Uncovering relationships between geospatial features and COVID-19 features is a comprehensive, confounding, cross-disciplinary and challenging topic, as the spread and effects of COVID-19 are related to many aspects of our lives, including socio-economic, cultural, and environmental features. Our research aims to provide an innovative data-driven method to uncover the relationships between the heterogeneous and cross-disciplinary geospatial features with COVID-19 features at the municipality scale in Germany. We exploit these relationships using supervised machine learning, explainable AI and spatial analysis in Germany from March 2020 to October 2021. First, we integrated multi-source data including social data, economic data, cultural data, air pollution data and COVID-19 features data into one spatiotemporally harmonised dataset. Second, we trained three machine learning models (a Support Vector Regressor, a Random Forest, and a Light Gradient Boosting Machine) on the integrated dataset to learn the relationships between the spatial features and the COVID-19 features. Third, we used Shapley Additive exPlanations (SHAP) to rank the relevance of each feature. After that, we illustrated the results by the visualised spatial differences within municipalities. The output delivers key information regarding the Covid hospitalisation rate with the control of NO2 concentration and education level in Germany with transferable methods.@en

Long-term exposure to ambient air pollution is one of the main public health concerns worldwide. Exposure to air pollution is highly related to a range of diseases including respiratory and cardiovascular diseases, such as lung cancers, asthma, diabetes, irregular heartbeat, stroke and obesity [1-3]. The outbreak of the pathogenic agent of coronavirus disease 19 (Covid-19) has led to a large number of deaths worldwide, and previous studies have pointed out how the long-term exposure to air pollution may have an impact on its high death rate [4]. Moreover, the hospitalization rate and infected population numbers are central indicators for lock-down policy-making, indicating whether the local medical system is able to handle the increasing infected population number through its available intensive care facilities. In fact, predicting hospitalization is vital for authorities and policymakers. We hereby hypothesize that high air pollutants concentration leads to a rise in the hospitalization rate under the influence of Covid-19 outbreaks. We attempt to predict such hospitalization numbers for past data by means of a task-specific optimized machine learning model, after we integrate social, economic, cultural, and other environmental features in future with an ongoing project we are conducting. While such a prediction model cannot directly be used for predicting the future development of the pandemic, analysing it still gives valuable insights on the influence of various environmental features had on it in the past.Air pollution is a mixture of a large number of chemical compounds such as CO2, CO, NOx, SO2, O3, heavy metals, and respirable particulate matter (PM2.5 and PM10); the main sources of such pollutants are identified as vehicle traffic, heating systems, and industrial plants [5]. Previous studies focused on the relationships between the variables of pandemic with the air pollutants information. Among all the air pollutants, NO2 and respirable articulate matter are highly related to the pandemic variables [6-8]. In our research, we extract the air pollutants information (CO, NO2, CH4, SO2) from the Sentinel-5P TROPOMI sensor, and integrate it with open-access data on Covid-19 features (mortality, infection rate, intensive care rate, etc). The air pollutant data is processed from the Sentinel-5P data catalog provided in Google Earth Engine. We therefore aim to ascertain the relationships between hospitalization and air pollutants concentration with the incidence of Covid-19. In particular, our ultimate research purpose is to develop a machine learning model to uncover the relationships between a mixture of features derived from air pollutants and Covid-19 related information, at municipality scales in Germany and the Netherlands. The relationships provide important clues on understanding how air pollution may affect on hospitalization rate and other features of Covid-19, through the evidence of potential low hospitalization or low mortality with better air quality. The output will deliver key information regarding public health effects and control of emission in Germany and the Netherlands. Specifically, on a temporal scale, we aggregated daily Covid-19 data and four air pollutant measures into weekly measures. On a spatial scale, the air pollutants were aggregated based on each municipality in Germany and the Netherlands to match the Covid-19 features. A choice of machine learning models were trained and evaluated on historical data (from March of 2020 to Oct of 2021), using features comprising weekly hospitalizations, death rate, and infected rate, tropospheric NO2 concentration, CO, SO2, CH4 concentrations. In addition, a post-processing analysis using machine-learning explainability methodologies was carried out to mine potential relationships between hospitalization attributes and specific air pollution concentration features. By processing municipalities as separate spatial entities, the results are intended to highlight hospitalization disparities and pollutants’ effect diversities among different geographic areas. By highlighting the relationships between air pollutant concentrations and incidence of Covid-19 with the hospitalization rate, and illustrating the hospitalization disparities among municipalities, our results provide key information regarding policymaking on urban emission control and public health at municipality level. When integrating other Covid-related features, our models could offer support to policymakers on effective lock-down decisions and health system management. Keywords: Air pollutant, Covid-19, supervised machine learning models, Google Earth Engine. Reference 1. Bernstein, J.A., et al., Health effects of air pollution. Journal of allergy and clinical immunology, 2004. 114(5): p. 1116-1123. 2. Brunekreef, B. and S.T. Holgate, Air pollution and health. The lancet, 2002. 360(9341): p. 1233-1242. 3. Strak, M., et al., Long-term exposure to particulate matter, NO2 and the oxidative potential of particulates and diabetes prevalence in a large national health survey. Environment international, 2017. 108: p. 228-236. 4. Ogen, Y., Assessing nitrogen dioxide (NO2) levels as a contributing factor to coronavirus (COVID-19) fatality. Science of The Total Environment, 2020. 726: p. 138605. 5. Vineis, P., et al., Air pollution and risk of lung cancer in a prospective study in Europe. International Journal of Cancer, 2006. 119(1): p. 169-174. 6. Gautam, S., COVID-19: air pollution remains low as people stay at home. Air Quality, Atmosphere & Health, 2020. 13: p. 853-857. 7. Vîrghileanu, M., et al., Nitrogen Dioxide (NO2) Pollution monitoring with Sentinel-5P satellite imagery over Europe during the coronavirus pandemic outbreak. Remote Sensing, 2020. 12(21): p. 3575. 8. Omrani, H., et al., Spatio-temporal data on the air pollutant nitrogen dioxide derived from Sentinel satellite for France. Data in Brief, 2020. 28: p. 105089.@en