Carbon uptake prediction mapping using GeoXAI (Geospatial eXplainable Artificial Intelligence)
김근한
Kim, Geunhan
With the announcement of the carbon neutral green growth basic plan, it became necessary to calculate carbon absorption at the city level. Because existing methods require a lot of time and budget, research was conducted to indirectly measure the biomass of carbon sinks using artificial intelligence technology and satellite images. However, black box models such as deep learning have high prediction accuracy, but have the disadvantage of making it difficult to understand the process and interpret the results, and the judgment criteria and process for AI prediction results must be verified, which led to the birth of XAI. Therefore, in this study, we aim to produce a carbon uptake prediction map of carbon sinks in Seoul based on GeoXAI, which applies XAI technology to GeoAI using spatial information. XGBoost was used as a machine learning technique, and SHAP was used as an XAI technique. We analyzed the impact of the vegetation index on carbon absorption, identified the vegetation index that has a significant impact on carbon absorption, and created a carbon uptake prediction map for Seoul. It is expected that the method presented in this study can be used to establish plans to achieve carbon neutrality in the future when establishing carbon-neutral green growth plans and urban master plans.
140601-16.pdf
