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

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 backwards 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.

[1]  Yoshua Bengio,et al.  Tackling Climate Change with Machine Learning , 2019, ACM Comput. Surv..

[2]  Alex Lopatka Meteorologists predict better weather forecasting with AI , 2019 .

[3]  Deborah Silver,et al.  Feature Visualization , 1994, Scientific Visualization.

[4]  Sebastian Ruder,et al.  An overview of gradient descent optimization algorithms , 2016, Vestnik komp'iuternykh i informatsionnykh tekhnologii.

[5]  Thomas Bolton,et al.  Applications of Deep Learning to Ocean Data Inference and Subgrid Parameterization , 2019, Journal of Advances in Modeling Earth Systems.

[6]  Peter John Huybers,et al.  Long-lead predictions of eastern United States hot days from Pacific sea surface temperatures , 2016 .

[7]  Alexander Binder,et al.  Explaining nonlinear classification decisions with deep Taylor decomposition , 2015, Pattern Recognit..

[8]  Wojciech Samek,et al.  Explaining and Interpreting LSTMs , 2019, Explainable AI.

[9]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[10]  Karthik Kashinath,et al.  Testing the Reliability of Interpretable Neural Networks in Geoscience Using the Madden-Julian Oscillation , 2019 .

[11]  Y. Nesterov A method for unconstrained convex minimization problem with the rate of convergence o(1/k^2) , 1983 .

[12]  Alexander Binder,et al.  On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation , 2015, PloS one.

[13]  J. Hurrell,et al.  Viewing Forced Climate Patterns Through an AI Lens , 2019, Geophysical Research Letters.

[14]  Arthur H. Rosenfeld,et al.  A New Estimate of the AverageEarth Surface Land TemperatureSpanning 1753 to 2011 , 2013 .

[15]  Wojciech Samek,et al.  Toward Interpretable Machine Learning: Transparent Deep Neural Networks and Beyond , 2020, ArXiv.

[16]  S. E. Haupt,et al.  Using Artificial Intelligence to Improve Real-Time Decision-Making for High-Impact Weather , 2017 .

[17]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[18]  Sotirios A. Tsaftaris,et al.  Understanding Deep Neural Networks for Regression in Leaf Counting , 2019, 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[19]  Chris North,et al.  Intelligent systems for geosciences , 2018, Communications of the ACM.

[20]  Chester F. Ropelewski,et al.  North American Precipitation and Temperature Patterns Associated with the El Niño/Southern Oscillation (ENSO) , 1986 .

[21]  Xin Wang,et al.  Different impacts of various El Niño events on the Indian Ocean Dipole , 2013, Climate Dynamics.

[22]  Matthew D. Collins,et al.  Climate predictability on interannual to decadal time scales: the initial value problem , 2002 .

[23]  Pierre Gentine,et al.  Deep learning to represent subgrid processes in climate models , 2018, Proceedings of the National Academy of Sciences.

[24]  Adam H. Monahan,et al.  Nonlinear Principal Component Analysis: Tropical Indo–Pacific Sea Surface Temperature and Sea Level Pressure , 2001 .

[25]  Simon Haykin,et al.  Neural Networks: A Comprehensive Foundation , 1998 .

[26]  Swadhin K. Behera,et al.  El Niño Modoki and its possible teleconnection , 2007 .

[27]  Klaus-Robert Müller,et al.  iNNvestigate neural networks! , 2018, J. Mach. Learn. Res..

[28]  Michael J. McPhaden,et al.  Increasing intensity of El Niño in the central-equatorial Pacific: INCREASING INTENSITY OF EL NIÑO , 2010 .

[29]  A. Gershunov ENSO Influence on Intraseasonal Extreme Rainfall and Temperature Frequencies in the Contiguous United States: Implications for Long-Range Predictability , 1998 .

[30]  Amy McGovern,et al.  Making the Black Box More Transparent: Understanding the Physical Implications of Machine Learning , 2019, Bulletin of the American Meteorological Society.

[31]  D. Smith,et al.  Multi‐year predictability of the tropical Atlantic atmosphere driven by the high latitude North Atlantic Ocean , 2011 .

[32]  E. Rasmusson,et al.  Meteorological Aspects of the El Ni�o/Southern Oscillation , 1983, Science.

[33]  P. O'Gorman,et al.  Using Machine Learning to Parameterize Moist Convection: Potential for Modeling of Climate, Climate Change, and Extreme Events , 2018, Journal of Advances in Modeling Earth Systems.

[34]  Francisco J. Doblas-Reyes,et al.  Seasonal climate predictability and forecasting: status and prospects , 2013 .

[35]  Jian Sun,et al.  Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification , 2015, 2015 IEEE International Conference on Computer Vision (ICCV).

[36]  Wojciech Samek,et al.  Explainable AI: Interpreting, Explaining and Visualizing Deep Learning , 2019, Explainable AI.

[37]  M. Gardner,et al.  Neural network modelling and prediction of hourly NOx and NO2 concentrations in urban air in London , 1999 .

[38]  M. Ting,et al.  Covariabilities of Winter U.S. Precipitation and Pacific Sea Surface Temperatures , 2000 .

[39]  Antonietta Capotondi,et al.  Predictability of US West Coast Ocean Temperatures is not solely due to ENSO , 2019, Scientific Reports.

[40]  S. Philander,et al.  El Niño Southern Oscillation phenomena , 1983, Nature.

[41]  Arthur H. Rosenfeld,et al.  A New Estimate of the AverageEarth Surface Land TemperatureSpanning 1753 to 2011 , 2013 .

[42]  Douglas W. Nychka,et al.  Interpretable Deep Learning for Spatial Analysis of Severe Hailstorms , 2019 .

[43]  Noah D. Brenowitz,et al.  Prognostic Validation of a Neural Network Unified Physics Parameterization , 2018, Geophysical Research Letters.

[44]  Maarten V. de Hoop,et al.  Machine learning for data-driven discovery in solid Earth geoscience , 2019, Science.

[45]  Hod Lipson,et al.  Understanding Neural Networks Through Deep Visualization , 2015, ArXiv.

[46]  Masayoshi Ishii,et al.  Centennial-Scale Sea Surface Temperature Analysis and Its Uncertainty , 2014 .

[47]  Stephen G. Penny,et al.  Artificial Intelligence May Be Key to Better Weather Forecasts , 2019, Eos.

[48]  Joachim Denzler,et al.  Deep learning and process understanding for data-driven Earth system science , 2019, Nature.

[49]  Jeong-Hwan Kim,et al.  Deep learning for multi-year ENSO forecasts , 2019, Nature.

[50]  Amy McGovern,et al.  Deep Learning for Spatially Explicit Prediction of Synoptic-Scale Fronts , 2019, Weather and Forecasting.

[51]  Geoffrey E. Hinton,et al.  Deep Learning , 2015, Nature.

[52]  D. Chalikov,et al.  New Approach to Calculation of Atmospheric Model Physics: Accurate and Fast Neural Network Emulation of Longwave Radiation in a Climate Model , 2005 .

[53]  Andrew Zisserman,et al.  Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps , 2013, ICLR.

[54]  Karthik Kashinath,et al.  Deep Learning for Scientific Inference from Geophysical Data: The Madden-Julian Oscillation as a Test Case , 2019, 1902.04621.

[55]  Anuj Karpatne,et al.  Machine Learning for the Geosciences: Challenges and Opportunities , 2017, IEEE Transactions on Knowledge and Data Engineering.

[56]  Alain Chedin,et al.  A Neural Network Approach for a Fast and Accurate Computation of a Longwave Radiative Budget , 1998 .

[57]  Jürgen Schmidhuber,et al.  Learning to forget: continual prediction with LSTM , 1999 .

[58]  Klaus Wolter,et al.  Short-Term Climate Extremes over the Continental United States and ENSO. Part I: Seasonal Temperatures , 1999 .

[59]  Arnt-Børre Salberg,et al.  Machine intelligence and the data-driven future of marine science , 2020, ICES Journal of Marine Science.

[60]  Yuan Zhang,et al.  Is Climate Variability over the North Pacific a Linear Response to ENSO , 1996 .

[61]  Wojciech Samek,et al.  Methods for interpreting and understanding deep neural networks , 2017, Digit. Signal Process..

[62]  Dietmar Dommenget,et al.  Analysis of the non-linearity in the pattern and time evolution of El Niño southern oscillation , 2013, Climate Dynamics.

[63]  Wojciech Samek,et al.  Explainable ai – preface , 2019 .

[64]  Noah D. Brenowitz,et al.  Spatially Extended Tests of a Neural Network Parametrization Trained by Coarse‐Graining , 2019, Journal of Advances in Modeling Earth Systems.

[65]  Qi Li,et al.  Artificial neural networks forecasting of PM2.5 pollution using air mass trajectory based geographic model and wavelet transformation , 2015 .

[66]  Alexander Binder,et al.  Unmasking Clever Hans predictors and assessing what machines really learn , 2019, Nature Communications.