Imme Ebert-Uphoff

Mark DeMaria, James L. Franklin, Galina Chirokova et al.

In just the past few years multiple data-driven Artificial Intelligence Weather Prediction (AIWP) models have been developed, with new versions appearing almost monthly. Given this rapid development, the applicability of these models to operational forecasting has yet to be adequately explored and documented. To assess their utility for operational tropical cyclone (TC) forecasting, the NHC verification procedure is used to evaluate seven-day track and intensity predictions for northern hemisphere TCs from May-November 2023. Four open-source AIWP models are considered (FourCastNetv1, FourCastNetv2-small, GraphCast-operational and Pangu-Weather). The AIWP track forecast errors and detection rates are comparable to those from the best-performing operational forecast models. However, the AIWP intensity forecast errors are larger than those of even the simplest intensity forecasts based on climatology and persistence. The AIWP models almost always reduce the TC intensity, especially within the first 24 h of the forecast, resulting in a substantial low bias. The contribution of the AIWP models to the NHC model consensus was also evaluated. The consensus track errors are reduced by up to 11% at the longer time periods. The five-day NHC official track forecasts have improved by about 2% per year since 2001, so this represents more than a five-year gain in accuracy. Despite substantial negative intensity biases, the AIWP models have a neutral impact on the intensity consensus. These results show that the current formulation of the AIWP models have promise for operational TC track forecasts, but improved bias corrections or model reformulations will be needed for accurate intensity forecasts.

Lander Ver Hoef, Henry Adams, Emily J. King et al.

Topological data analysis (TDA) is a tool from data science and mathematics that is beginning to make waves in environmental science. In this work, we seek to provide an intuitive and understandable introduction to a tool from TDA that is particularly useful for the analysis of imagery, namely persistent homology. We briefly discuss the theoretical background but focus primarily on understanding the output of this tool and discussing what information it can glean. To this end, we frame our discussion around a guiding example of classifying satellite images from the Sugar, Fish, Flower, and Gravel Dataset produced for the study of mesocale organization of clouds by Rasp et. al. in 2020 (arXiv:1906:01906). We demonstrate how persistent homology and its vectorization, persistence landscapes, can be used in a workflow with a simple machine learning algorithm to obtain good results, and explore in detail how we can explain this behavior in terms of image-level features. One of the core strengths of persistent homology is how interpretable it can be, so throughout this paper we discuss not just the patterns we find, but why those results are to be expected given what we know about the theory of persistent homology. Our goal is that a reader of this paper will leave with a better understanding of TDA and persistent homology, be able to identify problems and datasets of their own for which persistent homology could be helpful, and gain an understanding of results they obtain from applying the included GitHub example code.

Ryan Lagerquist, Imme Ebert-Uphoff

In the last decade, much work in atmospheric science has focused on spatial verification (SV) methods for gridded prediction, which overcome serious disadvantages of pixelwise verification. However, neural networks (NN) in atmospheric science are almost always trained to optimize pixelwise loss functions, even when ultimately assessed with SV methods. This establishes a disconnect between model verification during vs. after training. To address this issue, we develop spatially enhanced loss functions (SELF) and demonstrate their use for a real-world problem: predicting the occurrence of thunderstorms (henceforth, "convection") with NNs. In each SELF we use either a neighbourhood filter, which highlights convection at scales larger than a threshold, or a spectral filter (employing Fourier or wavelet decomposition), which is more flexible and highlights convection at scales between two thresholds. We use these filters to spatially enhance common verification scores, such as the Brier score. We train each NN with a different SELF and compare their performance at many scales of convection, from discrete storm cells to tropical cyclones. Among our many findings are that (a) for a low (high) risk threshold, the ideal SELF focuses on small (large) scales; (b) models trained with a pixelwise loss function perform surprisingly well; (c) however, models trained with a spectral filter produce much better-calibrated probabilities than a pixelwise model. We provide a general guide to using SELFs, including technical challenges and the final Python code, as well as demonstrating their use for the convection problem. To our knowledge this is the most in-depth guide to SELFs in the geosciences.

Akansha Singh Bansal, Yoonjin Lee, Kyle Hilburn et al.

Atmospheric processes involve both space and time. This is why human analysis of atmospheric imagery can often extract more information from animated loops of image sequences than from individual images. Automating such an analysis requires the ability to identify spatio-temporal patterns in image sequences which is a very challenging task, because of the endless possibilities of patterns in both space and time. In this paper we review different concepts and techniques that are useful to extract spatio-temporal context specifically for meteorological applications. In this survey we first motivate the need for these approaches in meteorology using two applications, solar forecasting and detecting convection from satellite imagery. Then we provide an overview of many different concepts and techniques that are helpful for the interpretation of meteorological image sequences, such as (1) feature engineering methods to strengthen the desired signal in the input, using meteorological knowledge, classic image processing, harmonic analysis and topological data analysis (2) explain how different convolution filters (2D/3D/LSTM-convolution) can be utilized strategically in convolutional neural network architectures to find patterns in both space and time (3) discuss the powerful new concept of 'attention' in neural networks and the powerful abilities it brings to the interpretation of image sequences (4) briefly survey strategies from unsupervised, self-supervised and transfer learning to reduce the need for large labeled datasets. We hope that presenting an overview of these tools - many of which are underutilized - will help accelerate progress in this area.

Ryan Lagerquist, Galina Chirokova, Robert DeMaria et al.

Determining the location of a tropical cyclone's (TC) surface circulation center -- "center-fixing" -- is a critical first step in the TC-forecasting process, affecting current/future estimates of track, intensity, and structure. Despite a recent increase in automated center-fixing methods, only one such method (ARCHER-2) is operational, and its best performance is achieved when using microwave or scatterometer data, which are often unavailable. We develop a deep-learning algorithm called GeoCenter; besides a few scalars in the operational Automated Tropical Cyclone Forecasting System, it relies only on geostationary infrared (IR) satellite imagery, which is available for all TC basins at high frequency (10 min) and low latency (< 10 min) during both day and night. GeoCenter ingests an animation (time series) of IR images, including 9 channels at lag times up to 4 hours. The animation is centered at a "first guess" location, offset from the true TC-center location by 48 km on average and sometimes > 100 km; GeoCenter is tasked with correcting this offset. On an independent testing dataset, GeoCenter achieves a mean/median/RMS (root mean square) error of 26.6/22.2/32.4 km for all systems, 24.7/20.8/30.0 km for tropical systems, and 14.6/12.5/17.3 km for category-2--5 hurricanes. These values are similar to ARCHER-2 errors with microwave or scatterometer data, and better than ARCHER-2 errors when only IR data are available. GeoCenter also performs skillful uncertainty quantification, producing a well calibrated ensemble of 150 TC-center locations. Furthermore, all predictors used by GeoCenter are available in real time, which would make GeoCenter easy to implement operationally every 10 min.

Randy J. Chase, Katherine Haynes, Lander Ver Hoef et al.

Clouds and precipitation are important for understanding weather and climate. Simulating clouds and precipitation with traditional numerical weather prediction is challenging because of the sub-grid parameterizations required. Machine learning has been explored for forecasting clouds and precipitation, but early machine learning methods often created blurry forecasts. In this paper we explore a newer method, named score-based diffusion, to nowcast (zero to three hour forecast) clouds and precipitation. We discuss the background and intuition of score-based diffusion models - thus providing a starting point for the community - while exploring the methodology's use for nowcasting geostationary infrared imagery. We experiment with three main types of diffusion models: a standard score-based diffusion model (Diff); a residual correction diffusion model (CorrDiff); and a latent diffusion model (LDM). Our results show that the diffusion models are able to not only advect existing clouds, but also generate and decay clouds, including convective initiation. These results are surprising because the forecasts are initiated with only the past 20 mins of infrared satellite imagery. A case study qualitatively shows the preservation of high resolution features longer into the forecast than a conventional mean-squared error trained U-Net. The best of the three diffusion models tested was the CorrDiff approach, outperforming all other diffusion models, the traditional U-Net, and a persistence forecast by one to two kelvin on root mean squared error. The diffusion models also enable out-of-the-box ensemble generation, which shows skillful calibration, with the spread of the ensemble correlating well to the error.

Nathan Mitchell, Lander Ver Hoef, Imme Ebert-Uphoff et al.

An Explainable Boosting Machine (EBM) is an interpretable machine learning (ML) algorithm that has benefits in high risk applications but has not yet found much use in atmospheric science. The overall goal of this work is twofold: (1) explore the use of EBMs, in combination with feature engineering, to obtain interpretable, physics-based machine learning algorithms for meteorological applications; (2) illustrate these methods for the detection of overshooting top (OTs) in satellite imagery. Specifically, we seek to simplify the process of OT detection by first using mathematical methods to extract key features, such as cloud texture using Gray-Level Co-occurrence Matrices, followed by applying an EBM. Our EBM focuses on the classification task of predicting OT regions, utilizing Channel 2 (visible imagery) and Channel 13 (infrared imagery) of the Advanced Baseline Imager sensor of the Geostationary Operational Environmental Satellite 16. Multi-Radar/Multi-Sensor system convection flags are used as labels to train the EBM model. Note, however, that detecting convection, while related, is different from detecting OTs. Once trained, the EBM was examined and minimally altered to more closely match strategies used by domain scientists to identify OTs. The result of our efforts is a fully interpretable ML algorithm that was developed in a human-machine collaboration. While the final model does not reach the accuracy of more complex approaches, it performs well and represents a significant step toward building fully interpretable ML algorithms for this and other meteorological applications.

Jason Stock, Kyle Hilburn, Imme Ebert-Uphoff et al.

We introduce a transformer-based neural network to generate high-resolution (3km) synthetic radar reflectivity fields at scale from geostationary satellite imagery. This work aims to enhance short-term convective-scale forecasts of high-impact weather events and aid in data assimilation for numerical weather prediction over the United States. Compared to convolutional approaches, which have limited receptive fields, our results show improved sharpness and higher accuracy across various composite reflectivity thresholds. Additional case studies over specific atmospheric phenomena support our quantitative findings, while a novel attribution method is introduced to guide domain experts in understanding model outputs.

Antonios Mamalakis, Elizabeth A. Barnes, Imme Ebert-Uphoff

Methods of eXplainable Artificial Intelligence (XAI) are used in geoscientific applications to gain insights into the decision-making strategy of Neural Networks (NNs) highlighting which features in the input contribute the most to a NN prediction. Here, we discuss our lesson learned that the task of attributing a prediction to the input does not have a single solution. Instead, the attribution results and their interpretation depend greatly on the considered baseline (sometimes referred to as reference point) that the XAI method utilizes; a fact that has been overlooked so far in the literature. This baseline can be chosen by the user or it is set by construction in the method s algorithm, often without the user being aware of that choice. We highlight that different baselines can lead to different insights for different science questions and, thus, should be chosen accordingly. To illustrate the impact of the baseline, we use a large ensemble of historical and future climate simulations forced with the SSP3-7.0 scenario and train a fully connected NN to predict the ensemble- and global-mean temperature (i.e., the forced global warming signal) given an annual temperature map from an individual ensemble member. We then use various XAI methods and different baselines to attribute the network predictions to the input. We show that attributions differ substantially when considering different baselines, as they correspond to answering different science questions. We conclude by discussing some important implications and considerations about the use of baselines in XAI research.

Antonios Mamalakis, Elizabeth A. Barnes, Imme Ebert-Uphoff

Convolutional neural networks (CNNs) have recently attracted great attention in geoscience due to their ability to capture non-linear system behavior and extract predictive spatiotemporal patterns. Given their black-box nature however, and the importance of prediction explainability, methods of explainable artificial intelligence (XAI) are gaining popularity as a means to explain the CNN decision-making strategy. Here, we establish an intercomparison of some of the most popular XAI methods and investigate their fidelity in explaining CNN decisions for geoscientific applications. Our goal is to raise awareness of the theoretical limitations of these methods and gain insight into the relative strengths and weaknesses to help guide best practices. The considered XAI methods are first applied to an idealized attribution benchmark, where the ground truth of explanation of the network is known a priori, to help objectively assess their performance. Secondly, we apply XAI to a climate-related prediction setting, namely to explain a CNN that is trained to predict the number of atmospheric rivers in daily snapshots of climate simulations. Our results highlight several important issues of XAI methods (e.g., gradient shattering, inability to distinguish the sign of attribution, ignorance to zero input) that have previously been overlooked in our field and, if not considered cautiously, may lead to a distorted picture of the CNN decision-making strategy. We envision that our analysis will motivate further investigation into XAI fidelity and will help towards a cautious implementation of XAI in geoscience, which can lead to further exploitation of CNNs and deep learning for prediction problems.

Amy McGovern, Imme Ebert-Uphoff, David John Gagne et al.

Given the growing use of Artificial Intelligence (AI) and machine learning (ML) methods across all aspects of environmental sciences, it is imperative that we initiate a discussion about the ethical and responsible use of AI. In fact, much can be learned from other domains where AI was introduced, often with the best of intentions, yet often led to unintended societal consequences, such as hard coding racial bias in the criminal justice system or increasing economic inequality through the financial system. A common misconception is that the environmental sciences are immune to such unintended consequences when AI is being used, as most data come from observations, and AI algorithms are based on mathematical formulas, which are often seen as objective. In this article, we argue the opposite can be the case. Using specific examples, we demonstrate many ways in which the use of AI can introduce similar consequences in the environmental sciences. This article will stimulate discussion and research efforts in this direction. As a community, we should avoid repeating any foreseeable mistakes made in other domains through the introduction of AI. In fact, with proper precautions, AI can be a great tool to help {\it reduce} climate and environmental injustice. We primarily focus on weather and climate examples but the conclusions apply broadly across the environmental sciences.

Imme Ebert-Uphoff, Ryan Lagerquist, Kyle Hilburn et al.

Neural networks are increasingly used in environmental science applications. Furthermore, neural network models are trained by minimizing a loss function, and it is crucial to choose the loss function very carefully for environmental science applications, as it determines what exactly is being optimized. Standard loss functions do not cover all the needs of the environmental sciences, which makes it important for scientists to be able to develop their own custom loss functions so that they can implement many of the classic performance measures already developed in environmental science, including measures developed for spatial model verification. However, there are very few resources available that cover the basics of custom loss function development comprehensively, and to the best of our knowledge none that focus on the needs of environmental scientists. This document seeks to fill this gap by providing a guide on how to write custom loss functions targeted toward environmental science applications. Topics include the basics of writing custom loss functions, common pitfalls, functions to use in loss functions, examples such as fractions skill score as loss function, how to incorporate physical constraints, discrete and soft discretization, and concepts such as focal, robust, and adaptive loss. While examples are currently provided in this guide for Python with Keras and the TensorFlow backend, the basic concepts also apply to other environments, such as Python with PyTorch. Similarly, while the sample loss functions provided here are from meteorology, these are just examples of how to create custom loss functions. Other fields in the environmental sciences have very similar needs for custom loss functions, e.g., for evaluating spatial forecasts effectively, and the concepts discussed here can be applied there as well. All code samples are provided in a GitHub repository.

Antonios Mamalakis, Imme Ebert-Uphoff, Elizabeth A. Barnes

Despite the increasingly successful application of neural networks to many problems in the geosciences, their complex and nonlinear structure makes the interpretation of their predictions difficult, which limits model trust and does not allow scientists to gain physical insights about the problem at hand. Many different methods have been introduced in the emerging field of eXplainable Artificial Intelligence (XAI), which aim at attributing the network s prediction to specific features in the input domain. XAI methods are usually assessed by using benchmark datasets (like MNIST or ImageNet for image classification). However, an objective, theoretically derived ground truth for the attribution is lacking for most of these datasets, making the assessment of XAI in many cases subjective. Also, benchmark datasets specifically designed for problems in geosciences are rare. Here, we provide a framework, based on the use of additively separable functions, to generate attribution benchmark datasets for regression problems for which the ground truth of the attribution is known a priori. We generate a large benchmark dataset and train a fully connected network to learn the underlying function that was used for simulation. We then compare estimated heatmaps from different XAI methods to the ground truth in order to identify examples where specific XAI methods perform well or poorly. We believe that attribution benchmarks as the ones introduced herein are of great importance for further application of neural networks in the geosciences, and for more objective assessment and accurate implementation of XAI methods, which will increase model trust and assist in discovering new science.

Imme Ebert-Uphoff, Kyle A. Hilburn

Neural networks have opened up many new opportunities to utilize remotely sensed images in meteorology. Common applications include image classification, e.g., to determine whether an image contains a tropical cyclone, and image translation, e.g., to emulate radar imagery for satellites that only have passive channels. However, there are yet many open questions regarding the use of neural networks in meteorology, such as best practices for evaluation, tuning and interpretation. This article highlights several strategies and practical considerations for neural network development that have not yet received much attention in the meteorological community, such as the concept of effective receptive fields, underutilized meteorological performance measures, and methods for NN interpretation, such as synthetic experiments and layer-wise relevance propagation. We also consider the process of neural network interpretation as a whole, recognizing it as an iterative scientist-driven discovery process, and breaking it down into individual steps that researchers can take. Finally, while most work on neural network interpretation in meteorology has so far focused on networks for image classification tasks, we expand the focus to also include networks for image translation.

Kyle A. Hilburn, Imme Ebert-Uphoff, Steven D. Miller

The objective of this research is to develop techniques for assimilating GOES-R Series observations in precipitating scenes for the purpose of improving short-term convective-scale forecasts of high impact weather hazards. Whereas one approach is radiance assimilation, the information content of GOES-R radiances from its Advanced Baseline Imager (ABI) saturates in precipitating scenes, and radiance assimilation does not make use of lightning observations from the GOES Lightning Mapper (GLM). Here, a convolutional neural network (CNN) is developed to transform GOES-R radiances and lightning into synthetic radar reflectivity fields to make use of existing radar assimilation techniques. We find that the ability of CNNs to utilize spatial context is essential for this application and offers breakthrough improvement in skill compared to traditional pixel-by-pixel based approaches. To understand the improved performance, we use a novel analysis methodology that combines several techniques, each providing different insights into the network's reasoning. Channel withholding experiments and spatial information withholding experiments are used to show that the CNN achieves skill at high reflectivity values from the information content in radiance gradients and the presence of lightning. The attribution method, layer-wise relevance propagation, demonstrates that the CNN uses radiance and lightning information synergistically, where lightning helps the CNN focus on which neighboring locations are most important. Synthetic inputs are used to quantify the sensitivity to radiance gradients, showing that sharper gradients produce a stronger response in predicted reflectivity. Finally, geostationary lightning observations are found to be uniquely valuable for their ability to pinpoint locations of strong radar echoes.

Benjamin A. Toms, Elizabeth A. Barnes, Imme Ebert-Uphoff

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.

Anuj Karpatne, Imme Ebert-Uphoff, Sai Ravela et al.

Geosciences is a field of great societal relevance that requires solutions to several urgent problems facing our humanity and the planet. As geosciences enters the era of big data, machine learning (ML) -- that has been widely successful in commercial domains -- offers immense potential to contribute to problems in geosciences. However, problems in geosciences have several unique challenges that are seldom found in traditional applications, requiring novel problem formulations and methodologies in machine learning. This article introduces researchers in the machine learning (ML) community to these challenges offered by geoscience problems and the opportunities that exist for advancing both machine learning and geosciences. We first highlight typical sources of geoscience data and describe their properties that make it challenging to use traditional machine learning techniques. We then describe some of the common categories of geoscience problems where machine learning can play a role, and discuss some of the existing efforts and promising directions for methodological development in machine learning. We conclude by discussing some of the emerging research themes in machine learning that are applicable across all problems in the geosciences, and the importance of a deep collaboration between machine learning and geosciences for synergistic advancements in both disciplines.

Jamal Golmohammadi, Imme Ebert-Uphoff, Sijie He et al.

In this paper, we consider the use of structure learning methods for probabilistic graphical models to identify statistical dependencies in high-dimensional physical processes. Such processes are often synthetically characterized using PDEs (partial differential equations) and are observed in a variety of natural phenomena, including geoscience data capturing atmospheric and hydrological phenomena. Classical structure learning approaches such as the PC algorithm and variants are challenging to apply due to their high computational and sample requirements. Modern approaches, often based on sparse regression and variants, do come with finite sample guarantees, but are usually highly sensitive to the choice of hyper-parameters, e.g., parameter $λ$ for sparsity inducing constraint or regularization. In this paper, we present ACLIME-ADMM, an efficient two-step algorithm for adaptive structure learning, which estimates an edge specific parameter $λ_{ij}$ in the first step, and uses these parameters to learn the structure in the second step. Both steps of our algorithm use (inexact) ADMM to solve suitable linear programs, and all iterations can be done in closed form in an efficient block parallel manner. We compare ACLIME-ADMM with baselines on both synthetic data simulated by partial differential equations (PDEs) that model advection-diffusion processes, and real data (50 years) of daily global geopotential heights to study information flow in the atmosphere. ACLIME-ADMM is shown to be efficient, stable, and competitive, usually better than the baselines especially on difficult problems. On real data, ACLIME-ADMM recovers the underlying structure of global atmospheric circulation, including switches in wind directions at the equator and tropics entirely from the data.

Imme Ebert-Uphoff, Yi Deng

Causal discovery algorithms based on probabilistic graphical models have emerged in geoscience applications for the identification and visualization of dynamical processes. The key idea is to learn the structure of a graphical model from observed spatio-temporal data, which indicates information flow, thus pathways of interactions, in the observed physical system. Studying those pathways allows geoscientists to learn subtle details about the underlying dynamical mechanisms governing our planet. Initial studies using this approach on real-world atmospheric data have shown great potential for scientific discovery. However, in these initial studies no ground truth was available, so that the resulting graphs have been evaluated only by whether a domain expert thinks they seemed physically plausible. This paper seeks to fill this gap. We develop a testbed that emulates two dynamical processes dominant in many geoscience applications, namely advection and diffusion, in a 2D grid. Then we apply the causal discovery based information tracking algorithms to the simulation data to study how well the algorithms work for different scenarios and to gain a better understanding of the physical meaning of the graph results, in particular of instantaneous connections. We make all data sets used in this study available to the community as a benchmark. Keywords: Information flow, graphical model, structure learning, causal discovery, geoscience.