key: cord-0989080-trf6zyd1 authors: Das, K.; Chatterjee, N. D. title: Examine the impact of weather and ambient air pollutant parameters on daily case of COVID-19 in India. date: 2020-06-09 journal: nan DOI: 10.1101/2020.06.08.20125401 sha: fc8baf7fb19a12eab5ab2360c4c21db02f68db60 doc_id: 989080 cord_uid: trf6zyd1 The present study presents a view on exploring the relationship pattern between COVID 19 daily cases with weather parameters and air pollutants in mainland India. We consider mean temperature, relative humidity, solar radiation, rainfall, wind speed, PM2.5, PM10, SO2, NO2 and CO as independent variable and daily COVID 19 cases as dependent variable for 18 states during 18th march to 30th April, 2020.After dividing the dataset for 0 to 10 day, 10 to 25 days and 0 to 44 days, the current study applied Akaike s Information Criteria (AIC) and Generalized Additive Model (GAM) to examine the kind of relationship between independent variables with COVID 19 cases. Initially GAM model result shows variables like temperature and solar radiation has positive relation (p<0.05) in 0 to 10 days study with daily cases. In 25 days dataset it significantly shows that temperature has positive relation above 23 degree centigrade, SO2 has a negative relationship and relative humidity has negative (between 30% to 45% and > 60%) and a positive relationship (45% to 60%) with COVID 19 cases (p=0.05). 44 days dataset has six parameters includes temperature as positive, relative humidity as negative (between 0 to 45%) and then positive (after >45%), NO2 as Positive (0 to 35 microgram/m3) followed by negative trend (after > 40 microgram/m3), SO2 and rainfall as negative relation. After sensitive analysis, it is found that weather variables like relative humidity, solar radiation and rainfall are more sensitive than temperature and wind speed. Whereas pollutants like NO2, PM2.5, PM10 and CO are more sensitive variables than SO2 in this study. In summary this study finds temperature, relative humidity, solar radiation, wind speed, SO2, PM2.5, and CO may be important factors associated with COVID 19 pandemic. Keywords: Weather parameter, Air pollutants, Daily COVID 19 cases, Akaike s Information Criteria (AIC), Generalized Additive Model (GAM) and Sensitive analysis. The severe acute respiratory coronavirus 2 (SARS-CoV-2) (Gorbalenya, 2020) is an infectious disease, initially found in Whan, China (Li et al., 2020) . Health Organization (WHO) has declared it pandemic worldwide on 11 th March, 2020, after spreading to several countries (Cucinotta et al., 2020) . COVID-19 infected patients have some typical symptoms including fever, myalgia, dry cough, pneumonia and throat sore (Şahin, 2020 (Şahin, . Huang et al., 2020 . As of 16 th may, 2020, a total 4,425,485 confirmed cases with 302,059 death has been registered so far (https://covid19.who.int/). Where United States of America (USA) is leading with 1.47 million conformed cases followed by Russia, United Kingdom, Spain and Italy. In India total active cases 53035 with 2752 death and 30,152 recovery reported till date of 16 th may 2020 (https://www.mohfw.gov.in/). These official data shows that recovery rate 35% and fatality stand at 3.2% in India. Being a densely populated area, India has alluring ground to be affected faster than other countries. But timely management and strict lockdown measures were able to reduce its predicted growth rate as per world level (Gupta et al., 2020) . In addition to human contact, there are many research works seeks to find the relationship of weather factors (Tosepu et al., 2020 , Zhu et al., 2020 and Shi et al., 2020 , air pollutants (Martelletti et al., 2020 , Ogen, 2020 and Conticini et al., 2020 , geographical factors (Gupta et al., 2020) and socio-economic factors (Qiu et al., 2019) with daily conform COVID-19 active cases as well as death due to this disease. Weather parameters like Temperature, Relative Humidity, Solar radiation, Rainfall, Wind speed (Chaudhuri et al., 2020 and Zhu et al., 2020) may enhance the transmission of corona positive cases. In addition to this ambient air pollutants like PM2.5, PM10, SO2, NO2 and CO (Bashir et al., 2020 and Yongjian et al.,) also could be associated with enhancing the process of COVID-19 cases. However recent study focuses on single or two factor based study like Temperature, relative humidity, solar radiation, SO2 and Particulate matter etc. Pollutants and weather parameters both have combine effect on cardiovascular and respiratory disease (Delamater et al., 2012 and Vanos et al., 2014) . Since sever acute respiratory coronavirus 2 (SARS-CoV-2) causing chronic damage and injury to the cardiovascular system (Zheng et al., 2020) it may have positive association with high concentration of Particulate Matter (PM) (Yongjian et al., 2020) . Again meteorological condition significantly influenced the concentration of PM (Zhao et al., 2014) . Therefore we assume that combine effect of weather parameters and pollutants will be a holistic way to find associative factors. The major objectives is to explore the relationship of weather parameters and air pollutants with COVID-19 cases. Unlike other study, we consider a state level rather than one city. Considering 14 days of moving average for temperature and relative humidity, each parameters were divided into three timeframe i.e. 1 to 10 days, 1 to 25 days and 1 to 44 days. This way of study will provide a holistic insight to study relationship of weather air pollutants with daily corona positive cases. appeal for micro level experiment and because of unavailability of meteorological station data lead to make it study on state level. Daily lab tested report of COVID 19 of 18 states has been collected from the Ministry of Health and Family Welfare (MoHFW) since seven days prior to first lock down day on 18 th March to 30 April, 2020. Meteorological and air pollutants data like daily mean temperature, relative humidity, solar radiation, rainfall, wind speed and Pollution data like mean daily concentration of Particulate Matter less than 2.5 micrometre (PM2.5), Particulate Matter less than 10 micrometre (PM10), Sulphur Dioxide (SO2), Nitrogen Dioxide (NO2) and Carbon Monoxide (CO) have been collected from Central Pollution Control Board (CPCB) website. A common issue is CPCB weather stations are not equally distributed all over India therefore mean values of all stations was calculated within a state is considered for final value of that state. Some of the states have fewer inactive stations that would again simulated by nearest station in neighbour state in respect to same physioclimatic condition. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 9, 2020. . https: //doi.org/10.1101 //doi.org/10. /2020 A . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint It is better choice to use 1-14 days of moving average in temperature and relative humidity because of the incubation period of COVID-19 range for 0 to 14 days (Zhu et al., 2020 , Duan et al., 2019 . Additionally smoothing spine functions of time (Ma et al., 2020 , Zeng et al., 2016 and Cheng et al., 2012 has been incorporated to coincide monotonic and non-monotonic pattern among COVID-19 cases with weather and pollution parameter CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint offers a more flexible model. AIC (Akaike's Information Criteria) was use as a measure for selecting those variables having some kind of association exist in between COVID-19 case and the parameters are used in this study to encounter model overfitting (Yen et al, 1998) . Where N is the number of observation, RSS is the residual sum of square, k is the number of parameters to fit plus 1 (k = β+1). AIC selects the optimum number of variables on the basis of changes in R-squared in a multivariate regression. Smaller AIC value is needed to select the best model (Posada et al., 2001) . We apply a step wise regression by considering both forward and backward direction method in R studio environment to identify variables with significant correlation with COVID cases. After selecting variables we further feed them into GAM model, defined as: Where log ( We also executed three way of sensitive analysis by dividing our total data set into 3 part (0-10 days, 0-25 days and 0-44days of lockdown). Another way of sensitive analysis is performed by changing degree of freedom (2 -9 df for time and other variable) in model execution time (Ma et al., 2020) . And most importantly repeat the model by excluding Maharashtra state as it has large number of case concentrate in last couple of days (Zhu et al., 2020). Since the model considers a relationship of weather parameters and pollutants with COVID ca ses, effect plots having visual interpretation of log transform daily case with other variables w ere produced. Significance level of 95% was adopted for all statistical analysis. Whole analysi s were performed in R software (version 4.0.0) with the help of "mgcv" package (version 1.8-. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint 31). For mapping purpose we have used ArcGis software (version 10.4.1). Spearman's correla tion has calculated to see the significance correlation among the variables of meteorological, p ollutant and COVID-19 cases. 3.1 Description of daily positive case, meteorological parameters and pollutants: Table 1 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint to moderate negative relationship with 95% confidence level, except temperature whereas wind speed and carbon monoxide (CO) are not significant. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint (Fig. 6 ). There is no significant relationship of relative humidity has been found with COVID-19 case in a 10 days span. It may be due to insufficient number of cases were included in some states. But in 25 days of study it was showing negative trend between 30 -45%, a positive linear trend between 45 -60% and again negative trend in >60%. 44 days of data set represent a negative trend between 0 -45% and a positive correlation >45%. But after sensitive analysis it denoted a sharp negative trend all over the curve. The exposer-response curve in Fig. 5c wind speed showed a significant nonlinear positive relationship whereas in spearman's correlation it was not significant. Wind speed between 1 to 2 m/s shows no relationship curve and >5m/s displayed a positive relation which is contrary of many research work (Ahmadi et al., 2020). In that case when we repeat the model by excluding Maharashtra, it has shown wind speed insignificant. Solar radiation in short period it could have effect on COVID-19 cases but in 44 days of study we have found insignificant relationship. Tough rainfall occurs in very limited areas with maximum 5.53mm with a mean of 0.096mm (Table 1) . Even though in 44 days of study it showed a negative relationship. After sensitive analysis rainfall became insignificant, solar radiation became significant and found a negative linear relation. Wind speed maintained a flat trend with the cases in most of the states having mean wind speed in between 0 -2.5 m/s (Fig. 6) . . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint In fig. 5 we have not found any significant relation of PM2.5, PM10, NO2, SO2 and CO with COVID-19 cases in 10 days study. But in 25 days there was only SO2 with decrease trend with slight positive after concentration of 35 µg/m 3 . A significant negative relationship of SO2 with COVID-19 cases has been recorded. It showed 0-10 mg/m 2 have flat and followed by a step decreased trend in concentration between 10-45 µg/m 3 . NO2 has a significant positive correlation in between 0-35 µg/m 3 and followed by a decrease trend after > 40 µg/m 3 (Fig. 5c ).Again it displayed a fresh negative correlation during sensitive analysis. In case of PM2.5 and PM10 it shows significant correlation with 6 degree of freedom therefore because of overfitting we eliminate it (Zhu et al., 2020 and Wang et al., 2018) . After sensitive analysis there is huge change appeared in association of pollutants in GAM model. PM2.5, and CO have become significant (Fig 6) . There was a positive relationship with PM2.5 showed through exposer-response curve, in most of the states, up to the concentration of 100 µg/m 3 . Same pattern followed by carbon monoxide (CO). . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint GAM result on 44 days data including Maharashtra state displayed Temperature, Relative humidity, NO2, SO2, wind speed and rainfall (Estimate 42.07, SD 2.66 and p<2e-16) as significant. For sensitive analysis, first we excluded Maharashtra state having a large number of case concentrate in last couple of days (Fig. 4) than other state for robust result (Zhu et al., 2020) . Fig. 6 present a robust study output (Estimate 30.75, SD 1.468 and p<2e-16) where PM2.5 , solar radiation and CO along with temperature, relative humidity, NO2 SO2 and wind speed have shown a significant association with COVID-19. With the help of this scenario it can be stated that pollutants like NO2, PM2.5, PM10 and CO are more sensitive variables than SO2 in this study. Weather variables like relative humidity and solar radiation and rainfall are more sensitive than temperature and wind speed. Pandemic COVID-19 became burden and its prolonged impact in every sphere of human life could be sensed even after successive decades. Our effort is to find out the association meteorological factors and pollutants with daily positive case of COVID-19. GAM model with 14 days moving average of temperature and showed a significant positive relationship with temperature, relative humidity, and wind speed whereas Relative Humidity, NO2, SO2, and rainfall and solar radiation showed a negative relation with daily case in 44 days of study. With the help of previous research work (Kumar, 2020 and Ahmadi et al., 2020) shows a positive and negative relation with temperature and relative humidity respectively. Similar to Chan et al., 2010, our study also shows high temperature leads to minimise the reported case. But it will need more experiment to establish a strong relationship with temperature. In the case of relative humidity two type of scenario has been seen. Firstly, landlocked States like Rajasthan, Madhya Pradesh, Uttar Pradesh Bihar, Punjab, and Himachal Pradesh (RH <57%) have shown a negative association with daily positive case. Secondly, Coastal states like Maharashtra, Kerala, Tamil Nadu and Andhra Pradesh with high relative humidity (>57%) having a positive association with cases. After excluding Maharashtra It seems a negative correlation in between 0 -90%, which is again validated by previous study (Kumar, 2020 , Qi et al., 2020 and Ma et al., 2020 . Again relative humidity may be statistically significant till it need more research to establish as a fact. In a study in Asyary et al., 2020 claim that Sunlight correlated significantly to recovery patient of COVID-19 because Vitamin D from sunlight can triggered body immune system. In our study we have seen a negative relation between infected cases with sunlight . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint exposer. There are states like Maharashtra, Kerala, Rajasthan Haryana, Delhi and Punjab enjoy solar radiation more as reported cases are also more. Therefore solar radiation will not be a establish factors can reduce COVID-19 transmission (Gupta et al., 2020) . Wind speed again pointed out a positive correlation can be validate with previous study in India (Gupta et al., 2020) . Many study found (Xie et al., 2019 , Tecer et al., and 2008 This study has many limitations like not inclusion of socio-cultural factors, gender, population density, movement of people and elevation etc. There is several factors can help to transmit this disease. Further laboratory and analytical studies are needed on these shortcomings. Our findings are there is a significant relationship between daily positive COVID-19 case with weather and pollution factors. Temperature and wind speed is positively associated with daily COVID-19 case count. Solar radiation has negative correlation. Positive association has been found PM2.5 and CO with COVID-19 confirmed cases and negative association with SO2 and NO2. Besides environmental factors, socio-economic factors, health infrustcture, literacy, number of person came from abroad, number of suspects in a states and their travel history can also enhance the accuracy of understanding actual influential factors in future study. We have no conflict of interest of this paper. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https: //doi.org/10.1101 //doi.org/10. /2020 Ogen, Y., 2020. Assessing nitrogen dioxide (NO2) levels as a contributing factor to the coronavirus (COVID-19) fatality rate. Science of the Total Environment, p.138605. Park, J.E., Son, W.S., Ryu, Y., Choi, S.B., Kwon, O. and Ahn, I., 2020 Tan, J., Mu, L., Huang, J., Yu, S., Chen, B. and Yin, J., 2005 . An initial investigation of the association between the SARS outbreak and weather: with the view of the environmental . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint Yongjian, Z., Jingu, X., Fengming, H. and Liqing, C., 2020. Association between short-term exposure to air pollution and COVID-19 infection: Evidence from China. Science of The Total Environment, p.138704. Zeng, Q., Li, G., Cui, Y., Jiang, G. and Pan, X., 2016. Estimating temperature-mortality exposure-response relationships and optimum ambient temperature at the multi-city level of . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 9, 2020. . https://doi.org/10.1101/2020.06.08.20125401 doi: medRxiv preprint Trend Analysis and Forecasting of COVID-19 outbreak in India. medRxiv Clinical features of patients infected with 2019 novel coronavirus in Wuhan, China. The lancet Diurnal temperature range and daily mortality in Shanghai Will COVID-19 pandemic diminish by summer-monsoon in India? Lesson from the first lockdown Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and corona virus disease-2019 (COVID-19): the epidemic and the challenges Substantial undocumented infection facilitates the rapid dissemination of novel coronavirus Systematic review and meta-analysis of the adverse health effects of ambient PM2. 5 and PM10 pollution in the Chinese population Short-term effects of air pollution on daily mortality and years of life lost in Nanjing Effects of temperature variation and humidity on the death of COVID-19 in Wuhan Air pollution and the novel Covid-19 disease: a putative disease risk factor. SN Comprehensive Clinical Medicine