To the Editor:
Higher temperature and ultraviolet (UV) index in Northern Europe have been reported as the most important meteorological protective factors for the transmission of influenza virus.1 On the other hand, a recent study in China suggests that higher temperature and UV radiation may not be associated with a decrease in the epidemics of coronavirus disease 2019 (COVID-19).2 To determine whether prevalence of COVID-19 is modulated by meteorological conditions, we herein conducted meta-regression of data from large US cities.
We selected 33 large US cities with a population of >500,000 in 2010 from US Census Bureau (http://www.census.gov). We obtained (1) integrated number of confirmed COVID-19 cases in the county (to which the city belongs) on 14 May 2020 from Johns Hopkins Coronavirus Resource Center (https://coronavirus.jhu.edu), (2) estimated population in 2019 in the county from US Census Bureau, and (3) monthly meteorological conditions in the city for 4 months (from January to April 2020) from National Weather Service (https://www.weather.gov), World Weather Online (https://www.worldweatheronline.com), and Global Solar Atlas (https://globalsolaratlas.info/map) (Table 1 ). As the meteorological conditions, (1) mean temperature (F), total precipitation (inch), mean wind speed (mph), mean sky cover, and mean relative humidity (%) were available from National Weather Service; (2) mean pressure (mb), mean UV index, and total sun hours were obtainable from World Weather Online; and (3) total solar direct normal irradiation (DNI) (kWh/m2) in the average year was procurable from Global Solar Atlas. Monthly data for the 4 months (mean pressure/UV index and total sun hours were available for 3 months, from January to March 2020) were averaged or cumulated. The COVID-19 prevalence was defined as the integrated number of COVID-19 cases divided by the population. Random-effects meta-regression was performed by means of OpenMetaAnalyst (http://www.cebm.brown.edu/openmeta/index.html). In a meta-regression graph, the COVID-19 prevalence (plotted as the logarithm transformed prevalence on the y-axis) was depicted as a function of a given factor(plotted as meteorological condition on the x-axis).
Table 1.
Extracted data in each large US city and county to which the city belongs
| City | State | County | Covid-19 prevalence in the county |
Meteorological parameter in the city |
||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Cases (n) | Population (n) | Prevalence | Mean temperature (F) | Total precipitation (inch) | Mean wind speed (mph) | Mean sky cover | Mean relative humidity (%) | |||
| Albuquerque | New Mexico | Bernalillo | 1,149 | 679,121 | 0.00169 | 47.1 | 1.65 | 8.4 | 0.45 | 43 |
| Austin | Texas | Travis | 2,345 | 1,273,954 | 0.00184 | 60.6 | 12.06 | 8.2 | 0.63 | 72 |
| Baltimore | Maryland | Baltimore | 4,290 | 827,370 | 0.00519 | 46.5 | 14.66 | 7.4 | 0.67 | 61 |
| Boston | Massachusetts | Suffolk | 15,881 | 803,907 | 0.01975 | 40.6 | 12.62 | 12.1 | 0.62 | 58 |
| Charlotte | North Carolina | Mecklenburg | 2,342 | 1,110,356 | 0.00211 | 53.7 | 21.33 | 7.3 | 0.62 | 63 |
| Chicago | Illinois | Cook | 58,457 | 5,150,233 | 0.01135 | 37.9 | 10.86 | 10.3 | – | – |
| Columbus | Ohio | Franklin | 4,227 | 1,316,756 | 0.00321 | 41.6 | 19.31 | 8.9 | 0.75 | 67 |
| Dallas | Texas | Dallas | 6,837 | 2,635,516 | 0.00259 | 57.0 | 17.53 | 11.1 | 0.65 | 70 |
| Denver | Colorado | Denver | 4,359 | 727,211 | 0.00599 | 38.0 | 2.82 | 10.1 | 0.57 | 57 |
| Detroit | Michigan | Wayne | 18,770 | 1,749,343 | 0.01073 | 37.5 | 11.35 | 9.6 | 0.81 | 73 |
| D.C. | D.C. | – | 6,736 | 705,749 | 0.00954 | 48.7 | 14.61 | 9.1 | 0.70 | 63 |
| El Paso | Texas | El Paso | 1,456 | 839,238 | 0.00173 | 57.1 | 3.07 | 8.7 | 0.45 | 40 |
| Fort Worth | Texas | Tarrant | 4,076 | 2,102,515 | 0.00194 | 57.0 | 17.53 | 11.1 | 0.65 | 70 |
| Houston | Texas | Harris | 8,817 | 4,713,325 | 0.00187 | 64.2 | 13.90 | 8.5 | 0.68 | 71 |
| Indianapolis | Indiana | Marion | 7,793 | 964,582 | 0.00808 | 41.4 | 15.61 | 10.5 | 0.74 | 73 |
| Jacksonville | Florida | Duval | 1,215 | 957,755 | 0.00127 | 65.2 | 6.36 | 7.4 | 0.60 | 68 |
| Las Vegas | Nevada | Clark | 5,045 | 2,266,715 | 0.00223 | 58.1 | 2.31 | 6.5 | 0.45 | 37 |
| Los Angeles | California | Los Angeles | 35,392 | 10,039,107 | 0.00353 | 61.0 | 7.17 | 7.1 | 0.48 | 62 |
| Louisville | Kentucky | Jefferson | 1,741 | 766,757 | 0.00227 | 47.9 | 16.28 | 8.6 | 0.75 | 62 |
| Memphis | Tennessee | Shelby | 3,542 | 937,166 | 0.00378 | 52.9 | 27.76 | 8.8 | 0.68 | 71 |
| Milwaukee | Wisconsin | Milwaukee | 4,387 | 945,726 | 0.00464 | 35.4 | 10.52 | 10.4 | 0.68 | 66 |
| Nashville | Tennessee | Davidson | 3,745 | 694,144 | 0.00540 | 51.0 | 23.11 | 7.8 | 0.70 | 64 |
| New York City | New York | New York City | 188,545 | 8,336,817 | 0.02262 | 44.4 | 12.74 | 6.5 | – | 57 |
| Oklahoma City | Oklahoma | Oklahoma | 1,013 | 797,434 | 0.00127 | 49.3 | 10.86 | 12.1 | – | – |
| Philadelphia | Pennsylvania | Philadelphia | 19,093 | 1,584,064 | 0.01205 | 45.1 | 12.79 | 9.8 | 0.67 | 60 |
| Phoenix | Arizona | Maricopa | 6,599 | 4,485,414 | 0.00147 | 63.6 | 3.55 | 5.9 | – | 42 |
| Portland | Oregon | Multnomah | 940 | 812,855 | 0.00116 | 47.7 | 12.35 | 7.2 | 0.68 | 71 |
| San Antonio | Texas | Bexar | 1,976 | 2,003,554 | 0.00099 | 63.0 | 7.35 | 8.3 | 0.65 | 66 |
| San Diego | California | San Diego | 5,391 | 3,338,330 | 0.00161 | 61.0 | 6.69 | 5.1 | 0.55 | 69 |
| San Francisco | California | San Francisco | 1,999 | 881,549 | 0.00227 | 55.0 | 3.86 | 9.3 | 0.54 | 69 |
| San José | California | Santa Clara | 2,391 | 1,927,852 | 0.00124 | 55.2 | 4.06 | 5.8 | 0.51 | 65 |
| Seattle | Washington | King | 7,290 | 2,252,782 | 0.00324 | 46.4 | 18.15 | 9.0 | 0.73 | 73 |
| Tucson | Arizona | Pima | 1,696 | 1,047,279 | 0.00162 | 59.7 | 2.09 | 6.8 | 0.11 | 40 |
| City | Meteorological parameter in the city |
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Mean pressure (mb) | Mean ultraviolet index | Total sun hours | Total solar direct normal irradiation (kWh/m2) | |||||||
| Albuquerque | 1018.0 | 3.0 | 803.5 | 839 | ||||||
| Austin | 1018.6 | 4.3 | 491 | 513 | ||||||
| Baltimore | 1019.5 | 1.3 | 564 | 506 | ||||||
| Boston | 1017.8 | 1.0 | 536 | 496 | ||||||
| Charlotte | 1020.7 | 2.3 | 532.5 | 582 | ||||||
| Chicago | 1019.2 | 1.3 | 418.5 | 428 | ||||||
| Columbus | 1019.6 | 1.3 | 390.5 | 405 | ||||||
| Dallas | 1018.8 | 4.0 | 521.5 | 550 | ||||||
| Denver | 1018.6 | 2.7 | 748 | 662 | ||||||
| Detroit | 1018.8 | 1.3 | 388 | 424 | ||||||
| D.C. | 1019.6 | 2.0 | 565 | 515 | ||||||
| El Paso | 1017.1 | 3.7 | 801 | 945 | ||||||
| Fort Worth | 1018.8 | 3.7 | 552.5 | 576 | ||||||
| Houston | 1019.2 | 4.7 | 499 | 480 | ||||||
| Indianapolis | 1019.7 | 1.7 | 436.5 | 432 | ||||||
| Jacksonville | 1020.9 | 4.7 | 688.5 | 629 | ||||||
| Las Vegas | 1017.5 | 4.3 | 820.5 | 813 | ||||||
| Los Angeles | 1018.4 | 5.0 | 717.5 | 682 | ||||||
| Louisville | 1020.0 | 1.7 | 428 | 443 | ||||||
| Memphis | 1020.1 | 3.0 | 469 | 503 | ||||||
| Milwaukee | 1019.1 | 1.3 | 435 | 475 | ||||||
| Nashville | 1020.5 | 2.0 | 435.5 | 466 | ||||||
| New York City | 1018.9 | 1.3 | 503 | 503 | ||||||
| Oklahoma City | 1018.6 | 3.0 | 615 | 610 | ||||||
| Philadelphia | 1019.3 | 1.7 | 560.5 | 503 | ||||||
| Phoenix | 1017.2 | 5.3 | 799.5 | 867 | ||||||
| Portland | 1019.9 | 2.3 | 349 | 292 | ||||||
| San Antonio | 1018.2 | 5.0 | 516 | 509 | ||||||
| San Diego | 1017.9 | 5.0 | 711.5 | 682 | ||||||
| San Francisco | 1020.3 | 4.0 | 721.5 | 573 | ||||||
| San José | 1020.3 | 4.7 | 722.5 | 598 | ||||||
| Seattle | 1018.2 | 2.3 | 349 | 310 | ||||||
| Tucson | 1017.2 | 4.7 | 776.5 | 928 | ||||||
Results of the meta-regression were summarized in Table 2 . A slope of the meta-regression line was significantly negative for mean temperature (coefficient, −0.069; P < .001; Fig 1 , upper panel), mean UV index (coefficient, −0.445; P < .001; Fig 1, middle panel), total sun hours (coefficient, –0.002; P = .028; Fig 1, lower panel), and total solar DNI (coefficient, –0.002; P = .023; Fig 2 , upper panel), which indicated that COVID-19 prevalence decreased significantly as temperature, UV index, sun hours, and solar DNI increased. Whereas, the slope was significantly positive for mean wind speed (coefficient, 0.174; P = .027; Fig 2, middle panel) and sky cover (coefficient, 2.220; P = .023; Fig 2, lower panel), which indicated that COVID-19 prevalence increased significantly as wind speed and sky cover increased.
Table 2.
Meta-regression summary
| Covariate | Coefficient |
P value | ||
|---|---|---|---|---|
| Lower bound | Upper bound | |||
| Mean temperature (F) | −0.069 | −0.093 | −0.045 | <.001 |
| Total precipitation (inch) | 0.038 | −0.004 | 0.081 | .075 |
| Mean wind speed (mph) | 0.174 | 0.020 | 0.328 | .027 |
| Mean sky cover | 2.220 | 0.313 | 4.128 | .023 |
| Mean relative humidity (%) | 0.007 | −0.020 | 0.035 | .613 |
| Mean pressure (mb) | 0.061 | −0.220 | 0.342 | .668 |
| Mean ultraviolet index | −0.445 | −0.585 | −0.306 | <.001 |
| Total sun hours | −0.002 | −0.004 | −0.000 | .028 |
| Total solar direct normal irradiation (kWh/m2) | −0.002 | −0.004 | −0.000 | .023 |
Fig 1.
Meta-regression graph depicting the COVID-19 prevalence (plotted as the logarithm transformed prevalence on the y-axis) as a function of a given factor (plotted as a meteorological condition on the x-axis). Upper panel, mean temperature; middle panel, mean UV index; lower panel, total sun hours.
Fig 2.
Meta-regression graph depicting the COVID-19 prevalence (plotted as the logarithm transformed prevalence on the y-axis) as a function of a given factor (plotted as a meteorological condition on the x-axis). Upper panel, total solar DNI; middle panel, mean wind speed; lower panel, sky cover.
The present meta-regression suggests that temperature, UV index, sun hours, and solar DNI may be negatively, and wind speed and sky cover may be positively associated with COVID-19 prevalence. Higher sun hours/solar DNI and lower sky cover are probably related to higher UV radiation. Despite the association of lower temperature and UV-index with the influenza transmission,1 no association of temperature and UV radiation with the COVID-19 epidemics has been reported,2 however, which may be denied by the present results of the association of higher temperature/UV index/sun hours/solar DNI and lower sky cover with lower COVID-19 prevalence. In conclusion, higher temperature/UV index/sun hours/solar DNI and lower wind speed/sky cover may be associated with lower COVID-19 prevalence (ie, lower temperature/UV index/sun hours/solar DNI and higher wind speed/sky cover may be associated with higher COVID-19 prevalence), which should be confirmed by further epidemiological researches adjusting for various risk and protective factors (in addition to meteorological conditions) of COVID-19.
Footnotes
Conflicts of interest: None to report.
References
- 1.Ianevski A, Zusinaite E, Shtaida N, et al. Low temperature and low UV indexes correlated with peaks of influenza virus activity in northern europe during 2010–2018. Viruses. 2019;11:207. doi: 10.3390/v11030207. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Yao Y, Pan J, Liu Z, et al. No association of COVID-19 transmission with temperature or UV radiation in Chinese cities. Eur Respir J. 2020;55 doi: 10.1183/13993003.00517-2020. [DOI] [PMC free article] [PubMed] [Google Scholar]