key: cord-0910855-z9aor6nc authors: Choma, J.; Mellado, B.; Lieberman, B.; Correa, F.; Maslo, C.; Naude, J.; Ruan, X.; Hayashi, K.; Monnakgotla, K.; Dahbi, S.-E.; Stevenson, F. D. title: Evaluating temperature and humidity gradients of COVID-19 infection rates in light of Non-Pharmaceutical Interventions date: 2020-07-25 journal: nan DOI: 10.1101/2020.07.20.20158071 sha: 88586a4daac17507ac07591a4c2a8efe6c99583e doc_id: 910855 cord_uid: z9aor6nc We evaluate potential temperature and humidity impact on the infection rate of COVID-19 with a data up to June 10th 2020, which comprises a large geographical footprint. It is critical to analyse data from different countries or regions at similar stages of the pandemic in order to avoid picking up false gradients. The degree of severity of NPIs is found to be a good gauge of the stage of the pandemic for individual countries. Data points are classified according to the stringency index of the NPIs in order to ensure that comparisons between countries are made on equal footing. We find that temperature and relative humidity gradients do not significantly deviate from the zero-gradient hypothesis. Upper limits on the absolute value of the gradients are set. The procedure chosen here yields 6{middle dot}10^-3{degrees}C^-1 and 3.3{middle dot}10^-3(%)^-1 upper limits on the absolute values of the temperature and relative humidity gradients, respectively, with a 95% Confidence Level. These findings do not preclude existence of seasonal effects and are indicative that these are likely to be nuanced. vary from country to country. If one analyses the infection rate in different countries at a fixed time it is possible to run 20 into the error of comparing outcomes at different stages of the pandemic. For instance, at a given time a country can 21 be at the early stages of the pandemic, which is characterized by low infection rates, while another can on be its way to reaching the apex with large infection rates. If both countries happen to display different temperatures or humidity 23 levels, the analysis will yield an apparent gradient. Also, if one analyses infection rates over prolonged periods of time 24 over which climatic conditions change significantly in a country, the analysis will also pick up an apparent gradient 25 in the absence of genuine dependencies. The "first wave" in Europe commenced towards the end of the winter and 26 was completed in the spring. At the time when the infection rates were large in some European countries in other 27 continents, such as South America which happened to be warmer and more humid, displayed lower infection rates. 28 From here One may be tempted to conclude on the presence of a gradient. 29 In order to reduce the probability of picking false gradients it is essential to compare infections rates for different 30 countries at similar stages of the progression of the pandemic. Given the large geographical footprint of the pandemic 31 attained so far it is possible to achieve a large enough differential in temperature and humidity in order to pin down 32 potential macroscopic gradients. This is achieved here by classifying the data according to the level of severity of 33 the non-pharmaceutical interventions (NPIs) applied in a country. We find that the severity of the total NPI policy 34 is a good metric to gauge the progression of the pandemic. It must also be noted that countries with a number of 35 positive cases below a certain threshold are not considered. This is motivated by the need to exclude countries that 36 find themselves at the early stages of the pandemic, which is a difficult period to model. Several measures have been In our work we are looking at countries with more than 5000 cases and thus far as of June 10, 2020 it a useful comparative parameter. It can also be used to show the progression of a pandemic in a country and to 70 determine an overall classification of a response as "tight" or "loose" control. The OxCGRT original version that was used for this paper takes into account the following NPIs in the stringency The OxCGRT stringency index is given by: where p J is defined by: with G J = 1 if the effect is general (and 0 otherwise), and N J is the cardinality of the intervention measure [22, 20] . Educational Facilities Closed U3: Non-essential Services Closed U4: Travel Severely Limited U5: Initial Workplace Closure U6: Banned Mass Gatherings The OxCGRT database contains data for 133 countries however no data is included for US states. In order to be 89 able to compare US states to countries, we mapped the known available levels of intervention in the US to match index is created. The following equation was developed: where v i is a number out of 100 indicating the extent to which each of the interventions are imposed. Due to lack of data on the Travel Severely Limited intervention on the IHME dashboard. It was required to source 99 US travel restrictions information from other US news sources [24, 25, 26]. Using the same logic used by OxCGRT 100 team the following equation for v i was introduced: where U i is ordinal and can vary from 0 to the cardinality of the specific intervention measure, N i . This is done to 102 incorporate levels of implementation of a specific intervention into the stringency calculation. Based on the data the 103 only intervention that requires levels of implementation is the U 4 intervention. The ordinal levels, 0-3, were allocated 104 for U 4 in order to include the relevant levels of implementation that were found to be applicable to the Travel Severely 105 Limited intervention. where 114 γ = dR dt 4 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 July 25, 2020. . https://doi.org/10.1101/2020.07.20.20158071 doi: medRxiv preprint Therefore, it is of the essence to make every effort to disentangle the intrinsic temporal evolution of a pandemic due to 124 the specific characteristics of a country from climate modulations that may potentially impact it. This is a challenging 125 exercise for a country in isolation. In the absence of a reliable overarching model, potential seasonality effects driven 126 by regional temperature and humidity need to be extracted iteratively by means of a global analysis that later could be 127 feed into a more localised scrutiny of the pandemic. Should significant temperature and humidity gradients be impacting the dynamics of the pandemic at the macro-129 scopic level, variations should be observed in a global and regional analysis of the data. In order to probe these (which was not certified by peer review) 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 July 25, 2020. lockdown measures more or less gradually with different levels of success. As discussed in Section 1, an attempt is made here to avoid picking up false temperature and humidity gradients 142 when comparing infection rates in countries or groups of countries that happen to be at different stages of the pan-143 demic. A first iteration is performed here by evaluating temperature and humidity gradients in groups of countries 144 that display "loose" or "tight" NPI control. This approach is effective as long the data displays sufficiently large dif- Given the differentials present in the data (see Section 4), this first iteration should be able to pick up gradients 149 greater or comparable to 10 −2 . In order to refine the measurement of the gradient an approach more sophisticated than 150 the discrete classification adopted here. For this purpose, parametric corrections would need to be implemented in the 151 data to account for the stringency and the efficiency of the NPIs. stringency index for the different states of the US is computed by the authors (see Section 2). This is done for the 156 period January 25 th 2020 to April 14 th , 2020. The epidemiological data was collected from the Johns Hopkins University data repository which contains the 158 global daily updates of the total cases, number of death and recovery cases [27] . In our study we focus on three cases. For the first case we consider all countries with more than 5000 total cases and US States with more than 2000 total 160 cases. This is enforced in order to ensure that the epidemiological parameters be stable, as these are calculated on a 161 daily basis. For the second case, which is called loose control, we look at points where infection rate is not 0 and the stringency 163 index is less than 50; lastly we have the tight control data sample, which focuses on the the period that starts two weeks 164 after the maximum stringency index is achieved. 165 Table 2 shows the of countries with more than 5000 cases and US States with more than 2000. The number of 166 data points before and after splitting into loose and tight NPI control are also given in consists of hourly and synoptic observations from over 35 000 stations worldwide [28] . These stations are identified 170 1 A data point is defined a vector of epidemiological parameters such that the estimated infection rate is not equal to zero. Data points are estimated on a daily basis. 6 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 July 25, 2020. . (10) Here we use the relative humidity. This is given by the ratio of the two pressures and is measured in percentage: The graphs in Figure 4 show the regression analysis of the transmission rate against temperature (top row ) and 185 relative humidity (bottom row) for the overall global data set discussed in Section 3. Results are shown for all three 186 7 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 July 25, 2020. . https://doi.org/10.1101/2020.07.20.20158071 doi: medRxiv preprint cases we have considered: the full data set, loose and tight controls, respectively. One appreciates the wide span 187 of temperatures and relative humidity displayed by the data sample used here. The differential in relative humidity 188 ranges from 90% to 80% for the full data set and the tight control sample, respectively. The corresponding differentials The procedure followed here (see Section 3) is intended to capture the leading terms in the underlying dynamics 211 driven by temperature or humidity. These terms appear small and seem compatible with the intrinsic uncertainties 212 derived from reporting and epidemiological parameters. Given that the size of the terms are below the threshold of the 213 analysis sensitivity, further implementation of more sophisticated functional forms of multi-dimensional approaches 214 are not warranted here. It is relevant to note that the weak correlation between temperature and humidity with the infection rate is observed 216 in the full data set, even before the classification according to the severity of the NPIs is implemented. This seems to 217 8 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) 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 July 25, 2020. Asia 2.17 · 10 −3 ± 4.31 · 10 −4 < 0.001 −4.17 · 10 −3 ± 2.79 · 10 −3 0.136 1.70 · 10 −3 ± 1.38 · 10 −3 0.218 Africa −1.37 · 10 −3 ± 5.61 · 10 −4 0.014 3.52 · 10 −3 ± 2.90 · 10 −3 0.226 −1.45 · 10 −3 ± 1.83 · 10 −3 0.427 Europe −1.60 · 10 −3 ± 1.91 · 10 −3 0.402 −9.96 · 10 −4 ± 1.81 · 10 −3 0.581 −3.00 · 10 −3 ± 4.52 · 10 −3 0.511 North America 3.80 · 10 −4 ± 5.98 · 10 −4 0.525 4.33 · 10 −3 ± 5.67 · 10 −4 < 0.001 −1.32 · 10 −3 ± 5.04 · 10 −4 0.008 South America −3.51 · 10 −3 ± 5.48 · 10 −3 0.005 −1.97 · 10 −2 ± 2.78 · 10 −3 < 0.001 −1.87 · 10 −3 ± 4.06 · 10 −3 0.644 Oceania 1.49 · 10 −3 ± 1.10 · 10 −3 0.177 −2.94 · 10 −2 ± 7.08 · 10 −3 < 0.001 1.55 · 10 −3 ± 2.39 · 10 −3 0.516 USA −5.75 · 10 −4 ± 2.51 · 10 −3 0.819 2.32 · 10 −3 ± 2.92 · 10 −3 0.427 −2.22 · 10 −3 ± 3.70 · 10 −3 0.548 Global 8.32 · 10 −4 ± 1.63 · 10 −3 0.607 −7.45 · 10 −4 ± 1.39 · 10 −3 0.593 3.96 · 10 −5 ± 3.00 · 10 −3 0.989 Asia −2.71 · 10 −4 ± 6.15 · 10 −4 0.660 −1.17 · 10 −3 ± 6.49 · 10 −4 0.070 −5.66 · 10 −4 ± 5.59 · 10 −4 0.312 Africa −3.80 · 10 −4 ± 1.57 · 10 −4 0.016 −3.17 · 10 −3 ± 7.90 · 10 −4 < 0.001 −5.56 · 10 −4 ± 6.19 · 10 −4 0.370 Europe −8.08 · 10 −4 ± 5.72 · 10 −4 0.158 −2.33 · 10 −4 ± 7.08 · 10 −4 0.742 −8.64 · 10 −4 ± 1.96 · 10 −3 0.659 North America −1.12 · 10 −3 ± 7.35 · 10 −4 0.095 −4.96 · 10 −3 ± 1.35 · 10 −3 < 0.001 −2.34 · 10 −3 ± 1.27 · 10 −3 0.067 South America −3.35 · 10 −3 ± 2.35 · 10 −3 0.155 −6.97 · 10 −3 ± 2.11 · 10 −3 < 0.001 3.19 · 10 −3 ± 3.52 · 10 −3 0.365 Oceania 1.31 · 10 −3 ± 5.16 · 10 −4 0.011 −3.57 · 10 −3 ± 1.65 · 10 −3 0.030 1.87 · 10 −3 ± 5.81 · 10 −4 0.001 USA 4.96 · 10 −4 ± 3.71 · 10 −3 0.894 6.86 · 10 −3 ± 4.46 · 10 −3 0.124 −2.24 · 10 −3 ± 5.63 · 10 −3 0.691 Global −8.47 · 10 −4 ± 5.96 · 10 −4 0.155 −4.40 · 10 −4 ± 6.22 · 10 −4 0.480 −4.04 · 10 −4 ± 1.65 · 10 −3 0.807 indicate that the size of the geographical footprint could play a significant role in removing biases. These biases could footing. 229 We find that temperature and relative humidity gradients do not significantly deviate from the zero-gradient hy-230 pothesis. This implies that changes in temperature and relative humidity do not seem to have an effect on the value 231 of the transmission rate or there is a small correlation between the transmission rate and temperature and relative 232 humidity. Our results seem to be in agreement with other studies that has concluded that there is no evidence as to 233 whether the spread of the COVID-19 is temperature dependent [16, 17, 18, 32] . In the absence of conclusive evidence of temperature or humidity gradients in the data, upper limits on the absolute Influenza pandemics of the 20 th century Avian influenza: Historical aspects Influenza: historical aspects of epidemics and pandemics Real-time epidemic forecasting for pandemic influenza High temperature and high humidity reduce the transmission of covid-19 Effects of temperature variation and humidity on the death of 249 covid-19 in wuhan, china Early transmission dynamics in wuhan, 251 china, of novel coronavirus-infected pneumonia WHO Director-General's opening remarks at the media briefing on COVID-19 -11 Coronavirus disease (COVID-19) Situation Report -150 A familial cluster of 259 pneumonia associated with the 2019 novel coronavirus indicating person-to-person transmission: a study of a family cluster Persistence of coronaviruses on inanimate surfaces and their inactivation with biocidal agents The effects of temperature and relative humidity on the viability of the sars 264 coronavirus Association between climate variables and 266 global transmission of sars-cov-2 Role of temperature and humidity in the modulation of the doubling time of covid-19 268 cases, medRxiv Impact of meteorological factors on the covid-19 270 transmission: A multi-city study in china Nexus between covid-19, temperature and exchange rate in wuhan city: New 272 findings from partial and multiple wavelet coherence The sensitivity and specificity analyses of ambient temperature and population size on the 274 transmission rate of the novel coronavirus (covid-19) in different provinces of iran No association of covid-19 transmission with temperature or uv radiation in 276 chinese cities Temperature significantly changes covid-19 transmission in (sub) tropical cities of brazil Worldwide effectiveness of various non-pharmaceutical intervention control strategies on the global covid-283 19 pandemic: A linearised control model 286 [23] University of washington, institute for health metrics and evaluation covid-19 projections Driving and travel restrictions across the united states U.s. to suspend most travel from europe as world scrambles to fight pandemic An interactive web-based dashboard to track covid-19 in real time ISD, Integrated Surface Database (ISD) Index of /data/global-hourly/archives/csv Department of Atmospheric Sciences (DAS) at the University of Illinois at Urbana-301 No evidence for temperature-dependence of the covid-19 epidemic