Physically Interpretable Neural Networks for the Geosciences: Applications to Earth System Variability

Abstract

Neural networks have become increasingly prevalent within the geosciences, although a common limitation of their usage has been a lack of methods to interpret what the networks learn and how they make decisions. As such, neural networks have often been used within the geosciences to most accurately identify a desired output given a set of inputs, with the interpretation of what the network learns used as a secondary metric to ensure the network is making the right decision for the right reason. Neural network interpretation techniques have become more advanced in recent years, however, and we therefore propose that the ultimate objective of using a neural network can also be the interpretation of what the network has learned rather than the output itself. We show that the interpretation of neural networks can enable the discovery of scientifically meaningful connections within geoscientific data. In particular, we use two methods for neural network interpretation called backward optimization and layerwise relevance propagation, both of which project the decision pathways of a network back onto the original input dimensions. To the best of our knowledge, LRP has not yet been applied to geoscientific research, and we believe it has great potential in this area. We show how these interpretation techniques can be used to reliably infer scientifically meaningful information from neural networks by applying them to common climate patterns. These results suggest that combining interpretable neural networks with novel scientific hypotheses will open the door to many new avenues in neural network‐related geoscience research.

Indicator Patterns of Forced Change Learned by an Artificial Neural Network

“…Toms et al. (2020) discuss the nuances of using backward optimization for geoscience applications, and we extend its use to interpret differences between climate models and the observations. Briefly, given a trained neural network, an input sample is incrementally adjusted towards the pattern most closely associated with a user‐defined prediction.…”

“…To do this, we utilize a neural network interpretation method called "layerwise relevance propagation" (LRP) to determine the most relevant regions of the input maps for the ANN's prediction (e.g., Bach et al, 2015;Montavon et al, 2017). Toms et al (2020) provide the first detailed discussion of how LRP can be used for interpretable neural networks in geoscience. We also provide the most relevant details of the method here.…”

“…Backward optimization can be used to gain a composite interpretation of the patterns contained within a trained neural network (Olah et al, 2017;Simonyan et al, 2013;Yosinski et al, 2015). Toms et al (2020) discuss the nuances of using backward optimization for geoscience applications, and we extend its use to interpret differences between climate models and the observations. Briefly, given a trained neural network, an input sample is incrementally adjusted towards the pattern most closely associated with a user-defined prediction.…”

A Systematic, Automated Approach for River Segmentation Tested on the Magdalena River (Colombia) and the Baker River (Chile)

“…Clearly, the outputs of AI algorithms heavily depend on images resolution and are hardly exportable to other settings. Finally, even if AI can potentially recognize stretches with similar patterns of characteristics, assigning them a given label, AI can hardly explicitly inform on the role of the attributes in leading to such a determination; advances exist, however, that go in this direction (e.g., Toms et al [20]).…”

Machine Learning-based Weather Support for the 2022 Winter Olympics

“…To improve the accuracy of forecasting, combining the numerical model and AI is an efficient approach (Reichstein et al, 2019;Toms et al, 2019). AI techniques can be used as model-post processing methods, as done in the present project.…”