How Hurricanes Irma and Maria affected population dynamics and nutrient content of Aedes aegypti in San Juan, PR, USA: socioeconomic and temporal factors

Abstract

Urban environments often contain mosquito species that are responsible for transmitting medically important pathogens to humans. Large disturbance events, like hurricanes, can devastate large urban areas, especially in the tropics, however little data exist for how these storms affect vector populations. During September 2017 Hurricanes Irma (category 5) and Maria (category 4) passed in proximity to the island of Puerto Rico, USA, causing significant damage to the built environment and significantly altering the abiotic environment including the removal of the plant canopy. We measured adult Aedes aegypti populations, the main vector of several pathogens, and larval containers across eight neighborhoods in San Juan, the capitol, that varied in socioeconomic status (SES) across eight sampling periods over 17 months following the storms. We also analyzed the nutrient content (%N, %C, C:N) and stable isotopes (δ¹⁵N, δ¹³C) from adults and isotopes from containers to assess how the nutrient environments changed post hurricanes. Mosquito population sizes were invariant throughout sampling, although more females were collected in lower SES neighborhoods that were more enriched in δ¹⁵N compared to higher SES locations. We did find that the storms altered the stoichiometric content of adults, with lower C:N values right after compared to a year later; larval containers showed an increase in δ¹⁵N through time. The lack of any interactive effects of the storms on specific neighborhoods suggests that Irma and Maria affected all locations equally, however, the storms altered the nutrient content of both adults and larval containers, a result with implications for pathogen transmission.

Introduction

Hurricanes have significant effects on habitat structure and ecosystem function, with changes in resource dynamics, especially due to canopy removal, often cited as a mechanism for changes in communities of organisms post disturbance (e.g. Shiels et al. 2010). Hurricanes, with their wind and rain, may influence ecosystems by determining the structure and function of the plant and animal communities exposed to them (Gardner et al. 2005, Lugo 2008). Although hurricane effects are fairly well studied in natural ecosystems (e.g. Lugo 2008, Brokaw et al. 2012, Liu et al. 2018), the effects of these large storms on ecosystem processes in cities has received less attention (Grimm et al. 2017, Méndez-Lázaro et al. 2018). As the intensity and frequency of large hurricanes are expected to increase due to global atmospheric warming, and as urban populations are predicted to increase from ∼55% of the world’s population to 68% by 2050, it is all the more imperative that we understand how large storms affect human dominated systems (Grimm et al. 2017, United Nations 2018, Knutson et al. 2020).

Mosquitoes are an important agent of human and animal disease (Yee et al. 2022). Cities represent a unique setting to study mosquitoes, given that humans in these places often create habitats (e.g. trash, storm drains), provide blood meals themselves or via other human-associated species (e.g. pets, urban birds), and alter their life history in response to urbanization processes like Urban Heat Island effects (e.g. Lee et al. 2020, Duval et al. 2023). Extreme landscape disturbances and flooding events in urban systems can temporarily increase availability of aquatic habitats for mosquito larvae, disrupt detrital inputs, and may provide a pulse of nutrient-rich water source as flood waters often intermix with wastewater (Yee et al. 2019). Moreover, changes in habitat may affect adult distributions via factors hypothesized to affect oviposition patterns, including light level, adult resting sites, and nectar sources (Rhodes et al. 2022). In particular, Aedes aegypti (yellow fever mosquito) is closely associated with human habitation in urban spaces, where it often can be found resting within dwellings. It actively seeks out humans for blood meals (De Benedictis et al. 2003, Barrera et al. 2011) and often bites people during the day, making it an effective vector for pathogens of human disease importance. In Puerto Rico, A. aegypti is the dominant urban mosquito responsible for the spread of arboviruses like dengue, chikungunya, and Zika (Sharp et al. 2013, 2014). Prior research has shown mosquito abundance and body nitrogen of A. aegypti are influenced by detrital content in larval habitats (Yee et al. 2015).

Few studies have explored how hurricane-induced changes across the landscape may affect mosquito populations (e.g. Mason and Cavalié 1964, Caillouët et al. 2008b, Barrera et al. 2019). Although many species of mosquitoes use large open-water aquatic habitats for development of their larvae, small aquatic isolated systems (i.e. containers), like discarded vehicle tires, tree holes, and buckets, are important sources of adults (Vezzani 2007, Yee 2008). For the larval stage, input of allochthonous detritus is the main source of energy for aquatic food webs. Allochthonous detritus breaks down via the action of microorganisms and is transformed into new microorganism cells, which are then ingested by foraging larvae. A nitrogen and carbon stoichiometry approach, measured in both the food and the consumer, has yielded new insights into the mechanisms for spatial and temporal patterns of container mosquito species, as well as elucidated variation in larval survival, development times, and mass (Kaufman et al. 2010, Winters and Yee 2012, Yee et al. 2015); as mass and fecundity are highly correlated, this latter measurement can inform us about effects on population growth as well (Yee et al. 2007). Although prior work has established the basic mechanisms responsible for the distribution and survival of urban mosquitoes, we do not know how large ecosystem perturbations associated with hurricanes would affect these dynamics, nor how they may vary with respect to socioeconomic factors.

During September 2017 significant storms affected the island of Puerto Rico. On September fifth, Hurricane Irma, a category 5 storm passed within 50 km of the island, generating 193 kph wind speeds and dropping up to 25 cm of rainfall [FEMA (Federal Emergency Management Association) 2018]. Two weeks later on September 20th, Hurricane Maria, a category 4 storm, passed directly over the main island of Puerto Rico, producing winds of 225 kph and rainfall in excess of 89 cm. These storms also caused extensive flooding, where floodwaters were persistent for weeks after the hurricanes and temporarily eliminated canopy cover. But the recovery from the hurricanes had clear ecosystem effects that persisted for months after they passed. Branoff (2020) found that canopy cover in coastal mangrove forests in Puerto Rico was still reduced to 51% of prior cover 1 month after Irma and Maria. These destructive storms provided an important opportunity to identify indirect effects like linking extreme landscape disturbance and flooding events to disease ecology. This includes changes in detritus inputs that affect nutrient availability and production of mosquitoes across time as the systems recovered to pre-hurricane conditions. With climate change increasing the frequency of storms (Knutson et al. 2020, Romero et al. 2020), these effects will be increasingly important to understand.

In addition to changes in adult mosquito populations, using a stoichiometric and isotopic data approach could also allow us to identify a link in changes in detrital inputs to mosquito production and adult and larval container nutrient content. Besides nutrients (i.e. %N, %C), stable isotopes (δ¹⁵N, δ¹³C) provided us a record of assimilation, processing, and recycling of nutrients, such that samples collected from different locations can have unique isotope signatures (Oczkowski et al. 2009, 2011). We used these measurements to determine if differences among neighborhoods may indicate how the nutrients were affected by and changed after the storms. For instance, some San Juan neighborhoods receive annual inputs of sewage via flooding events and contain more adult A. aegypti, and higher levels of nitrogen in adults and containers, whereas neighborhoods without flooding do not show these nutrient levels (Yee et al. 2019). These types of differences are affiliated with socioeconomic status (SES) and may be altered or exacerbated by large storm events, however how these patterns would be affected by large magnitude storms is unknown. Canopy cover may also be affected by these storms, and as detrital inputs affect mosquito production it has the capacity of changing adult populations and nutrient dynamics.

Given that low socio-economical populations are more prone to damage due to disasters like hurricanes, especially in the tropics [Ishizawa and Miranda 2016, SAMSHA 2017], we sought to build on recent studies that document the negative effects of low SES on human health effects from hurricanes by exploring additional indirect consequences associated with mosquito prevalence and mosquito-borne diseases. More specifically, we aimed to determine whether there were spatial or temporal effects of the hurricanes on the nutrient chemistry and population dynamics of A. aegypti, and their nearby container habitats, and to assess whether these effects varied with degree of urbanization. To address these questions, we collected nutrient and stable isotope data on populations of A. aegypti in eight neighborhoods across seven sampling periods that began 4 months after Hurricane Maria and ended 16 months later. Given the large and widespread effect these storms had on these neighborhoods, we hypothesized that litter inputs via changes in canopy after the storm would vary across neighborhoods, producing variable effects on larval container environments, and on to adult numbers and their nutrient content. Canopy cover has been shown to vary by SES in urban systems including Puerto Rico (Martinuzzi et al. 2018). These storms provided us with a rare opportunity to investigate both the short- and long-term effects of large atmospheric disturbances on the production of a medically important mosquito in a dense tropical urban city.

Materials and methods

Study area

The island of Puerto Rico has a subtropical, maritime climate with wet and dry seasons occurring from May to October, and November to April, respectively; mosquitoes are most abundant during the wet season (Barrera et al. 2011). The 78 municipalities that make up the island represent a variety of biomes, from urbanized areas, wetlands, tropical forest, and dry forests, with the cities also containing urban forest fragments and managed green spaces (Smith et al. 2009). Our work was conducted in the capital and largest city, San Juan (18° 27′ N, 66° 05′ W), on the north-central coast of the island, which is home to ∼350  000 people (United States Census Bureau 2018).

Sampling took place in eight neighborhoods across three municipalities (Carolina, Cataño, San Juan) (Fig. 1). Neighborhoods varied in their SES (defined metrics including human density, education, income, and abandonment), and past work has shown that adult female mosquito abundance, across a range of mosquito species, varied based on SES (Scavo et al. 2021), and that, prior to the hurricanes, larval containers and adult nutrient patterns varied with SES (Yee et al. 2019). In depth details of this sampling approach and neighborhood categorization can be found in Scavo et al. (2021) (Table 1). Briefly, we selected these neighborhoods based on past work in the area (Yee et al. 2019). We categorized these locations along a gradient of SES and environmental factors using cluster analysis, as established by Scavo et al. (2021). Neighborhood heterogeneity was quantified by measuring human (i.e. median household income, population density, college-level educational attainment, unemployment, health insurance coverage, and percentage of households below the poverty line, level of abandonment, litter incidence) and environmental (i.e. park size and amount of litter/trash, distance to the nearest water body, water body presence) factors. Although not necessarily direct metrics of mosquito habitat and abundance, these factors were used as indicators of general SES, which we used to explore potential linkages between SES and nutrient content of mosquitos and their habitats. Land cover data were not available after the storm so were not included as a variable. Cluster analysis of the eight neighborhoods produced five groupings along a gradient of SES and environmental factors (e.g. Scavo et al. 2021) (Table 1). Villa Venecia (VV) was the highest SES community and was the only gated community. Rio Piedras (RP), Vista Mar (VM), El Comandante (EC), and Puerto Nuevo (PN) were in the moderate SES group, with the latter two being unique based on their proximity to a large, forested park. The lowest SES neighborhoods were Torrecilla (TO), Cataño (CA), and Martin Peña (MP) (Fig. 1). Both MP and CA were generally characterized by closely placed housing and semi-frequent flooding from canals within the communities (Yee et al. 2019); TO was unique in its proximity to mangrove forests and saltwater habitats.

Mosquito and container sampling

Adult mosquito sampling started on 26 January 2018, ∼4 months after Hurricane Maria (20 September 2017). Subsequent samples took place six times (22 March 2018, 24 May 2018, 22 July 2018, 11 October 2018, 10 January 2019, and 11 May 2019). Time between sampling varied from 55 to 121 days (mean = 78 days) and each sampling period lasted no more than 10 days. This approach allowed us to sample both during the rainy season (May, July, October 2018, May 2019) and dry season (January, March 2018, January 2019). For each sampling trip we collected the same data categories: adult mosquitoes via traps and water samples from larval containers.

Adult A. aegypti were collected using BG Sentinel 2 traps (Biogents, Regensburg, Germany) baited with human scented BG lures (Biogents, Regensburg, Germany). These traps are designed to attract anthropophilic mosquitoes like A. aegypti (Maciel-de-Freitas et al. 2006). During each sampling period, six traps per neighborhood were placed in association with residences. This plan would have yielded 336 trapping events (8 neighborhoods × 6 traps per neighborhood × 7 sampling periods), however 14 traps malfunctioned (often because of a dead battery), which reduced our data down to 322 trapping events (there was no pattern to these malfunctions with respect to time or neighborhood). Locations within neighborhoods were chosen based on willingness and availability of residents to participate and with the caveat that the traps at residences were at least 200 m apart from one another to maximize mosquito capture due to adult flight distances (Harrington et al. 2005). After 48 h, any captured adults were killed by placing them on ice, and then sorted to species. Data for species besides A. aegypti can be found elsewhere (Scavo et al. 2021). After removing heads for viral analysis, female bodies were placed in individual 0.25 dram glass vials with a cotton stopper and dried for ≥48 h at 50°C.

During each sampling period we searched for larval containers within neighborhoods and sampled any that had larvae present. To find containers two to three researchers did visual inspections from the roads in front of properties, in any public spaces in those neighborhoods (e.g. parks), and on properties for which we received permission to place traps. Although this likely led to missing containers in other areas (e.g. on properties where we did not have access), it did mean that containers were often associated close to adult trapping locations. For each container we measured canopy cover, total water volume (ml), container type, and took filtered water samples that were later analyzed for isotope analysis of δ¹³C and δ¹⁵N in the Suspended Particulate Organic Matter (SPOM). Canopy cover was measured using a hand-held densiometer ∼1.5 m above the ground (values ranged from 0, no cover, to 39, total cover). We collected three 50 ml water samples from each accessible larval container. Water samples were filtered on site through 2.5 cm glass fiber filters (Whatman GF/F), with filters immediately placed on ice and then in a drying oven within a few hours of collection each day. Syringes and filter holders were rinsed several times with distilled water between samples and researchers wore latex gloves when processing. Total water volume (ml) was measured by removing the entire contents and placing all liquid in a graduated cylinder. Containers included discarded cans, buckets, plastic containers, as well as water held on tarps, and in bird baths and discarded automobile tires. Some containers were sampled more than once over the project, and some appeared later during sampling. American Airlines lost the baggage that contained adult mosquitoes and filters from the October 2018 samples, and thus only adult mosquito population data was available for that time period (however counts of females per trap had already been conducted before we left the island and thus those data remained).

Nutrient and stable isotope analysis

We measured % carbon and % nitrogen for each female A. aegypti adult captured, excluding those with obvious blood or sugar meals or who were gravid. Stable isotope values (i.e. δ¹³C, δ¹⁵N) were measured for both adult mosquitoes and SPOM. Replicate values from all filters obtained from a container were averaged to produce a single value. Stable isotopes from SPOM and adult mosquitoes were measured using an Elementar Vario Micro elemental analyzer connected to a continuous-flow Isoprime 100 isotope ratio mass spectrometer (Elementar Americas, Mt. Laurel, NJ) located at the U.S. Environmental Protection Agency in Narragansett, Rhode Island, USA. Replicate analyses of isotopic standard reference materials USGS 40 (δ¹³C = −26.39%; δ¹⁵N = −4.52%) and USGS 41 (δ¹³C = 37.63%; δ¹⁵N = 47.57%) were used to normalize isotopic values of working standards (blue mussel homogenate) to the air (δ¹⁵N) and Vienna Pee Dee Belemnite (δ¹³C) scales (Paul et al. 2007). Isotope values are expressed in δ notation following the formula δX (%) = [(R_sample/R_standard) − 1], where X is ¹³C or ¹⁵N and R is ¹³C/¹²C or ¹⁵N/¹⁴N isotopic ratio, respectively. Working standards were analyzed after every 24 samples to monitor instrument performance and check data normalization. The precision of the laboratory standards was better than ±0.3% for C and N. The sample particulate N and C content was calculated from area/mass relationships, determined during stable isotope analysis, for replicate standards of known N and C (Oczkowski et al. 2009, 2011). The relationship between the N and C contents of the standards and their analytical peak areas were used to determine the N and C contents of the samples via the peak areas of the samples. For adult mosquitoes, N or C was divided by total sample mass to determine the %N or %C. We also calculated the ratio of %C and %N (i.e. C:N) for each adult.

Dengue infection

For all adult females captured we removed heads and placed them individually into 1.5 ml snap cap vials in RNAlater (ThermoFisher Scientific). Briefly, RNA was extracted from the female heads with the RNeasy MinElute Kit on homogenized tissue and purified using the RNase-Free DNase Set. RNA concentration and purity were measured using a Thermo Scientific™ NanoDrop™ One Microvolume UV-Vis Spectrophotometer. All qPCR assays were performed using iTAQ polymerase supermix for probe-based assays (Bio-Rad). The DENV capsid gene and Vero cellular β-actin gene primer and probe sequences were adapted according to previous publications (Paul et al. 2014).

Statistical analyses

Total abundance of female A. aegypti were analyzed using a Repeated Measures ANOVA across the seven time periods across the eight neighborhoods. As raw data did not meet assumptions, we used a Ln(x + 1) transformation that normalized the data. Traps that failed were excluded from analysis.

Adult female nutrient values (%N, %C, C:N) and stable isotopes (δ¹³C, δ¹⁵N) were dependent variables in a nested multivariate analysis of variance (MANOVA), with neighborhood (8) and sampling period (6) as independent variables. Individual traps were nested within neighborhood and sampling period. Because of this, and because we often had to relocate trap locations within neighborhoods, we avoided a repeated measure approach. A square root transformation on absolute values for nutrients and stable isotope values was used to meet assumptions. Standardized canonical coefficients (SCC) were used to identify important dependent variables contributing to multivariate effects (Scheiner 2001). For instances of significant main effects or interactions, mean separation was used with a Tukey adjustment to control for comparison-wide error rates.

Container isotopes (δ¹³C, δ¹⁵N) were analyzed using MANOVA across neighborhoods (8) and sampling periods (5) with canopy cover as a covariate. The loss of samples from our October 2018 sampling reduced the number of sampling periods compared to adult population data. As the covariate was not involved in any significant interactions with main effects those interactions were dropped, and the analysis was rerun. Much like the analysis of adult nutrients, because we often did not sample the same containers, we decided to treat sampling periods as independent. Raw data met assumptions of normality and heteroscedasticity. All analyses were performed in SAS (2016).

Results

Populations of A. aegypti varied significantly with sampling period (F_6,108 = 2.15, P = 0.050) and neighborhood (F_6,18 = 2.85, P = 0.039) but not their interaction (F_36,108 = 0.86, P = 0.691). Based on contrasts between concurrent sampling periods, there were significantly fewer mosquitoes in March 2018 compared to January and May 2018 with all other dates showing no differences (Fig. 2A). Adults were more abundant in TO vs. PN, EC, RP, VM, and VV, with CA and MP being intermediate (Fig. 2B).

For adult nutrients and isotopes, MANOVA produced significant effects of sampling period (Pillai’s Trace_25,1135 = 2.33, P < 0.001) and neighborhood (Pillai’s Trace_35,1135 = 0.21, P = 0.049), but not their interaction (Pillai’s Trace_175,1135 = 0.60, P = 0.852). Based on SCCs, differences in sampling period were mostly due to %N (SCC = −4.75), %C (5.48), and C:N (−6.28) compared to δ¹⁵N (0.06) and δ¹³C (0.20). Females had the lowest %C in January 2018 and May 2019 compared to October 2019 and January 2019, with other dates intermediate (Fig. 3A). For %N, females had the lowest values in May 2019 and March 2018 compared to October 2018, with other periods intermediate (Fig. 3B). The C:N for females was lowest in January 2018 compared to the last two sampling periods, with other periods intermediate (Fig. 3C).

For the neighborhood effect, differences were mostly due to δ¹³C (SCC = 0.400) and δ¹⁵N (SCC = 0.789) with nutrient values all having SCC values <0.35. Adult female δ¹³C values were generally lowest (most negative and thus least enriched) at MP vs. higher values (more enrichment) in CA, PN, and RP, with other locations intermediate (Fig. 4A). For δ¹⁵N values, the highest enrichment (most positive) was in PN compared to TO, RP, VM, and VV, followed by CA and MP compared to VV (but similar to PN), with other contrasts non-significant (Fig. 4B).

Containers were non-uniformly distributed both across and within neighborhoods and varied with sampling period. Overall, we collected 61 containers representing 13 different larval container types (i.e. bird bath, bucket, can, cooler, cooler lid, cup, fountain, garbage can, plant pot, tire, tray, tub, and vehicle tire). Of these, tires (n = 13) and buckets (n = 13) were the most common. Container abundance varied across neighborhoods, with MP (16), RP (15), and TO (12), having the most, and PN (3), VM (3), and VV (0) having the least. Container volume varied from 10 ml to 72.5 l. Canopy cover (mean value 25.1 ± 1.73) did not affect isotope values (Pillai’s Trace_2,36 = 0.43, P = 0.654), however sampling period (Pillai’s Trace_8,74 = 2.87, P = 0.008) and neighborhood (Pillai’s Trace_12,74 = 2.46, P = 0.009) were significant; their interaction (Pillai’s Trace_24,74 = 1.49, P = 0.100) was not significant. Isotopes values both contributed to the significant sampling period effect (δ¹⁵N SCC = 1.287, δ¹³C SCC = 0.981), however for the neighborhood effect δ¹⁵N (SCC = 0.491) contributed less than δ¹³C (SCC = 1.674).

Across time values of δ¹³C decreased (less enrichment) from January through May 2018, with increased enrichment after those dates (Fig. 5A). An almost opposite pattern emerged for δ¹⁵N across time, with enrichment higher in the later three sampling periods compared to the first (Fig. 5B).

For the neighborhood effect we found variation in δ¹³C, with the lowest enrichment (more negative values) in CA compared to TO, EC, and RP, with other locations having intermediate values (Fig. 6A); δ¹⁵N did not vary across containers across neighborhoods (Fig. 6B).

No mosquitoes were positive for dengue (DENV-2) infection.

Discussion

We set out to answer three primary questions about how mosquitoes may have been affected by the two large storms in urban San Juan, PR: Were there spatial or temporal effects of the hurricanes on the population dynamics of A. aegypti?, Were there spatial or temporal effects on nutrient stoichiometry, nutrient content, and stable isotopes of adult females and larval containers?, and Did container nutrients vary with period and neighborhood and was the nutrient content reflective of adults? Overall, we hypothesized that litter inputs via changes in canopy after the storm would vary across neighborhoods, producing variable effects on larval container environments, and that this would affect adult mosquito numbers and their nutrient content. However, although the effect of the storms on mosquito populations and their nutrient content was often pronounced, the effects of the hurricanes and socioeconomic factors acted separately. The lack of any interactive effects of time since Hurricane Irma and Maria on different SES neighborhoods in our analyses may simply point to the spatial extent of the storms, which caused extensive damage to the canopy across the entire island (Leitold et al. 2022). These hurricanes deposited leaf and woody material equal to or greater than the annual equivalent amount usually deposited in four measured forests on the island (Liu et al. 2018). Similarly, urban yards in San Juan were found to have a 27% reduction of standing stems of vegetation as well as 31% mortality after these storms (Olivero-Lora et al. 2022). However, as we were unable to assess pre-storm canopy data and compare it to post-storm levels, this factor may not have varied at the neighborhood-level.

For our first question we found that populations of female A. aegypti were not affected by the storms, with population sizes being largely consistent across the sampling locations over time. The decline in numbers of adult females in March 2018 is unexplained but it does not appear to be seasonal as May 2019 numbers were consistent with other months within the dry season. One explanation for a lack of response to the hurricanes is that our sampling did not begin until 4 months after Maria, and thus we may have missed short-term effects on mosquito populations. Barrera et al. (2019) found a short-term doubling of female A. aegypti numbers after these hurricanes in Caguas city (south of San Juan), which peaked between October and December 2017, before dropping to lower levels by January. Sampling by Barrera et al. (2019) stopped approximately when our sampling started, so it does appear that we missed the short-term population increase that they describe. This suggests that the effects of large storms may have an ephemeral effect on mosquito populations, however our work identified longer term effects on adult and container nutrient content. We do note that no wide-spread insecticidal spraying was conducted before or after the storm in San Juan, and thus changes to populations of A. aegypti are unlikely to be attributed to organized human control measures.

Populations of A. aegypti were generally higher in lower SES neighborhoods, which is consistent with past studies of mosquito populations on the island (Yee et al. 2019, Scavo et al. 2021) and more generally across urban spaces (Whiteman et al. 2020, Yitbarek et al. 2023). Several reasons exist for why more adult mosquitoes may be found in low SES locations, including those related to education about mosquito sources, lack of control infrastructure, housing abandonment (where yards may facilitate adult mosquito resting and production of larvae due to trash), and high human densities (Scavo et al. 2021). We included some environmental measures into SES, and some (e.g. distance to water bodies) may have also affected adult populations. However, we note that if this is correct, then those environmental factors included in SES were also varying in ways consistent with human factors (e.g. education, income). Although low SES is correlated to higher vector populations in urban areas (e.g. Little et al. 2017), this relationship is not always consistent. Whiteman et al. (2020) found that almost half of all studies investigating relationships between SES and occurrence of medically important Aedes in urban spaces found either a negative or null relationship. This result points to deficiencies in our understanding of how human spaces may influence vector populations. We do note that in most locations through time the number of A. aegypti females likely exceeded those thought to be necessary for transmission of pathogens among humans in Puerto Rico (Barrera et al. 2017, Sharp et al. 2019), although we and others (Barrera et al. 2019) found no evidence of active transmission.

For our second question addressing variation in adult A. aegypti and container particulate nutrient content, we did identify some noteworthy patterns. Although adult %C and %N were variable through time, C:N displayed a strong increasing trend from our first sampling after the storms. Adults had higher C:N a year after we started sampling (16 months since the storms), which may indicate that this was approximately how long nutrients from the storm via litter inputs took to return to pre-storm levels. We collected similar nutrient data from adult mosquitoes in the 2 years prior to 2017 in three of the neighborhoods (MP, RP, TO). In June of 2016 values for C:N were much higher (mean C:N = 5.23 ± 0.08, n = 106) than right after Maria (from the same neighborhoods, mean C:N = 4.56 ± 0.07, n = 183), however in June 2017 they were similar (mean C:N = 4.76 ± 0.06, n = 89). As our sampling effort was lower and these data were taken only in the wet season, it is difficult to know if they are representative of nutrient values of A. aegypti in San Juan prior to Hurricanes Irma and Maria. However, C:N values were lower at the first measurement after the storm compared to 1 year later, indicating the potential for this large disturbance to have affected the nutrient balance within mosquitoes. Over the course of the study we found 61 containers, and this number may reflect our approach to locate them only in freely accessible locations. It is possible that there were more containers in cryptic or inaccessible places (e.g. backyards), however given the nature of how neighborhoods are arranged in San Juan, we do think that the sampled containers are representative of all containers in the surrounding areas. Although these results warrant further exploration, ours is the first study that has examined how nutrients and stoichiometry vary with time in a large urban center.

Comparisons of nutrient content between lab and field studies are rare for mosquitoes but could allow for a better understanding of the forces regulating mosquito populations in nature. For example, even the best lab diets (animal based that are high in limiting N) had shown overall lower values for %N (Yee et al. 2015) than those found here. This suggests that wild diets contain more nitrogen, which has been shown to affect pathogen transmission. Specifically, Paige et al. (2019) grew A. aegypti in the lab under 10 ratios of plant and animal detritus that produced a variety of nitrogen diets. Adults from these experiments were challenged with Zika virus (ZIKV), and females that were grown with less nitrogen showed higher proportion of disseminated infection, whereas those in animal detritus, which had generally high nitrogen, had lower overall infection rates (Paige et al. 2019). The mechanism for this is still unclear, but raises a tantalizing role of larval nutrition, especially nitrogen, for pathogen transmission. Although higher N content leads to lower dissemination of ZIKV and values of body nitrogen above 4% appears to show a decline in dissemination (Paige et al. 2019), all locations sampled in this study through time produced values >9% body nitrogen, so it appears unlikely that the effect of low nitrogen would manifest on pathogen transmission. We do note that no arboviruses (dengue, chikungunya, ZIKV) were detected during the months prior to our work in San Juan (Barrera et al. 2019) or dengue by us. The role of adult nitrogen content and arboviruses is clearly a topic worth exploring more.

As mentioned above, adult populations in lower SES locations and isotope values also followed this pattern. Specifically, the lowest mean value of δ¹³C was in MP (low SES) compared to two middle SES and one low SES neighborhood. However, enrichment of δ¹⁵N was highest in two of the lowest SES locations (CA, MP) and one of the middle SES (PN) compared to the highest (VV); nutrient enrichment generally followed a pattern of lowest enrichment in highest SES. Nitrogen is critical to larval development but is often limiting in containers that receive large inputs of plant material (Carpenter 1983, Kaufman and Walker 2006, Yee and Juliano 2006). A pattern of higher nutrient input, perhaps via flooding events, has been shown in these same neighborhoods (Yee et al. 2019).

For our final question, we wanted to know if the particulate matter nutrient values in larval containers were related to adult nutrients. Here, containers showed a different pattern across neighborhoods than adults. Specifically, whereas adults had high values of δ¹⁵N in low SES locations, there was no trend in δ¹⁵N across containers in these same locations. Neighborhoods did vary in δ¹³C, but patterns were not consistent with differences in SES. This difference among isotope values may suggest a disconnect between input of particulate nutrients (likely from the canopy) and the nutrient content of adults captured near containers, however data on this are lacking. Isotopes in larvae within containers were correlated with adults within neighborhoods in San Juan (Yee et al. 2019), although we could not formally test this relationship here. We also note very low δ¹³C in containers. Previous studies recorded values in containers in San Juan ranged from −20‰ to −28‰ (Yee et al. 2019), which is consistent with lab and field studies that measured δ¹³C of A. aegypti in the −20‰ to −30‰ range (Young et al. 2014, Yee et al. 2015). Thus, the finding that container δ¹³C values were −30‰ to −50‰ was unexpected. We do not have a good explanation for this result, but low values may be related to factors like water residence time or carbon source contributions. We also note that some container particulate matter contributions may have occurred after the two extreme weather events, as compared to some detrital material that was present throughout our sampling. Given the episodic and opportunistic nature of our sampling, we have no way of knowing how container age or longevity may have affected our results, however given that nutrients varied with SES it does appear that some factors (e.g. detrital inputs) did vary.

Nutrient values between adult females and container particulate matter did not seem to be as strongly correlated as they were in previous studies (Yee et al. 2019). Overall, we did not observe the expected trophic offset, especially at moderate SES locations like PN and VM, where container particulate matter δ¹⁵N values were higher (mean = 11.0‰, 12.2‰, respectively) than adult females (mean = 10.6‰, 9.4‰, respectively) (Sterner and Elser 2002). Other locations, especially low SES neighborhoods like CA, MP, and TO, all showed higher δ¹⁵N values for females compared to containers. If females are flying shorter distances in low SES locations (<150 m, e.g. Harrington et al. 2005), due to the presence of more containers, then we might expect a closer correspondence between container and adult nutrient values. There were more containers in low SES locations overall (36 total in MP, CA, and TO) compared to moderate SES (25 total across EC, PN, RP, and VM) with no containers in the sole high SES neighborhood (VV). Our sampling of containers was largely opportunistic, wherein we would locate containers close to the street and often outside of private property. Thus, in moderate and high SES neighborhoods, where the streets were generally freer of trash, containers that produced the adults we captured may not have been included in our sampling. The importance of trash as a production site for adult mosquitoes in urban locations is well established (Little et al. 2017, Yitbarek et al. 2023), however our work suggests that individual containers can have distinct isotope values, particularly in low SES neighborhoods. This could potentially be useful for linking mosquitos back to their food source or production sites but also presents a challenge when trying to scale individual container data to whole neighborhoods or regions. Isotope sources did vary widely at the individual container level, with ranges on the order of 8 ‰ for N and 28 ‰ for C, however variability in the adult females was much lower. Although there were slight, but significant isotope differences in the adult females among sites, these variations were not strictly associated with the containers in the vicinity. The scale at which mosquitoes vary may be influenced by factors that may be relevant at the larval habitat but not at the dispersal range of adult populations.

Large storms like hurricanes are projected to increase in magnitude and frequency over the coming century (Knutson et al. 2020, Romero et al. 2020). How these storms may affect both infrastructure in urban systems and the natural components of the environment is still being debated (Grimm et al. 2017). However, the data for how hurricanes have disrupted vector populations is scant, but it does suggest that storms can have direct consequences for mosquitoes and mosquito-borne disease dynamics. For instance, flooding from Hurricane Katrina had a positive effect on populations of Culex quinquefasciatus (southern house mosquito) (Caillouët et al. 2008a). This increase likely led to the documented doubling in the number of reported cases of West Nile neuroinvasive disease (WNND) in the hurricane-affected regions of Louisiana (Caillouët et al. 2008b). Hurricane Florence, and the ensuing flooding, is thought responsible for up to 75 000 cases of malaria in Haiti in 1963 (Mason and Cavalié 1964). Flooding via hurricanes may enhance mosquito-borne disease via several mechanisms, including producing more mosquito larval production sites, causing vectors and hosts to move into closer contact (due to inundation of dry land for humans and mosquitoes), and changes in vector-host dynamics [reviewed in Hunter (2003)]; flooding may also extend the time when mosquitoes are abundant, and thus increase transmission periods (Tall and Gatton 2020). Heavy rains events are thought responsible for an epidemic of St Louis encephalitis in Florida via stimulation of blood-feeding by Culex nigripalpus (Day and Curtis 1989). We have shown that nutrient content of adults and containers were affected by these large storms, however populations of adult mosquitoes did not vary in any long-term way. We do not have other examples to compare to, but given the nature of hurricanes there is a possibility that any effects of nutrients may be overshadowed by the creation of new larval habitats for mosquitoes caused by damage and distribution of trash in the environment. Datasets like this one will be important so we may assess and understand the effect of storms on vector populations and pathogen transmission. These storms are but another manifestation that global warming is likely to have on mosquito-borne disease outbreaks.

Acknowledgements

We thank C. Dean, N. Fijman, K. Noa, and T. Sanchez for assistance with field sampling. We also thank J. Zimmerman and the staff at El Verde Field Station for logistical support. We are extremely thankful for the assistance of the many helpful residents of the communities where sampling took place. Special thanks to R. Barrera and G. Felix and the Center for Disease Control Dengue Branch in San Juan for loaning us mosquito traps and batteries. This report has been reviewed technically by the US EPA’s Office of Research and Development, Center for Environmental Measurement and Monitoring, Atlantic Coastal Environmental Sciences Division. The views expressed in this article are those of the authors and do not necessarily reflect the views or policies of the US Environmental Protection Agency (EPA). The EPA does not endorse any commercial products, services, or enterprises. Any use of trade, firm or product names is for descriptive purposes only and does not imply endorsement by the US Government.

Author contributions

Donald A. Yee (Conceptualization [lead], Data curation [supporting], Formal analysis [lead], Funding acquisition [lead], Investigation [lead], Methodology [supporting], Project administration [lead], Resources [lead], Supervision [lead], Visualization [supporting], Writing—original draft [lead], Writing—review & editing [equal]), Nicole A. Scavo (Data curation [equal], Investigation [lead], Writing—review & editing [equal]), Limarie J. Reyes-Torres (Data curation [equal], Investigation [equal], Writing—review & editing [equal]), and Autumn Oczkowski (Conceptualization [supporting], Data curation [supporting], Formal analysis [supporting], Methodology [supporting], Writing—review & editing [equal])

Conflict of interest: None declared.

Funding

This research was supported by the National Science Foundation (DEB-1806122 to D.A.Y.).

Data availability

Data will be available via a data repository should the paper be accepted for publication.

References