key: cord-0751330-53r9xvkm authors: Huang, J.; Liu, X.; Zhang, L.; Yang, K.; Chen, Y.; Huang, Z.; Liu, C.; Lian, X.; Wang, D. title: Projecting the second outbreaks for global COVID-19 pandemic date: 2020-07-16 journal: nan DOI: 10.1101/2020.07.15.20154161 sha: 2b3b51ce1ec4b28acdfb89539af7964fdb4c35d3 doc_id: 751330 cord_uid: 53r9xvkm COVID-19 is now in an epidemic phase, with a second outbreak likely to appear at any time. The intensity and timing of a second outbreak is a common concern worldwide. In this study, we made scenario projections of the potential second outbreak of COVID-19 using a statistical-epidemiology model, which considers both the impact of seasonal changes in meteorological elements and human social behaviors such as protests and city unblocking. Recent street protests in the United States and other countries are identified as a hidden trigger and amplifier of the second outbreak. The scale and intensity of subsequent COVID-19 outbreaks in the U.S. cities where the epidemic is under initial control are projected to be much greater than those of the first outbreak. For countries without reported protests, lifting the COVID-19 related restrictions prematurely would accelerate the spread of the disease and place mounting pressure on the local medical system that is already overloaded. We anticipate these projections will support public health planning and policymaking by governments and international organizations. outbreak. For countries without reported protests, lifting the COVID-19 related 23 restrictions prematurely would accelerate the spread of the disease and place mounting 24 pressure on the local medical system that is already overloaded. We anticipate these 25 projections will support public health planning and policymaking by governments and 26 international organizations. . CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. . https://doi.org/10.1101/2020.07.15.20154161 doi: medRxiv preprint 1 Introduction 28 Recently, the COVID-19 pandemic has spread rapidly and poses a dire threat to global 29 public health, which claimed over 0.49 million lives, along with 9.8 million confirmed cases 30 as of June 28 th 1 . Beyond the spread itself, the outbreak may have far-reaching consequences, 31 negatively affecting the economic development worldwide and posing a series of long-standing 32 social problems 2,3 . There is an urgent need for a global prediction system that can provide 33 scientific guidelines for the World Health Organization and international decision-makers to 34 implement effective containment measures capable of curbing the spread of COVID-19 4 . 35 Researchers worldwide have developed various models with mathematical and statistical 36 methods, including stochastic simulations, lognormal distribution 5 , machine learning, and 37 artificial intelligence 6 . Among them, the susceptible-infectious-removed infectious disease 38 model (SIR) is the most widely used 7-9 . However, this simple model is built under a series of 39 idealized assumptions, which may limit the accuracy and reliability of the prediction. In order 40 to obtain the prediction results with higher credibility, more complex models with fewer 41 assumptions should be developed so as to simulate the actual situations in a more realistic 42 manner 10 . 43 Although it is difficult to establish an accurate epidemiological model describing the 44 spread of a pandemic, the reported global pandemic data contain particular solutions to the 45 mathematical equations incorporated in epidemiological models 3,6 . It is theoretically possible 46 to remedy the defects of prior epidemiological models by introducing the latest pandemic data 47 and hence improve the pandemic prediction 2,4,6 . Based on this idea, we have developed a 48 Global Prediction System of the COVID-19 Pandemic (GPCP) 11 . The system develops a 49 modified version of the SIR model and determines the parameters through historical data 50 fitting 12,13 , which allows it to make targeted predictions for various countries and obtain better 51 prediction results. The first version of GPCP (CPCP-1) can capture the major features of the 52 daily number of confirmed new cases and provides reliable predictions. However, the 53 prediction of GPCP-1 is only valid for one month 11 . 54 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 16, 2020. The sum of the six categories is equal to the total population (N) at any time. 79 S + P + E + I + Q + R + D = N 80 We modified the model by introducing the timing of community reopening collected from 81 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 16, 2020. . https://doi.org/10.1101/2020.07.15.20154161 doi: medRxiv preprint 4 news reports. If the timing collected from news reports is not explicit enough as an input to our 82 model, the timing will be indicated by the daily new cases on the day of reopening (dQc). As 83 the number of newly confirmed cases on a given day falls lower than dQc, local authority begins 84 to lift or loose the lockdowns. 85 In addition, the temporal variation of transmission rate due to changes in local temperature 86 as well as human behaviors are considered. Generally, the transmission rate ( ( ) ) can be 87 expressed by the following equations: 104 . CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. In order to enhance the stability of the traditional least square method (Gauss-Newton 120 algorithm), we use an improved damped least square method called Levenberg-Marquardt 121 . CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. . https://doi.org/10.1101/2020.07.15.20154161 doi: medRxiv preprint 6 algorithm 17 . This method inserts a damping coefficient into the Gauss-Newton method when 122 calculating the Hessian matrix. The benefit of introducing this damping coefficient is that it 123 can converge very quickly in the steepest direction in many cases even when the initial solution 124 is very far from ideal values, which makes the parameter determination more robust 18 CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. In addition to the United States, protests of a certain scale also broke out in other countries. 184 Using similar parameterization of the protests, Table 3 (Figure 3) , the second outbreak is likely to peak during August. Under the impact of protests, 188 . CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. 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 16, 2020. Table 3 we can see that the peak time and size of the second outbreak varies from 218 countries to countries, due to different levels of interventions measures, environmental 219 . CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a 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 16, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted July 16, 2020. COVID-19 Data repository by the Center for Systems Science and Engineering The socio-economic implications of the coronavirus pandemic 275 (COVID-19): a review Coronavirus disease 2019 situation report 51 11th 277 Forecasting the novel coronavirus COVID-19 Incubation period and other epidemiological characteristics of 281 2019 novel coronavirus infections with right truncation: a statistical analysis of 282 publicly available case data Predicting the growth and trend of COVID-19 284 pandemic using machine learning and cloud computing CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted China: a modelling study Phase-adjusted estimation of the number of coronavirus disease 2019 290 cases in Wuhan Modified SEIR and AI prediction of the epidemics trend of COVID-19 in China under public health interventions SEIR Modeling of the Italian Epidemic of SARS-295 CoV-2 Using Computational Swarm Intelligence Global prediction system for COVID-19 pandemic Inversion of nonlinear dynamical model from the observation 1993: An analogue-dynamical long-range 301 numerical weather prediction system incorporating historical evolution Epidemic analysis of COVID-305 19 in China by dynamical modeling Generalized SEIR epidemic model (fitting and computation) Optimal temperature zone for the dispersal of COVID-19 Methods for non-linear least squares 312 problems International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted Relative effectiveness of the trust-region algorithm with precise secund 314 order derivatives Images of police using violence against peaceful protesters are going viral Tear gas: an 319 epidemiological and mechanistic reassessment From containment to mitigation of COVID-19 in the US High population densities catalyse the spread of COVID-19 Mass gathering events and reducing further global spread of 326 COVID-19: a political and public health dilemma Rapid spread of zika virus in the Americas -implications for public 329 ealth preparedness for mass gatherings at the 2016 Brazil Olympic Games Correlation between weather and COVID-19 pandemic in Jakarta Environmental predictors of seasonal influenza epidemics across 334 temperate and tropical climates Coronavirus disease (COVID-19) advice for the 337 public International license It is made available under a 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 16, 2020. 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 16, 2020. . https://doi.org/10.1101/2020.07.15.20154161 doi: medRxiv preprint