Association between Climate Parameters and COVID-19 Pandemic in Jordan


 This research aims to study the association between the daily confirmed COVID − 19 cases in the three major cities of Jordan; Amman, Zarqa, and Irbid and climate indicators to include the average daily temperature (0C), wind speed (m/s), relative humidity (%), pressure (kPa), and the concertation of the four pollutants (CO, NO2, PM10 and SO2). The data were obtained from both the World Air Quality Project website and the Jordanian Ministry of Environment. In this work, the multiple linear regression and feedforward artificial neural network were used for data analysis. It was concluded that the multiple linear regression and feedforward artificial neural network had the capability to forecast the COVID-19 confirmed cases in the case studies; Amman, Irbid and Zarqa. Finally, global sensitivity analysis using Sobol analysis indicated that pressure in Amman and Zarqa and the concentration of NO2 in Irbid has a high rate of positive cases that supports the virus's spread.


Introduction
Since its detection in China in 2019, coronavirus, has been subjected to intense studies in many aspects to limit its transmission or suppressing it. Recently, research work was conducted to investigate the effect of metrological condition (the quality of air) and metrological data, (i.e. humidity, ambient temperature, solar radiation) on COVID-19 transmission around the world. Xie and Zhu (2020) used a generalized additive model to investigate how the active cases of COVID-19 is in uenced by the mean ambient temperature in 122 cities in China, they concluded that COVID-19 active cases increase linearly with ambient temperature with a threshold of 3°C. They also concluded that there is no evidence supporting the fact that the positive cases will decline in warmer conditions. Bashir et al. (2020) analyzed the relation of COVID-19 cases with metrological data in New York including average temperature, minimum temperature, maximum temperature, rainfall, average humidity, wind speed, and air quality. They concluded that, the minimum and maximum temperatures together with air quality are signi cantly related to the number of COVID-19 positive cases.
Méndez-Arriaga (2020) used Statistical analysis to investigate the association between the daily local COVID-19 positive cases and both climate characteristics and the regional meteorological data in 31 states and the capital city Mexico from February 29 to March 31, 2020. He concluded that there is no concrete association between temperature and positive cases, while precipitation is positively associated to positive cases. Lian et al. (2020) conducted a study on the effect of lockdown caused by COVID-19 on air quality in Wuhan, china. They found that most common pollutants concentration were signi cantly reduced during the lockdown, with the average monthly air quality index was 59.7, which is 33.9% less than that before the lockdown.
Zambrano-Monserrate et al. (2020) Studied the effects of COVID-19 on the environment in China, USA, Italy, and Spain. They stated that contingency measures are related with clean environment including good air quality, noise reduction and clean beaches. However, they concluded that there is a reduction in recycling and increase in waste that leads to contamination of water and land.
Collivignarelli (2020) Studied the effect of the lockdown on air quality in Milan Metropolitan / Italy. He concluded that the lockdown caused a signi cant reduction of pollutants concentration (PM 10 , PM 2.5 , BC, benzene, CO, and NO x ) which is mainly due to the reduced vehicular tra c. While, O3 demonstrated a major increase, possibly, due to the minor NO concentration. Furthermore, He found that the lockdown led to a major drop in the concentration of SO2 in Milan city.
Pirouz et al, (2020) investigated the effect of climate on active cases of COVID-19 using (multivariate linear regression (MLR) in an attempt to propose a prediction model. They considered including the parameters of relative humidity, daily average temperature, and wind speed, with some urban parameters such as population density. Their analysis showed a positive effect of the proposed model on the con rmed cases and they concluded that Their ndings may be applied by considering several variables that exhibit the exact delay to new con rmed cases of COVID-19. Abdelhafez et al. (2021), performed a study to correlate the COVID-19 active cases with metrological components to include: relative humidity (%), the average daily temperature ( 0 C), maximum ambient temperature ( 0 C), pressure (kPa) wind speed (m/s), and average daily solar radiation (W/m 2 ). In their work, they used the Spearman correlation test for data analysis. They concluded that the most effective weather parameter on the active cases of COVID-19 is the average daily solar radiation in the initial wave transmission, while all other tests for other parameters failed. Furthermore, in the second wave transmission, the maximum temperature was found to be the most effective weather parameter. Finally, using a global sensitivity analysis using Sobol analysis it was found that the daily solar radiation has a high rate of active cases that support the virus's survival in both wave transmissions.
In this work, Arti cial neural network (ANN) and multiple linear regression are used to predict COVID-19 active cases in the three cities in Jordan by using climate indicators as an input. Also, sensitivity analysis between variables (climate indicators) will be conducted using Sobol method.

Study Area
Jordan is located within latitudes 29°19' N and 32°35' N (southern region up to 31°12' N, middle region up to 32°05' N, and northern region up to 32°35' N) (Batieha et al. 2011). According to the Jordanian department of statistics the area of Jordan is 89,342 km 2 , and an estimated number of populations around eleven million.
To study the association between climate indicators and COVID-19 pandemic in Jordan, three cities were selected for the analysis based on the availability of data. The Jordanian Ministry of Environment has only 3 data stations, these stations are located in the three major cities of Jordan. Amman, the capital city, the administrative and the commercial center with an area of 1700 km 2 and a population of

Data Collection
The daily data of con rmed COVID − 19 cases in Amman, Zarqa and Irbid in Jordan over the period from 2 March to 31 December, 2020, was obtained from the Jordanian Ministry of health, while the climate indicators including relative humidity (%), the average daily temperature ( 0 C), pressure (kPa), wind speed (m/s), and the concertation of the four pollutants (CO, NO 2 , PM 10 and SO 2 ) in the three cities in Jordan was obtained from World Air Quality Project website (https://aqicn.org/) and the Jordanian Ministry of Environment.

Results And Discussion
Since the announcement made by WHO that COVID-19 is a pandemic, the number of active cases is continuously escalating worldwide (Cucinotta and Vanelli 2020). O cially, on March 2nd 2020, the rst con rmed COVD-19 case was declared by the Jordanian Ministry of Health (Abdelhafez et al. 2021). The number of cases escalated to 294,494 on January 31st, 2021, though the number of recovered cases was 270,551, 20,109 cases were still under treatment. The total death numbers were 3834 cases. Figures (1), (2) and (3) show the number of con rmed COVID-19 cases over the period from 2nd March to 31st December 2020 in Amman, Irbid and Zarqa respectively.

Multiple Linear Regression
A multiple linear regression model has been implemented for predicting con rmed daily cases of the COVID-19 given climate indicators (average daily temperature, relative humidity, wind speed, pressure and the concertation of the four pollutants (CO, NO 2 , PM 10 and SO 2 )) in the study. This study uses the linear approach for modeling the relationship between a scalar dependent variable y and one or more independent variables or (explanatory variables) denoted X (Freedman 2009). In this study 305 samples were used to obtain the following Linear Equations: Y 1 = 6.816X 1 + 11.999X 2 -30.618X 3 + 72.170X 4 -0.318X 5 + 38.049X 6 + 40.228X 7 + 6.796X 8 -74096.85 Equ.
(1) Y 2 = 0.086X 1 + 0.043X 2 -0.142X 3 + 0.031X 4 -0.443X 5 + 0.219X 6 -0.290X 7  Where Y 1 , Y 2 , Y 3 are the number of COVID-19 con rmed cases in Amman, Irbid and Zarqa respectively. X 1 is average daily temperature, X 2 is relative humidity, X 3 is wind speed, X 4 is pressure, X 5 is the concentration of CO, X 6 is the concentration of NO 2 , X 7 is the concentration of SO 2 and X 8 is the concentration of PM 10 .
Table (1) shows the performance of the multiple linear regression, the mean absolute error (MAE) for Amman over the testing set is 0.327, which indicates that on average, the predicted values using multiple linear regression depart from the true value by 47.66% and the coe cient of correlation (R) for Amman is equal to 0.638.
The performance of the multiple linear regression, MAE for Irbid over the testing set is 0.246, which indicates that on average, the predicted values using multiple linear regression depart from the true value by 67.33% and R for Irbid is equal to 0.366.
The performance of the multiple linear regression, MAE for Zarqa over the testing set is 0.391, which indicates that on average, the predicted values using multiple linear regression depart from the true value by 57.63% and R for Zarqa is equal to 0.500.

Feedforward Arti cial Neural Network
Feedforward model was designed and tested using the MATLAB software. This model was used to correlate the climate indicators with the COVID-19 active cases in Amman, Irbid and Zarqa, of Jordan.
More details and description about the proposed three models of ANN is found in (Al-Naami et al., 2014).
ANN network with neuron numbers (8, 20, 1) were designed and tested within the MATLAB. A total data consists of 305 model samples were attained from previously experimental data and is used as the input of ANN network. Furthermore, a total of 70% of this data is used for training, 15% is used for validation and 15% is used for testing. In the hidden layer, tangent sigmoid function is applied, and linear transfer function is used in the output layer. In this analysis and based on many trails and errors, 20 was selected as the number of the hidden layer. The proposed network was trained using Levenberg-Marquardt (trainlm) algorithm.

Sobol Sensitivity Test
This test is based on the decomposition of model output variance into summands of variances of the input parameters in increasing dimensionality (Saltelli et al. 1999 andSobol 2001). Furthermore, its analysis, which may be carried out via the global sensitivity analysis toolbox in MATLAB. is very useful in quantifying the interactions between parameters, and can be attained by computing the rst-order, second-order, higher-order, and the overall sensitivity indices (Abdelhafez et al. 2021). In this work, Sobol sensitivity test was conducted to Investigate the potential of variables in the prediction of the climate indicators (the concertation of the four pollutants (CO, NO 2 , PM 10 and SO 2 ), and average daily temperature, relative humidity, wind speed, and pressure), Global sensitivity analysis toolbox in MATLAB is used to conduct the Sobol sensitivity analysis.
Figures (7), (8) and (9) show the signi cance and sensitivity analysis of the Sobol method for climate indicators in Amman, Irbid and Zarqa. The results show that: Pressure is the most important variable in prediction of COVID-19 con rmed cases in Amman, the concentration of NO 2 is the most important variable in prediction of COVID-19 con rmed cases in Irbid, Pressure is the most important variable in prediction of COVID-19 con rmed cases in Zarqa.

Conclusion
In this study climate indicators variables including average daily temperature, relative humidity, wind speed, pressure and the concertation of the four pollutants (CO, NO 2 , PM 10 and SO 2 ) were correlated with COVID-19 active cases in Amman, Irbid and Zarqa, of Jordan. This was achieved by using multiple linear regression and feedforward arti cial neural network.
It was found that the multiple linear regression and feedforward arti cial neural network had the capability to forecast the COVID-19 con rmed cases in Amman, Irbid and Zarqa, of Jordan In the examination of the importance of variables of global sensitivity analysis by using Sobol analysis it was shown that the pressure plays an important role in the COVID-19 outbreak in Amman and Zarqa. While the concentration of NO 2 plays an important role in the COVID-19 outbreak in Irbid.

Declarations
Ethical Approval Not applicable

Consent to Participate
Not applicable

Consent to Publish
Not applicable