The focus of this workshop is the use of machine learning to help address climate change, encompassing mitigation efforts (reducing greenhouse gas emissions), adaptation measures (preparing for unavoidable consequences), and climate science (our understanding of the climate and future climate predictions). The scope of the workshop includes climate-relevant applications of machine learning to the power sector, buildings and transportation infrastructure, agriculture and land use, extreme event prediction, disaster response, climate policy, and climate finance. The goals of the workshop are: (1) to showcase high-impact applications of ML to climate change mitigation, adaptation, and climate science, (2) to showcase novel and interesting problem settings and challenges for ML techniques, (3) to encourage fruitful collaboration between the ML community and a diverse set of researchers and practitioners from climate change-related fields, and (4) to promote dialogue with decision-makers in the private and public sectors to ensure that the work presented leads to responsible and meaningful deployment.
Tue 6:15 a.m. - 6:30 a.m.
|
Opening Remarks
SlidesLive Video » |
🔗 |
Tue 6:30 a.m. - 7:15 a.m.
|
Daron Acemoglu: Is AI the Solution to Climate Change?
(
Keynote talk
)
SlidesLive Video » Title: Is AI the Solution to Climate Change? Abstract: Many pin their hopes on the tech sector and advances in Artificial Intelligence (AI) and other digital technologies in combating climate change. In this talk, I argue against this optimistic view. I first provide evidence on how AI has been used in US businesses, pointing out that it has had few of the promised benefits and has instead continued the process of inequality-increasing and wage-reducing automation. I then review the evidence on advances in renewable energy, arguing that it has responded strongly to subsidies and prices of fossil fuels, but these advances have slowed down lately. There appears to be no alternative to increasing carbon taxes significantly and investing in renewable and green technologies. There is no evidence that big tech companies have played a leading role, and most existing evidence suggests that big energy companies have typically undermined efforts to switch to renewables. Bio: Daron Acemoglu is an Institute Professor at MIT and an elected fellow of the National Academy of Sciences, American Philosophical Society, the British Academy of Sciences, the Turkish Academy of Sciences, the American Academy of Arts and Sciences, the Econometric Society, the European Economic Association, and the Society of Labor Economists. He is also a member of the Group of Thirty. He is the author of five books, including New York Times bestseller Why Nations Fail: Power, Prosperity, and Poverty (joint with James A. Robinson), Introduction to Modern Economic Growth, and The Narrow Corridor: States, Societies, and the Fate of Liberty (with James A. Robinson). His academic work covers a wide range of areas, including political economy, economic development, economic growth, technological change, inequality, labor economics and economics of networks. Daron Acemoglu has received the inaugural T. W. Shultz Prize from the University of Chicago in 2004, and the inaugural Sherwin Rosen Award for outstanding contribution to labor economics in 2004, Distinguished Science Award from the Turkish Sciences Association in 2006, the John von Neumann Award, Rajk College, Budapest in 2007, the Carnegie Fellowship in 2017, the Jean-Jacques Laffont Prize in 2018, the Global Economy Prize in 2019, and the CME Mathematical and Statistical Research Institute prize in 2021. He was awarded the John Bates Clark Medal in 2005, the Erwin Plein Nemmers Prize in 2012, and the 2016 BBVA Frontiers of Knowledge Award. He holds Honorary Doctorates from the University of Utrecht, the Bosporus University, University of Athens, Bilkent University, the University of Bath, Ecole Normale Superieure, Saclay Paris, and the London Business School. |
Kamer Acemoglu 🔗 |
Tue 7:15 a.m. - 7:25 a.m.
|
Resolving Super Fine-Resolution SIF via Coarsely-Supervised U-Net Regression
(
Spotlight
)
link »
SlidesLive Video » Climate change presents challenges to crop productivity, such as increasing the likelihood of heat stress and drought. Solar-Induced Chlorophyll Fluorescence (SIF) is a powerful way to monitor how crop productivity and photosynthesis are affected by changing climatic conditions. However, satellite SIF observations are only available at a coarse spatial resolution (e.g. 3-5km) in most places, making it difficult to determine how individual crop types or farms are doing. This poses a challenging coarsely-supervised regression task; at training time, we only have access to SIF labels at a coarse resolution (3 km), yet we want to predict SIF at a very fine spatial resolution (30 meters), a 100x increase. We do have some fine-resolution input features (such as Landsat reflectance) that are correlated with SIF, but the nature of the correlation is unknown. To address this, we propose Coarsely-Supervised Regression U-Net (CSR-U-Net), a novel approach to train a U-Net for this coarse supervision setting. CSR-U-Net takes in a fine-resolution input image, and outputs a SIF prediction for each pixel; the average of the pixel predictions is trained to equal the true coarse-resolution SIF for the entire image. Even though this is a very weak form of supervision, CSR-U-Net can still learn to predict accurately, due to its inherent localization abilities, plus additional enhancements that facilitate the incorporation of scientific prior knowledge. CSR-U-Net can resolve fine-grained variations in SIF more accurately than existing averaging-based approaches, which ignore fine-resolution spatial variation during training. CSR-U-Net could also be useful for a wide range of "downscaling'" problems in climate science, such as increasing the resolution of global climate models. |
Joshua Fan · Di Chen · Jiaming Wen · Ying Sun · Carla Gomes 🔗 |
Tue 7:25 a.m. - 7:34 a.m.
|
Detecting Abandoned Oil Wells Using Machine Learning and Semantic Segmentation
(
Spotlight
)
link »
SlidesLive Video » Around the world, there are millions of unplugged abandoned oil and gas wells, leaking methane into the atmosphere. The locations of many of these wells, as well as their greenhouse gas emissions impacts, are unknown. Machine learning methods in computer vision and remote sensing, such as semantic segmentation, have made it possible to quickly analyze large amounts of satellite imagery to detect salient information. This project aims to automatically identify undocumented oil and gas wells in the province of Alberta, Canada to aid in documentation, estimation of emissions and maintenance of high-emitting wells. |
Michelle Lin · David Rolnick 🔗 |
Tue 7:34 a.m. - 7:38 a.m.
|
Semi-Supervised Classification and Segmentation on High Resolution Aerial Images
(
Spotlight
)
link »
SlidesLive Video » FloodNet is a high-resolution image dataset acquired by a small UAV platform, DJI Mavic Pro quadcopters, after Hurricane Harvey. The dataset presents a unique challenge of advancing the damage assessment process for post-disaster scenarios using unlabeled and limited labeled dataset. We propose a solution to address their classification and semantic segmentation challenge. We approach this problem by generating pseudo labels for both classification and segmentation during training and slowly incrementing the amount by which the pseudo label loss affects the final loss. Using this semi-supervised method of training helped us improve our baseline supervised loss by a huge margin for classification, allowing the model to generalize and perform better on the validation and test splits of the dataset. In this paper, we compare and contrast the various methods and models for image classification and semantic segmentation on the FloodNet dataset. |
Sahil Khose · Abhiraj Tiwari · Ankita Ghosh 🔗 |
Tue 7:38 a.m. - 7:46 a.m.
|
A GNN-RNN Approach for Harnessing Geospatial and Temporal Information: Application to Crop Yield Prediction
(
Spotlight
)
link »
SlidesLive Video » Climate change poses new challenges to agricultural production, as crop yields are extremely sensitive to climatic variation. Accurately predicting the effects of weather patterns on crop yield is crucial for addressing issues such as food insecurity, supply stability, and economic planning. Recently, there have been many attempts to use machine learning models for crop yield prediction. However, these models either restrict their tasks to a relatively small region or a short time-period (e.g. a few years), which makes them hard to generalize spatially and temporally. They also view each location as an i.i.d sample, ignoring spatial correlations in the data. In this paper, we introduce a novel graph-based recurrent neural network for crop yield prediction, which incorporates both geographical and temporal structure. Our method is trained, validated, and tested on over 2000 counties from 41 states in the US mainland, covering years from 1981 to 2019. As far as we know, this is the first machine learning method that embeds geographical knowledge in crop yield prediction and predicts crop yields at the county level nationwide. Experimental results show that our proposed method consistently outperforms a wide variety of existing state-of-the-art methods, validating the effectiveness of geospatial and temporal information. |
Joshua Fan · Junwen Bai · Zhiyun Li · Ariel Ortiz-Bobea · Carla Gomes 🔗 |
Tue 7:45 a.m. - 8:30 a.m.
|
Poster Session 1
(
Poster Session
)
Gather.Town Rooms:
Tutorials Track: |
🔗 |
Tue 8:30 a.m. - 9:30 a.m.
|
Discussion Panel 1: Decision Making
(
Discussion Panel
)
SlidesLive Video » |
Lieve M.L Helsen · Lynn Kaack · João M. Costa Sousa · Eliane Ubalijoro 🔗 |
Tue 9:30 a.m. - 9:40 a.m.
|
Two-phase training mitigates class imbalance for camera trap image classification with CNNs
(
Spotlight
)
link »
SlidesLive Video » By leveraging deep learning to automatically classify camera trap images, ecologists can monitor biodiversity conservation efforts and the effects of climate change on ecosystems more efficiently. Due to the imbalanced class-distribution of camera trap datasets, current models are biased towards the majority classes. As a result, they obtain good performance for a few majority classes but poor performance for many minority classes. We used two-phase training to increase the performance for these minority classes. We trained, next to a baseline model, four models that implemented a different versions of two-phase training on a subset of the highly imbalanced Snapshot Serengeti dataset. Our results suggest that two-phase training can improve performance for many minority classes, with limited loss in performance for the other classes. We find that two-phase training based on majority undersampling increases class-specific F1-scores up to 3.0%. We also find that two-phase training outperforms using only oversampling or undersampling by 6.1% in F1-score on average. Finally, we find that a combination of over- and undersampling leads to a better performance than using them individually. |
Ruben Cartuyvels 🔗 |
Tue 9:40 a.m. - 9:49 a.m.
|
Predicting Atlantic Multidecadal Variability
(
Spotlight
)
link »
SlidesLive Video » Atlantic Multidecadal Variability (AMV) describes variations of North Atlantic sea surface temperature with a typical cycle of between 60 and 70 years. AMV strongly impacts local climate over North America and Europe, therefore prediction of AMV, especially the extreme values, is of great societal utility for understanding and responding to regional climate change. This work tests multiple machine learning models to improve the state of AMV prediction from maps of sea surface temperature, salinity, and sea level pressure in the North Atlantic region. We use data from the Community Earth System Model 1 Large Ensemble Project, a state-of-the-art climate model with 3,440 years of data. Our results demonstrate that all of the models we use outperform the traditional persistence forecast baseline. Predicting the AMV is important for identifying future extreme temperatures and precipitation as well as hurricane activity, in Europe and North America up to 25 years in advance. |
Glenn Liu · Peidong Wang · Matthew Beveridge · Young-Oh Kwon · Iddo Drori 🔗 |
Tue 9:49 a.m. - 9:59 a.m.
|
Learned Benchmarks for Subseasonal Forecasting
(
Spotlight
)
link »
SlidesLive Video » We develop a subseasonal forecasting toolkit of simple learned benchmark models that outperform both operational practice and state-of-the-art machine learning and deep learning methods. Our new models include (a) Climatology++, an adaptive alternative to climatology that, for precipitation, is 9% more accurate and 250\% more skillful than the United States operational Climate Forecasting System (CFSv2); (b) CFSv2++, a learned CFSv2 correction that improves temperature and precipitation accuracy by 7-8% and skill by 50-275%; and (c) Persistence++, an augmented persistence model that combines CFSv2 forecasts with lagged measurements to improve temperature and precipitation accuracy by 6-9% and skill by 40-130%. Across the contiguous U.S., these models consistently outperform standard meteorological baselines, state-of-the-art learning methods, and the European Centre for Medium-Range Weather Forecasts ensemble. Overall, we find that augmenting traditional forecasting approaches with learned enhancements yields an effective and computationally inexpensive strategy for building the next generation of subseasonal forecasting benchmarks. |
Soukayna Mouatadid · Paulo Orenstein · Genevieve Flaspohler · Miruna Oprescu · Judah Cohen · Franklyn Wang · Sean Knight · Maria Geogdzhayeva · Sam Levang · Ernest Fraenkel · Lester Mackey
|
Tue 10:00 a.m. - 10:45 a.m.
|
Anima Anandkumar: Role of AI in predicting and mitigating climate change
(
Keynote talk
)
SlidesLive Video » Title: Role of AI in predicting and mitigating climate change Abstract: Predicting extreme weather events in a warming world at fine scales is a grand challenge faced by climate scientists. Policy makers and society at large depend on reliable predictions to plan for the disastrous impact of climate change and develop effective adaptation strategies. Deep learning (DL) offers novel methods that are potentially more accurate and orders of magnitude faster than traditional weather and climate models for predicting extreme events. The Fourier Neural Operator (FNO), a novel deep learning method has shown promising results for predicting complex systems, such as spatio-temporal chaos, turbulence, and weather phenomena. I will give an overview of the method as well as our recent results. Bio: Anima Anandkumar is a Bren Professor at Caltech and Director of ML Research at NVIDIA. She was previously a Principal Scientist at Amazon Web Services. She has received several honors such as Alfred. P. Sloan Fellowship, NSF Career Award, Young investigator awards from DoD, and Faculty Fellowships from Microsoft, Google, Facebook, and Adobe. She is part of the World Economic Forum's Expert Network. She is passionate about designing principled AI algorithms and applying them in interdisciplinary applications. Her research focus is on unsupervised AI, optimization, and tensor methods. |
Anima Anandkumar 🔗 |
Tue 10:45 a.m. - 10:55 a.m.
|
ClimART: A Benchmark Dataset for Emulating Atmospheric Radiative Transfer in Weather and Climate Models
(
Spotlight
)
link »
SlidesLive Video » Numerical simulations of Earth's weather and climate require substantial amounts of computation. This has led to a growing interest in replacing subroutines that explicitly compute physical processes with approximate machine learning (ML) methods that are fast at inference time. Within weather and climate models, atmospheric radiative transfer (RT) calculations are especially expensive. This has made them a popular target for neural network-based emulators. However, prior work is hard to compare due to the lack of a comprehensive dataset and standardized best practices for ML benchmarking. To fill this gap, we introduce the \climart dataset, with more than \emph{10 million samples from present, pre-industrial, and future climate conditions}. ClimART poses several methodological challenges for the ML community, such as multiple out-of-distribution test sets, underlying domain physics, and a trade-off between accuracy and inference speed. We also present several novel baselines that indicate shortcomings of the datasets and network architectures used in prior work. |
Salva Rühling Cachay · Venkatesh Ramesh · Jason N. S. Cole · Howard Barker · David Rolnick 🔗 |
Tue 10:55 a.m. - 11:05 a.m.
|
DeepQuake: Artificial Intelligence for Earthquake Forecasting Using Fine-Grained Climate Data
(
Spotlight
)
link »
SlidesLive Video » Earthquakes are one of the most catastrophic natural disasters, making accurate, fine-grained, and real-time earthquake forecasting extremely important for the safety and security of human lives. In this work, we propose DeepQuake, a hybrid physics and deep learning model for fine-grained earthquake forecasting using time-series data of the horizontal displacement of earth’s surface measured from continuously operating Global Positioning System (cGPS) data. Recent studies using cGPS data have established a link between transient deformation within earth's crust to climate variables. DeepQuake’s physics-based pre-processing algorithm extracts relevant features including the x, y, and xy components of strain in earth’s crust, capturing earth’s elastic response to these climate variables, and feeds it into a deep learning neural network to predict key earthquake variables such as the time, location, magnitude, and depth of a future earthquake. Results across California show promising correlations between cGPS derived strain patterns and the earthquake catalog ground truth for a given location and time. |
Yash Narayan 🔗 |
Tue 11:05 a.m. - 11:18 a.m.
|
Hurricane Forecasting: A Novel Multimodal Machine Learning Framework
(
Spotlight
)
link »
SlidesLive Video » This paper describes a machine learning (ML) framework for tropical cyclone intensity and track forecasting, combining multiple distinct ML techniques and utilizing diverse data sources. Our framework, which we refer to as Hurricast (HURR), is built upon the combination of distinct data processing techniques using gradient-boosted trees and novel encoder-decoder architectures, including CNN, GRU and Transformers components. We propose a deep-learning feature extractor methodology to mix spatial-temporal data with statistical data efficiently. Our multimodal framework unleashes the potential of making forecasts based on a wide range of data sources, including historical storm data, and visual data such as reanalysis atmospheric images. We evaluate our models with current operational forecasts in North Atlantic (NA) and Eastern Pacific (EP) basins on 2016-2019 for 24-hour lead time, and show our models consistently outperform statistical-dynamical models and compete with the best dynamical models. Furthermore, the inclusion of Hurricast into an operational forecast consensus model leads to a significant improvement of 5% - 15% over NHC's official forecast, thus highlighting the complementary properties with existing approaches. |
Léonard Boussioux · Cynthia Zeng · Dimitris Bertsimas · Théo Guenais 🔗 |
Tue 11:20 a.m. - 11:30 a.m.
|
Emissions-aware electricity network expansion planning via implicit differentiation
(
Spotlight
)
link »
SlidesLive Video » We consider a variant of the classical problem of designing or expanding an electricity network. Instead of minimizing only investment and production costs, however, we seek to minimize some mixture of cost and greenhouse gas emissions, even if the underlying dispatch model does not tax emissions. This enables grid planners to directly minimize consumption-based emissions, when expanding or modifying the grid, regardless of whether or not the carbon market incorporates a carbon tax. We solve this problem using gradient descent with implicit differentiation, a technique recently popularized in machine learning. To demonstrate the method, we optimize transmission and storage resources on the IEEE 14-bus test network and compare our solution to one generated by standard planning with a carbon tax. Our solution significantly reduces emissions for the same levelized cost of electricity. |
Anthony Degleris · Lucas Fuentes · Abbas El Gamal · Ram Rajagopal 🔗 |
Tue 11:30 a.m. - 12:15 p.m.
|
Tianzhen Hong: Machine Learning for Smart Buildings: Applications and Perspectives
(
Keynote talk
)
SlidesLive Video » Title: Machine Learning for Smart Buildings: Applications and Perspectives Abstract: Fueled by big data, powerful computing, and advanced algorithms, machine learning has been explored and applied to smart buildings and has demonstrated its potential to enhance building performance. This talk presents an overview of how machine learning has been applied across different stages of building life cycle with a focus on building design, operation, and control. A few applications using machine learning will be presented. Challenges and opportunities of applying machine learning to buildings research will be discussed also. Bio: Dr. Tianzhen Hong is a Senior Scientist and Deputy Head of the Building Technologies Department of LBNL. He leads the Urban Systems Group and a team with research on data, methods, computing, occupant behavior, and policy for design and operation of low energy buildings and sustainable urban systems. He is an IBPSA Fellow and ASHRAE Fellow. He received B.Eng. and Ph.D. in HVACR, and B.Sc. in Applied Mathematics from Tsinghua University, China. |
Tianzhen Hong 🔗 |
Tue 12:15 p.m. - 1:00 p.m.
|
Poster Session 2
(
Poster Session
)
Gather.Town rooms:
|
|
Tue 1:00 p.m. - 2:00 p.m.
|
Discussion Panel: Data
(
Discussion Panel
)
SlidesLive Video » This panel will be on the data and innovation gaps and opportunities needed to support large-scale projects at the intersection of climate change and AI, and what approaches should be taken, in different contexts (academia, industry, government, and intersections among them) to address the data challenges. |
Bethany Lusch · Kakani Katija · brookie guzder-williams · Tricia Martinez 🔗 |
Tue 2:00 p.m. - 2:10 p.m.
|
Tutorials track intro
(
Track intro
)
SlidesLive Video » |
🔗 |
Tue 2:10 p.m. - 2:20 p.m.
|
A day in a sustainable life
(
Tutorial
)
link »
SlidesLive Video » In this notebook, we show the reader how to use an electrical battery to minimize the operational carbon intensity of a building. The central idea is to charge the battery when the carbon intensity of the grid energy mix is low, and vice versa. The same methodology is used in practice to optimise for a number of different objective functions, including energy costs. Taking the hypothetical case of Pi, an eco-conscious and tech-savvy householder in the UK, we walk the reader through getting carbon intensity data, and how to use this with a number of different optimisation algorithms to decarbonise. Starting off with easy-to-understand, brute force search, we establish a baseline for subsequent (hopefully smarter) optimization algorithms. This should come naturally, since in their day job Pi is a data scientist where they often use grid and random search to tune hyperparameters of ML models. The second optimization algorithm we explore is a genetic algorithm, which belongs to the class of derivative free optimizers and is consequently extremely versatile. However, the flexibility of these algorithms comes at the cost of computational speed and effort. In many situations, it makes sense to utilize an optimization method which can make use of the special structure in the problem. As the final step, we see how Pi can optimally solve the problem of minimizing their carbon intensity by formulating it as a linear program. Along the way, we also keep an eye out for some of the most important challenges that arise in practice. |
Hussain Kazmi 🔗 |
Tue 2:20 p.m. - 2:30 p.m.
|
Open Catalyst Project: An Introduction to ML applied to Molecular Simulations
(
Tutorial
)
link »
SlidesLive Video » As the world continues to battle energy scarcity and climate change, the future of our energy infrastructure is a growing challenge. Renewable energy technologies offer the opportunity to drive efficient carbon-neutral means for energy storage and generation. Doing so, however, requires the discovery of efficient and economic catalysts (materials) to accelerate associated chemical processes. A common approach in discovering high performance catalysts is using molecular simulations. Specifically, each simulation models the interaction of a catalyst surface with molecules that are commonly seen in electrochemical reactions. By predicting these interactions accurately, the catalyst's impact on the overall rate of a chemical reaction may be estimated. The Open Catalyst Project (OCP) aims to develop new ML methods and models to accelerate the catalyst simulation process for renewable energy technologies and improve our ability to predict properties across catalyst composition. The initial release of the Open Catalyst 2020 (OC20) dataset presented the largest open dataset of molecular combinations, spanning 55 unique elements and over 130M+ data points. We will present a comprehensive tutorial of the Open Catalyst Project repository, including (1) Accessing & visualizing the dataset, (2) Overview of the various tasks, (3) Training graph neural network (GNN) models, (4) Developing your own model for OCP, (5) Running ML-driven simulations, and (6) Visualizing the results. Primary tools include PyTorch and PyTorch Geometric. No background in chemistry is assumed. Following this tutorial we hope to better equip attendees with a basic understanding of the data and repository. |
Muhammed Shuaibi 🔗 |
Tue 2:30 p.m. - 3:15 p.m.
|
Amy McGovern: Developing Trustworthy AI for Weather and Climate
(
Keynote talk
)
link »
SlidesLive Video » Title: Developing Trustworthy AI for Weather and Climate Abstract: In this talk we give an overview of the work of the NSF AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography. For this talk, we will focus on our applications to weather and climate predictions, including convective hazards and extreme heat. We also briefly review the need for the development of principles for ethical and responsible AI for weather and climate. Bio: Lloyd G. and Joyce Austin Presidential Professor, School of Computer Science and School of Meteorology Director, NSF AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography University of Oklahoma |
Amy McGovern 🔗 |
Tue 3:15 p.m. - 3:30 p.m.
|
Closing remarks and awards
(
Closing remarks
)
SlidesLive Video » |
🔗 |
Tue 3:30 p.m. - 4:15 p.m.
|
Poster Session 3
(
Poster Session
)
Gather.Town Rooms:
|
|
Tue 4:15 p.m. - 5:00 p.m.
|
Gather.Town networking
(
Networking
)
|
🔗 |