RTM_surrogate_modeling

In this project, I developed a framework to emulate the MODTRAN Radiative Transfer Model (RTM), which is widely used for atmospheric correction of optical satellite imagery. To achieve this, top-of-atmosphere (TOA) and bottom-of-atmosphere (BOA) spectroscopy data were used to train multiple machine learning models. The results showed that a deep learning autoencoder outperformed other models, achieving an R² of 98.23%, a 3.16% improvement over Random Forest.
For the visual outputs of this project, visit my website.