Can neural networks think like geomorphologists? @gregtucker.bsky.social and I trained a neural network to invert the streampower-diffusion model, and then opened the black box. It learned how valley spacing and drainage density connect to process. Out now in GRL: doi.org/10.1029/2026...
doi.org
Can Neural Networks Think Like Geomorphologists?
A simple neural network can infer parameters from synthetic topography generated by a process-based numerical model of landscape evolution The network can be queried to identify a sensible relati...