Elsevier

Ecological Modelling

Volume 272, 24 January 2014, Pages 301-310
Ecological Modelling

Environmental forcing and density-dependent controls of Culex pipiens abundance in a temperate climate (Northeastern Italy)

https://doi.org/10.1016/j.ecolmodel.2013.10.019Get rights and content

Highlights

  • Both endogenous and exogenous controls are important in Culex pipiens population dynamics.

  • Culex pipiens abundance displays clear evidence of density dependence at the weekly time scale.

  • Carrying capacity and maximum per capita growth rate change spatially with land use.

  • Models driven by soil moisture and rainfall result in similar predictive skills.

Abstract

New and old mosquito-borne diseases have emerged and re-emerged in temperate regions over the recent past, but a mechanistic understanding of mosquito population dynamics, a fundamental step toward disease control, remains elusive. We propose here a Gompertz-based approach to address two obstacles to the development of vector dynamics models in temperate regions: (i) the inclusion of endogenous processes (e.g. density limitation, delayed responses, etc.) and the evaluation of their relative importance vs. exogenous environmental forcings; (ii) the inclusion of realistic descriptions of hydrologic processes and the evaluation of soil moisture as a more direct driver of mosquito population dynamics. The new model is based on a hierarchical state-space structure and is applied to the description of the abundance of Culex pipiens – a West Nile Virus vector – in the Po River Delta region (Northeastern Italy), using weekly mosquito abundance observations at more than 20 sites in the period May–September in 2010 and 2011. The hierarchical structure provides an efficient way of fully exploiting the information from a large network of observation sites. We find that Cx. pipiens abundance has significant density dependence at the one-week scale, which is coherent with its larval developmental time during the summer. This result points to the importance of endogenous population dynamics, most often neglected in mosquito population models, usually simply driven by exogenous environmental forcings. Among exogenous controls, temperature, daylight hours, and soil moisture were found to be most influential. Use of precipitation or soil moisture to force the model leads to very similar predictive skills. The negative correlation of soil moisture and mosquito population may be attributed to the abundance of water in the region (e.g. due to irrigation) and the preference for eutrophic habitats by Cx. pipiens. Variations among sites were highly correlated with land-use factors. The carrying capacity is seen to decrease with the distance to the nearest rice field, while the maximum population growth rate was positively related with the Normalized Difference Vegetation Index, a proxy of vegetation cover. The model shows a satisfactory performance in explaining the variation of mosquito abundance over a horizon of 1 week, particularly as far as peak timing and magnitude are concerned. Large rates of change of population abundance remain difficult to predict, as in other existing models, pointing to persisting gaps in our understanding of the mechanisms regulating mosquito population dynamics.

Introduction

Emerging and re-emerging mosquito-borne diseases in temperate regions have become progressively more common in recent years as ever increasing human and vector mobility creates conditions which are favorable for their diffusion (Enserink, 2008, Randolph and Rogers, 2010). This is a source of concern, particularly in areas with high human population density characterized by low immunity and uncertain vector-host relationships (Hayes et al., 2005, Enserink, 2008, Lafferty, 2009, Randolph and Rogers, 2010). Well-known and relevant examples include the West Nile Virus (WNV) diffusion in the United States and Europe, mostly associated with the Culex genus (Hayes et al., 2005), and the increasingly frequent and numerous Chikungunya and Dengue fever cases in Europe, spread by Aedes albopictus (Enserink, 2008).

A fundamental step toward understanding and predicting mosquito-borne disease transmission in temperate climates requires the development of a mechanistic modeling framework of vector population dynamics (Stenseth et al., 2002, Jones et al., 2008). However, work on modeling mosquito population dynamics in temperate climates remains relatively limited. It is often based on empirical regressions of mosquito population on environmental drivers, and usually assumes a fast response of population abundance to meteorological forcings. This assumption is likely to hold in environments where there is a clear-cut, on/off, water control on the mosquito life cycle, such as in areas where rainfall is strongly seasonal. On the contrary, temperate areas are characterized by sometimes subtle links between mosquito population dynamics and highly heterogeneous environmental drivers (e.g. due to a lack of sharp transitions between dry/wet rainfall regimes). Temperate regions also usually exhibit a high human density and degree of anthropogenic control on water availability (e.g. irrigation, diffuse water impoundments, etc.). As a result, reliable models of mosquito abundance remain elusive in temperate areas (Shaman et al., 2006, Linard et al., 2009). In this context, it is important to provide a realistic representation of water dynamics in the soil and at its surface, jointly controlled by several hydrologic processes, such as infiltration, evapotranspiration and runoff generation. However, most approaches to modeling water controls on mosquito population dynamics simply use rainfall as a direct driver (e.g. Yang et al., 2008b, Chaves et al., 2012) and relatively few contributions use explicit hydrological models (Shaman et al., 2006, Bomblies et al., 2008, Day and Shaman, 2008, Montosi et al., 2012) and/or satellite remote sensing (Chuang et al., 2012, Midekisa et al., 2012) to drive mosquito population models.

The importance of endogenous controls, such as density dependence and delayed population responses (Fowler, 1981, Turchin, 1990, Sibly et al., 2005) should also not be underestimated. Only recently they have been recognized as important regulatory mechanisms of mosquito population and they have been rarely incorporated in modeling formulations (Yang et al., 2008a, Yang et al., 2008b, Chaves et al., 2012). We hypothesize that endogenous controls may be particularly important in humid temperate climates, where water is less often limiting and the interaction of multiple environmental controls may result in a wider distribution of adult emergence times with respect to those occurring under more homogeneous environmental conditions (e.g. in tropical climates).

From a population dynamical perspective, previous works on mosquito modeling typically made use of either detailed entomological approaches or of empirical approaches (LaDeau et al., 2011). The first approach uses detailed descriptions of the mosquito life cycle and consequently involves a relatively large number of parameters (Bomblies et al., 2008, Linard et al., 2009, Cailly et al., 2012). These approaches are effective in the presence of a large amount of data to constrain model parameters and when controlling factors are known in detail. But very often this is not the case, because of sampling limitations (of both mosquito abundance and atmospheric forcings) and of the difficulty of transferring to field conditions parameter values obtained from lab experiments. Furthermore, noise is always present both in measurements and in population-dynamic processes, but detailed formulations of mosquito life cycles do not typically allow for stochastic components, thus presenting a quite fundamental mismatch between model assumptions and actual population behavior.

Empirical models, at the other end of the spectrum, may be able to explain a large proportion of the observed variability by simply using meteorological and habitat information. However, forecasting skills are typically poor beyond the conditions under which the model is calibrated, because of the intrinsic lack of a foundation in the governing physical mechanisms (LaDeau et al., 2011). Furthermore, empirical models do not allow insights into the relative importance of controlling mechanisms because such mechanisms are not explicitly resolved in the model structure. A balance is thus desirable between a very detailed model, posing fundamental problems of parameter identification, and completely empirical models, lacking a robust process basis.

Based on the above considerations, the overall goals of the present work are: (i) explore the relative importance of endogenous (e.g. density dependence) vs. exogenous (meteorological and hydrological forcing) factors. We explore in particular the relationship between per capita growth rate and population abundance, the magnitude of the delays with which population responses become manifested, and how they are influenced by environmental factors; (ii) evaluate if and how the use of a full-fledged hydrological model of infiltration, percolation, and runoff to estimate soil moisture affects the model ability to explain the variability in mosquito population abundance in a temperate climate; (iii) develop a sound hierarchical state-space modeling approach (Clark, 2007, Gelman and Hill, 2007) to optimally exploit spatially-distributed observations and minimize the impact of mosquito population sampling schemes which do not fully represent all scales of spatial and temporal heterogeneity.

Section snippets

Study area and data collection

Western Europe in general is identified as a hotspot for the emergence and re-emergence of vector-borne diseases (Jones et al., 2008), spread by 96 endemic mosquito species. Of these species, 62 were found in Italy in 1999 (Snow and Ramsdale, 1999), and this number has probably increased significantly since then. The Po river delta in Northeastern Italy, where our study site is located (Fig. 1), is in particular characterized by a high mosquito abundance, due to favorable climatic and

Results

In general, mosquito abundance in both years started from about zero in May, reached a peak in late June-early July and then gradually decreased in August and September (Fig. 3). The variation among sites was very significant. The number of captured Cx. pipiens in a given night ranged between 0 and 4000. However, consistence was observed across the two years: sites displaying high abundance maintained this characteristic in both years.

The model performance was first evaluated in terms of

Discussion and conclusions

Our analysis of the data through a new model of mosquito population dynamics shows that endogenous and exogenous factors are both important regulators of mosquito abundance and that they interact in a complex way. Density dependence was observed under laboratory conditions for Cx. pipiens larvae: higher larvae density led to larger mortality of the immature stages and longer development times (Vinogradova, 2000). However, the relevance of density dependence mechanisms in the field had not been

Acknowledgements

We acknowledge support by the Nicholas School of Environment (graduate support to YJ, and funding to MM and SS.) and by the “Progetto Pilota di Lotta alle Zanzare nella provincia di Rovigo” funded by Regione Veneto and by Parco Delta del Po. Valuable comments were provided by Dr. William Pan, Dr. Song Qian and Dr. Massimiliano Ignaccolo. We thank Dr. Mario Putti, for making available the Cathy model code. We thank Consorzio di Bonifica del Delta del Po, and in particular Dott. Matteo Bozzolan,

References (54)

  • G.E.P. Box et al.

    Time Series Analysis: Forecasting and Control

    (2008)
  • M. Camporese et al.

    Surface-subsurface flow modeling with path-based runoff routing, boundary condition-based coupling, and assimilation of multisource observation data

    Water Resources Research

    (2010)
  • L.F. Chaves et al.

    Nonlinear impacts of climatic variability on the density-dependent regulation of an insect vector of disease

    Global Change Biology

    (2012)
  • J.S. Clark

    Models for Ecological Data: An Introduction

    (2007)
  • J.S. Clark et al.

    Population time series: process variability, observation errors, missing values, lags, and hidden states

    Ecology

    (2004)
  • A.N. Clements

    The Biology of Mosquitoes

    (1992)
  • W.J. Crans

    Culex pipiens Linnaeus

    (2013)
  • J.F. Day et al.

    Using hydrologic conditions to forecast the risk of focal and epidemic arboviral transmission in peninsular Florida

    Journal of Medical Entomology

    (2008)
  • B. de Meillon et al.

    Time of arrival of gravid Culex pipiens fatigans at an oviposition site, the oviposition cycle and the relationship between time of feeding and time of oviposition

    Bulletin of the World Health Organization

    (1967)
  • D.J. Dohm et al.

    Effect of environmental temperature on the ability of Culex pipiens (Diptera: Culicidae) to transmit West Nile virus

    Journal of Medical Entomology

    (2002)
  • M. Enserink

    A mosquito goes global

    Science

    (2008)
  • V. Ermert et al.

    Development of a new version of the Liverpool Malaria Model. I. Refining the parameter settings and mathematical formulation of basic processes based on a literature review

    Malaria Journal

    (2011)
  • C.W. Fowler

    Density dependence as related to life-history strategy

    Ecology

    (1981)
  • A. Gelman et al.

    Data Analysis Using Regression and Multilevel/Hierarchical Models

    (2007)
  • G.L. Hamer et al.

    Culex pipiens (Diptera: Culicidae): a bridge vector of West Nile virus to humans

    Journal of Medical Entomology

    (2008)
  • E.B. Hayes et al.

    Epidemiology and transmission dynamics of West Nile Virus disease

    Emerging Infectious Diseases

    (2005)
  • IPCC

    Climate Change 2007: Synthesis Report Intergovernmental Panel on Climate Change

    (2007)
  • Cited by (23)

    • Environmental drivers, climate change and emergent diseases transmitted by mosquitoes and their vectors in southern Europe: A systematic review

      2020, Environmental Research
      Citation Excerpt :

      Nine studies used the normalized difference vegetation index (NDVI), a variable related to vegetation presence and photosynthetic activity (Bisanzio et al., 2011; Calzolari et al., 2015; Conte et al., 2015b; Jian et al., 2014; Marcantonio et al., 2015; Rosa et al., 2014; Semenza et al., 2016; Tran et al., 2014; Valiakos et al., 2014). Three studies found a positive relationship between NDVI and vector density (Bisanzio et al., 2011; Calzolari et al., 2015; Jian et al., 2014). Three of the 7 studies that made predictions used it as a key predictor (Bisanzio et al., 2011; Calzolari et al., 2015; Conte et al., 2015b).

    View all citing articles on Scopus
    View full text