Elsevier

Ecological Modelling

Volume 456, 15 September 2021, 109658
Ecological Modelling

Modelling arthropod active dispersal using Partial differential equations: the case of the mosquito Aedes albopictus

https://doi.org/10.1016/j.ecolmodel.2021.109658Get rights and content

Highlights

  • We propose a novel modelling framework to estimate insect dispersal.

  • We consider time, space and insect survival in one single mathematical equation.

  • Insect flight range can be computed as function of the time.

Abstract

Dispersal is an important driver for animal population dynamics. Insect dispersal is conventionally assessed by Mark-Release-Recapture (MRR) experiments, whose results are usually analyzed by regression or Bayesian approaches which do not incorporate relevant parameters affecting this behavior, such as time dependence and mortality. Here we present an advanced mathematical-statistical method based on partial differential equations (PDEs) to predict dispersal based on MRR data, taking into consideration time, space, and daily mortality. As a case study, the model is applied to estimate the dispersal of the mosquito vector Aedes albopictus using data from three field MRR experiments. We used a two-dimensional PDE heat equation, a normal bivariate distribution, where we incorporated the survival and capture processes. We developed a stochastic model by specifying a likelihood function, with Poisson distribution, to calibrate the model free parameters, including the diffusion coefficient. We then computed quantities of interested as function of space and time, such as the area travelled in unit time. Results show that the PDE approach allowed to compute time dependent measurement of dispersal. In the case study, the model well reproduces the observed recapture process as 86%, 78% and 84% of the experimental observations lie within the 95% CI of the model predictions in the three releases, respectively. The estimated mean values diffusion coefficient are 1,800 (95% CI: 1,704–1 896), 960 (95% CI: 912- 1 128), 552 (95% CI 432–1 080) m2/day for MRR1, MRR2 and MRR3, respectively. The incorporation of time, space, and daily mortality in a single equation provides a more realistic representation of the dispersal process than conventional Bayesian methods and can be easily adapted to estimate the dispersal of insect species of public health and economic relevance. A more realistic prediction of vector species movement will improve the modelling of diseases spread and the effectiveness of control strategies against vectors and pests.

Introduction

Animal dispersal refers to movements away from the place of birth towards another location for reproduction. The main drivers of dispersal are related to the avoidance of kin competition and inbreeding and escaping deteriorating environmental conditions (Bowler and Benton, 2005). In the case of insects, assessing the active dispersal range might be of crucial importance particularly in the case of species which damage agricultural productions, unsettle ecosystems, and threat human health. Among more than 6 million species of insects known, only less than 100 are either important pests for major crops or relevant vectors of human and/or animal diseases. The deep knowledge of these species dispersal is instrumental to develop proper integrated pest management plans which maximize cost-effectiveness of interventions and protect the environment in a sustainable way.

Although many theoretical models are available, empirical studies are generally lacking due to the difficulties of linking observations to the quantification of dispersion (Tesson and Edelaar, 2013). Advanced satellite radio telemetry and acoustic are giving new opportunities to study dispersion of big size animals such as large mammals or sharks (Cagnacci et al., 2010; Spaet et al., 2020), but are less useful to quantify the dispersal of small animals. In the case of insects, Mark-release-recapture (MRR) is the most widely used technique for quantifying dispersal: specimens are first collected/reared and marked, then released from a single site and subsequently recaptured through traps placed at different distances in a given study area (Pollock et al., 1990).

The most common statistical approaches used to estimate dispersal from MRR data are regression techniques which aim to estimate the mean distance travelled (MDT) and the flight range (FRx) of a specified fraction x of the population, rather than the diffusion process of individual marked specimens. Other methods to estimate insect dispersal follow a Bayesian framework which explicitly models the diffusion process (Villela et al., 2015). For instance, the hierarchical Bayesian model proposed by Villela et al. (2015) for the mosquito Aedes aegypti takes advantage from the flexibility of the Bayesian approach and expands the frequentist approach by including three components: two probabilistic models, describing the spatial distribution of specimens and the daily survival of marked and native individuals, and an observation model describing the sampling process.

Partial Differential Equations (PDEs) represent a standard mathematical method to model diffusion processes, such as the gas dynamics and heat distribution (Borthwick et al., 2016). In ecology, PDEs have been applied to study spatial-temporal dispersal of animal populations in a continuous domain (Bassett et al., 2017; Kareiva et al., 1990), such as the home-range dynamics of meerkats (Suricata suricatta) (Bateman et al., 2015) and the dispersal of butterflies (Ovaskainen, 2004).

The aim of this work is to provide a PDE-based analytical method to estimate insect dispersion based on MRR field data. This method, compared to previously quoted ones, allows to estimate the dispersion taking into account the daily mortality of marked release insects and the days after release (time) in a single mathematical equation. In particular, we applied the proposed modeling framework to estimate the dispersal of blood-fed females of the tiger mosquito, Aedes albopictus, during the egg laying phase. This species represents a significant public health burden due to its capacity to transmit exotic arboviruses, such as dengue (DENV) and chikungunya (CHIKV), capable of induce serious diseases in humans (Zeller et al., 2016). The species was the primary responsible of the thousands of DENV cases recorded in the southwest of the Indian Ocean in 2015–2018 (Vincent et al., 2019) and of the first autochthonous cases of both viruses in Europe (Marrama Rakotoarivony and Schaffner, 2012), where it caused two large CHIKV outbreaks with hundreds of human cases (Italy 2007 and 2017; Caputo et al., 2020).

We expect that the proposed approach can be applied to quantify dispersal, and hence improve control of diseases transmitted by Ae. albopictus and by other mosquito species of global relevance (such as the major arbovirus vector, Ae. aegypti, and malaria vector species), as well as of agricultural pest species.

Section snippets

Modelling

Our model expands the one proposed in (Lutambi et al., 2013). Precisely, we included in the main equation the mortality and capture processes of the species of interest. Thus, the equation assumes the following form:Mt=D(2Mx2+2My2)M(μ+β)Where μ is the mortality rate, β is capture rate, (x, y) represent location coordinates as distance (in meters) along the x and y spatial axis from a given origin (x0, y0) respectively, t is the time (i.e., days or hours), M(x,y,t) is the density of the

Results

In this section we present: (i) the estimated values for the diffusion coefficients (D1, D2, D3 for MRR1, MRR2 and MRR3 respectively), the correction factor (ζ) and the daily mortality rate (μ1,μ2,μ3 for the first, second and third semi-field experiments); (ii) the simulation of the dispersal process of the marked mosquitoes during five consecutive days after release; (iii) a validation of the mathematical-statistical model proposed here using the analytical solution of equation [2].

Discussion

Here we presented a PDE-based stochastic framework to estimate insect dispersal based on MRR data. We tested it in a specific case-study to overcome the limitations of the commonly used analytical approaches, i.e. regression analysis (Marini et al., 2019) and hierarchical Bayesian models (Villela et al., 2015). Indeed, the big advantage in the use of PDEs is the possibility of including time, space, and daily mortality in only one mathematical equation, thus providing a more realistic

Conclusions

The results here obtained are useful for the definition of the optimal buffer on which to focus emergency mosquito-borne virus control interventions (i.e. deployment of adulticides insecticides aimed at eliminating potentially infected mosquitoes in the area surrounding the residence of an arbovirus infected person). This information is crucial for public authorities, as it has already been shown that enlarging the size of the area to be treated and reducing the time interval between infective

Data availability statement

The R code and the data are available at https://github.com/Chia1992/Partial-Differential-Equation

CRediT Author Statement

Chiara Virgillito: Methodology, Data curation, Writing. Mattia Manica: Methodology, Data curation, Writing. Giovanni Marini: Methodology, Data curation, Writing. Beniamino Caputo: Collecting data, Writing. Alessandra della Torre: Collecting data, Writing-Reviewing. Roberto Rosà: Methodology, Data curation, Writing-Reviewing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (44)

  • F. Cagnacci et al.

    Animal ecology meets GPS-based radiotelemetry: A perfect storm of opportunities and challenges

    Philos. Trans. R. Soc. B Biol. Sci.

    (2010)
  • B. Caputo et al.

    A comparative analysis of the 2007 and 2017 Italian chikungunya outbreaks and implication for public health response

    PLoS Negl. Trop. Dis.

    (2020)
  • B. Cummins et al.

    A spatial model of mosquito host-seeking behavior

    PLoS Comput. Biol.

    (2012)
  • J. Ekström

    Computation of Multivariate Normal and t Probabilities

    J. Stat. Softw.

    (2010)
  • P.S. Epopa et al.

    The use of sequential mark-release-recapture experiments to estimate population size, survival and dispersal of male mosquitoes of the Anopheles gambiae complex in Bana, a west African humid savannah village

    Parasites and Vectors

    (2017)
  • L. Facchinelli et al.

    Development of a novel sticky trap for container-breeding mosquitoes and evaluation of its sampling properties to monitor urban populations of Aedes albopictus

    Med. Vet. Entomol.

    (2007)
  • A. Gelman et al.

    Bayesian data analysis, third edition, Bayesian Data Analysis

    Third Edition

    (2013)
  • G. Guzzetta et al.

    Spatial modes for transmission of chikungunya virus during a large chikungunya outbreak in Italy: A modeling analysis

    BMC Med

    (2020)
  • P. Kareiva

    Population dynamics in spatially complex environments: theory and data

    Philos. Trans. - R. Soc. London, B

    (1990)
  • B.G.J. Knols et al.

    Odors Influence Mosquito Behavior

    Sci. Med.

    (1997)
  • C.Y. Lee et al.

    Estimating Flight Distance of Platypus koryoensis (Coleoptera: Curculionidae) by Mark-Release-Recapture and Its Validation by Field Observation

    J. Econ. Entomol.

    (2019)
  • S.L. Li et al.

    A review of models applied to the geographic spread of Zika virus

    Trans. R. Soc. Trop. Med. Hyg.

    (2021)
  • Cited by (2)

    View full text