A spatiotemporal machine learning approach to forecasting COVID-19 incidence at the county level in the USA.

With COVID-19 affecting every country globally and changing everyday life, the ability to forecast the spread of the disease is more important than any previous epidemic. The conventional methods of disease-spread modeling, compartmental models, are based on the assumption of spatiotemporal homogeneity of the spread of the virus, which may cause forecasting to underperform, especially at high spatial resolutions. In this paper, we approach the forecasting task with an alternative technique-spatiotemporal machine learning. We present COVID-LSTM, a data-driven model based on a long short-term memory deep learning architecture for forecasting COVID-19 incidence at the county level in the USA. We use the weekly number of new positive cases as temporal input, and hand-engineered spatial features from Facebook movement and connectedness datasets to capture the spread of the disease in time and space. COVID-LSTM outperforms the COVID-19 Forecast Hub's Ensemble model (COVIDhub-ensemble) on our 17-week evaluation period, making it the first model to be more accurate than the COVIDhub-ensemble over one or more forecast periods. Over the 4-week forecast horizon, our model is on average 50 cases per county more accurate than the COVIDhub-ensemble. We highlight that the underutilization of data-driven forecasting of disease spread prior to COVID-19 is likely due to the lack of sufficient data available for previous diseases, in addition to the recent advances in machine learning methods for spatiotemporal forecasting. We discuss the impediments to the wider uptake of data-driven forecasting, and whether it is likely that more deep learning-based models will be used in the future.

A Geo-AI-based ensemble mixed spatial prediction model with fine spatial-temporal resolution for estimating daytime/nighttime/daily average ozone concentrations variations in Taiwan.

High levels of ground level ozone (O3) are associated with detrimental health concerns. Most of the studies only focused on daily average and daytime trends due to the presence of sunlight that initiates its formation. However, atmospheric chemical reactions occur all day, thus, nighttime concentrations should be given equal importance. In this study, geospatial-artificial intelligence (Geo-AI) which combined kriging, land use regression (LUR), machine learning, an ensemble learning, was applied to develop ensemble mixed spatial models (EMSMs) for daily, daytime, and nighttime periods. These models were used to estimate the long-term O3 spatio-temporal variations using a two-decade worth of in-situ measurements, meteorological parameters, geospatial predictors, and social and season-dependent factors. From the traditional LUR approach, the performance of EMSMs improved by 60% (daytime), 49% (nighttime), and 57% (daily). The resulting daily, daytime, and nighttime EMSMs had a high explanatory power with and adjusted R2 of 0.91, 0.91, and 0.88, respectively. Estimation maps were produced to examine the changes before and during the implementation of nationwide COVID-19 restrictions. These results provide accurate estimates and its diurnal variation that will support pollution control measure and epidemiological studies.

A Geospatial Artificial Intelligence and satellite-based earth observation cognitive system in response to COVID-19.

The pandemic emergency caused by the spread of COVID-19 has stressed the importance of promptly identifying new epidemic clusters and patterns, to ensure the implementation of local risk containment measures and provide the needed healthcare to the population. In this framework, artificial intelligence, GIS, geospatial analysis and space assets can play a crucial role. Social media analytics can be used to trigger Earth Observation (EO) satellite acquisitions over potential new areas of human aggregation. Similarly, EO satellites can be used jointly with social media analytics to systematically monitor well-known areas of aggregation (green urban areas, public markets, etc.). The information that can be obtained from the Earth Cognitive System 4 COVID-19 (ECO4CO) are both predictive, aiming to identify possible new clusters of outbreaks, and at the same time supervisorial, by monitoring infrastructures (i.e. traffic jams, parking lots) or specific categories (i.e. teenagers, doctors, teachers, etc.). In this perspective, the technologies described in this paper will allow us to detect critical areas where individuals can be involved in risky aggregation clusters. The ECO4CO data lake will be integrated with ad hoc data obtained by health care structures to understand trends and dynamics, to assess criticalities with respect to medical response and supplies, and to test possibilities useful to tackle potential future emergencies. The System will also provide geographical information on the spread of the infection which will allow an appropriate context-specific public health response to the epidemic. This project has been co-funded by the European Space Agency under its Business Applications programme.