen; Metals – any metal content that has been quantified from water samples; Other – all other parameters that are not DO, pH, conductivity, turbidity, nitrogen (of any form) and phosphorus (of any form). Some examples are alkalinity, total hardness, biochemical oxygen demand, chemical oxygen demand, etc.

A summary risk of bias assessment is shown in Table 3. Most studies had an overall high risk of bias (n = 48, 61%) and these deficiencies were commonly attributed to possible confounders (i.e., insufficient control for external elements affecting MPA). Several studies also had an unclear risk of bias rating (n = 27, 34%), due mainly to insufficient reporting of water sampling methods, and possible selection bias in the site and sample selection. Only a few studies (n = 8, 10%) received an overall low risk of bias rating. Details on the risk of bias ratings and relevant outcomes for each study are available in Supplementary Dataset S1.

Table 3.

Summary of risk of bias assessments of the outcomes within the 79 included studies

Criteria	No. of unique outcome assessmentsa	No. (%)a,b
		Yes (Low risk)	Not clear (Unclear risk)	No (High risk)
Were sites selected in a way which makes them comparable across groups and/or unlikely to influence the outcome?	79	44 (56%)	33 (42%)	2 (3%)
Were confounders appropriately identified and accounted for?	79	30 (38%)	5 (6%)	38 (48%)
Was exposure measurement conducted in a valid and reliable manner?	79	20 (25%)	56 (71%)	3 (4%)
Was outcome assessment conducted in a valid and reliable manner?	79	52 (66%)	26 (33%)	1 (1%)
Were exclusions from analysis reported?	79	72 (91%)	3 (4%)	4 (5%)
Did the authors report all intended outcomes?	79	62 (78%)	2 (3%)	15 (19%)
Was the study free of other problems that could put it at a high risk of bias?	79	78 (99%)	0 (%)	1 (1%)
Overall risk-of-bias for each outcome (within-study summary assessment)	83	8 (10%)	27 (34%)	48 (61%)

^aAll percentages were calculated using the total number of included relevant studies (n = 79), thus percentages and sum of counts can exceed 100% and 79, respectively, when more than 1 unique outcome assessment is attributed to a study.

^bRisk of bias definitions: Low = bias unlikely to modify reported outcomes; Unclear = unclear if bias will modify reported outcomes; High = bias significantly reduces confidence in results of outcomes.

Effects of pH and Alkalinity on MPA

There was a significant positive pooled correlation of pH on MPA (r = 0.10, 95% CI: 0–0.20, P = 0.05, n = 132) (Fig. 3) with a median number of sampling sites of 56 (Mdn = 56). We found substantial heterogeneity between studies not caused by sampling error (I² = 97%) and a large prediction interval (95% PI: −0.73 to 0.92). Species-specific correlations (Fig. 3) within the Aedes genus revealed significant positive correlations for 2 species with pH (Ae. aegypti and Ae. albopictus), while 1 species exhibited a significant negative correlation (Ae. camptorhynchus). Correlations observed within the Anopheles genus demonstrated significant positive correlations for 4 species (An. albimanus, An. culicifacies Giles, An. funestus, and An. gambiae), and 4 species displayed significant negative correlations (An. barbirostris van der Wulp, An. crucians, An. labranchiae, and An. peditaeniatus). Within the Culex genus, 2 species displayed significant positive correlations (Cx. bitaeniorhynchus Giles and Cx. vishnui), while no species showed significant negative correlations. The impact of pH on other species, including both direction and magnitude, is illustrated in Fig. 3.

Fig. 3.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for pH effects on MPA. The number of effect sizes is denoted by n. Confidence intervals positioned to the right (without crossing the dotted line), to the left (without crossing the dotted line), or overlapping with the dotted line, indicate whether mosquito species were positively, negatively, or not affected by pH. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Ar, Armigeres; Cx, Culex; Cs, Culiseta; Mn, Mansonia; Oc, Ochlerotatus; Tr, Tripteroides; Tx, Toxorhynchites.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for pH and MPA in arid regions (r = 0.03, 95% CI: −0.21 to 0.26, P = 0.79, n = 33, Mdn = 40, I² = 98%, 95% PI: −0.86 to 0.88). Whereas the equatorial subgroup showed a highly heterogenous significant positive pooled correlation (r = 0.25, 95% CI: 0.08–0.40, P = 0.004, n = 59, Mdn = 84, I² = 98%, 95% PI: −0.77 to 0.91). The temperate subgroup showed a moderately heterogenous significant negative pooled correlation (r = −0.10, 95% CI: −0.19 to −0.015, P = 0.02, n = 38, Mdn = 30, I² = 71%, 95% PI: −0.48 to 0.31). The continental subgroup showed a highly heterogenous nonsignificant near-null pooled correlation (r = −0.01, 95% CI: −0.19 to 0.24, P = 0.91, n = 7, Mdn = 230, I² = 93%, 95% PI: −0.50 to 0.52).

There was a nonsignificant near-null pooled correlation of alkalinity on MPA (r = 0.02, 95% CI: −0.20 to 0.24, P = 0.85, n = 36, Mdn = 30) (Fig. 4). We found substantial heterogeneity between studies not caused by sampling error (I² = 95%) and a large prediction interval (95% PI: −0.85 to 0.87). Species-specific correlations (Fig. 4) within the Aedes genus revealed a significant negative correlation for 1 species with alkalinity (Ae. aegypti), while no species exhibited significant positive correlations. Correlations observed within the Anopheles genus demonstrated significant positive correlations for 3 species (An. barbirostris, An. peditaeniatus, and An. vagus), and 1 species displayed a significant negative correlation (An. stephensi). Within the Culex genus, 2 species displayed significant positive correlations (Cx. gelidus Theobald and Cx. pipiens), and 1 species showed a significant negative correlation (Cx. vishnui). The impact of alkalinity on other species, including both direction and magnitude, is illustrated in Fig. 4.

Fig. 4.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for alkalinity effects on MPA. The number of effect sizes is denoted by n. Confidence intervals positioned to the right (without crossing the dotted line), to the left (without crossing the dotted line), or overlapping with the dotted line, indicate whether mosquito species were positively, negatively, or not affected by alkalinity. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Cx, Culex; Cs, Culiseta; Mn, Mansonia; Tx, Toxorhynchites.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for alkalinity and MPA in arid regions (r = 0.07, 95% CI: −0.43 to 0.53, P = 0.80, n = 2, Mdn = 10, I² = 0%, 95% PI: −0.43 to 0.53). The equatorial subgroup showed a highly heterogenous nonsignificant near-null pooled correlation (r = −0.04, 95% CI: −0.43 to 0.17, P = 0.87, n = 19, Mdn = 84, I² = 98%, 95% PI: −0.95 to 0.95). Whereas the temperate subgroup showed a homogeneous significant positive pooled correlation (r = 0.12, 95% CI: 0.01–0.23, P = 0.03, n = 15, Mdn = 30, I² = 22%, 95% PI: −0.11 to 0.34). The continental subgroup showed a highly heterogenous nonsignificant near-null pooled correlation (r = −0.01, 95% CI: −0.19 to 0.24, P = 0.91, n = 230, Mdn = 84, I² = 93%, 95% PI: −0.50 to 0.52). The continental subgroup (n = 0) was not meta-analyzed.

Effects of Turbidity on MPA

From the turbidity dataset, there was a significant positive pooled correlation on MPA (r = 0.26, 95% CI: 0.13–0.37, P < 0.0001, n = 92, Mdn = 77) (Fig. 5). We found substantial heterogeneity between studies not caused by sampling error (I² = 99%) and a large prediction interval (95% PI: −0.73 to 0.90). Species-specific correlations (Fig. 5) within the Aedes genus revealed significant positive correlations for 2 species with turbidity (Ae. aegypti and Ae. albopictus), while 1 species exhibited a significant negative correlation (Ae. camptorhynchus). Correlations observed within the Anopheles genus demonstrated significant positive correlations for 5 species (An. albimanus, An. arabiensis, An. cinereus Theobald,An. culicifacies, and An. stephensi), and 2 species displayed significant negative correlations (An. labranchiae and An. sinensis). Within the Culex genus, 1 species displayed a significant positive correlation (Cx. bitaeniorhynchus), while no species showed significant negative correlations. The impact of turbidity on other species, including both direction and magnitude, is illustrated in Fig. 5.

Fig. 5.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for turbidity effects on MPA. The number of effect sizes is denoted by n. whether mosquito species were positively, negatively, or not affected by turbidity. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Cx, Culex; Cs, Culiseta; Mn, Mansonia; Oc, Ochlerotatus; Tx, Toxorhynchites.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for turbidity and MPA in arid regions (r = 0.44, 95% CI: 0.17–0.64, P = 0.002, n = 26, Mdn = 47, I² = 99%, 95% PI: −0.75 to 0.96). Whereas the equatorial subgroup showed a highly heterogenous significant positive pooled correlation (r = 0.33, 95% CI: 0.11–0.52, P = 0.004, n = 34, Mdn = 84, I² = 99%, 95% PI: −0.76 to 0.93). The temperate subgroup showed a highly heterogenous nonsignificant near-null pooled correlation (r = −0.02, 95% CI: −0.15 to 0.010, P = 0.73, n = 27, Mdn = 30, I² = 76%, 95% PI: −0.52 to 0.49). The continental subgroup showed a highly heterogenous nonsignificant positive pooled correlation (r = 0.11, 95% CI: −0.17 to 0.38, P = 0.43, n = 7, Mdn = 230, I² = 97%, 95% PI: −0.59 to 0.72).

Effects of Conductivity on MPA

There was a significant positive pooled correlation of conductivity on MPA (r = 0.18, 95% CI: 0.06–0.30, P = 0.005, n = 74, Mdn = 40) (Fig. 6). We found substantial heterogeneity between studies not caused by sampling error (I² = 98%) and a large prediction interval (95% PI: −0.70 to 0.84). Species-specific correlations (Fig. 6) within the Aedes genus revealed significant positive correlations for 2 species with conductivity (Ae. aegypti and Ae. albopictus), while no species exhibited significant negative correlations. Correlations observed within the Anopheles genus demonstrated a significant positive correlation for 1 species (An. funestus), while 3 species displayed significant negative correlations (An. argyritarsis, An. pseudopunctipennis, and An. sinensis). Within the Culex genus, 1 species displayed a significant positive correlation (Cx. vishnui), and 1 species showed a significant negative correlation (Cx. quinquefasciatus). The impact of conductivity on other species, including both direction and magnitude, is illustrated in Fig. 6.

Fig. 6.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for conductivity effects on MPA. The number of effect sizes is denoted by n. whether mosquito species were positively, negatively, or not affected by conductivity. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Cx, Culex; Cs, Culiseta; Coq, Coquillettidia; Mn, Mansonia; Tr, Tripteroides; Tx, Toxorhynchites.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for conductivity and MPA in arid regions (r = 0.12, 95% CI: −0.12 to 0.35, P = 0.32, n = 17, Mdn = 40, I² = 99%, 95% PI: −0.68 to 0.79). Whereas the equatorial subgroup showed a highly heterogenous significant positive pooled correlation (r = 0.40, 95% CI: 0.13–0.62, P = 0.005, n = 28, Mdn = 44, I² = 98%, 95% PI: −0.81 to 0.96). The temperate subgroup showed a moderately heterogenous nonsignificant near-null pooled correlation (r = 0.04, 95% CI: −0.03 to 0.12, P = 0.26, n = 25, Mdn = 30, I² = 54%, 95% PI: −0.22 to 0.30). The continental subgroup showed a highly heterogenous nonsignificant near-null pooled correlation (r = −0.06, 95% CI: −0.22 to 0.09, P = 0.42, n = 6, Mdn = 230, I² = 89%, 95% PI: −0.43 to 0.32).

Effects of Dissolved Oxygen on MPA

There was a significant positive pooled correlation of dissolved oxygen on MPA (r = 0.32, 95% CI: 0.18 to 0.44, P < 0.0001, n = 79, Mdn = 72) (Fig. 7). We found substantial heterogeneity between studies not caused by sampling error (I² = 98%) and a large prediction interval (95% PI: −0.72 to 0.92). Species-specific correlations (Fig. 7) within the Aedes genus revealed a significant positive correlation for 1 species with dissolved oxygen (Ae. aegypti), while no species exhibited significant negative correlations. Correlations observed within the Anopheles genus demonstrated significant positive correlations for 7 species (An. barbirostris, An. culicifacies, An. oswaldoi Peryassú, An. peditaeniatus, An. pseudopunctipennis, An. punctimacula, and An. subpictus), while no species displayed significant negative correlations. Within the Culex genus, 2 species displayed significant positive correlations (Cx. gelidus and Cx. vishnui), and 2 species showed significant negative correlations (Cx. hortensis and Cx. hutchinsoni Barraud). The impact of dissolved oxygen on other species, including both direction and magnitude, is illustrated in Fig. 7.

Fig. 7.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for dissolved oxygen effects on MPA. The number of effect sizes is denoted by n. whether mosquito species were positively, negatively, or not affected by dissolved oxygen. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Ar, Armigeres; Cx, Culex; Cs, Culiseta; Coq, Coquillettidia; Mn, Mansonia; Tx, Toxorhynchites.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for dissolved oxygen and MPA in arid regions (r = 0.18, 95% CI: −0.02 to 0.37, P = 0.08, n = 16, Mdn = 143.5, I² = 97%, 95% PI: −0.69 to 0.97). Whereas the equatorial subgroup showed a highly heterogenous significant positive pooled correlation (r = 0.54, 95% CI: 0.34 to 0.70, P < 0.0001, n = 34, Mdn = 72, I² = 97%, 95% PI: −0.69 to 0.97). The temperate subgroup showed a highly heterogenous nonsignificant near-null pooled correlation (r = −0.05, 95% CI: −0.21 to 0.60, P = 0.53, n = 24, Mdn = 33, I² = 90%, 95% PI: −0.65 to 0.58). The continental subgroup showed a highly heterogenous significant positive pooled correlation (r = 0.17, 95% CI: 0.02–0.32, P = 0.03, n = 6, Mdn = 230, I² = 89%, 95% PI: −0.22 to 0.52).

Effects of Nutrients on MPA

First, from the studies investigating nitrogen, there was a significant positive pooled correlation on MPA (r = 0.21, 95% CI: 0.11–0.32, P < 0.0001, n = 97, Mdn = 33) (Fig. 8). We found substantial heterogeneity between studies not caused by sampling error (I² = 97%) and a large prediction interval (95% PI: −0.66 to 0.84). Species-specific correlations (Fig. 8) within the Aedes genus revealed no significant positive or negative correlations. Correlations observed within the Anopheles genus demonstrated significant positive correlations for 4 species (An. culicifacies, An. funestus, An. stephensi, and An. subpictus), while no species displayed significant negative correlations. Within the Culex genus, 3 species displayed significant positive correlations (Cx. quinquefasciatus, Cx. tritaeniorhynchus Giles, and Cx. vishnui), while 1 species showed a significant negative correlation (Cx. bitaeniorhynchus). The impact of nitrogen on other species, including both direction and magnitude, is illustrated in Fig. 8.

Fig. 8.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for nitrogen effects on MPA. The number of effect sizes is denoted by n. whether mosquito species were positively, negatively, or not affected by nitrogen. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Ar, Armigeres; Cx, Culex; Cs, Culiseta; Mn, Mansonia; Tx, Toxorhynchites.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for nitrogen and MPA in arid regions (r = 0.39, 95% CI: 0.10–0.61, P = 0.009, n = 11, Mdn = 176, I² = 99%, 95% PI: −0.51 to 0.88). Whereas the equatorial subgroup showed a highly heterogenous significant positive pooled correlation (r = 0.47, 95% CI: 0.23 to 0.65, P = 0.003, n = 26, Mdn = 84, I² = 97%, 95% PI: −0.71 to 0.96). The temperate subgroup showed a moderately heterogenous nonsignificant positive pooled correlation (r = 0.08, 95% CI: −0.01 to 0.017, P = 0.09, n = 40, Mdn = 30, I² = 68%, 95% PI: −0.41 to 0.53). The continental subgroup showed a highly heterogenous nonsignificant negative pooled correlation (r = −0.08, 95% CI: −0.40 to 0.26, P = 0.64, n = 9, Mdn = 230, I² = 98%, 95% PI: −0.82 to 0.76).

As for the phosphorus dataset, there was a significant positive pooled correlation on MPA (r = 0.24, 95% CI: 0.16–0.32, P < 0.0001, n = 43, Mdn = 30) (Fig. 9). We found substantial heterogeneity between studies not caused by sampling error (I² = 90%) and a large prediction interval (95% PI: −0.19 to 0.59). Species-specific correlations with phosphorus (Fig. 9) showed significant positive correlations for 2 Aedes species (Ae. aegypti and Ae. albopictus), 4 Anopheles species (An. culicifacies, An. gambiae, An. stephensi, and An. subpictus), and 1 Culex species (Cx. quinquefasciatus). No species displayed significant negative correlations. The impact of phosphorus on other species, including both direction and magnitude, is illustrated in Fig. 9.

Fig. 9.

Meta-analytic means per mosquito species and estimated pooled correlation based on the Pearson r coefficient (±95% CI) in a random-effects meta-analysis for phosphorus effects on MPA. The number of effect sizes is denoted by n. mosquito species were positively, negatively, or not affected by phosphorus. “All sampled”, studies including all sampled mosquito for analyses; Ae, Aedes; An, Anopheles; Cx, Culex; Cs, Culiseta.

Subgroup analysis by climate showed highly heterogenous nonsignificant near-null overall correlations for phosphorus and MPA in arid regions (r = 0.42, 95% CI: 0.28–0.54, P = <0.0001, n = 7, Mdn = 1063.5, I² = 97%, 95% PI: 0.04–0.69). The equatorial subgroup showed a highly heterogenous nonsignificant positive pooled correlation (r = 0.30, 95% CI: −0.05 to 0.58, P = 0.09, n = 4, Mdn = 140, I² = 93%, 95% PI: −0.43 to 0.79). Whereas the temperate subgroup showed a moderately heterogenous significant positive pooled correlation (r = 0.14, 95% CI: 0.05–0.23, P = 0.002, n = 29, Mdn = 18, I² = 31%, 95% PI: −0.14 to 0.40). The continental subgroup was not meta-analyzed (n = 1).

Publication Bias

Based on Egger’s regression, only the pooled correlations of turbidity (intercept = 0.46, 95% CI: 0.25–0.67, P = 0.02) and phosphorus (intercept = 0.49, 95% CI: 0.35–0.63, P = 0.001) showed statistical significance and potential publication bias, while the trim-and-final analysis estimated 22 and 16 effect sizes missing on the right side of the distribution of their effect sizes, respectively (Supplementary Table S3). This analysis increased the pooled correlation of both properties (Supplementary Table S3). However, the high heterogeneity between effect sizes and the lack of evidence of significant publication bias in the other meta-analyses indicates the potential skewing of the newly estimated correlations for the overall turbidity and phosphorus meta-analyses (Peters et al. 2007). This suggests there was no evidence of important publication bias in the overall set of included studies. Other sources for the newly estimated correlations (Supplementary Table S3) and high heterogeneity between effect sizes must be investigated, such as the quality of studies, differences in study designs, and the biological implications of the observed pooled correlations.

Other Water Quality Properties

All other properties, including sulfate and hardness, for example, were not sufficiently reported in our final list of articles for potential meta-analysis. These other water quality properties had < 25 total effect sizes reported. Nevertheless, as observed in the 7 properties meta-analyzed, the strength and direction of relationships varied between studies and species for all other captured water quality properties. A full list of effect sizes for all properties is available in Supplementary Dataset S1.

Discussion

The effects of physicochemical characteristics of mosquito OIM sites on MPA have been studied in most parts of the world, with a significant focus on areas with important potential for MBD outbreaks. However, no study to-date has applied meta-analytical approaches to investigate and synthesize species-specific tendencies at a global scale. Encompassing all mosquito species and several generalized geographical and climatic settings, this systematic review identified an overall positive link between MPA and several easy-to-characterize water quality properties. However, these effects varied when accounting for climate zones and mosquito species. Considering the high levels of heterogeneity observed in our pooled estimates, this suggests the observed overall correlation of some water quality properties may be species-specific, climate-specific, and/or proxy of other factors influencing MPA.

As expected, the most reported water quality properties were pH, alkalinity, turbidity, electrical conductivity, dissolved oxygen, and nutrient concentrations. To attempt to uncover the ecological and biological significance of our results, we considered the role of these water quality properties on the ecology, development, OIM, and survivorship characteristics of the reported mosquito species.

The Role of pH and Alkalinity on MPA

Water pH was a frequently investigated property, likely because of its significance in assessing suitable conditions for aquatic life and its ease of measurement (Bell 1971, Thurston et al. 1981, Wiseman et al. 2010). A circumneutral level of pH has been shown to be crucial for bioregulatory processes in aquatic species (Gensemer et al. 2018). Laboratory assays for mosquito pH tolerance have shown that aedine larvae from various genera can have successful larval and pupal development when pH levels range between 4 and 11, while pH levels below 3 and above 11 negatively affected the survival of these species (Clark et al. 2004). In addition, the post-emergence longevity of Cx. quinquefasciatus, an important vector for Zika and West Nile viruses, has been shown to be most successful in pH ranging between 6 and 8, whereas larvae growth rate significantly reduced in pH levels outside of 7 (Ukubuiwe et al. 2020); although this species maintained survivorship in pH values ranging from 5 to 9. These controlled laboratory experiments suggest a triangular distribution for the effects of pH on MPA, where MPA is positively correlated with pH levels between 6 and 8, while values outside that range result in negative relationships. Considering the variability of correlations at the species-level, we posit that the implications of correlations captured across studies may be spurious, and that controlled laboratory assays may be better suited for determining pH preferences of mosquito species. Furthermore, the high heterogeneity between effect sizes, including in the climatic subgroups, indicates other factors may have been at play for the resulting MPA measures. These may include the type of OIM habitat, where container species, such as Ae. aegypti and Ae. albopictus, were sampled from various artificial habitats (e.g., discarded tires, flowerpots, and sewers) which can have more dynamic pH conditions compared to permanent water sources. Also, considering that pH is known to fluctuate diurnally, studies that have measured the pH of larval habitats rarely controlled for the time of sampling. Other confounding elements such as climate, pollution, temperature, and altitude are known drivers for the variation of pH in rainwater, which constitute a large fraction of larvae habitats (Ruiz et al. 2010, Mohammed and Chadee 2011). These environmental factors have demonstrated themselves as primary drivers of MPA independently of pH (Bai et al. 2013, Brugueras et al. 2020, Nambunga et al. 2020). Consequently, it becomes imperative to recognize the concurrent influence of these environmental factors on both pH and MPA, potentially introducing confounding bias when elucidating the direct impacts of pH on MPA. It is also key to acknowledge that the pH levels of water sources also exert an influence on the presence and abundance of microorganisms upon which mosquito larvae and immatures rely for sustenance (Baker et al. 1982). This observation suggests a potential connection between pH variations and the dietary availability for mosquitoes, rather than a direct influence on their bioregulatory demands. Altogether, if studies are investigating habitat preferences, results from direct measures of association between pH and MPA must be carefully assessed. Our meta-analyses reinforce the notion that suitable pH conditions vary by species, while the heterogeneity in effects sizes between climatic subgroups suggest that pH preferences may be affected by the type of available water habitats and their permanence, as well as precipitation patterns. Specifically, the pooled correlation was near-null in the arid subgroup (r = 0.03) and the pooled correlation for the temperate subgroup was weakly negative (r = −0.10). As these regions receive significantly less rainfall than tropical localities, this may in turn reduce the impact of rainfall related pH instabilities on MPA.

Some research attention was attributed to the relationships between alkalinity and MPA (i.e., water’s resistance to acidification), however our meta-analysis showed that pooled correlations yielded no significant direction of association with MPA (Fig. 4). When comparing the species-specific pooled correlations of pH and alkalinity, Ae. aegypti and Cx. vishnui had significantly positive pooled correlations with pH and significantly negative pooled correlations with alkalinity, while we observed the opposite for An. barbirostris and An. peditaeniatus. Studies have shown that many mosquito species have mechanisms to acclimate to extreme levels of both pH and alkalinity (Clark et al. 2007, Multini et al. 2021), but our results further hint that some species may not be as adaptable. We also observed a significantly positive and homogenous pooled correlation for the temperate subgroup. This could be indicative of the heavy rainfall patterns of these regions decreasing overall alkalinity over time (Zeng et al. 2020), limiting certain mosquito species to OIM habitats with milder levels of alkalinity, such as puddles and ponds (Torreias et al. 2010). However, our analyses showed high heterogeneity for all other alkalinity meta-analyses, which signals that the same proxies discussed for pH are most likely affecting the relationships between alkalinity and MPA.

The Role of Turbidity on MPA

Turbidity refers to the degree of cloudiness of water, which in turn affects the amount and distance of light that can traverse the column. For some species, access to sunlight for immature development has shown to be an important ecological characteristic for oviposition preferences. For example, species such as Ae. albopictus and Ae. aegypti opt for oviposition in container habitats in part due to the clarity of water emanating from recent precipitation events (Juliano et al. 2004), while An. arabiensis and An. gambiae s.l. larvae are especially prevalent on the surface of pools and paddy fields where access to direct sunlight is uninterrupted (Gimnig et al. 2001). However, except for An. gambiae s.l. which had a large interval of correlations, the meta-analytical means of these clear-water species showed significantly positive correlations. This does not necessarily contradict the typical OIM characteristics of these species, as higher levels of turbidity can influence MPA positively by limiting the visibility of larvae to predators (Kweka et al. 2011) and by increasing larval development time via an increase in water temperature (i.e., decreased albedo) (Leisnham et al. 2004). Therefore, these significant positive relationships may be reflective of confounding albedo modifications. Additionally, turbidity can also stem from natural processes such as algae blooms, which have been shown to be part of the dietary regimes of some mosquitoes. Although species outside of the Toxorhynchites, Psorophora Fabricius, and Uranotaenia Lynch Arribálzaga genera do not rely on algae as primary food sources, some habitats may be lacking in bacterial and protozoan food sources that are preferred by most mosquito larvae which may promote their consumption of algae for development (Clements 1992). The increased pooled correlation captured in the arid subgroup analysis could be explained by lack of rainfall and surface water limiting the amount of organic matter that can be introduced into the aquatic environment. This can lead to fewer microorganisms for larvae consumption, and therefore, a reliability on algae for feed (Pointing and Belnap 2012, De Senerpont Domis et al. 2013, Carvajal-Lago et al. 2021). In contrast, it is also essential to consider that the impact of turbidity on MPA might be intertwined with visibility factors that influence predatorial feeding on eggs and immatures (Ortega et al. 2020). As such, turbidity conditions may not directly influence mosquito OIM, but rather manifest as an indirect outcome stemming from reduced predatorial activity.

The Role of Conductivity on MPA

Electrical conductivity, also know as specific conductance, is measured to assess the ability of water to conduct electricity; it is typically used as a proxy for salt concentrations. Conductance capabilities are linked to the concentration of ions present in the water column and may be induced by drivers such as hardness, salinity, and metal concentration. Consequently, elevated levels of conductivity have been associated with decreased water quality (Banna et al. 2014). Like pH, the variability of pooled estimates per species for the effects of conductivity on MPA may be justified by a triangular distribution. To help corroborate this, we considered an experiment by Mamai et al. (2021) where the authors tested the effects of conductivity levels on the rearing productivity of Aedes mosquitoes. They observed that conductivity levels above 368 µS/cm negatively impacted pupal development, yet the rate of larvae to pupae development increased in parallel to the rise in conductivity. Similar outcomes have been observed with Cx. quinquefasciatus (Ukubuiwe et al. 2020). The authors of both experiments suggested that the increase in conductivity appeared to influence the molting and metabolic rate of larvae, but once at the pupal stage, the mosquitoes were exposed to an excess in biofilm which resulted in overfeeding and subsequent mortality. They proposed that ion concentrations could impact the microbial and bacterial community composition in both the larval diet and water environment over time. In fact, Goller & Romeo (2008) have shown that higher levels of conductivity can drive biofilm development. This, in turn, may affect larval growth by limiting nutrient availability as a result of bacterial competition. Hence, we suspect that these trade-offs explain the significant positive and negative pooled correlations of the meta-analyses (Fig. 6), while the nonsignificant relationships may stem from confounding factors elucidated earlier. These factors encompass modifications in land use, which have been demonstrated to induce changes in sediment runoff, as well as variations in temperature and rainfall within OIM habitats, all of which have exhibited direct influences on both conductivity and MPA independently (Mainuri and Owino 2013, Rakotoarinia et al. 2022).

The Role of Dissolved Oxygen on MPA

As reviewed by Clements et al. (Clements 1992), mosquito larvae generally consume atmospheric oxygen for development and survival, but they can rely on dissolved oxygen in certain low-oxygen habitats. The pooled correlation for dissolved oxygen effects on MPA was significantly positive, possibly due to the large number of positive pooled correlations captured in the Anopheles genus (Fig. 7). Specifically, we observed a pattern of positive correlations for anopheline and aedine species whereas culicines were relatively unaffected by dissolved oxygen. Although dissolved oxygen has shown to be an important water quality property for the presence and abundance of multiple genera, its impact varies when considering the species-specific habitat niches. Container species like Ae. aegypti will colonize stagnant OIM sources in which oxygen resources are largely derived from aquatic plants and algae (Suryadi et al. 2019), while Anopheles species, such as An. arabiensis and An. subpictus, are prevalent in ponds, swamps, and irrigation ditches with similar oxygen sources (Muturi et al. 2008, Ratnasari et al. 2020). This corroborates with the results of an experiment by Yamada et al. (2020) where they quantified the role of oxygen depletion on an Ae. aegypti, Ae. albopictus, and An. arabiensis pupae and reported that all species depleted the dissolved oxygen under 0.5% in less than 30 min in artificial containers. In contrast, culicine species, including Cx. pipiens s.l., Cx. tarsalis, and Cx. quinquefasciatus, have been observed to colonize water bodies with reduced dissolved oxygen levels (Vinogradova 2000, Muturi et al. 2009). This ecological characteristic allows them to outcompete other aquatic species for ecosystem resources, such as common water fleas, while also evading predation from larvivorous fish, which have both shown to unsuccessfully develop in less oxygenated waters (Cech et al. 1985, Nebeker et al. 1992). However, a study measured the isolated effects of dissolved oxygen on the survival and development time of Cx. pipiens and found that lower levels of dissolved oxygen in water led to decreased larval survival rates and increased development time even when provided with atmospheric oxygen (Silberbush et al. 2015), suggesting that this property remains crucial in some culicine species. While outcomes from the dissolved oxygen meta-analysis can further substantiate previously reported niche characteristics of mosquito species, the high heterogeneity between effect sizes must be considered. This may be in part due to dissolved oxygen levels being associated with the other water quality properties explored in this study as well as other MPA drivers such as temperature and precipitation. Nevertheless, we underscore the significance of assessing dissolved oxygen in larval habitats to assess potential productive colonization, particularly in arid and tropical regions where dissolved oxygen can often serve as the primary source of oxygen.

The Role of Nutrients on MPA

One of the primary hypotheses proposed for the effects of nutrients on MPA is that increased nutrients provide additional access to thriving algae and microorganisms for developing mosquitoes, while promoting aquatic species growth for added habitat provision, and therefore, shelter from predation (Clements 1992, Carvajal-Lago et al. 2021). Pooled correlations revealed a significant positive association between both nitrogen and phosphorus and MPA. Pooled correlations by species showed once again that the strength and significance of correlations were species-specific, however only a single significant negative association was found in both meta-analyses, where Cx. bitaeniorhynchus was negatively impacted by nitrogen concentrations (Fig. 8). Interestingly, the meta-analytical means for the effects of phosphorus on MPA showed that no species had a significant negative relationship with the various forms of phosphorus (Fig. 9). It is important to note that the high heterogeneity found within the meta-analyses may be due to various factors influencing nutrient necessities, including immature feeding habits, habitat characteristics, and life history traits. This suggests that different mosquito species have varying nutrient type and abundance necessities for their development and reproduction, and the effect of nutrient availability on MPA may depend on the mosquito species present in the habitat. For example, some mosquito species require high levels of phosphorus for optimal development, such as Ae. aegypti, which has been shown to benefit from phosphorus enrichment in artificial containers (Clements 1992, Darriet and Corbel 2008, Carvajal-Lago et al. 2021). Leisnham et al. (2004) noted that other species may require higher levels of nitrogen, such as Culex mosquitoes, which have shown to prosper in sources with high nitrogen levels due to their ability to feed on blooming bacteria and algae, although the authors noted that larvae could not successfully develop in waters with extreme detrital loads; perhaps as a response to the organisms being damaged physically by moving material in the water course. Factors such as temperature and humidity may also play a role in shaping the nutrient requirements of different mosquito species and their response to nutrient availability (Clements 1992, Brugueras et al. 2020). In addition, nutrient concentrations may be influenced by agricultural runoff and rainfall patterns, leading to higher nutrient levels in some areas more than others (Zanon et al. 2020). Furthermore, the climatic subgroup analyses also revealed that the pooled correlations for nitrogen were significantly increased in both the arid and tropical climate subgroup analyses. It can be inferred that the effects of nitrogen on MPA may be more pronounced in environments with limited water resources, such as arid regions, where mosquito larvae and pupae may be more reliant on nutrients from the water body due to a lack of alternative food sources (e.g., algae). These results highlight the importance of considering species-specific nutrient necessities based on life stage, climate, and landscape. This proposed stratification would provide a more robust framework for understanding the underlying processes driving the association between nutrient availability and MPA, while providing evidence for the use of effective water management systems to limit the proliferation of MBD vectors.

Recommendations and Limitations

We identified several biases and shortcomings associated with the identified studies which may have affected the overall conclusions of this meta-analysis. Firstly, the geographical distribution of the included studies was largely focused in Asian and African countries (n = 64, 81%). This focused distribution is engendered by the amplified risks of MBD outbreaks in these regions (Norris 2004, Tolle 2009, Bai et al. 2013, Okanga et al. 2013). However, we feel the small number of relevant studies in North America, South America and Europe is not proportional to these regions’ concern surrounding MBD outbreaks (Lanciotti et al. 2007, Waits et al. 2018, Brugueras et al. 2020). Furthermore, within-study biases were identified in a few areas. Some studies did not clearly state the rationale behind selection or allocation of sampling sites as well as methods for quantifying water quality properties. We recommend future studies in this area to improve transparency by reporting on the following: reasoning and justification behind the selection of study sites; instruments and analytical approaches used throughout the study and across groups with multiple readings or continuous monitoring; whether trained personnel were recruited for measurements using calibrated and pre-tested instruments as well as for data collection and tabulation (Millsap and Everson 1993, Fitzpatrick et al. 2009). Also, many studies in this review did not account for external factors affecting MPA. Meta-analytical results for the effects of each water quality property on MPA has suggested a need for multi-factorial assessments to further understand the preferences for oviposition as many properties likely interactively impact MPA. Moreover, the independent analysis of water quality properties on MPA veils the autocorrelative nature of relationships between these indicators, hindering on establishing the strongest predictors for MPA. Some examples include autocorrelations between pH and alkalinity (Saalidong et al. 2022) as well as dissolved oxygen and organic nutrient loading (Dodds 2006). We recommend the identification and control of such influences through matching, stratification, multivariable analysis, or other approaches (Skelly et al. 2012). Lastly, we suggest the inclusion of temporal lags in any multivariate modeling as the effect of MPA drivers can vary based on mosquito life history traits (Rakotoarinia et al. 2022).

Some limitations were present in our review. For instance, language bias was present as only publications in English, French, and Spanish were included for review, which ultimately excluded 14 potentially relevant articles (Fig. 1). There was also potential bias in excluding predatory journals per Beall’s List (2020), since this list may be biased as well (Kimotho 2019). Additionally, we recognize that our search strategy focused on MPA which limited the inclusion of studies exploring drivers for mosquito development traits (e.g., molting period and survival thresholds). However, these traits were investigated in our discussion to uncover the underlying mechanisms for the effects of water quality on MPA. Finally, we acknowledge that the high heterogeneity among correlations could be indicative of external factors impacting MPA, and these must be considered when interpreting our findings.

Conclusion

Globally, studies aggregating information on the ecological characterization of vector species habitats help to provide evidence on the effect of climatic and environmental variables related to habitat preferences of disease vectors, and therefore, provide insights on factors promoting their expansion and persistence (Servadio et al. 2018, Brugueras et al. 2020, Perrin et al. 2022). We synthesized global data and determined whether the water quality of OIM habitats in various climates played a role in mosquito occurrence and density, while considering the preferable range of the investigated water quality properties at the species, genera, and mosquito level. Based on our synthesis, we have the following conclusions and observations:

• There was a significant positive pooled correlation between MPA and pH, turbidity, electrical conductivity, dissolved oxygen, and nutrients.

• Correlations per species revealed that suitable ranges for pH, alkalinity, turbidity, electrical conductivity, and dissolved oxygen are species- and/or genus-specific.

• The high heterogeneity between effect sizes suggests that other abiotic and biotic factors may be influencing the impact of these properties on MPA.

• Climate regime has shown to influence the strength and direction of pH, alkalinity, turbidity, electrical conductivity, dissolved oxygen, and nutrient effects on MPA. Yet climate zonation must be interpreted in the context of standard of living of urban populations living within them. Countries with a high Human Development Index (HDI) may be able to provide critical resources that directly or indirectly manage MPA and associated disease risks, in relation to countries with a lower HDI, and many of the countries with a high HDI occur in the more temperate regions of the world.

• Urban vector species have shown to be most adaptable to a wider range of values for water quality properties.

By considering the key factors highlighted in this review, future research can strengthen existing models of vector species expansion in diverse landscapes and serve as a fundamental basis for further investigations into the effects of water quality properties on the spread of vectors. Such insights could help support urban and rural water quality management, encourage improved agricultural practices and waste production to prevent vector OIM, and advocate for stronger water management policies to control the spread of MBDs.

Acknowledgments

We would like to give our most sincere thanks to Mark Sunohara and Emilia Craiovan for their support in conceptualization. We are grateful for the help from Susan Young who helped structure our search strategy. We would also like to thank all authors whose studies were included in this review for their publications.

Funding

Funding for this project was possible thanks to Agriculture and Agri-Food Canada and Carleton University.

Data Availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.