Geographical quantification of the seasonality of transmission of COVID-19 in human populations as a function of the variability of temperatures

The occurrence of cases of COVID-19 suggests that it will likely become seasonally endemic in human populations. We seek to provide a quantification of the seasonality of the occurrence and severity of COVID-19 cases in human populations. Using global data, we show that the spatiotemporal distribution of COVID-19 cases is a function of distinct seasons and climates. We investigated this at the county and the country scale using a comparison of seasonal means, correlation analyses using ambient air temperatures and dew point temperatures, and multiple linear regression techniques. We found that most locations had the highest incidence of COVID-19 during winter compared to other seasons. Regions closer to the equator had a higher incidence of COVID-19 during the summer than regions further from the equator. Regions close to the equator, where mean annual temperatures have less variance compared to those further from the equator, had smaller differences between seasonal COVID-19 incidence. Correlation and regression analyses showed that ambient air and dew point temperatures were significantly associated with COVID-19 incidence. Our results suggest that temperature and the environment are influential factors to understand the transmission of COVID-19 within the human population. This research provides empirical evidence that temperature changes are a strong indicator of seasonal COVID-19 outbreaks, and as such it will aid in planning for future outbreaks and for mitigating their impacts.


Introduction
The ongoing COVID-19 pandemic reported over 6 million fatalities worldwide and over 700 million confirmed cases [1], as of 27 March 2023.Despite vaccinations for COVID-19 being introduced in late 2020 [2], COVID-19 persists as a threat to global public health.However, the landscape of the pandemic is very different than it was when vaccinations were introduced and much different from the initial outbreak in late 2019.The emergence of new COVID-19 variants has changed the primary concern for public health agencies.The Omicron variant has accounted for over 98% of publicly available analyzed sequences since February 2022 [3].The Omicron variant is transmitted more easily compared to previous variants [4].
With an even more transmissible virus, understanding the routes of transmission becomes important in this phase of the pandemic.One possible route of COVID-19 transmission is via aerosols [5,6].If this pathway exists, then two factors that would likely have an impact on COVID-19 transmission are ambient air temperatures and dew point temperatures because it has been previously observed that a combination of low dew point temperatures and low ambient air temperatures was directly proportional to the survival of aerosolized viruses [7][8][9][10][11].This implies that low ambient air and dew point temperatures would increase the likelihood of aerosolized COVID-19 virus transmission if it behaves like other aerosolized viruses.A recent is possible that these factors, or other meteorological factors, may significantly contribute to COVID-19 seasonality, but the current lack of evidence in the literature suggests otherwise.
This study aims to understand the seasonality of temperature on transmission of COVID-19 and prevalence as a function of seasons and climates both regionally within the US and globally.The approach undertaken examined the influence that ambient air temperatures and dew point temperatures have on generating and sustaining the seasonal pattern of disease outbreaks.The impact of the results from this study is likely to help design COVID-19 mitigation and management policies in the near future and has the potential to shape COVID-19 vaccination planning, production, administration, and efficacy in accordance with these seasonal patterns.

Method
Seasonality trends in the United States and globally were analyzed.The US was divided into 10 distinct regions based on regional differences in climate, as shown in figure 1(a).The global data was split into three regions based on climate to determine if warm and cold countries experienced a difference in the number of cases depending on the season (figure 1(b)).For the global regions, the tropics include countries between the Tropics of Cancer and Capricorn.The subtropics include countries between the Tropics of Cancer and Capricorn and 34 degrees North or South [19].The temperate countries have the remaining countries.The placement for each country was done based on the latitude of the country's centroid.
Daily COVID-19 case data was obtained from the John Hopkins University datasets available on GitHub [20].Daily COVID-19 reported cases for each county and country were separated by season from 1 March 2020, to 28 February 2022.Winter was defined from 1 December to 28/29 February, spring from 1 March to 31 May, summer from 1 June to 31 August, and fall from 1 September to 30 November.These dates are based on the meteorological seasons [21].For the global analysis, the meteorological seasons were flipped for the southern hemisphere (June, July, and August cases were assigned as winter cases, March, April, and May cases were assigned as fall, December, January, and February cases were assigned as summer cases and September, October, and November cases were assigned as spring cases).For each county and country, the datasets for daily COVID-19 reported cases for each season were compared to each other using a z-test for the difference in the mean values.Equation (1) shows how the z-value was calculated for each comparison.This was done for all possible comparisons to investigate which seasons had significantly more daily COVID-19 cases than other seasons within a specific county or country To study the seasonal impact of ambient air temperature and dew point temperature on COVID-19 cases, a correlation analysis was conducted.The objective of this analysis was to determine if the conditions contributing to COVID-19 transmission hypothesized in Usmani et al (with ambient air temperatures outside the 17 • C-24 • C range and dew point temperatures below 0 • C leading to outbreaks) were significantly associated with COVID-19 cases at the beginning of COVID-19 outbreaks, or when COVID-19 cases first started increasing over multiple days to an epidemiological peak.The objective was also to determine if the subsidence of these outbreaks, or when COVID-19 cases decreased over multiple days back down to an approximate average following the epidemiological peak, was significantly associated with the environmental conditions retreating to the hypothesized ranges of reduced COVID-19 transmission.
Ambient air temperature and dew point temperature data were collected for each county through the online PRISM datasets [22].However, for the global scale analysis, ambient air temperature and dew point temperature data were collected using daily MERRA-2 data [23].Historical temperature data was considered up until the end of January 2022.For each county and country, the obtained temperature data was examined against a 7 day moving average of daily COVID-19 cases with up to a 14 day lag for temperatures.Each parameter was lagged up to 14 d for the correlation analysis to examine the environment under which the transmission would have occurred in.COVID-19 incubation times are said to rarely exceed 14 d [24], so a maximum of 14 d lag was deemed appropriate.A 7 day moving average was used so that inconsistent data reporting, such as some counties only reporting cases every three to five days, would not impact results.Equation (2) shows how the moving average was calculated

Cases used for Day
C n = Number of cases on Day n .
One county from each of the ten US regions was randomly selected to be studied.In order to examine if temperatures had an impact on COVID-19 prevalence, each county was divided up into specific start and end points for the correlation analysis based on when a COVID-19 outbreak had occurred and when the same outbreak had ended.These start and end points were determined manually.The start point was determined to be the date on which cases were at their lowest point before they started to increase consistently (over multiple days) to an epidemiological peak.The endpoint was determined to be the date on which cases reached a low point after the epidemiological peak, often identified by daily cases staying nearly the same or increasing again.
Correlation coefficients were calculated both for the beginning of each outbreak and the subsidence of each outbreak.The number of peaks for each county varies, and the number of peaks was also manually identified by identifying points on each county's epidemiological curve where cases clearly increased over multiple days to a daily caseload much higher than the average for each county and then decreased over multiple days back down to an approximate average for each county.Figure 2 provides visual clarification of the start and end dates for outbreak beginnings and subsidence using the example of Person, North Carolina, which is one of the counties selected for this analysis.An outbreak beginning would use data from the first date of the outbreak until the date of the epidemiological peak, and the outbreak subsidence would use data from the epidemiological peak to the last date of outbreak subsidence.
Pearson's correlation coefficient was calculated for the beginning of each outbreak and the subsidence of each outbreak, as shown in equation (3).Four parameters were compared to the moving average of COVID-19 cases in this analysis.Parameter 1 is daily average ambient air temperature.Parameter 2 is daily average dew point temperature.Parameter 3 and Parameter 4 are based on the temperature ranges previously studied in Usmani et al [12] Parameter 3 is a measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range in absolute value.Parameter 4 is a measure of how far daily average dew point temperature is below 0 • C. Parameters 3 and 4 are included to see if using the specified temperature range from Usmani et al [12] will have more significant correlations compared to using average air and dew point temperature x i = values of the parameter in the sample,y i = values of COVID − 19 cases in the sample x t−h = mean of the values of the parameter for period t minus lag h y t = mean of the values of the COVID − 19 cases for hypo period t.
In order to determine the parameters of significant association to COVID-19 cases, it was decided that any significant correlation value (p < 0.05) within the 0-14 d period would constitute that parameter as having a significant association with COVID-19 cases for that outbreak.This means that if Parameter 1 had a significant correlation value for 14 d of lag, but no significance for any other amount of lag, it was still counted as a significant association with COVID-19 cases.This was done because the number of scenarios and the exact amount of lag that produced significant correlation values were inconsistent, and it is impossible to know on which days transmission occurred.The correlation coefficient sign also had to match with what was expected for it to be considered significant.For instance, a negative coefficient would be expected in the colder months for average air temperature compared to cases, while a positive coefficient would be expected in the warmer months.This was done to calculate the total number of outbreaks that matched the conditions of the proposed hypothesis compared to the total number of outbreaks.Some outbreaks may have had a significant opposite correlation sign to what was expected, but they were not counted to this total and do not impact the results shown.
To examine the seasonal impact of ambient air temperature and dew point temperature on COVID-19 cases in different countries, nine countries were selected.A good geographic distribution would be best for this analysis, so at least one country from each continent was selected to study.Varying latitudes were also important, so many continents had two countries selected to examine the spatial difference between countries.Two countries very close to the equator were selected to see how minor seasonal variability impacted results.At least one country from each climate region and each hemisphere was selected.Most countries selected are in the northern hemisphere due to the majority of global landmass being in the northern hemisphere.Each selected country had an average computed for daily temperatures by averaging the temperatures of all the daily data points within the borders.The beginnings and ends of outbreaks were analyzed in the same way that they were for individual counties, and the same four parameters were examined and lagged up to 14 d against a seven-day moving average of daily COVID-19 cases.
To examine ambient air temperature and dew point temperature correlations more robustly, all countries and counties were examined in the following analysis.To see if warm periods in specific locations had high correlations between ambient air temperature and daily COVID-19 cases, mean ambient air temperatures were compared to daily COVID-19 cases on days when the mean ambient air temperature for a country or county was above 24 • C. If the temperature were below this value, a value of 0 was assigned for daily cases on that day.Pearson's correlation coefficient was calculated for every county and country with average ambient air temperatures lagging between 0 and 14 d before cases for the period of 1 March 2020, to 31 January 2022.To see if cold periods in specific locations also produced these correlations, the same analysis was done using COVID-19 daily cases that occurred on days when mean ambient air temperatures were below 17 • C.
The same process was done to examine the relationships between high absolute humidity and low absolute humidity periods across different locations, except the daily COVID-19 cases were limited to days in which mean dew point temperatures were above 0 • C to examine high absolute humidity conditions and below 0 • C to examine low absolute humidity conditions.
To examine the impact that latitude and general climate have on these results, a running percentage of counties and countries with significant correlation between the temperatures and cases was used.For the counties, it was done starting from the lowest latitude county to the highest latitude county.For instance, if the county with the lowest latitude was 10 • , then that would be the first county used in the total and the only county used in that current total.If it showed significant correlation in any of the 0-14 d of lag, then the percentage would start at 100% because 100% of all counties currently being counted showed significant correlation.The next county counted would be the one with the next lowest latitude, and the number of counties with significant correlation would be divided by the total number of counties being counted.This would persist until all counties were counted.The same was done for all countries, but the northern and southern hemispheres were split into two different running percentages.However, the starting place differed depending on the parameter because a visual parabolic relationship was desired to be shown.For high ambient air temperatures and high dew point temperatures, the starting point was the equator for both hemispheres, while the starting point for low ambient air temperatures and low dew point temperatures was the southern-most or northern-most country.
To examine how ambient air temperatures and dew point temperatures interact together to influence COVID-19 transmission and to see if both of these parameters are significant, the best (highest R 2 value) multiple linear regression models were developed for each county and country using COVID-19 case data and a combination of temperature parameters.The linear regression models had to use two parameters, with one parameter including a measure of ambient air temperature (parameter 1 or 3) and one parameter including dew point temperature (parameter 2 or 4).

Results
The primary objective of the analysis was to study the seasonal trends of COVID-19 regionally and globally and to determine if warmer and colder countries differ in the number of cases, such that it can be related to modalities of ambient temperatures.Table 1 shows the results of the county comparison tests expressed as a percentage of counties in each region that had significantly (p < 0.05) more daily cases of COVID-19 in one season than another.All the regions had most of their counties with significantly more cases in winter compared to any other season.Only the West North Central, East North Central, and Rocky Mountains regions had some counties with significantly more cases in fall than in winter.Most counties in each region had fall as the season with the second most cases except for the Southeast, which had summer as second.Only the Southeast, West South Central, and California regions had at least 10% of counties with significantly more cases in summer than in fall.Spring was ranked last among the regions, except for the Northeast, which had more counties with significantly more cases in spring than in summer.
Table 1 also shows the results for the country seasonal comparisons.The temperate and tropical regions had most of their countries with significantly more cases in winter compared to any other season.The subtropical region had a higher percentage of countries with significantly more cases in summer compared to winter by 7•7% of countries.Both the tropical and subtropical regions had a higher percentage of countries with significantly more cases in summer than in fall or spring.Summer ranked the lowest among seasons for the temperate countries, with fall ranking second and spring ranking third.Spring was ranked last among the subtropical and tropical regions, and fall was ranked third for both.The key results from this analysis show that, on county scales primarily in the USA, most of the cases were observed during winter season.However, warmer regions in the USA had a significant number of cases during summer season.The temperate countries experienced many more cases in winter when compared to other seasons, while the subtropics and tropics had the most significant caseloads in winter and summer seasons.
The results of the correlation analysis for the ten counties are shown in figure 3. Again, the objective of this analysis was to determine if the conditions contributing to COVID-19 transmission hypothesized by Usmani et al were present and showed a significant correlation to the beginning of COVID-19 outbreaks and to determine if the subsidence of these outbreaks was also significantly associated with the environmental conditions subsiding, meaning significant correlations discussed here are not independent of correlation sign.Significant correlations were more common in outbreak beginnings compared to outbreak subsidence.Parameter 1 (average ambient air temperature) had significant correlation to 24 out of 30 total outbreak beginnings and 13 out of 20 outbreak subsidence in all counties.Parameter 2 (average dew point temperature) had significant correlation to 23 out of 30 outbreak beginnings and 9 out of 20 outbreak subsidence in all counties.Parameter 3 (absolute value measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range) had significant correlation to 21 out of 30 outbreak beginnings and 13 out of 20 outbreak subsidence in all counties.Parameter 4 (measure of how far daily average dew point temperature is below 0 • C) had significant correlation to 15 out of the 22 outbreak beginnings with dew point temperatures below 0 • C and 2 out of the 13 outbreak subsidence with dew point temperatures below 0 • C. The outbreak beginnings and subsidence without any dates with dew point temperatures below 0 • C were not counted in this total for parameter 4 or in figure 3.
The results of the correlation analysis for the nine countries are shown in figure 4. Significant correlations were more common in outbreak beginnings compared to outbreak subsidence.Parameter 1 (average ambient air temperature) had significant correlation to 28 out of 35 total outbreak beginnings and 17 out of 28 outbreak subsidence in all countries.Parameter 2 (average dew point temperature) had significant correlation to 27 out of 35 outbreak beginnings and 18 out of 28 outbreak subsidence in all countries.Parameter 3 (absolute value measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range) had significant correlation to 22 out of 35 outbreak beginnings and 13 out of 28 outbreak subsidence in all countries.Parameter 4 (measure of how far daily average dew point temperature is below 0 • C) had significant correlation to 11 out of the 18 outbreak beginnings with dew point temperatures below 0 • C and 9 out of the 15 outbreak subsidence with dew point temperatures below 0 • C. The outbreak beginnings and subsidence without any dates with dew point temperatures below 0 • C were not counted in this total for parameter 4 here or in figure 4. The main results here are that outbreak beginnings had more instances of significant correlation than outbreak subsidence for both counties and countries, the average temperature parameters performed slightly better than the absolute value measures, and the majority of outbreak beginnings for both counties and countries had significant correlation to all four parameters.
Figure 5 shows the graphical results of the running percentages for all four correlation scenarios at a county scale, and figure 6 shows the same but for a country scale.The objective of this analysis was to show a more robust relationship between COVID-19 cases and periods of especially high or especially low temperatures across latitudes.Linear trendlines and R 2 values were computed and shown for all graphs.The  running percentage generally decreases as latitude goes further from 0 in the scenario of high ambient air temperatures for both the countries and counties.The running percentage generally increases as latitude goes further from 0 for the scenario of low ambient air temperatures.The running percentage generally decreases as latitude increases for the scenario of high dew point temperatures for both counties and countries in both hemispheres.Finally, the running percentage generally increases as latitude goes further from 0 in the scenario of low dew point temperatures.The key result here is that the more extreme latitude counties and countries showed more correlations to COVID-19 cases during the colder conditions with ambient temperatures, while the counties and countries closer to the equator showed high correlations to COVID-19 cases during warmer conditions (with ambient temperatures).Additionally, the average correlation value across all counties and countries was computed for each day of lag as well as the percentage of counties and countries that had significant correlations on that day of lag.These are shown in table 2 for the counties and table 3 for the countries.Generally, the percentage of counties and countries with significant correlations was higher with fewer days of lag and with low ambient air and dew point temperatures compared to high ambient air and dew point temperatures.The correlation coefficients were also generally higher among low ambient air and dew point temperatures compared to high ambient air and dew point temperatures.
The objective of the linear regression analysis was to determine if both ambient air temperature and dew point temperature were significant parameters in their relationship to COVID-19 cases.Table 4 shows the results of the best linear regression model for each county, including which parameters in each model were used and if the coefficients for each parameter in the model were significant.The average R 2 value was 0•203 with a standard deviation of 0•099 for all counties.Parameters 2 and 3 created the best models in 7 out of 10 counties, and both of the used parameters were significant in 7 out of 10 models with the dew point The results for the country linear regression models are in table 4. The average R 2 value was 0•144 with a standard deviation of 0•054 for all countries.Parameters 2 and 3 created the best models in 5 out of 9 countries, and both of the used parameters were significant in 4 out of 9 models with the dew point temperature parameter being insignificant in 4 models and the ambient air temperature parameter being insignificant in 1.The main results for this analysis are that the combination of average dew point temperatures and the absolute value measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range created the best linear regression models in most instances, and the parameter that showed insignificance most often were the dew point temperature parameters.

Discussion
When examining the results of the seasonal analysis, most counties in the United States experience more COVID-19 cases in winter than in any other season.This supports the hypothesis that COVID-19 transmission will be highest in cold, dry environments.Only two regions (the Rocky Mountains and West North Central) contained more than 1% of counties with significantly more cases in another season when compared to winter, and they are two of the coldest regions in the United States.Both regions showed significantly more cases in fall compared to winter.This is likely because temperatures decrease earlier in the year in those counties compared to other counties in the United States, suggesting those counties will have an environment to support a COVID-19 outbreak earlier in the year.This would also indicate that the time when temperatures start decreasing is more important than how low temperatures get when a COVID-19 outbreak begins, which is supported by the results of the correlation analysis.
While winter leads among all seasons in terms of reported COVID-19 cases for all regions, the other seasons varied in ranking from region to region.All except one region had fall as the season with the second most cases.It was expected that fall would have significantly more cases than summer and spring because temperatures will be lower; however, this is only true in colder regions.It is expected that in the warm regions of the United States, high temperatures during the summer will increase the likelihood of COVID-19 transmission.The southeast, which is among the warm regions of the United States, was the only region that had a higher percentage of counties with significantly more COVID-19 cases in summer than in fall.Additionally, the warm regions of West South Central and California had over 10% of counties with significantly more COVID-19 cases in summer than fall and had a much lower percentage of counties with significantly more cases in fall than summer compared to other regions in the United States.
The Northeast region did not have spring ranked as the season with the least number of COVID-19 cases; instead, summer was the season with the least cases.This could be because the Northeast is a very cold region with temperatures in summer that are not much different than in spring.This would also explain why the only other regions to have than 15% of counties with significantly more cases in spring than in summer are the cold regions of the Rocky Mountains, West North Central, and East North Central.
In the global seasonal analysis, the results are very similar in the warmer and colder countries to the warmer and colder US regions, with some slight differences.Winter was the season with the most COVID-19 cases in temperate countries, followed by fall, spring, and then summer.Tropical countries also had winter as the season with the highest number of COVID-19 cases, but summer was very close behind, followed by fall and then spring.There was less than a 10% difference between tropical countries with significantly more cases in winter than in summer and countries with significantly more cases in summer than in winter.Tropical countries are the warmest countries in the world, so it is expected that the summer temperatures will be high enough to support outbreaks.However, annual temperatures in some tropical countries very close to the equator vary only a few degrees Celsius on average [25], so there should not be much of a difference in outbreak potential at different points in the year in these countries.This may be why many of the comparisons for the tropical countries do not show a high percentage of countries having significantly more cases in one season than another.Some tropical countries do have larger variations, which does account for some of the differences in seasonal variation.The highest percentages obtained in the comparisons for the tropics were when other seasons are compared to spring, which might be due to spring 2020 being the beginning of the pandemic.It is also likely that the lack of temperature variation near the equator is equally responsible for the poor results in Indonesia for the correlation analysis.
The subtropics did not have winter as the season with the most COVID-19 cases.Instead, the season with the most cases was summer.In the subtropics, temperatures get cold enough in the winter to support outbreaks, and temperatures get warm enough in the summer to support outbreaks.It would be expected that the winter outbreaks would be worse because of the influence of both the ambient and built environment.However, the difference between the summer and winter comparisons is less than seven percent, so the discrepancy is relatively small.Subtropical countries had the third most cases in fall and the least cases in spring, which is the same as the results for the tropics and expected.
The results of the outbreak correlation analyses indicate that the timing of temperature changes is a better predictor of when outbreaks start and subside than the timing of when temperatures rise or fall below a certain threshold, as seen by better performance of parameters 1 and 2 (the average values of ambient air temperature and dew point temperature, respectively).However, the linear regression models showed that using the absolute value beyond the 17 • C to 24 • C range of ambient air temperature (parameter 3) led to a better model than using the average ambient air temperature.This would indicate that the severity of a  Indonesia had the same R 2 when using both parameter 1 and parameter 3, which is why both are shown.
COVID-19 outbreak may be better predicted by examining how far air temperatures are outside this threshold, but the timing of the outbreaks may be better predicted by looking at when average temperatures change.The same is not true for dew point temperatures, as average dew point temperatures (parameter 2) performed better in the linear regression models than measuring the dew point temperature below 0 • C (parameter 4).Although, this might be because dew point temperatures are not below 0 • C for much of the year, causing this parameter to perform poorly in a model.This might also be an indication that the temperature thresholds vary from region to region or that the selected thresholds need to be modified.Since the linear regression models showed that both parameters of dew point temperature and ambient air temperature were significant in most models, there is evidence that the interaction of both contributes to COVID-19 transmission, not just one of the two factors.When parameters were insignificant in the models, it was mostly the dew point temperature parameters.This is likely because the linear regression models examined COVID-19 cases across several seasons, and dew point temperatures are not expected to be as significant a contributor as ambient air temperatures for summer outbreaks as they would be for winter outbreaks.
All four parameters exhibited a significant correlation to at least 60% of COVID-19 outbreak beginnings, with parameter 1 a showing significant correlation to 80% of outbreak beginnings in both the US counties and the countries.This is evidence that temperature trends are strong indicators of COVID-19 outbreak beginnings in any climate.However, each parameter's performance was unsatisfactory for COVID-19 outbreak subsidence.This could be due to several possibilities, but the most likely explanation is that most COVID-19 outbreaks will naturally subside once the epidemiological curve has reached its maximum carrying capacity in the population.Once the virus has spread rapidly, the number of people infected each day will have to decrease once more of the vulnerable population become infected.Therefore, the subsidence of rapid outbreaks was not predicted well by environmental parameters.Another likely explanation is that the outbreaks subsided due to human interventions, such as containment measures and other non-pharmaceutical interventions, or human behavior changing in response to the outbreak, both of which would reduce the transmission throughout a community more quickly than environmental parameters would predict.
The county-level results of the correlation analysis showed that the higher latitude counties generally had low values of correlation between COVID-19 cases occurring on high ambient air temperature days and mean ambient air temperatures.This provides evidence that COVID-19 transmission in warmer, tropical climates will be more influenced by high ambient air temperatures compared to colder climates.For the low dew point temperature correlations, there was no evidence of a linear increase or decrease across latitudes since the R 2 value was very low.This is likely because COVID-19 transmission is not expected to be impacted differently across climates by high dew point temperatures.For low ambient air temperature and low dew point temperature correlations across counties, there is not enough evidence to support that COVID-19 cases will be influenced by either parameter differently at a county (or localized) level.However, the United States is overwhelmingly a temperate climate zone with few counties in the lower latitudes, therefore it likely provides an explanation as to why the correlation seems highly significant at all latitudes.
For the country analysis, the trend for high ambient air temperatures showed that the percentage increased as latitudes approached near the equator.This is perhaps expected, as there should be a more significant correlation in the warmer climates for this scenario.For the low ambient air temperatures, the percentages increased as latitudes moved further from the equator.High dew point temperatures showed an increase in percentage as latitude increased for both the southern and northern hemispheres.As mentioned previously, high dew point temperatures are not expected to be influential on COVID-19 transmission with varying climates.Finally, the low dew point temperature correlations showed that percentages increased moving further from the equator, as they did in the low ambient air temperature scenario and are the expected results based on the initial hypothesis.This is in general agreement with results between the county and country-level analyses also provides strong evidence for our hypothesis.
As for the correlations generated from this analysis, the negative correlations among the low ambient air temperature and low dew point temperature conditions were expected, as were the positive correlations between high ambient air temperatures.The correlations for the high dew point temperature scenario were by far the lowest compared to the other scenarios for both a country and county scale.The correlations for the low ambient air temperatures and low dew point temperatures were also higher than the high ambient air temperature correlations.This supports the hypothesis of the outbreaks being more severe in colder conditions compared to warmer conditions, with the ambient and built environment supporting the outbreak in the cold conditions compared to just the built environment in the warm conditions.
It is important to note that significant correlations between temperatures and COVID-19 cases during outbreaks may be due to several factors, not just those proposed by the Usmani et al [12] hypothesis.Weather's impacts on COVID-19 transmission and detection can vary [26], with changes in weather having possible impacts on COVID-19 detection for up to a month or more.Furthermore, different locations can experience different variations.Because of this, it is very difficult to assign a specific reason for significant correlation values without additional analysis and evidence.However, compounding the correlation analysis results with the other evidence gathered in this study does help to support the posited hypothesis, but additional studies would be required to examine the exact transmission conditions occurring before outbreaks to provide critical evidence for this hypothesis.
There is an argument to be made that observed COVID-19 seasonality may be due to the emergence of COVID-19 variants and subsequent outbreaks.In the United States, for instance, variant emergence did not follow the seasonal nature that was observed in this study throughout.The Alpha variant was discovered in the United States in December of 2020 [27] but did not make up more than 15% of COVID-19 cases in the United States until spring of 2021 [28], over a year after the pandemic started and seasonality was first observed.The Delta variant was first identified in spring of 2021 (March) [29] but did not begin to make up more than 50% of analyzed sequences until the summer of 2021 (June) [28], eventually making up over 99% of cases by late August [28].Summer outbreaks related to changes in ambient air temperatures likely allowed the variant to spread more rapidly, which is why it was unable to make up a significant percent of analyzed sequences in the spring when ambient air temperatures were not increasing.
Contrastingly, the Omicron variant was first identified in late November of 2021 [30] and made up 99% of analyzed sequences by the end of January 2022 [28].This rapid emergence of Omicron is assisted by its increased transmissibility, but it is also likely ambient air temperatures and dew point temperatures being supportive of COVID-19 aerosolization in the winter allowed for such a quick spread of this variant, especially considering other variants remained a small percent of analyzed sequences for months before making up a majority of analyzed sequences.With all variant emergences in the United States considered, the general trend appears to be that the emergence of variants does not follow a seasonal pattern, but the significant transmission of these variants does seem to follow a seasonal trend.Therefore, it may be inaccurate to claim that COVID-19 seasonality is driven by the emergence of variants.

Conclusion
COVID-19 has been a constant threat since its emergence and still the most obvious pattern in the reported incidences is not properly explained or understood.Therefore, we present the findings on the environmental influence on the transmission of COVID-19 in the human population.We argue that COVID-19 cases increase when ambient air temperatures and dew point temperatures support aerosolization and interactive human behavior, which is supported by winter being the season in which most COVID-19 cases were reported.The reported COVID-19 cases also increase in warmer months once the temperature starts to increase, particularly in warm regions, and the hypothesis of human interactions increasing in the built environment during these conditions, while still untested, appears to be a reasonable explanation of the patterns observed here.
The extent of seasonal outbreak predictability does greatly depend on factors not related to temperatures, such as non-pharmaceutical interventions, like lockdowns, mask mandates, and more.These interventions could help reduce transmission and the likelihood of seasonal outbreaks.Accounting for these, along with other useful data, such as vaccination data, may be useful in future studies to determine if seasonal outbreaks are still likely when the proper precautions are present.However, based on the evidence in this study, it appears that temperatures are, indeed, a very strong indicator of seasonal outbreaks, and may be useful in development of anticipatory decision making for understanding where and when an outbreak is likely.
While this study only examined as far as the end of February 2022, the seasonal trends of COVID-19 do not appear to be subsiding, with numerous outbreaks already occurring in the summer of 2022, even though surveillance of COVID-19 is declining [31].This observation along with evidence that COVID-19 caseloads will be experienced in seasonal patterns at both a regional and global scale based on ambient air temperatures and dew point temperatures in all climate conditions may be suggestive of a seasonal nature of COVID-19 emerging across spatial scales.
The findings of this study suggest that COVID-19 mitigation measures should be designed to mirror this seasonal pattern.Providing recommendations in a seasonal time frame may even increase participation among the public by reducing their fatigue with these policies for half of the year.
) x1 = mean daily cases of season 1, x2 = mean daily cases of season 2, σ 1 = standard deviation of cases of season 1, σ 2 = standard deviation of cases of season 2, n 1 = number of days in season 1,n 2 = number of days in season 2.

Figure 1 .
Figure 1.(a) The geographic regions created for the regional analysis, (b) the geographic regions created for the global analysis.

Figure 2 .
Figure 2. A visual representation of how COVID-19 case outbreaks were divided in this analysis for Person County, North Carolina.

Figure 3 .
Figure 3.The results from the correlation analysis for all ten randomly selected counties for all four parameters.The totals for the selected counties are shown at the bottom of the figure.Parameter 1 is average ambient air temperature, parameter 2 is average dew point temperature, parameter 3 is the absolute value measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range, and parameter 4 is an absolute value measure of how far daily average dew point temperature is below 0 • C.

Figure 4 .
Figure 4.The results from the correlation analysis for all nine randomly selected countries for all four parameters.The totals for the selected countries are shown at the bottom right of the figure.Parameter 1 is average ambient air temperature, parameter 2 is average dew point temperature, parameter 3 is the absolute value measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range, and parameter 4 is an absolute value measure of how far daily average dew point temperature is below 0 • C.

Figure 5 .
Figure 5.Running percentage of United States counties with significant correlation in the four correlation scenarios by increasing latitude.

Figure 6 .
Figure 6.Running percentage of countries with significant correlation in the four correlation scenarios.

3 .
Average correlations and percentage of significance for all United States counties for all days of lagged temperatures in each correlation scenario.Average correlations and percentage of significance for all countries for all days of lagged temperatures in each correlation scenario.

Figure 7
is included to visually represent when different COVID-19 variants emerged and became prevalent in the United States.
Figure 7  has vertical lines that indicate the dates of discovery for major COVID-19 variants.These variants include Alpha, Delta, and Omicron.The horizontal lines and text (e.g.'Alpha') show the time from the first date of that variant's discovery in the United States to the date of the first discovery of the next major variant.

Figure 7 .
Figure 7. Dates of first discovery in the United States for major COVID-19 variants compared to COVID-19 cases in Person County, North Carolina.

Table 1 .
The results of the seasonal comparison analysis for Spring 2020-Winter 2022.

Table 4 .
The results from the linear regression models.This parameter showed significance in the linear regression model.Parameter 1 is average ambient air temperature, parameter 2 is average dew point temperature, parameter 3 is the absolute value measure of how far the daily average ambient air temperature is away from the 17 • C to 24 • C range, and parameter 4 is an absolute value measure of how far daily average dew point temperature is below 0 • C. a