Modeling Mosquito Activity Built on Mosquito Population Dynamics: A Simulation Study
Background: West Nile virus (WNv) continues to be one of the most destructive mosquito borne diseases in the world, and Saskatchewan has experienced the highest incidence rates for WNv in North America. Its primary transmitters are mosquitoes, with Culex tarsalis serving as the main vector in Saskatchewan. For this reason, mosquito population dynamics is an important determinant of WNv risk. Weather factors, in turn, exert a pronounced impact on mosquito populations. It is important to understand the environmental factors playing a crucial role in oscillations of the mosquito population. It is also important to construct a model or create a method which can monitor and accurately estimate the overall dynamics of the mosquito population. Methods: In this study, a Probability Generating model is developed to simulate the mosquito observation counts, making use of a pre-existing System Dynamics Model to simulate a mosquito population. A MCMC method was further used to draw samples from a posterior distribution for Bayesian inference and analyse how frequency of observation of mosquito trap counts can improve performance of our model or method. Purpose of study: This study mainly focuses on investigating the feasibility of estimating the regression coefficients of the logistic regression model for the parameters (β) by using the proposed computational method. Meanwhile, we consider comparing the performance of this method with analysis under different sampling frequencies. Results: The results of the Probability Generating model depicts the distribution of the simulated observation data (y_i) over our study region (city of Saskatoon) seasonally, which suggests the environmental variables have a significant effect in driving variations in mosquito populations under the simulation experiments; the results of the three different sampling frequencies suggest that the current frequency (weekly) of measuring counts of trapped mosquitos is insufficient for reliable estimation of the parameters (β) for the durations examined. Conclusion: In this study, we formulated a probabilistic model from a combination of a reasonably complex dynamic model and a probabilistic generating model. Additionally, we have investigated the frequency of collecting real-world data associated with the accuracy of the model and revealed the importance of sampling mosquito population every day for reliably estimating parameter values, rather than pursuing the standard of sampling mosquito population every week.
DegreeMaster of Science (M.Sc.)
DepartmentMathematics and Statistics
SupervisorLiu, Junxin; Osgood, Nathaniel
CommitteeKhan, Shahedul; Samei, Ebrahim; Epp, Tasha
Copyright DateAugust 2017
Keywords: West Nile virus (WNv)
System Dynamics Model (SDM)
Probability Generating model
Markov Chain Monte Carlo (MCMC)
Highest Posterior Density (HPD)Interval.