Development of a deep learning emulator of source-receptor footprints for computationally efficient flux inversions
Development of a deep learning emulator of source-receptor footprints for computationally efficient flux inversions
