West Nile virus (WNV) was first reported in 1999 in New York City, NY, USA. By 2000 the disease has spread throughout the northeastern USA [1
]. The virus reached Louisiana in the fall of 2001, when a dead crow in Jefferson Parish was identified as being infected with WNV [4
]. By 2003, WNV infections occurred in 60 of the 64 Louisiana’s parishes. In the New Orleans metropolitan areas (Orleans and Jefferson Parishes) focal transmission activity occurs principally during mid-July [4
Say, as the main vector, with Cx. salinarius
Coquillett possibly acting as a secondary vector, were incriminated in the WNV outbreak in southern Louisiana during 2002 [5
]. The former mosquito species, with a feeding preference for mammals, was responsible for enzootic/epidemic transmission, especially in urban and sub-urban settings [7
]. The primary mosquito vector showed biological and ecological resilience in space and time based on the available environmental resources. This resilience may influence the spatio-temporal distribution of the WNV vector, which may or may not bring them to the vicinity of both reservoir host(s) and human populations. Eventually this will affect the amplification and transmission cycles of WNV in areas under risk.
In New Orleans, the confluence of availability of competent mosquito vector(s), susceptible reservoir host(s), suitable natural systems and climate for both mosquitoes and host(s) enabled the autochthonous transmission of WNV with hundreds of human cases and major mortality of wild native and exotic birds [2
]. Nonetheless, the transmission dynamics of WNV in terms of space and time in relationship to the biology, ecology of mosquito vector(s), and their biophysical systems remains unclear. In fact, the distribution, blood-feeding preference, flight range and vectorial capacity of mosquito vectors are very critical inputs for predicting the transmission cycle of this disease.
Furthermore, mosquito vectors often shift their feeding preference seasonally or spatially, depending on the availability of the blood meal source. For example, Cx. quinquefasciatus
showed an opportunistic preference for blood meal. In peninsular Florida, it is responsible for an epizootic cycle and sustaining the virus circulation within reservoir host bird(s) [15
]. However, it has been incriminated with the enzootic/epidemic transmission cycle of WNV in urban and sub-urban areas in Louisiana due to feeding preference to humans and other mammals [4
Currently, most species distribution models for mosquitoes are based on hydrological and meteorological data [15
]. Some models include socio-environmental predictors in terms of vegetation or urban and sub-urban areas [23
]. With respect to WNV, models have used either used data points of WNV cases and mosquito vectors instead of the flight range of the mosquito vectors around their hosts or predicted the distribution risk of WNV on regional scale. Prediction models for WNV and Zika virus (ZIKV) transmission potential were generated for their mosquito vectors in regard to their flight range around recorded positive cases highlighting their response to surrounding biophysical systems such as climate and non-climate factors [24
]. Although previous models are useful, their findings did not adequately account for the comprehensive response of vector-host contact (VHC) ratios to climate and non-climate variables such as land use-land cover (LULC) and Digital Elevation Models (DEM) and the overall influence on arbovirus transmission potential [26
]. Mosquito density reflect neither the likelihood of biting risk, which is caused by mosquito vector, nor the transmission potential as a function of biting rate. The VHC explains the ratio between collected mosquito density and human population census, which reflects areas under risk of increased biting rate by mosquito vector.
The lack of available vaccines for WNV and consistent development of insecticide resistance for mosquito vector populations jeopardize public health in affected areas. Additionally, the focal and sporadic locally-transmitted cases justify the necessity to generate prediction models that identifies areas under risk of infective biting rates in order to target during surveillance and control activities. In our model, we evaluated the spatio-temporal distribution of VHC ratios in response to: (i) future climate scenarios during 2011–2030, (ii) LULC, (iii) socioeconomic, and (iv) DEM systems. Our correlative models were generated within the flight range of WNV vector in the city of New Orleans, LA (NOLA). The spatio-temporal VHC ratios were estimated utilizing data records on female gravid mosquito and human population census per block during 2015. The spatio-temporal resilience of VHC ratios to their predicted biophysical systems was characterized. This allowed developing prediction risk maps for the WNV vector presence using the Maximum Entropy (MaxEnt) tool, emphasizing the human population under risk of infective mosquito bites. Since the local economy in NOLA is primarily driven by tourism, management of arbovirus diseases has a significant economic implications. Arbovirus transmission has the potential to jeopardize the tourism industry, making NOLA surveillance and control programs very important to the economic and ecological health of the city.
In our model, we generated habitat suitability estimates for the spatio-temporal likelihood of VHC ratios in NOLA. Since the relationship between mosquito population density and human hosts is important in determining the infective biting rate and transmission risk of arboviruses, density of Cx. quinquefasciatus
was linked to the human population census in order to generate risk maps for areas under risk of increased VHC. Additionally, we predicted the habitat suitability for the likelihood of VHC within flight ranges of Cx. quinquefasciatus
(~5-km) around their sampling sites [29
]. The significant explanatory variables were evaluated and selected using minimum AICc values using stepwise RM. The potentiality of the current prediction model was proven by the high AUC and R2
values produced by MaxEnt and RM. These thresholds indicate that occurrence records were likely assigned a higher probability of presence than background sites. Additionally, the generated risk map was validated using 18 independent field collected sampling points and tested for mosquito WNV infection rates during the season.
The human population in NOLA is centralized in the western areas of the city. However, the likelihood of VHC ratios demonstrated heterogeneous distributions in this side (Figure 4
). Although the monthly predicting variables showed some variations, especially the climate, in terms of their percent contribution, the number of these variables declined gradually toward the end of the season. However, the changes in these predicting variables had a consistent influence on the distribution of VHC likelihood of the same high risk areas in the west side of the city. This may reflect the temporal resilience of this mosquito vector to their predicting climate variables in these habitats. This resilience gives the WNV vector the ability to develop and survive in close vicinity to WNV reservoir bird host(s), which was confirmed by the positive WNV mosquito pools during our study [29
]. However, multi-year mosquito data are recommended to be included in further investigations.
During June and early August 2002, WNV was identified in pools of Cx
mosquitoes in southeastern Louisiana with the possibility of Cx
acting as a secondary vector [6
]. Although other mosquito vectors were incriminated in amplification and transmission potentials, the selective feeding preference on both human and avian blood and vectorial capacity experiments emphasized that Cx. quinquefasciatus
is the competent vector in transmitting WNV in LA. The Cx. quinquefasciatus
mosquito is well known as exophilic and exophagic and the breeding habitats range from ditches, woodland pools, and freshwater marshes of a semi-permanent or permanent nature [25
]. As much as Cx. quinquefasciatus
maintains and amplifies WNV within reservoir host bird(s) [15
], it is responsible for the urban transmission cycle of WNV in southern and southeastern parts of the USA [17
In the spatial analysis, the RM demonstrated a significant association between NFWL and VHC ratios (R2
= 82, p
< 0.01). The NFWL habitats were dominant in the eastern side of the city with significantly low human population census. Although this side has not been extensively sampled (No. traps = 1), it is worthy to be highlighted in further investigations to understand the temporal association between WNV vector and reservoir host(s). Other LULC related variables such as TD, OUBL and RU showed a reduced contribution in predicting the likelihood of VHC. These LULC habitats provide both sugar and blood meals, and are favorable to WNV maintenance by enhancing maintenance and amplification phases between mosquito vector and their nesting/roosting reservoir bird hosts, especially the passerines [17
]. This finding was confirmed by the selective feeding preference of this mosquito vector in NOLA, their contribution in both enzootic and epidemic transmission cycle of WNV [5
], and the extended transmission season due to the milder climatic conditions of the Gulf Coast as manifested in mosquito WNV infection rates (Figure 3
). Similar reports of increases in mosquito WNV infection rates have been made in 2005 and 2006 in Chicago [58
]. Moreover, both the Jackknife test in MaxEnt (28.9%) and temporal analysis emphasized the association between LULC types during amplification and early transmission phases (May–August) (Table 3
and Table 5
Although no WNV human cases were included in our model, the transmission potential still exists due to the high VHC ratios during September and high suitable habitats. In addition, the lack of data on vector competence, survival rate, and the gonotrophic cycle period for NOLA prevented inclusion of this information in our model. However, the high risk probability was generated utilizing WNV minimum infection rates (0.8/1000) [5
The climate variables shared reduced contributions in predicting likelihood of VHC in the spatial analysis [55
]. However, climate was demonstrated as the key predicting factor in determining temporal distribution of WNV vectors. The varied correlations between temperature-related variables and VHC explain the association of WNV mosquito distribution with warmer areas during the coldest months. This was confirmed by the negative correlation between VHC and Bio6, which represents the minimum temp of the coldest months. Similarly, Bio6 predicts timing and distribution of the nesting/roosting bird hosts during early spring. In a field study, severe winter caused delay in birds nesting, which explains the variation in timing of host feeding shifts. The VHC showed a negative correlation with precipitation during the driest (Bio17) and warmest (Bio18) quarters [25
]. This negative correlation may be attributed to the flush of limited numbers of mosquitoes from breeding habitats due to rainfall during dry and wet seasons. Subsequently, this reduced the abundance/distribution of WNV mosquito vector during these periods.
In the temporal analysis, we attempted to characterize the ecological resilience of WNV mosquito vector in response to seasonal temperature and precipitation related variables, in association with other LULC variables. This resilience was demonstrated by a decrease in the number of variables that predicted VHC ratios during the late season. The lagged influence of mean temperature of the coldest quarter (Bio11) had a positive association with the increased development/distribution of WNV vector during April–September [25
]. This was manifested as maximum prediction probability of VHC ratios during September (R2
= 58.49, p
< 0.01). This lagged influence may cause the increase in the development rate of mosquito vectors in affected areas during the amplification phase of the WNV pathogen during April–May. Meanwhile, neotropical bird migrants, mainly passerines, tend to nest during April–May or roost through July, which is crucial for the amplification phase [55
]. Accordingly, this may accelerate the disease transmission during June–September. Moreover, blood meal preference may shift from birds to mammals including humans that are temporally associated with distribution of nesting/roosting birds, thereby enhancing human risk of arboviruses [55
]. This shift in mosquito feeding preference may be influenced by the temperature-related variables during dry and wet seasons, i.e., negative correlations between VHC and temperature-related variables during dry and wet seasons. The increase in these temperatures reflects the reduction in water bodies that may allow breeding and nesting habitats for mosquito vectors and birds, respectively. Similarly, the negative influence of precipitation variables during April–September, especially annual precipitation (Bio12) and precipitation of the driest month (Bio14), shared a reduced influence toward the prediction of distributions of Cx. quinquefasciatus.
This may help stimulate blood feeding during these times. The seasonal precipitation (Bio15) during April–August increased both available water habitats and distribution of WNV vector.
Our findings showed heterogeneity in the temporal distribution of Cx. quinquefasciatus
in response to LULC during May–August. This reflects the influence of some LULC classes on monthly flux and distribution patterns of mosquito populations. The heterogeneity in monthly distributions of mosquito vectors can have a large impact on virus amplification during the early season, especially when it is in close vicinity to reservoir bird hosts. Although no WNV positive mosquito pools were reported during April and May, NFWL was positively correlated with VHC during May (R2
= 72.6). This may reflect the association between nesting/roosting reservoir bird hosts and VHC ratios in May. During June–August, WNV seemed to build inside mosquito bodies to the detectable level that can cause potential transmission. Both OUBL and RU settings were associated with the onset of mosquito WNV infection rates during June. Other models reported that WNV transmission was either associated with forested and urban land [18
] or socioeconomic status [61
]. However, these correlations may vary spatially or temporally [62
Some previous studies improved our understanding about biology and ecology of Cx. quinquefasciatus
and epidemiology [16
]. Nevertheless, their findings did not highlight the interaction between different systems and their overall influence on mosquito vector distribution at a local scale. Moreover, these models predicted the geographic distribution of WNV vectors utilizing mosquito density as sampling points rather than the flight range area and vector-host contact ratios.
Land geomorphology and topography were used in predicting suitable habitats for mosquito breeding [25
]. Four potential indicators were rigorously investigated: aspect ratio, slope, land surface curvature and hill shade [25
]. The sampled gravid mosquito vectors reflect the proximity of water habitats to collected samples. Although these indicators showed potentiality in predicting WNV mosquito vectors in other areas [25
], only relative high altitude demonstrated temporal influence on increased VHC ratios during May and June.
In the current study we modeled the spatio-temporal distribution of VHC ratios in response to future climate scenario, LULC, human population census, and DEM. Vector-host contact (VHC) ratios were estimated as a potential entomological indicator for the likelihood of biting rate and transmission potential of WNV. The VHC ratios were estimated within 5-km buffer zones around mosquito sampling sites representing their average flight range utilizing mosquito density and population census/house block. The likelihood of VHC ratio was first predicted in response to the biophysical systems using stepwise multiple regression model (RM). Accordingly, we used the significant predicting variables from RM to highlight the spatio-temporal distribution of areas under risk of increased VHC emphasizing the likelihood of infective mosquito bites.
The interaction between these different biophysical systems showed heterogeneous influences on the spatio-temporal distribution of VHC ratios. In the spatial analysis, 12 variables were associated with the distribution of VHC ratios, and NFWL showed the highest prediction gain (R2 = 81.62). Although NFWL has not been sampled extensively during our study (1), to complete our objective, this variable needs to be highlighted rigorously in a separate investigation. Seasonal precipitation- and annual temperature-related variables shared reduced significant associations with VHC ratios. The average likelihood of predicting very high risk areas was ~107 km2, which is ~9.87% of the total area of NOLA. Although neither the temporal distribution of VHC ratios nor their estimates were significantly changed from month-to-month (except during September), their monthly response showed resilience to the number and type of the influential factors. The highest VHC ratio was reported during September, which was associated with the peak positive WNV mosquito pools. Seasonal temperature-related variables showed the highest influence on monthly likelihood of VHC in comparison with seasonal precipitation, LULC and DEM variables. This finding was confirmed by both RM (R2 = 58.49) and Jackknife’s test (88.5%). The influence of LULC on likelihood of VHC was demonstrated during May–August. The NFWL showed a positive association with the increased VHC ratio during May, with no positive WNV mosquito pools. Meanwhile, both OUBL and RU settings were associated with the onset of mosquito WNV infection rates during June. This may reflect the distribution of WNV vector in close vicinity to reservoir host(s) during May, during the virus amplification phase, and the virus beginning to build up inside mosquito bodies during June–August. During September–December, reduced numbers of positive WNV mosquito pools were recorded, which may explain the reduced viremia in wild Cx. quinquefasciatus.
The independent field collected sampling points were consistent with both likelihood of VHC ratios and spatio-temporal distribution of increased VHC ratios. However, multi-year mosquito, spatial projections of LULC and human population census data are recommended to be included in further investigations. Moreover, due to data limitation of reservoir host(s), human cases, mosquito vector survival rates, vector capacity parameters, and gonotrophic periods, we did not have the chance to include these variables in our study.