key: cord-0903196-dx1fwngb
authors: Hossain, Md. Sabbir; Ahmed, Sulaiman; Uddin, Md. Jamal
title: Impact of weather on COVID-19 transmission in South Asian Countries: an application of the ARIMAX model
date: 2020-11-02
journal: Sci Total Environ
DOI: 10.1016/j.scitotenv.2020.143315
sha: 7c03ca88f55365e74b1235f96317666094fe4c9f
doc_id: 903196
cord_uid: dx1fwngb

We aimed to examine the impact of weather on COVID-19 confirmed cases in South Asian countries, namely, Afghanistan, Bangladesh, India, Pakistan, and Sri Lanka. Data on daily confirmed cases, together with weather parameters, were collected from the first day of COVID confirmed cases in each country to 31 August 2020. The weather parameters were Rainfall (mm), relative humidity (%), maximum and minimum temperature (°C), surface pressure (kpa), maximum air pollutants matter PM 2.5 (μg/m3) and maximum wind speed (m/s). Data were analyzed for each investigated countries separately by using the Autoregressive Integrated Moving Average with Explanatory Variables (ARIMAX) model. We found that maximum wind speed had significant negative impact on COVID-19 transmission in India (-209.45, 95% confidence interval (CI): -369.13, -49.77) and Sri Lanka (-2.77, 95% CI: -4.77, -0.77). Apart from India, temperature had mixed effects (i.e., positive or negative) in four countries in South Asia. For example, maximum temperature had negative impact (-30.52, 95% CI: -60.24, -0.78) in Bangladesh and positive impact (5.10, 95% CI: 0.06, 10.14) in Afghanistan. Whereas rainfall had negative effects (-48.64, 95% CI: -80.17, -17.09) in India and mixed effects in Pakistan. Besides, maximum air pollutants matter PM 2.5 was negatively associated with the confirmed cases of COVID-19. In conclusion, maximum wind speed, rainfall, air pollutants (maximum PM 2.5) and temperature are four variables that could play a vital role in the transmission of COVID-19. Although there is a mixed conclusion regarding weather parameters and COVID-19 transmission, we recommend developing environmental policies regarding the transmission of COVID-19 in South Asian countries.

Novel coronavirus (COVID-19) is a respiratory disease caused by severe acute coronavirus syndrome 2 (SARS-CoV-2) (F. . Due to its highly human -to -human transmission nature (Bogoch et al., 2020; C. Wang et al., 2020) the disease spread rapidly throughout the world and was declared a global pandemic by world health organization (WHO) on 11 March 2020. Over 25 million confirmed cases (25, 094, 338) and 844,060 COVID-19 deaths were reported around the world until 31 August 2020 (World Health Organization, 2020) .

Several studies suggested that weather parameters like temperature, rainfall, humidity, wind speed, and air pollutants may influence the transmission of the COVID-19 (Ahmadi et al., 2020;  Al-Rousan and Al-Najjar, 2020; Bashir et al., 2020; Jüni et al., 2020; Liu et al., 2020; Qi et al., 2020; Sobral et al., 2020; Tosepu et al., 2020; Zoran et al., 2020) . For example, Liu et al. (2020) , Qi et al.(2020) , Sobral et al. (2020) and Y. demonstrated that the transmission of COVID-19 infection is suppressed as temperature increases in China and some other regions around the world. Moreover, Al-Rousan and Al-Najjar (2020), Tosepu et al. (2020) and Zoran et al. (2020) reported a significant positive correlation in China, Indonesia and Italy while in Iran, Ahmadi et al. (2020) and in 144 geopolitical areas worldwide, Jüni et al. (2020) reported no significant correlation between temperature and COVID-19 transmission. Furthermore, Liu et al. (2020) , Qi et al. (2020) and Y. stated humidity as one of the important weather parameters that significantly reduce the virus transmission of COVID-19. But some studies reported no significant impact of humidity on the COVID-19 epidemic (Bashir et al., 2020; Tosepu et al., 2020) . Also, Ahmadi et al. (2020) , Sobral et al. (2020) and Zoran et al. (2020) investigated the impact of rainfall, air pollutants, and wind speed on COVID-19 and found that rainfall and air pollutants had a positive J o u r n a l P r e -p r o o f impact while wind speed had a significant negative impact on COVID-19. However, Poirier et al. (2020) reported that weather parameters alone could not decline the coronavirus transmission.

Several statistical methods such as correlation, regression analysis, generalized additive model Qi et al., 2020; Xie and Zhu, 2020) , and generalized linear model , are extensively applied to assess the impact of environmental factors on COVID-19 transmissibility. However, among those methods, most of the studies used correlation and regression analysis techniques (Guo et al., 2020; Jüni et al., 2020; Şahin, 2020; Sobral et al., 2020; Tosepu et al., 2020; Zoran et al., 2020) . These studies have used the regression model for time series data but have not determined the stationary condition of the data, although stationary is the precondition for implementing any regression analysis techniques with time-series data (Chen et al., 2004) . As a result, these studies could have produced incorrect estimates due to misspecification of the applied models. A well-defined time series technique, such as the Autoregressive Integrated Moving Average with Explanatory Variables (ARIMAX) model, is therefore essential to apply in order to eliminate the long-term trend of COVID-19 epidemic and to consider environmental factors as external regressors too. In addition, due to lack of data availability, most of the study was conducted at the birthplace of COVID-19,

This study was accumulated different types of data, including daily COVID-19, climate, and air pollutants datasets from the first unequal date of COVID-19 confirmed cases to 31 August 2020 of five South Asian countries, such as Afghanistan, Bangladesh, India, Pakistan, and Sri Lanka. J o u r n a l P r e -p r o o f °C and annual rainfall averaging about 255 millimeters (Wikipedia, 2020b) . Sri Lanka's climate is tropical and consists of distinct wet and dry seasons (World Travel Guide, 2020). The geographical location of five south Asian countries with cumulative COVID-19 confirmed cases up to 31 August was presented in Fig. 1 .

We used the ARIMAX model to evaluate the relationship between daily COVID-19 confirmed cases and daily climatic variables. The steps of the whole process are listed as below (Yan et al., 2017) :

First, the ARIMA model was developed to the time series of confirmed cases from 1 st day of confirmed cases to 31 August 2020 for each country. We used the Bayesian Information Criteria (BIC) to identify the best model for each country. The detailed procedure of the ARIMA model was found in supplementary materials (S1).

Second, cross-correlation function (CCF) was used to explore the relationship between climatic variables and COVID-19 confirmed cases for each country separately. The dependent (COVID-19 confirmed cases) and independent variables (climate variables) were pre-whitened by the previously fitted ARIMA models. Pre-whitening is an important technique for seeing which lag of the independent variable affects the dependent variable. Pre-whitening, the method of eliminating or reducing short-term stochastic persistence widely applied to the study of a variety of geophysical variables in time series (Razavi and Vogel, 2018) .

Finally, climate variables selected through step 2 were incorporated as covariates into the ARIMAX model.

The maximum likelihood method was applied for the estimation of the parameters. To assess if the residual series was white noise, the Ljung-Box Q test was performed. We used R software version 3.6.1 to analyze the data (Team, 2013) . To perform this analysis, MASS, Hmisc, pastecs, forecast, tseries, & lmtest packages were used. Figure 1 shows that until 31 August 2020 the highest and lowest number of confirmed COVID-19 cases were recorded in India (3,691,200), and Sri Lanka (3,050), respectively. Also, Fig. 2 shows that the trend line of COVID-19 confirmed cases between Afghanistan, Bangladesh, and

Pakistan is very similar, after 100 days of transmission these countries shows a decreasing trend.

India have the highest rate of transmission compared to others and the transmission rate is continuously increasing from the beginning which makes India as a new hotspot of COVID-19 but the line of the pattern for Pakistan and Sri Lanka shows considerable inconsistency. Fig. A.3) Figure 3 shows that the cross-correlation between the pre-whitened weather variables and COVID-19 confirmed cases at lags 0 to 12 days. Here only positive lags would be considered because the positive value indicated that climatic factors could affect COVID-19 confirmed cases a certain period later. For Afghanistan, except minimum temperature at lag 5 and maximum temperature at lag 12 days, all other climatic variables failed to prove the statistically J o u r n a l P r e -p r o o f significant correlation with COVID-19 confirmed cases at different lags (Fig. 3) . Maximum temperature at lag 3 (Fig. 3) , rainfall and maximum wind speed at lag 6 ( Fig. A.1 

Bangladesh and for India rainfall at lag 6, wind speed at lag 8 (Fig. 3) and maximum particulate matter PM 2.5 at lag 2 (Fig. A.1) had a significant correlation with confirmed cases. Similarly, for Sri Lanka, and Pakistan various weather parameters at different lags have found a significant correlation with COVID-19 confirmed cases (Fig. 3, Fig. A.1) .

At different lags, attempts were made to integrate the above-mentioned climate variables as covariates into the ARIMAX model. For example, for Afghanistan, the ARIMAX model was set up separately by considering each of the covariates; maximum temperature at lag 12, and minimum temperature at lag 5. Likewise, the previously selected variables at different lags were considered to be covariates in the ARIMAX model for certain countries and only those were provided in Table 3 , which shows a significant major effect on COVID-19. 

We examined the effects of weather on COVID-19 transmission in South Asian countries using the ARIMAX model approach. Our study shows that weather variables have significant positive or negative effects on COVID-19 infections. However, the findings don't seem to be consistent across the selected countries in South Asia. For example, the maximum wind speed was significantly negatively correlated with the COVID-19 transmission only in India and Sri Lanka.

In addition, the temperature had significant positive and negative impact on the transmission in four countries in South Asia, excluding India.

We found that maximum wind speed had a significant negative impact on COVID-19 transmission which is in line with some previous studies (Ahmadi et al., 2020; Islam et al., 2020; Zoran et al., 2020) . However, this finding was opposed to some recent studies (Bashir et al., 2020; Menebo, 2020) where they claimed that there was no correlation between wind speed and COVID-19 transmission. Moreover, in the previous outbreak of influenza virus or severe acute respiratory syndrome (SARS), or middle east respiratory syndrome coronavirus (MERS-CoV), the wind speed regarded as one of the key factors that promote the transmission of SARS, MERS and influenza (Altamimi and Ahmed, 2020; Chan et al., 2011; Chong et al., 2020; Peci et al., 2019) . Moreover, previous studies (Doremalen et al., 2020; Bourouiba, 2016) claimed that COVID-19 may be airborne which can stabilize in aerosols for up to 3 hours and increase the risk of transmission through aerosols. The risk of COVID-19 spread might be more in closed places with low wind speed because in low wind speed particle density of infectious droplets is much higher which can favor the spread of COVID-19 virus in the environment.

Our analyses also showed that temperature had a significant positive and negative impact on the transmissibility of COVID-19 in four south Asian countries except for India. Menebo, (2020), Bashir et al. (2020) and (A. Gupta et al., 2020) also reported similar findings that temperature had a significant positive correlation with COVID-19. Furthermore, some other studies reported inverse relationship (Guo et al., 2020; Islam et al., 2020; Rahman et al., 2020; Shi et al., 2020; or no relation (Jamil et al., 2020; Poirier et al., 2020) between temperature and incidence of COVID-19. As the south Asian countries have high population density with low income including less aware of the transmission, people often break the lockdown for their livelihood on shiny days. And this may be one of the reasons for the positive relationship between temperature and confirmed cases. In that case, temperature plays one of the key indirect factors for COVID-19 transmission.

We did not find any association of COVID-19 transmission with relative humidity and surface pressure. Shi et al. (2020) also obtained similar findings that there was no significant correlation between COVID-19 incidence and absolute humidity. However, some other studies (S. Gupta et al., 2020; Ma et al., 2020; found opposite results that there exists an association of COVID-19 with relative humidity and pressure.

Air pollutants included particulate matter with aerodynamic diameter µg per m3

(maximum PM 2.5) had a significant negative impact on COVID-19 only in Pakistan. Ma et al. (2020) found similar findings in Wuhan, China that the mortality counts due to COVID-19 were negatively correlated with PM 2.5 and PM 10. Zoran et al. (2020) and Bashir et al. (2020) reported opposite findings (positive correlation) that a high level of air pollution has a significant impact on the increased rates of confirmed COVID-19 cases. It is obvious that higher J o u r n a l P r e -p r o o f Journal Pre-proof concentrations of air pollution could increase the risk of respiratory virus infection (Horne et al. 2018 ). But interestingly, our results showed that COVID confirmed cases were associated with a decrease in the average concentration of air pollutants (PM 2.5), which could be a reflection of higher rainfall at the same time, as rainfall washes away pollutants from the atmosphere and for Pakistan, we also reported the significant effect of rainfall. Similar type of argument was also made by Silva et al. (2014) where they showed that Severe Acute Respiratory Infection (SARI) cases were negatively associated with air pollutants. Hence, we argue that the effect of air pollutants on COVID transmission depends on the impact of other relevant factors (rainfall) and

the effect varies depending on city, region, and the specific pollutants under investigation.

Rainfall had a significant negative impact on COVID-19 transmission in India and Pakistan, also had positive impact in Pakistan (at lag 12) which is in line with (A. Gupta et al., 2020; Menebo, 2020; Sobral et al., 2020) and opposite with some recently published studies (Ahmadi et al., 2020; Bashir et al., 2020; Tosepu et al., 2020) . The possible reason for the negative correlation is that rainfall rate contributes to the accumulation and washout process of aerosols and microbial bio-aerosols (Bacteria, viruses, fungi) implying that viruses could not have longer residence times in the atmosphere and, consequently will not able to disperse further.

Another hypothetical justification might be that people often stay home on rainy days, which also could reduce the transmission. We also found that rainfall at lag 12 had an impact on Fig. 3 . Cross-correlation between COVID-19 confirmed cases and weather variables 5

Investigation of effective climatology parameters on COVID-19 outbreak in Iran

The correlation between the spread of COVID-19 infections and weather variables in 30 Chinese provinces and the impact of Chinese government mitigation plans

Climate factors and incidence of Middle East respiratory syndrome coronavirus

Correlation between climate indicators and COVID-19 pandemic in New York

Potential for global spread of a novel coronavirus from

The effects of temperature and relative humidity on the viability of the SARS coronavirus

Yield variability as influenced by climate: a statistical investigation

Association between meteorological variations and activities of influenza A and B across different climate zones: a multi-region modelling analysis across the globe

Discovery Bangladesh [WWW Document

Aerosol and Surface Stability of SARS-CoV-2 as Compared with SARS-CoV-1

Transmissibility of COVID-19 and its association with temperature and humidity

Shortterm elevation of fine particulate matter air pollution and acute lower respiratory infection

Novel Coronavirus (COVID-19) Cases Data [WWW Document

Temperature, humidity, and wind speed are associated with lower Covid-19 incidence

No Evidence for Temperature-Dependence of the COVID-19 Epidemic. medRxiv 2020.03.29

Impact of climate and public health interventions on the COVID-19 pandemic: A prospective cohort study

Meteorological factors correlate with transmission of 2019-nCoV: Proof of incidence of novel coronavirus pneumonia in Hubei Province

Impact of meteorological factors on the COVID-19 transmission: A multi-city study in China

Effects of temperature variation and humidity on the death of COVID-19 in Wuhan

Temperature and precipitation associate with Covid-19 new daily cases : A correlation study between weather and Covid-19 pandemic in Oslo

The POWER Project [WWW Document

Effects of absolute humidity, relative humidity, temperature, and wind speed on influenza

COVID-19 transmission in Mainland China is associated with temperature and humidity: A time-series analysis

A Retrospective Analysis of Influence of Environmental / Air Temperature and Relative Humidity on SARS-CoV-2 Outbreak. Prepr

Prewhitening of hydroclimatic time series ? Implications for inferred change and variability across time scales

Impact of weather on COVID-19 pandemic in Turkey

The impact of temperature and absolute humidity on the coronavirus disease 2019 (COVID-19) outbreak -evidence from China

Respiratory viral infections and effects of meteorological parameters and air pollution in adults with respiratory symptoms admitted to the emergency room

Association between climate variables and global transmission oF SARS-CoV-2

Swedish Committee for Afghanistan (SCA) [WWW Document

Estimation of the Transmission Risk of the 2019-nCoV and Its Implication for Public Health Interventions

R: A language and environment for statistical computing

Correlation between weather and Covid-19 pandemic in Jakarta

~:text=A tropical rainy climate governs,that fall under this group.&text=2) In India a tropical

Climate_of_Pakistan#:~:text=Pakistan has four seasons%3A a,period of October and November

World Air Quality Index project

World Health Organization, 2020. Coronavirus Disease (COVID-19)

World Travel Guide [WWW Document

A new coronavirus associated with human respiratory disease in China

Effects of temperature and humidity on the daily new cases and new deaths of COVID-19 in 166 countries

Association between ambient temperature and COVID-19 infection in 122 cities from China

Impact of meteorological factors on the incidence of bacillary dysentery in Beijing, China : A time series analysis

Assessing the relationship between surface levels of PM2 . 5 and PM10 particulate matter impact on COVID-19 in

August 2020. 1 Here, ARIMA (2,1,2) means autoregressive term of order 2, moving average term of order 2 and first difference was taken to make the series stationary. 2 All the selected best ARIMA models need to satisfy the three conditions: lowest BIC, significant parameter estimation and no autocorrelation in residual analysis. Though ARIMA (0,1,3) has lowest BIC value, it provides insignificant parameter estimation. That's why we move for 2 nd lowest BIC value and ARIMA (2,1,2) satisfies all the three conditions. 3 For India 2 nd difference was taken to make the COVID confirmed series trend and mean stationary. Note: PM 2.5 means maximum atmospheric particulate matter with diameter less than 2.5 and * means significant at 5% level J o u r n a l P r e -p r o o f