Atmospheric Relative Humidity Can Increase the Spread of COVID-19


 COVID-19 has been spreading around the world since the end of 2019, and there is no sign of a slowdown. Previous studies on seasonality of similar infectious diseases have hinted that meteorological factors may influence COVID-19 outbreaks as well1. Here we show, based on data collected in 132 cities of China, that relative humidity, as an essential meteorological indicator, is positively correlated with the growth rate of incidence of COVID-19, which contradicts previous research findings. Our result suggests an increasing risk of COVID-19 cases as summer and rain seasons arrive in many places of the world. They also help countries and regions to formulate pandemic prevention and control measures and policies according to local meteorological characteristics.

Today, it has inflicted more than 7 million people and caused more than 400,000 deaths worldwide. Studies have shown that various public health measures, including handwashing, face masking, social distancing, and restricting the movement, have contributed significantly to control the spread [6][7][8][9] . However, with no effective treatments available and vaccines still months away, scientists around the world are racing to develop models that can predict if and when the second wave will come. Given the evidence that there are mitigating effects of warmer temperatures, humidity and sunlight on other infectious diseases such as pneumonia and flu, there have been speculations that weather factors may play a similar role on COVID-19 1,10 . Here we analyzed the early outbreak data in China and its correlation with meteorological factors. Our results provide evidence that atmospheric humidity facilitates rather than hinders the infection, contrary to a common belief that warm and humid weather can slow down the spread of the virus 1, 2 .
Previous studies on the relationship between atmospheric humidity and COVID-19 pandemic have shown that the higher the humidity is, the more favorable it is to control the pandamic 2,11 . However, these conclusions are contrary to the previous findings of the relationship between atmospheric humidity, pneumonia, and influenza [12][13][14] . We believe that there are many methodological problems in previous studies on the relationship between Page 3/10 atmospheric humidity and COVID-19 pandemic, which may lead to unreliable conclusions.
First, some studies only observed data for three days 2 , which is not enough to reflect the relationship between atmospheric humidity and the long-term development of the pandemic (overall incidence rate). Second, most of the studies used meteorological data at the provincial or even national level 11,15 , which is not reasonable, because the regional meteorological differences within a province or country are often large, while the average meteorological characteristics of two neighboring provinces or countries may be very close. This method may cover up the real relationship between atmospheric humidity and COVID-19 pandemic. Third, previous studies used the number of new cases per day in a period of time that was arbitrarily chosen 2,11 , without taking into account the number of initial cases and the highest number of cases in a region, and therefore could not objectively reflect the largest incidence rate in the region. Fourth, some studies included data after outbreaks began to declin 16 , which may confuse atmospheric factors with human management measures and therapeutic effects. The methods adopted in this study overcome the above problems.
We collected data from 132 Chinese cities to test the impact of natural climate factors on the pandemic's growth rate (see Method for details). We selected the pandemic data (the number of increased cases per day) of each city in the fastest-growing regions of the pandemic since January 23rd (the day of closure of Wuhan) to March (when the confirmed COVID-19 cases of all cities first reached the peak). Besides, we collected the average daily atmospheric temperature, average relative humidity, and average air pressure of the 132 cities during the period. We also collected the GDP data (2019), population density data (2018) and the distance to Wuhan of each city.
We used general linear regression for the analysis. First, we took GDP, population density, distance, temperature, temperature range and air pressure as independent variables, and the pandemic growth rate as the dependent variable for analysis (see Formula 1). The results showed that the distance negatively related to the growth rate of Page 4/10 the pandemic (B=-0.34, p=0.001). Temperature had a marginally significant correlation with the growth rate of the pandemic (B=-0.21, p=0.07). Second, we took GDP, population density, distance, temperature, temperature range, atmospheric pressure, and atmospheric relative humidity as independent variables and the pandemic growth rate as the dependent variable for analysis (see Formula 2). The results showed that atmospheric relative humidity significantly predicted the growth rate of the pandemic (B=0.40, p=0.001) (see Table 1), and the higher the relative humidity in cities, the higher the local incidence see Figure 1). Temperature was negatively associated with the growth rate of the pandemic (B=-0.26, p=0.02). The distance was negatively correlated to the pandemic growth rate (B=-0.32, p=0.001).
The findings put forward a warning to the pandemic prevention and control in special cities and regions: cities with high relative humidity are more likely to spread the COVID-19 pandemic, and special precautions should be taken.
It should be noted that this study found that relative humidity positively predicted the transmission of COVID-19, which is inconsistent with previous studies 2, 11 . The reason may be that the previous study only selected three days of data, making its conclusion very limited. The period of relative humidity data selected in this study is much more extended, so our conclusion should be universal.
Our findings provide new insights into the relationship between climate and the COVID-19 pandemic. In particular, with the coming of summer in the northern hemisphere, the atmospheric relative humidity in many cities will further increase, which may aggravate the spread of COVID-19. All urban administrative departments and medical institutions shall take corresponding prevention and control measures in advance.

Method
To investigate the effect of climate on the pandemic, we collected COVID-19 pandemic data from 132 Chinese cities for analysis. The principle of selecting cities is to include all municipalities directly under the central government and the top five cities in each province in terms of confirmed cases. If there are less than five cities in the province with the COVID-19 pandemic, all of these cities with confirmed cases were selected (for example, only two cities in Qinghai province have confirmed COVID-19 cases, so data from both cities are used for analysis). The epidemic data were collected from the National Health Commission of China 7 , Sina News 19 , and Git-Hub 18 .
Since the spreading of the pandemic may be affected by population mobility, we choose the date of Wuhan city closure (January 23th) as the starting point of our data collection to control the population mobility. As long-distance may weaken the spreading of the pandemic, we also controlled for the distance from each city to Wuhan. To better reflect the pandemic's spreading tendency, we use the date when the number of confirmed cases in the city reached the highest point for the first time as the endpoint of data collection in this city to calculate the epidemic growth rate of the city. The formula is as follows:   Figure 1 1a

Figures
The effect of relative humidity and the growth rate of COVID-19 | The horizontal axis represents the average air relative humidity of each city in the last three months (90 days). The vertical axis represents the growth rate of COVID-19; it means an average multiple of the initial value per day. 1b The relationship between relative humidity and the growth rate of COVID-19 | The color of the histogram represents the