Assessing the efficacy of artificial intelligence based city-scale blue green infrastructure mapping using Google Earth Engine in the Bangkok metropolitan region

Urbanization disrupts natural water, energy, and nutrient cycles, but integrating green and blue infrastructures (BGI) can mitigate these effects by facilitating processes like evapotranspiration, soil water infiltration, and root nutrient absorption. In Bangkok Metropolitan Region (BMR), escalating urbanization poses challenges to these natural cycles. Mapping Land Use and Land Cover (LULC) and identifying green infrastructure locations are crucial for effective urban planning, sustainable development, and environmental conservation amidst rapid urban growth. Continuous monitoring of dynamic urban areas is time-consuming, labor-intensive, and costly. Previous research primarily focused on land use and land cover classification followed by BGI identification. An automated classification system focusing solely on BGI can greatly enhance efficiency, accuracy, and cost-effectiveness in land classification for urban planning and decision-making. However, automating this system remains a significant challenge for the remote sensing community. Therefore, the research is first to develop a cloud based artificial intelligence tools such as Smile Random Forest and Smile CART integration with Google Earth Engine (GEE) to identify BGI for BMR. For this analysis, we have used open-sources and mostly used satellite images (e.g., Landsat and Sentinel) and consider to analyze seasonal changes for waterbodies, natural vegetations and human intervened vegetations with developing cloud-based artificial intelligence (AI). Surprisingly, Landsat-9 data demonstrated superior accuracy compared to Sentinel-2, indicating that the advanced technology of Landsat 9 may be more effective for BGI classification using AI. The study revealed a most distinct transition from deep green to green infrastructures during the transition from summer to monsoon season, whereas significant changes in blue infrastructure occurred between the monsoon and winter seasons. Seasonal variations in BGI are complex and influenced by factors such as the types of BGI implemented and the nuances of local climatic conditions. These advancements could provide precise insights for urban managers and policymakers, offering valuable tools to identify and understand BGI dynamics across various urban scales. • Cloud-based artificial intelligence (AI) was developed for blue and green infrastructure mapping. • Landsat images exhibited superior performance over Sentinel-2. • The RF classifier outperformed the CART for both Sentinel-2 & Landsat-9. • AI demonstrates capability of mapping accuracy for deep greens, grasslands, and shrubs. • Maximum blue and green infrastructure transition observed in summer to monsoon.