Determining role of air temperature in predicting and controlling COVID-19 risk levels anywhere anytime using multiple modelling analyses

COVID-19 is a pandemic with no cure. There is an urgent need for low-cost interventions. Macroclimate work through microclimate. In many situations, man-made microclimate, such as air conditioning, may override the effect of natural macroclimate in determining the pathogenicity of SARS-CoV-2. This study aimed to determine if there is a ‘safe’ temperature that is comfortable to human beings while significantly inhibitory for SARS-CoV-2 pathogenicity. Data on monthly new deaths or new cases per million population (MDPM or MCPM) and monthly cumulated days with more cases than the previous day (DI) from March 2 to June 15, 2020 were collected from all 118 countries with population over five million. Monthly average AT negatively correlated with the transmission parameters. A significant decrease in transmission was observed when AT reached above 20 oC. Monthly average (not average high) AT of countries with MDPM <2, MCPM<10, or DI<=7 was found to be between 24.54 and 25.90 oC (25.00 oC on average) with average standard error of 4.97. Thus, average AT <20, 20-25, >25 oC were considered as high, medium, and low risk AT. Furthermore, MDPM in countries with AT <20 oC were 80.93, 50.23, 13.52, and 7.72 times of those Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 30 June 2020 doi:10.20944/preprints202006.0373.v1 © 2020 by the author(s). Distributed under a Creative Commons CC BY license. 2 in countries with AT >25 oC in March, April, May, and June 1-15, respectively. MDPM high-risk rates in countries with AT >25 oC were 0, 6.25, 14.55, and 9.84%, and the low-risk rates were 100, 83.33, 52.73, and 81.97%, respectively. In countries with AT <20 oC, the trends were opposite. Setting indoor temperature to 25 oC could decrease the need of social distancing for containing SARS-CoV-2 transmission. Cooling indoor temperature too low may be a reason of COVID-19 outbreak in some high AT countries. Authorities and the general population can evaluate COVID-19 risk level and manipulate microclimate to reduce the risk anywhere anytime based on local day average AT.


Introduction
Novel coronavirus disease (COVID-19) has killed over 420,000 individuals globally. It may never go away, even with a vaccine. Severe COVID-19 is mainly an adult respiratory distress syndrome (ARDS) caused by SARS-CoV-2 virus. Therapeutic options for either viral diseases or ARDS are limited. Thus, till the development and availability of the vaccine or acquirement of herd immunity, decreasing the survival time of SARS-CoV-2 in the environment and preventing the spread of virus to susceptible people are the most effective approaches to prevent the disease 1 .
In the battle between the causative agent of the disease and body's defense mechanism, the outcomes vary, ranging from disease-free, asymptomatic state to mild, severe, or deathly illness. While the basic characteristics of the disease are determined by the causative agent, the pathogenicity of disease is affected by numerous natural or social conditions, also called precipitating factors. In many situations, the precipitating factors determine the onset and development of a disease. For an infection disease like COVID-19, precipitating factors may affect a pathogen at various stages including reproduction, survival, or reaching host. When a precipitating factor becomes strong enough, it may dominantly control the virus' pathogenicity.
Social distancing has been demonstrated as an effective approach in containing SARS-CoV-2 transmission 1. However, it is unsustainable because of its burden on economy and daily life activities 2,3. Therefore, a less costly and more acceptable approach is urgently needed.
Ambient temperature (AT) is a seasonal and controllable natural factor. AT has been found to be associated with transmission of, influenza, and many other viruses 4,5. For example, seasonal cycles are known to play a crucial role in the transmission of the common cold and flu. They usually reach epidemic in winters. The transmission by SARS-CoV-1 and SARS-CoV-2 also occur in the winters. Several studies have revealed that AT is negatively associated with COVID-19 transmission. It is possible that higher temperature is needed to contain SARS-CoV-2 pathogenicity than its pier coronaviruses.
Most of our lives in the developed world are spent indoors. Overlap between SARS- In the present study, we aimed to identify a temperature comfortable to human beings while significantly inhibitory for SARS-CoV-2 transmission which may lower the risk of COVID-19 transmission. In addition, we hypothesized that the AT may be used to classify risk level of a region and general people may predict and control COVID-19 risk based on local AT. This will also facilitate government bodies to use this criteria to identify, plan, respond to, and reduce the impact of COVID-19 in their entities.
Printable heat maps for monthly (January to December) COVID-19 risk levels of all countries, major world cities, and subnational entities of representative countries can be provided for convenience. Surfing seasonal change of AT and conditioning temperature indoor would be a low cost and easy to implement approach in containing COVID-19 pandemic. Of course, when other precipitating factors are extremely strong too, the effectiveness of controlling AT could be limited. Ventilation and sanitizing the air with ultraviolet light in nonbusiness hours may be additionally effective.

Data collection
The relationship between monthly average ambient temperature and three epidemic were not in the above lists but belong to top 50 most popular cities in the world were added (https://www.worldatlas.com/citypops.htm).

Establishment of ambient temperature criteria for high, medium, and low risks
For each of the above four parameters, the average AT of the low-risk countries in March, April, May, and the first half of June were calculated. Average AT related to MDPM for March were excluded because many countries did not have any mortality cases yet. The remaining 11 AT were averaged (AT-M) and standard deviation (AT-SD) was calculated. Because AT is negatively correlated to epidemic according to our analysis and existing literature, AT higher than AT-M was considered low risk. The AT between AT-M and AT-M minus mean of AT-SD was considered medium risk. AT lower than AT-M minus mean of AT-SD was considered high risk.

Evaluation of accuracy of the risk classification criteria retrospectively
Based on the AT classification criteria, each country was allocated into corresponding categories of high, medium, and low risk. Match rates for March, April, May, and the first half of June were calculated separately. Reflection points of fitting curves between AT and each of the three epidemic parameters were observed to see if they were consistent with AT criteria. The match rates between AT-based risk allocation result and real risk levels based on MDPM were calculated. As previously stated, MDPM<2 is considered low risk. Here, MDPM>5 was considered high risk, ranking COVID-19 as the top 20 th death cause in the world just after breast cancer. MDPM between two-five was considered as medium risk. The reason of using MDPM is that, among the three parameters analyzed, MDPM is the most reliable data. In addition to percentage match rate, corresponding MDPM of high, medium, and low risk AT were compared.

Monthly risk sheets for all countries, major world cities and subnational divisions of representative countries
According to AT classification criteria, monthly (January to December) COVID-19 risk  Table 1).

Continued curve fitting showed that the epidemic parameters decreased significantly at 20 ºC
Nonlinear curve fitting between AT and the three epidemic parameters in March, April, May, and June 1 to 15 were performed separately. As shown in Figure 1, all the 12 nonlinear curve lines had a reflection point at approximately 20 ºC that showed significantly decrease in the epidemic incidence (Fig. 1).

Average AT and average standard deviation of low risk countries
Monthly average AT and standard deviation of three categories of low risk countries in  The monthly average AT of 25.00 ºC consistently determined by three categories of low risk countries. Thus, 25 ºC served as standard of low risk. Because AT is negatively related to risk level, AT above 25 ºC was considered low risk. One standard deviation (4.97) down from the low risk AT was considered medium risk. Thus, temperature between 20-25ºC was considered medium risk. Temperature below 20 ºC was high risk AT. Consistently, all curve fitting results showed that the epidemic incidence decreased significantly when AT was above 20 ºC.

The risk level classification system accurately represented the real COVID-19 pandemic in the past three months
Using our risk classifying system to analyze March, April, May, and June 1 to 15 data retrospectively, the MDPM high-risk rates in countries with AT >25 ºC were 0.00 (0/42),  . 2b). In May, daily average air temperatures approaches 20 ºC and above in many north hemisphere countries whereas it tends to decrease to 20 ºC and below in many south hemisphere countries. Line chart of semi-monthly data between AT and MDPM or MCPM showed that the new deaths and cases caused by COVID-19 decreased significantly in many north hemisphere countries, especially in European countries and the United States, and increased significantly in south hemisphere countries, especially in Brazil, Chile, and Peru (Fig. 3). Furthermore, one can predict risk level anytime anywhere using our AT classification criteria.

Discussion
All observational and modeling studies to date indicated that the SARS-CoV-2 could produce a substantial outbreak regardless of the season 5,6. The purpose of this study was to determine an AT that is comfortable for human as well as diminishing for SARS- countries was especially useful in isolating the relationship between AT and COVID-19 transmission.
Using multiple epidemic parameters and analysis methods, the relationship between AT and epidemic dynamics were analyzed. All the results consistently showed AT above 25 ºC as low risk for SARS-CoV-2 transmission. Nonlinear curve fittings of all three epidemic parameter showed 20 ºC as the temperature that significantly decreased SARS-CoV-2 transmission. Retrospectively analysis showed that the low risk rates in countries with AT >25 ºC were high. In contrast, countries with AT <20 ºC were more at high-risk. The situation of countries with AT between 20-25 ºC was closer to that observed in countries with AT >25 ºC. Most countries with AT >25 ºC that were not in low risk had some regions with AT below 25 ºC. However, in United Arab Emirates, AT was higher than 25 ºC throughout the country in both April and May while MDPM was observed to be in high risk category in both the months. The reason for this discrepancy is not known although we noticed that it differs from other countries with AT >25 ºC in two aspects that may affect the transmission dynamics. It is a high -income country and has more visitors and migrant workers than their own nationals.
Importantly, although the total cases and deaths per million population is high, CFR of the country is low (<0.85%). Similarity, several other countries with AT >25 ºC such as, Singapore and Oman, though were not in the list of high risk in term of MDPM, had high MCPM. In addition, these countries had even lower CFR (<0.50%).
The accuracy was also proven by two other ways. First, MDPM in countries with AT <20 ºC was 7.72 to 80.93 times more than that in countries with AT >25 ºC. Second, in May, SARS-CoV-2 transmission dynamic decreased in many north hemisphere countries because AT increased to 20 ºC or above. The transmission increased in many south hemisphere countries because AT decreased to 20 ºC or less. Furthermore, the accuracy of our conclusion may be undermined by countries such as China, Japan and Korea because strict social interventions in these countries distorted the data by turning high risk countries into low risk countries. Thus, using AT >25 ºC as a cut-off criterion for low risk is reliable. The safe temperature is independent of environmental temperature and viral mutation because the safe temperature mined from data of all the three months are the same.
We aimed to determine if there is an AT range that significantly contained SARS-CoV-2 pathogenicity. However, our study is not without limitation. First, because only single factor is analyzed, other natural and social factors not integrated into our model may affect the results, if they are in the extreme conditions. Social distancing has been proven to be effective in controlling SARS-CoV-2 pathogenicity. Extreme short social distance, such as those living in slums, may facilitate the virus transmission before high AT inhibits SARS-CoV-2 pathogenicity. Humidity also negative impacts the SARS-CoV-2 pathogenicity. In desert countries, the possible low risk AT may be higher.
Second, AT may vary region-wise within a country or even within a subnational entity.
Thus, people should evaluate risk level based on the real local AT. Besides, people migrating within a country may flatten transmission dynamic within a country. Thus, in a country with national monthly average temperature around 20 or 25 ºC, low AT areas may have more impact on national transmission dynamics. In addition, the effect of AT below 0ºC is not evaluated because the number of countries with AT < 0ºC is too less to be analyzed. Our criteria will have to be tailored to local conditions and updated as more accurate data become available. When other precipitating factors, such as social distance and humidity, become extremely favor to SARS-CoV-2 pathogenicity, the effectiveness of controlling AT could be limited. Even with all these limitations, our classification system is practical because all the above scenarios that distract the of mild and symptomatic infections, high AT season may be the best season for achieving herd immunity.
The significance of our study is beyond SARS-CoV-2. The result indicated outbreak or re-outbreak of a viral disease may be caused by moving of human, animal co-host or freight carrying the virus from low risk area to high risk area, from an area with R0 (the basic reproduction number) smaller than 1 to an area with R0 larger than 1.
Establishing potential virus list and vaccine bank may be a necessary way to prevent many local communicable viral diseases from becoming pandemic. Thus, our results also provide clue in preventing new viral pandemic in the future.

Contributors
SC conceived and designed the study, and drafted the manuscript. YR collected data and performed data analyses. Both SC and YR contributed to the interpretation of data and read and approved the final manuscript.

Declaration of interests
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.

Data sharing
Our data are accessible to researchers upon reasonable request for data sharing to the corresponding author.
new cases per million population, or the total number of days with more cases than previous days in March, April, May, and June 1 to 15 were performed separately for all 118 countries with population over five million.