Theory and Conditions for AI-Based Inversion Paradigm of Geophysical Parameters Using Energy Balance

To construct a universal artificial intelligence (AI) model for geophysical parameter inversion, this study proposes a new remote sensing parameter inversion paradigm theory by changing cognition to unify physical, statistics and AI methods. Using the energy balance equation, we demonstrate that establishing a closed system of physical inversion equations between input and output variables in deep learning is the foundation for forming parameter inversion paradigms. On this basis, a generalized statistical method is constructed to guide the acquisition of representative solutions required for deep learning, thereby achieving the coupling of physical and statistical methods. Meanwhile, paradigm judgment conditions can improve the accuracy of data collection. Theoretical derivation and experimental results indicate that deep learning achieves physical consistency and generalization in remote sensing parameter inversion. This theory lays the foundation for developing AI parameter inversion model based on energy balance, and can also optimize the design of remote sensing sensors.