Gramener, a Straive Company, has developed GeoBox, a solution that employs spatial data and advanced modeling techniques to understand and mitigate the impacts of urban heat islands (UHIs). UHIs, areas within cities that register significantly higher temperatures than rural areas, can lead to a variety of environmental and health issues including increased energy consumption, poor air quality, and heat-related illnesses.
GeoBox allows users to analyze public geospatial data using its potent API, which both streamlines the process and saves time and resources. This helps communities identify UHI hotspots and develop data-driven mitigation strategies. The platform’s insights are presented in user-friendly formats such as raster, GeoJSON, and Excel, facilitating implementation. With GeoBox, measures for sustainable urban development like improving air quality, reducing energy consumption, and creating healthier environments can be effectively planned and implemented.
GeoBox taps into the geospatial capabilities of Amazon SageMaker for earth observation analysis and for utilizing satellite imagery to produce UHI insights. SageMaker simplifies the tasks for data scientists and machine learning engineers with its pre-trained machine learning models and the ability to enrich large-scale geospatial datasets efficiently.
Geobox’s analysis method involves estimating land surface temperature (LST) at a granular level and identifying the key factors influencing UHIs. It uses a structured workflow with the primary focus being to target UHIs in a Canadian city.
Phase one, the data pipeline phase, relies on data captured by the Landsat 8 satellite. Phase two involves exploratory analysis that helps in understanding the city’s structure and development. The final phase, analytics model, employs machine learning models to gain insights into LST and its relationship with other influencing factors.
To predict LST trends, Gramener used various data sets and the SageMaker geospatial notebook. The data was retrieved using the SearchRasterDataCollection API to acquire Landsat 8 satellite data.
Computation work was handled by Amazon SageMaker Processing with the geospatial container, facilitating flexible scaling of clusters based on task sizes. After necessary calculations, the data was aggregated into a grid of about 100 meters in size.
Spatial modeling was conducted using methods like linear regression and spatial fixed effects. Exponential smoothing – a method that applies weighted averages for time-series forecasting – was used to predict future LST values. The forecasted results were visualized using SageMaker’s geospatial notebook.
The development of GeoBox signifies a huge leap towards sustainable urban development planning. With SageMaker, Gramener has been able to reduce UHI analysis time from weeks to hours, better equipping their clients with timely interventions for managing UHIs.
