With the established of the recent Taiwan Air IoTs Wide Array Network （TAIWAN）, up to 10,000 micro sensors are deployed until the end of 2021. The data from the sensor network provide good spatial and temporal values for data analysis. However, due to the costs of maintenance of sensors, the locations of the deployment of sensors mostly cluster near the industrial area where the air is more prone to pollution. There is a severe scarcity in environmental monitoring of residential environment. The spatial discontinuity of air quality data has become an issue that requires immediate solution. This research aims at integrating multiple data sources like environmental information and data from IoT sensors. We propose to use the latest Geographical and Temporal Weighted Regression Kriging（GTWRK）and deep learning algorithm to develop an AQI spatial estimation model with super spatial-temporal resolution. With verification from field observation, we compare the performance of both model in real life application. By optimizing the model parameters, we fine-tune the model for the best model accuracy. Finally, we sum up the two model with the spatial distribution of the model residuals and offer the most optimized suggestion for sensor deployment. This research has developed a geographical and temporal weighted regression model. In order to preserve time-series characteristics of air quality, the model employs unidirectional spatial-temporal weights for simulation. We implement three spatial statistical methods GWR, GTWR and GTWRK for AQI modeling. The results demonstrate that GTWRK model has better spatial estimation and lower residual. The average residual error is only -0.02 (μg/m3). It indicates that GWRK is helpful to modify the model residual.
|Effective start/end date||2022/08/01 → 2023/07/31|
- Geographical and Temporal Weighted Regression
- spatial estimation model
- Air Quality Index
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.