Wetland characteristics linked to broad-scale patterns in Culiseta melanura abundance and eastern equine encephalitis virus infection

Abstract

Background: Eastern equine encephalitis virus (EEEV) is an expanding mosquito-borne threat to humans and domestic animal populations in the northeastern United States. Outbreaks of EEEV are challenging to predict due to spatial and temporal uncertainty in the abundance and viral infection of Cs. melanura, the principal enzootic vector. EEEV activity may be closely linked to wetlands because they provide essential habitat for mosquito vectors and avian reservoir hosts. However, wetlands are not homogeneous and can vary by vegetation, connectivity, size, and inundation patterns. Wetlands may also have different effects on EEEV transmission depending on the assessed spatial scale. We investigated associations between wetland characteristics and Cs. melanura abundance and infection with EEEV at multiple spatial scales in Connecticut, USA.

Results: Our findings indicate that wetland vegetative characteristics have strong associations with Cs. melanura abundance. Deciduous and evergreen forested wetlands were associated with higher Cs. melanura abundance, likely because these wetlands provide suitable subterranean habitat for Cs. melanura development. In contrast, Cs. melanura abundance was negatively associated with emergent and scrub/shrub wetlands, and wetland connectivity to streams. These relationships were generally strongest at broad spatial scales. Additionally, the relationships between wetland characteristics and EEEV infection in Cs. melanura were generally weak. However, Cs. melanura abundance was strongly associated with EEEV infection, suggesting that wetland-associated changes in abundance may be indirectly linked to EEEV infection in Cs. melanura. Finally, we found that wet hydrological conditions during the transmission season and during the fall/winter preceding the transmission season were associated with higher Cs. melanura abundance and EEEV infection, indicating that wet conditions are favorable for EEEV transmission.

Conclusions: These results expand the broad-scale understanding of the effects of wetlands on EEEV transmission and help to reduce the spatial and temporal uncertainty associated with EEEV outbreaks.

Keywords: Connectivity; Culiseta melanura; Drought; Eastern equine encephalitis virus; Hydrology; Vegetation; Wetlands.

Fig. 1

Study extent, sample locations, vector abundance and vector EEEV infection rate for all sampling years (2001–2014). Black points mark each of the mosquito trapping locations. The size of the blue circle above the black point represents mean Cs. melanura abundance per sampling night at the sampling location and the size of the red circle below the black point reflects the mean EEEV infection rate of Cs. melanura at the sampling location

Fig. 2

Relative importance of spatial scales and temporal lags. Each plot highlights the particular monthly lags in hydrological wetness conditions (PHDI) and the spatial scales (for wetland variables) that are included in the best performing models (determined by AIC). Each black point represents a different model explaining either Cs. melanura abundance (a, b) or EEEV presence/absence (c). Each model has a unique combination of spatial scale for the wetland variable (represented by point size) and monthly lag in hydrological conditions (PHDI; x-axis value). A different wetland variable is included in each plot: (a) forested semi-permanent wetlands; (b) forested wetlands; and (c) the abundance of forested semi-permanent wetland relative to other wetland types. The y-axis lists AIC scores for each model centered on the mean AIC score of all the models. A lower centered AIC score for a model suggests better performance for that spatial scale and monthly lag combination. The background color shows the interpolated relative importance of a particular monthly lag averaged across all the models included in the plot. Red bands indicate monthly lags when hydrological conditions have the highest relative importance

Fig. 3

Generalized additive model (GAM) response curves depicting the relationship between mean Cs. melanura abundance per sampling night and 6 explanatory variables: (a) deciduous forested wetland; (b) evergreen forested wetland; (c) emergent wetland; (d) scrub/shrub wetland; (e) stream connectivity; and (f) impervious surfaces (Table 2). The percent of the total model deviance explained by each variable (%dev.) and associated P-values are listed in the upper right of each plot. Grey bands represent 95% confidence intervals (1.96*SE) on the estimated Cs. melanura abundance based on GAM predictions

Fig. 4

Contour plots showing the effects of interactions between the proportional area of forested semi-permanent wetland (a) or all types of forested wetlands (b, c), and hydrological conditions (PHDI) on Cs. melanura abundance. (a) and (b) show this relationship during the transmission season (0–1 month lag) and (c) depicts the previous fall/winter (8 month lag). Blue represents low vector abundance and yellow represents high abundance. The range of abundance values is listed in the upper right of each plot

Fig. 5

Generalized additive model (GAM) response curves depicting the relationship between the log odds of an EEEV positive Cs. melanura pool and 7 explanatory variables: (a) deciduous forested wetland; (b) evergreen forested wetland; (c) emergent wetland; (d) scrub/shrub wetland; (e) forested wetland size; (f) impervious surfaces; and (g) mean Cs. melanura abundance (Table 3). The percent of the total model deviance explained by each variable (%dev.) and associated P-values are listed in the upper right of each plot. Grey bands represent 95% confidence intervals (1.96*SE) on the estimated log odds of EEEV presence based on GAM predictions

Fig. 6

Contour plots showing the effects of interactions between the relative area of semi-permanent wetland and hydrological conditions (PHDI) on the log odds of an EEEV positive Cs. melanura pool. (a) shows this relationship during the previous fall/winter (10 month lag) and (b) depicts the relationship during the transmission season (0 month lag). Blue represents low log odds of EEEV presence, yellow represents high log odds of EEEV presence, and the value range is listed in the upper right of both plots