key: cord-0707171-d8qul4hg
authors: Kalippurayil Moozhipurath, R.
title: Role of Weather Factors in COVID-19 Deaths in Tropical Climate: A Data-Driven Study Focused on Brazil
date: 2020-09-14
journal: nan
DOI: 10.1101/2020.09.13.20193532
sha: eab3196fa139bf7c67202578b1fa12ef602c4535
doc_id: 707171
cord_uid: d8qul4hg

Background: Brazil reported 123,780 deaths across 27 administrative regions, making it the second-worst affected country after the US in terms of COVID-19 deaths as of 3 September 2020. Understanding the role of weather factors in COVID-19 in Brazil is helpful in the long-term mitigation strategy of COVID-19 in other tropical countries because Brazil experienced early large-scale outbreak among tropical countries. Recent COVID-19 studies indicate that relevant weather factors such as temperature, humidity, UV Index (UVI), precipitation, ozone, pollution and cloud cover may influence the spread of COVID-19. Yet, the magnitude and direction of those associations remain inconclusive. Furthermore, there is only limited research exploring the impact of these weather factors in a tropical country like Brazil. In this observational study, we outline the roles of 7 relevant weather factors including temperature, humidity, UVI, precipitation, ozone, pollution (visibility) and cloud cover in COVID-19 deaths in Brazil. Methods: We use a log-linear fixed-effects model to a panel dataset of 27 administrative regions in Brazil across 182 days (n=3882) and analyze the role of relevant weather factors by using daily cumulative COVID-19 deaths in Brazil as the dependent variable. We carry out robustness checks using case-fatality-rate (CFR) as the dependent variable. Findings: We control for all time-fixed and various time-varying region-specific factors confounding factors. We observe a significant negative association of COVID-19 daily deaths growth rate in Brazil with weather factors - UVI, temperature, ozone and cloud cover. Specifically, a unit increase in UVI, maximum temperature, and ozone independently associate with 6.0 percentage points [p<0.001], 1.8 percentage points [p<0.01] and 0.3 percentage points [p<0. 1] decline in COVID-19 deaths growth rate. Further, a unit percentage increase in cloud cover associates with a decline of 0.148 percentage points [p<0.05] in COVID-19 deaths growth rate. Surprisingly, contrary to other studies, we do not find evidence of any association between COVID-19 daily deaths growth rate and humidity, visibility and precipitation. We find our results to be consistent even when we use the CFR as the dependent variable. Interpretation: We find independent protective roles of UVI, temperature, ozone and cloud cover in mitigating COVID-19 deaths, even in a tropical country like Brazil. We observe these results to be consistent across various model specifications, especially for UVI and cloud cover, even after incorporating additional time-varying weather parameters such as dewpoint, pressure, wind speed and wind gust. These results could guide health-related policy decision making in Brazil as well as similar tropical countries.

Brazil reported 123,780 deaths across 27 administrative regions, making it the secondworst affected country after the US in terms of COVID-19 deaths as of 3 September 2020 1 4 . Therefore, the results of such a study in Brazil can guide further healthrelated policy decision making in similar tropical countries.

To the best of our knowledge, so far, no empirical study has comprehensively explored the role of these different relevant weather factors in COVID-19 deaths covering 27 different administrative regions in Brazil, controlling for various other time-varying factors well as region-specific time-constant factors. In this observational study, we empirically describe the role of these weather factors in COVID-19 deaths in Brazil. We observe a significant negative association of daily growth rate of cumulative COVID-19 deaths (COVID-19 daily deaths growth rate) in Brazil with weather factors -UVI, temperature, ozone and cloud cover, indicting their respective protective roles. Surprisingly, contrary to other studies, we do not find any association between COVID-19 daily deaths growth rate and humidity, visibility and precipitation. We find our results to be consistent even when we use the case-fatality-rate (CFR) as the dependent variable.

Studies indicate that the impact of weather factors on respiratory viral infections and deaths comes from three major mechanisms 5 Prior studies show that weather factors 6 such as humidity 7 , UVI 3 , temperature 8 and precipitation 8,9 play a substantial role in the transmission of viruses such as influenza. Early All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint studies indicate that weather factors such as temperature 10 and humidity 10, 11 may influence the spread of COVID-19 and plausibly deaths. Temperature and humidity primarily impact the survival period of the virus on different surfaces influencing its transmission [12] [13] [14] . Thus, their protective role may be due to their impact on the viability of viruses as they affect the viral membranes 5 . On the contrary, temperature and humidity also influence the behaviour of people. For example, a higher temperature may also increase their likelihood of outdoor activities and thereby increasing the possibility of transmission.

Moisture-based weather factors like precipitation may play a role in converting dry environment into moist and cold conditions, making it more conducive for the survival and transmission of viruses 15 . However, precipitation and cloud cover also influence the behaviour of people, limiting their likelihood of outdoor activities and thereby potentially impacting the transmission of COVID-19.

Further, air pollution 16 may be a significant risk factor for respiratory infection as it affects the immune system 17 , as well as denser particulate matter, could carry coronavirus for more extended periods as well as across larger distances 17 . Studies also indicate a significant association of air pollution and COVID-19 death rate 17 . Furthermore, studies show that the presence of air pollutants like ozone is also associated with an increased risk of ARDS (Acute Respiratory Distress Syndrome) 18 , which is a significant risk factor in COVID-19. On the contrary, some studies also indicate a protective role of ozone as it reduces the transmission of viral diseases like influenza 19 and plausibly COVID-19 20 , plausibly due to anti-viral effects 20 .

UV Radiation, another weather factor, plays a protective role in two ways. UV Radiation inactivates viruses in fomite transmission, thereby mitigating the viral spread and deaths 21 .

UV Radiation, specifically UVB also plays a significant role in vitamin D skin synthesis 22-26 . Even in a sunny and tropical country like Brazil, studies indicate that Vitamin D deficiency is common 27 . UVB Radiation and the likelihood of skin exposure & skin synthesis vary All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in (OH)2D], an active vitamin D form, plays an essential role in modulating innate and adaptive immune systems 37, 38 , renin-angiotensin system (RAS) [38] [39] [40] as well as in modulating the inflammatory response 37, 38 . Emerging COVID-19 studies indicate that vitamin D deficiency might be a risk factor not only for incidence 34,35 , but also for severity 41 42 and mortality 36, 43, 44 .

Therefore, based on the prior studies, we consider temperature, humidity, UV Index, precipitation, ozone, visibility and cloud cover to be the relevant factors to consider for exploring the role of weather in COVID-19.

We anticipate that studies exploring the role of weather in COVID-19 need to consider all the above relevant weather factors, as considering only a subset of these factors may confound and limit the reliability of results 45 . Early studies also indicate that a tropical country like Brazil shows some substantial variation from higher latitude countries with low temperature and low UV Index or those with periods under winter 4 . We argue that conclusions of studies with over-representation of locations with a lower temperature or lower UV 45 either due to the study period or due to specific regions may have limited applicability in a tropical country like Brazil. Furthermore, prior studies indicate that an extended study period helps in strengthening the evidence for the association between these factors and COVID-19 4 . Therefore, it is vital to consider relevant weather parameters as well as the regional characteristics and the duration of the study.

Studies also indicate the importance of considering lags, as there is usually a substantial delay in the onset of symptoms, testing, hospitalisation and deaths 45, 46 . Furthermore, humans All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint can store vitamin D produced via the UV Index to be used later 32 . According to prior studies, time-constant factors such as population, genetic profile, culture, density, age structure, gender, underlying disease, co-morbidities, skin pigmentation, gender, mobility, underlying conditions, diet or socio-economic factors are relevant to consider in COVID-19 32,33,45 . As COVID-19 is a novel disease, we anticipate that controlling for other weather factors such as dew point, pressure, wind speed, and wind gust may increase the robustness of the conclusions. With this rationale, we derive twelve criteria for evaluating the relevant literature exploring the role of weather factors in COVID-19, as shown in Table 1 .

We provide an overview of the evaluation of various studies focusing on the role of the above relevant weather parameters in COVID-19, as shown in Table 1 Further, studies exploring the impact of these weather factors on a tropical country like Brazil are limited. A few studies focused on Brazil find a negative association with temperature 2,47 ; whereas others show a positive association with temperature 4 and humidity 4, 56 or no association with temperature 56 . Further, as shown in Table 1 , the majority of the Brazilian studies used data for a shorter period covering only the time until the start of Brazilian winter. Furthermore, no studies in Brazil has so far considered weather factors such as UV Index, ozone, visibility and cloud cover.

The reasons for the above inconclusive results are unclear 45 . As mentioned in Table 1 , most of these studies considered only a subset of these weather factors. They did not take into All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint account the lags, which might potentially limit the reliability of the results 45 . Further, the majority of these studies did not take into account the above time constant confounding factors. Other limitations include using Reproduction Number (R0) of COVID-19 as the dependent variable, whose estimation includes several assumptions 45 . Majority of the studies had over-representation of locations with a lower temperature or lower UV 45 either due to various studies conducted during northern hemisphere winter/spring or due to the focus on regions in the higher latitudes in the northern hemisphere.

Extending the study by Moozhipurath, Kraft, and Skiera (2020) 32 i.e., the protective role of UVB radiation in COVID-19 is valid for a tropical country like Brazil, with higher UVI and limited variation in UVI and temperature. Contrary to the above studies, we also use a more comprehensive weather dataset focusing on Brazil, a tropical country in Southern Hemisphere with additional weather factors such as wind speed, dew point, pressure and wind gust. We also use a more extended period of data.

In this observational study, we empirically describe the role of these seven relevant weather factors in COVID-19 deaths in Brazil. Contrary to the majority of other studies, we focus on the within-country variation located primarily in the Southern Hemisphere tropical All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint region, consider time lags and control for all time-constant and various time-varying confounding factors. Further, we consider all relevant weather factors as well as analyze extended periods of data. In robustness checks, we also incorporate additional time-varying weather parameters such as dewpoint, pressure, wind speed and wind gust as well as governmental regulations to ensure the robustness of our results.

All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint based on the above latitude and longitude information. Subsequently, we merged these data to construct the final dataset at an administrative region level, as outlined in Table 2 preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint UVI were on average, 75.9% and 7.4, respectively. Precipitation, ozone, visibility and cloud cover were on average 0.14 mmph, 15.7 KM, 257 DU and 0.45 or 45% respectively. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint 

We perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint methodology and the interpretation of these associations in section 1 and section 2 in the Supplementary Appendix.

We summarize our main results in Table 4 . After controlling for all time-constant and various time-varying confounding factors, we note a substantial and significant negative association of daily growth rates of cumulative COVID-19 deaths (COVID-19 daily deaths growth rate) with UVI, cloud cover, maximum temperature and ozone, with a moving average window of 42 days (Model 1 in Table 4 ). Specifically, a permanent unit increase of of daily growth rates of COVID-19 deaths as compared to the average daily growth rates over two weeks (14 days). As these results indicate, the above associations are substantial in relative terms, over a period of two weeks. Model 2 of Table 4 outlines that the results are consistent, even when we use CFR as the dependent variable. The estimated associations in Model 2 are weaker than those of Model 1 as the weather factors not only affect the COVID-19 deaths but also the infections, as outlined in section 2 of this manuscript.

We check the robustness of our results in Table S1 in Supplementary Appendix. We include additional time-varying weather parameters such as dewpoint, pressure, wind speed All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint and wind gust. We find our results to be consistent across various model specifications, especially for UVI and cloud cover, even after incorporating these additional time-varying weather parameters. Note: +: p < 0.10, *: p < 0.05, **: p < 0.01. t-statistics based on robust standard errors in parentheses. FE stands for region fixed-effects, TRSE stands for time region-specific effects.

We observe a significant negative association of COVID-19 daily growth rate in Brazil with weather factors -UVI, temperature, ozone and cloud cover. Specifically, a unit increase All rights reserved. No reuse allowed without permission.

perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in Our results are consistent with various theoretical mechanisms through which weather factors may affect immunity, human behaviour as well as viral survival and transmission 5 .

Prior studies indicate that in higher latitudes (above 35 o latitude) during winter, ozone absorbs most of the UVB 57 and reduces the protective role of UVB. However, it is surprising that even in a tropical country like Brazil with high average UV Index and most of the geographic area below 33°S, UVB radiation plays a crucial mitigating role in COVID-19. These results are consistent with the emerging observational and clinical findings related to the protective role of UVB radiation due to its role in the inactivation of viruses as well as vitamin D synthesis 32, 33, 34, 35, 36 . This finding can guide the policy decisions, for examplesensible sunlight exposure and vitamin D supplementationeven in tropical countries.

We anticipate that temperature may play a protective role in COVID-19 in Brazil since higher temperature reduces the spread of COVID-19 and thereby the likelihood of deaths plausibly due to its effect on viral survival and transmission. These results are consistent with the majority of the existing studies on other viruses such as SARS-CoV 58 . In contrast with All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint some prior studies, we find a protective role of ozone in COVID- 19 20 in Brazil 45 . This protective role may be plausibly due to its anti-viral effects 20 . Finally, the protective role of cloud cover in Brazil may be due to the changes in the behaviour of peoplefor examplemore time spent indoors and thereby reducing the likelihood of transmission of COVID-19

outdoors. In contrast to many other studies 45,10,15,50,51,46,52 , we do not find any association between COVID-19 daily deaths growth rate and other weather factors such as humidity, visibility and precipitation in Brazil. We observe our results to be consistent across various model specifications, especially for UVI and cloud cover, even after incorporating additional time-varying weather parameters such as dewpoint, pressure, wind speed and wind gust.

We control for all time-constant and region-specific factors and different time-varying confounding factors 32,33 . Nevertheless, we admit that we may not be able to exclude other time-varying confounding factors, which might affect our results 32,33 . Our study uses the latest COVID-19 data with one of the comprehensive datasets available on weather in Brazil at an administrative region level.

In general, tropical countries do not undergo substantial variation in temperature and UVB radiation across seasons. However, our results indicate that even in such tropical countries, a consistent drop in UVI or temperature due to change of seasons (e.g., monsoon, rainy season, winter) may be associated with a substantial increase in the daily death growth rate of COVID-19. These results also serve as an early warning for tropical regions in the Northern Hemisphere which is transitioning to autumn/winter. Furthermore, our study indicates that with the arrival of summer (higher temperature and higher UVI), tropical countries in the southern hemisphere like Brazil may face a decline in COVID-19 deaths. We further acknowledge that our study results cannot be a substitute for health guidance for Brazil.

However, we hope that the results of our study will guide further ecological studies and guide policy decision making in Brazil and other similar tropical countries. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint

RKM is a PhD student at Goethe University, Frankfurt. He is a full-time employee of a multinational chemical company involved in vitamin D business and holds the shares of the company. This study is intended to contribute to the ongoing COVID-19 crisis and is not sponsored by his company. The views expressed in the paper are those of the authors and do not represent that of any organization. No other relationships or activities that could appear to have influenced the submitted work.

We would like to thank Lennart Kraft for his immense contribution to this paper and for providing inputs to this paper. We would like to acknowledge Sharath Mandya Krishna, for his immense contribution to this paper -for providing inputs and assisting with data collection, data transformation and data engineering. We thank Matthew Little for his inputs and his assistance in the review. We would also like to acknowledge Magdalena Ceklarz for her valuable contributions to our paper and the discussions about COVID-19 at different points in time.

RKM conceptualised the research idea, conducted literature research, designed theoretical framework and collected COVID-19 data. RKM analyzed and interpreted the results and wrote the article.

This study is not sponsored by any organisation. The corresponding author had full access to all the data and had final responsibility for the submission decision. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint

Correspondence and requests for materials should be addressed to Rahul Kalippurayil Moozhipurath (rahulkm85@gmail.com).

The data used in the study are sourced from publicly available sources. Data regarding COVID-19 are obtained on 25 August 2020 from the data released by the Ministry of Health, Brazil and can be accessed at https://covid.saude.gov.br/. Data regarding weather is obtained from Dark Sky on the 25 August, 2020 and can be accessed at https://darksky.net/. We will make specific data set used in this study available for any future research. Interested researchers can contact Rahul Kalippurayil Moozhipurath via email to get access to the data. All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in

The copyright holder for this this version posted September 14, 2020. . https://doi.org/10.1101/2020.09.13.20193532 doi: medRxiv preprint

An interactive web-based dashboard to track COVID-19 in real time

Temperature significantly changes COVID-19 transmission in (sub)tropical cities of Brazil

Köppen's climate classification map for Brazil

Evidence that high temperatures and intermediate relative humidity might favor the spread of COVID-19 in tropical climate: A case study for the most affected Brazilian cities

Seasonality of respiratory viral infections

Influenza forecasting in human populations: A scoping review

Shortcomings of vitamin D-based model simulations of seasonal influenza

Low temperature and low UV indexes correlated with peaks of influenza virus activity in Northern Europe during

Dynamics of influenza in tropical Africa: Temperature, humidity, and cocirculating (sub)types. Influenza Other Respir

Temperature, humidity and latitude analysis to predict potential spread and seasonality for COVID-19

Impact of climate and public health interventions on the COVID-19 pandemic: a prospective cohort study

Effects of air temperature and relative humidity on coronavirus survival on surfaces

Stability of SARS-CoV-2 in different environmental conditions

Aerosol and surface stability of SARS-CoV-2 as compared with SARS-CoV-1

Impact of meteorological factors on COVID-19 pandemic: Evidence from top 20 countries with confirmed cases

Does air pollution influence COVID-19 outbreaks?

Association between short-term exposure to air pollution and COVID-19 infection: Evidence from China

Impact of long-term exposures to ambient PM2.5 and ozone on ARDS Risk for older adults in the United States

Ambient ozone and influenza transmissibility in Hong Kong

The ambient ozone and COVID-19 transmissibility in China: A data-driven ecological study of 154 cities

Inactivation of influenza virus by solar radiation

Modulation of the immune system by UV radiation: more than just the effects of vitamin D?

Prostate cancer risk and exposure to ultraviolet radiation: further support for the protective effect of sunlight

All rights reserved. No reuse allowed without permission. perpetuity

preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in 24. Grant, W. B. An estimate of premature cancer mortality in the US due to inadequate doses of solar ultraviolet-B radiation

An ecologic study of the role of solar UV-B radiation in reducing the risk of cancer using cancer mortality data, dietary supply data, and latitude for European countries in Biologic Effects of Light

Ultraviolet light may contribute to geographic and racial blood pressure differences

Prevalence of vitamin D insufficiency in Brazilian adolescents

Vitamin D deficiency

Vitamin D in preventive medicine: are we ignoring the evidence?

Vitamin D insufficiency among free-living healthy young adults

Plasma concentrations of 25-hydroxyvitamin D in meat eaters, fish eaters, vegetarians and vegans: results from the EPIC-Oxford study

25-Hydroxyvitamin D concentrations are lower in patients with positive PCR for SARS-CoV-2

Vitamin D sufficiency reduced risk for morbidity and mortality in COVID-19 patients

Evidence that vitamin D supplementation could reduce risk of influenza and COVID-19 infections and deaths

Immunologic effects of vitamin D on human health and disease

Vitamin D receptor activation regulates microglia polarization and oxidative stress in spontaneously hypertensive rats and angiotensin II-exposed microglial cells: Role of renin-angiotensin system

Vitamin D alleviates lipopolysaccharide-induced acute lung injury via regulation of the renin-angiotensin system

Role of vitamin D in pathogenesis and severity of COVID-19 infection

Vitamin D supplementation could possibly improve clinical outcomes of patients infected with coronavirus-2019 (COVID-19)

The role of vitamin D in the prevention of coronavirus disease 2019 infection and mortality

Evidence supports a causal role for vitamin D status in COVID-19 outcomes

Temperature and precipitation associate with COVID-19 new daily cases: A correlation study between weather and Covid-19 pandemic in Oslo

Relationship between COVID-19 and weather: Case study in a tropical country

Correlation between climate indicators and COVID-19 pandemic

A mechanism-based parameterisation scheme to investigate the association between transmission rate of COVID-19 and meteorological factors on plains in China

Role of temperature and humidity in the modulation of the doubling time of COVID-19 cases

Effects of meteorological conditions and air pollution on COVID-19 transmission: Evidence from 219 Chinese cities

Analysis of meteorological conditions and prediction of epidemic trend of 2019-nCoV infection in 2020

A spatio-temporal analysis for exploring the effect of temperature on COVID-19 early evolution in Spain

Exponential phase of COVID-19 expansion is driven by airport connections

No association of COVID-19 transmission with temperature or UV radiation in Chinese cities

Correlation between weather, population size and COVID-19 pandemic: a study of Brazilian capitals

Vitamin D: importance in the prevention of cancers, type 1 diabetes, heart disease, and osteoporosis

The effects of temperature and relative humidity on the viability of the SARS coronavirus

UV radiation and the skin

The effect of solar UVB doses and vitamin D production, skin cancer action spectra, and smoking in explaining links between skin cancers and solid tumours