Estimating the Mosquito Density in Guangzhou City, China

Mosquito is a vector of many diseases. Predicting the trend of mosquito density is important for early warning and control of mosquito diseases. In this paper, we fit a discrete time mosquito model developed by Gong et al. in 2011, which considers the immature and adult stages, and weather dependent model parameters, to the Breteau Index and Bite Index data for Aedes aegypti in Guangzhou city, China in 2014, as well as the weather data for average temperature, precipitation, evaporation and daylight for the same period. We estimated the model parameters using the Markov Chain Monte-Carlo (MCMC) method. We find that many parameters are not identifiable. We revise and simplify the model so that the parameters of our new model are identifiable. Our results indicate that the model predicted mosquito prevalence agrees well with data. We then use the fitted parameter values against the Breteau Index and Bite Index data for Guangzhou city in 2017 and 2018, and show that the estimated parameter values are applicable for other seasons.

the posterior distributions of the parameters. At last, we will apply the model parameters to the BI and Bite based on the Sharpe and DeMichele equation (Rueda et al., 1990), where the constants are determined 63 empirically. The survival rate of the adults µ t is assumed to be a bell curve of the temperature T t , i.e., The diapause rate of the adults F t is assumed to be a linear function of the day length H t (in hours), i.e., Maturation rate assuming no temperature inactivation of the critical enzyme. H A Enthalpy of activation of the reaction that is catalyzed by the enzyme (cal mol −1 ). H H Change of Enthalpy with high temperature inactivation of the enzyme (cal mol −1 ). T H Temperature where 50% of the enzyme is inactivated by high temperature ( • C). µ 1 Survival rate at optimal temperature for immatures. T 01 Optimal temperature for survival of immatures ( • C). v 1 Variance of function. µ 2 Survival rate at optimal temperature for adults. T 02 Optimal temperature for survival of adults ( • C). v 2 Variance of function. k Decay rate of diapausing rate with day length.

Mosquito density data 66
The mosquito monitoring data for Guangzhou city were obtained from the Guangzhou Center for Disease The population density is related to counts, which typically assumed to be a Poisson random variable 88 and has a variance similar to the mean. However, the BI X t and the Bite Index Y t are not true counts. So, 89 we assume that they are normally distributed with a mean A t and variance A t for the adults, and a mean 90 J t p and vairance J t p for the immatures, i.e., where the parameter  Table 2. We used 4 chains, 10,000 iterations with the first 6,000 iterations discarded as a burn-in period.

99
The posterior distributions for the initial conditions for the 12 districts, model parameters for all districts,  of the prior distribution. In addition, the parameter v 2 of the adult survival rate µ t is also unidentifiable 104 because its posterior density kept rising even when increasing the widths of the prior distribution.

105
A new maturation rate The maturation rate d t defined in (1f) is a single humped function of the 106 temperature T t . For simplicity, we replace it with a bell curve: where T 03 is the peak and v 3 is the spread of the bell curve.

108
A new adult survival rate From (1g), the adult survival rate is approximately a constant (independent 109 of the temperature T t ) if v 2 becomes very large, which is suggested by our fitting results. Thus, we assume A new diapause rate Based on the work reported by Spielman (Spielman, 2001), the diapause rate was a 112 function of decreasing hours of daylight H t , ranging from 0 to 1. We assume that it decreases exponentially 113 with the daylight H t , The simulation results of improved model We fitted the basic model (1)(a,b) with (3)(a,b,c) to the 115 same data. All the parameters are identifiable. The prior distributions of parameters were shown in Table   116 3. The posterior distributions are given in Figure 4, Figure 5 and Figure 6, the mean and the confidence 117 intervals of the model parameters (excluding the initial densities and the ratio of the capture probabilities) 118 are listed in Table 4.  Table 4.

135
In this paper, we modified the discrete time mosquito model with weather factors developed by Gong et al. very likely that the model parameters that we estimated in this paper may be applicable to other regions as 148 well, but this needs to be validated using the mosquito data for the regions of interest.

149
Our results provide a quantification of the weather factors on the mosquito population dynamics. This 150 model can be coupled with disease models to provide a tool for evaluating the risk of mosquito-borne diseases 151 such as dengue infection.