One of the significant aspects of climate change is the globally rising temperature. Extreme heat events, including heatwaves and urban heat island, are producing life-threatening conditions, overheating rivers, plants, and wildlife. Localized weather can be quite different from the regional weather forecast. Localized heat forecast can help identify the regions which are prone to overheating and target warnings to citizens on potential heatwaves and provide aid to residents in time.
The increasing availability of IoT and satellite observation can provide an excellent complement to the numerical prediction regarding local uncertainty. The emergence of the Fifth Generation (5G) of mobile technologies and its potential impacts to IoT will bring enormous benefits to localized weather observation with higher data transmission speed and more connected networks. 5G and IoT will continue to grow. With the explosive growth of mobile traffic and new technologies continuing to connect more and more devices online, 5G and IoT will play a vital role in our future, monitoring and assessing smart energy systems and environmental change. Challenges exist in 1) the intelligent processing of IoT data regarding its huge volume and complex data structure, 2) the knowledge integration of IoT data and satellite observations, and 3) the integration of heterogeneous information into weather prediction.
This project is proposed to develop a framework to 1) integrate IoT, satellite observations, and numerical simulations in both spatial and temporal dimensions using probabilistic spatial-temporal graphs, and 2) improve the short-term prediction of localized weather using machine and deep learning techniques. The proposed project will test the framework in major U.S. cities using surface temperature as the target weather variable. The expected outcomes of the project include the framework of IoT data processing and fusion, and the localized weather forecasts.
Resulting Publications
- Using Long Short-Term Memory (LSTM) and Internet of Things (IoT) for Localized Surface Temperature Forecasting in an Urban Environment
Yu, M., Xu, F., Hu, W., Sun, J. & Cervone, G., 2021, In: IEEE Access. 9, p. 137406-137418 13 p. - A comparative study of deep learning-based time-series forecasting techniques for fine-scale urban extreme heat prediction using Internet of Things observations
Yu, M., Shen, T. and Cervone, G., 2022. Nanotechnology-Based Smart Remote Sensing Networks for Disaster Prevention (pp. 253-271). Elsevier. - A high spatiotemporal resolution framework for urban temperature prediction using IoT data
Yang, J., Yu, M., Liu, Q., Li, Y., Duffy, D.Q. and Yang, C., 2022. Computers & Geosciences, 159, p.104991.
Resulting Presentations
- Forecasting localized surface temperature in an urban environment using 5G, Internet of Things (IoT) and deep learning approaches
Yu, M. (June 2022). 2022 EarthCube Annual Meeting. - Using Long Short-Term Memory (LSTM) and Internet of Things (IoT) for localized temperature forecasting in an urban environment
Yu, M. (Author and Presenter), Xu, F., Hu, W., Sun, J., & Cervone, G. (September 17, 2020). American Geophysical Union (AGU) Fall Meeting, Virtual, Accepted. National.
