Computer modeling

Subterms

    Latest story

    88 Shares119 Views

    Making climate models relevant for local decision-makers

    by

    Climate models are a key technology in predicting the impacts of climate change. By running simulations of the Earth’s climate, scientists and policymakers can estimate conditions like sea level rise, flooding, and rising temperatures, and make decisions about how to appropriately respond. But current climate models struggle to provide this information quickly or affordably enough to be useful on smaller scales, such as the size of a city. Now, authors of a new open-access paper published in the Journal of Advances in Modeling Earth Systems have found a method to leverage machine learning to utilize the benefits of current climate models, while reducing the computational costs needed to run them. “It turns the traditional wisdom on its head,” says Sai Ravela, a principal research scientist in MIT’s Department of Earth, Atmospheric and Planetary Sciences (EAPS) who wrote the paper with EAPS postdoc Anamitra Saha. Traditional wisdomIn climate modeling, downscaling is the process of using a global climate model with coarse resolution to generate finer details over smaller regions. Imagine a digital picture: A global model is a large picture of the world with a low number of pixels. To downscale, you zoom in on just the section of the photo you want to look at — for example, Boston. But because the original picture was low resolution, the new version is blurry; it doesn’t give enough detail to be particularly useful. “If you go from coarse resolution to fine resolution, you have to add information somehow,” explains Saha. Downscaling attempts to add that information back in by filling in the missing pixels. “That addition of information can happen two ways: Either it can come from theory, or it can come from data.” Conventional downscaling often involves using models built on physics (such as the process of air rising, cooling, and condensing, or the landscape of the area), and supplementing it with statistical data taken from historical observations. But this method is computationally taxing: It takes a lot of time and computing power to run, while also being expensive. A little bit of both In their new paper, Saha and Ravela have figured out a way to add the data another way. They’ve employed a technique in machine learning called adversarial learning. It uses two machines: One generates data to go into our photo. But the other machine judges the sample by comparing it to actual data. If it thinks the image is fake, then the first machine has to try again until it convinces the second machine. The end-goal of the process is to create super-resolution data. Using machine learning techniques like adversarial learning is not a new idea in climate modeling; where it currently struggles is its inability to handle large amounts of basic physics, like conservation laws. The researchers discovered that simplifying the physics going in and supplementing it with statistics from the historical data was enough to generate the results they needed. “If you augment machine learning with some information from the statistics and simplified physics both, then suddenly, it’s magical,” says Ravela. He and Saha started with estimating extreme rainfall amounts by removing more complex physics equations and focusing on water vapor and land topography. They then generated general rainfall patterns for mountainous Denver and flat Chicago alike, applying historical accounts to correct the output. “It’s giving us extremes, like the physics does, at a much lower cost. And it’s giving us similar speeds to statistics, but at much higher resolution.” Another unexpected benefit of the results was how little training data was needed. “The fact that that only a little bit of physics and little bit of statistics was enough to improve the performance of the ML [machine learning] model … was actually not obvious from the beginning,” says Saha. It only takes a few hours to train, and can produce results in minutes, an improvement over the months other models take to run. Quantifying risk quicklyBeing able to run the models quickly and often is a key requirement for stakeholders such as insurance companies and local policymakers. Ravela gives the example of Bangladesh: By seeing how extreme weather events will impact the country, decisions about what crops should be grown or where populations should migrate to can be made considering a very broad range of conditions and uncertainties as soon as possible.“We can’t wait months or years to be able to quantify this risk,” he says. “You need to look out way into the future and at a large number of uncertainties to be able to say what might be a good decision.”While the current model only looks at extreme precipitation, training it to examine other critical events, such as tropical storms, winds, and temperature, is the next step of the project. With a more robust model, Ravela is hoping to apply it to other places like Boston and Puerto Rico as part of a Climate Grand Challenges project.“We’re very excited both by the methodology that we put together, as well as the potential applications that it could lead to,” he says.  More

    More stories

    • 150 Shares169 Views

      in Environment

      Two MIT PhD students awarded J-WAFS fellowships for their research on water

      by

      Since 2014, the Abdul Latif Jameel Water and Food Systems Lab (J-WAFS) has advanced interdisciplinary research aimed at solving the world’s most pressing water and food security challenges to meet human needs. In 2017, J-WAFS established the Rasikbhai L. Meswani Water Solutions Fellowship and the J-WAFS Graduate Student Fellowship. These fellowships provide support to outstanding MIT graduate students who are pursuing research that has the potential to improve water and food systems around the world. Recently, J-WAFS awarded the 2024-25 fellowships to Jonathan Bessette and Akash Ball, two MIT PhD students dedicated to addressing water scarcity by enhancing desalination and purification processes. This work is of important relevance since the world’s freshwater supply has been steadily depleting due to the effects of climate change. In fact, one-third of the global population lacks access to safe drinking water. Bessette and Ball are focused on designing innovative solutions to enhance the resilience and sustainability of global water systems. To support their endeavors, J-WAFS will provide each recipient with funding for one academic semester for continued research and related activities.“This year, we received many strong fellowship applications,” says J-WAFS executive director Renee J. Robins. “Bessette and Ball both stood out, even in a very competitive pool of candidates. The award of the J-WAFS fellowships to these two students underscores our confidence in their potential to bring transformative solutions to global water challenges.”2024-25 Rasikbhai L. Meswani Fellowship for Water SolutionsThe Rasikbhai L. Meswani Fellowship for Water Solutions is a doctoral fellowship for students pursuing research related to water and water supply at MIT. The fellowship is made possible by Elina and Nikhil Meswani and family. Jonathan Bessette is a doctoral student in the Global Engineering and Research (GEAR) Center within the Department of Mechanical Engineering at MIT, advised by Professor Amos Winter. His research is focused on water treatment systems for the developing world, mainly desalination, or the process in which salts are removed from water. Currently, Bessette is working on designing and constructing a low-cost, deployable, community-scale desalination system for humanitarian crises.In arid and semi-arid regions, groundwater often serves as the sole water source, despite its common salinity issues. Many remote and developing areas lack reliable centralized power and water systems, making brackish groundwater desalination a vital, sustainable solution for global water scarcity. “An overlooked need for desalination is inland groundwater aquifers, rather than in coastal areas,” says Bessette. “This is because much of the population lives far enough from a coast that seawater desalination could never reach them. My work involves designing low-cost, sustainable, renewable-powered desalination technologies for highly constrained situations, such as drinking water for remote communities,” he adds.To achieve this goal, Bessette developed a batteryless, renewable electrodialysis desalination system. The technology is energy-efficient, conserves water, and is particularly suited for challenging environments, as it is decentralized and sustainable. The system offers significant advantages over the conventional reverse osmosis method, especially in terms of reduced energy consumption for treating brackish water. Highlighting Bessette’s capacity for engineering insight, his advisor noted the “simple and elegant solution” that Bessette and a staff engineer, Shane Pratt, devised that negated the need for the system to have large batteries. Bessette is now focusing on simplifying the system’s architecture to make it more reliable and cost-effective for deployment in remote areas.Growing up in upstate New York, Bessette completed a bachelor’s degree at the State University of New York at Buffalo. As an undergrad, he taught middle and high school students in low-income areas of Buffalo about engineering and sustainability. However, he cited his junior-year travel to India and his experience there measuring water contaminants in rural sites as cementing his dedication to a career addressing food, water, and sanitation challenges. In addition to his doctoral research, his commitment to these goals is further evidenced by another project he is pursuing, funded by a J-WAFS India grant, that uses low-cost, remote sensors to better understand water fetching practices. Bessette is conducting this work with fellow MIT student Gokul Sampath in order to help families in rural India gain access to safe drinking water.2024-25 J-WAFS Graduate Student Fellowship for Water and Food SolutionsThe J-WAFS Graduate Student Fellowship is supported by the J-WAFS Research Affiliate Program, which offers companies the opportunity to engage with MIT on water and food research. Current fellowship support was provided by two J-WAFS Research Affiliates: Xylem, a leading U.S.-based provider of water treatment and infrastructure solutions, and GoAigua, a Spanish company at the forefront of digital transformation in the water industry through innovative solutions. Akash Ball is a doctoral candidate in the Department of Chemical Engineering, advised by Professor Heather Kulik. His research focuses on the computational discovery of novel functional materials for energy-efficient ion separation membranes with high selectivity. Advanced membranes like these are increasingly needed for applications such as water desalination, battery recycling, and removal of heavy metals from industrial wastewater. “Climate change, water pollution, and scarce freshwater reserves cause severe water distress for about 4 billion people annually, with 2 billion in India and China’s semiarid regions,” Ball notes. “One potential solution to this global water predicament is the desalination of seawater, since seawater accounts for 97 percent of all water on Earth.”Although several commercial reverse osmosis membranes are currently available, these membranes suffer several problems, like slow water permeation, permeability-selectivity trade-off, and high fabrication costs. Metal-organic frameworks (MOFs) are porous crystalline materials that are promising candidates for highly selective ion separation with fast water transport due to high surface area, the presence of different pore windows, and the tunability of chemical functionality.In the Kulik lab, Ball is developing a systematic understanding of how MOF chemistry and pore geometry affect water transport and ion rejection rates. By the end of his PhD, Ball plans to identify existing, best-performing MOFs with unparalleled water uptake using machine learning models, propose novel hypothetical MOFs tailored to specific ion separations from water, and discover experimental design rules that enable the synthesis of next-generation membranes.  Ball’s advisor praised the creativity he brings to his research, and his leadership skills that benefit her whole lab. Before coming to MIT, Ball obtained a master’s degree in chemical engineering from the Indian Institute of Technology (IIT) Bombay and a bachelor’s degree in chemical engineering from Jadavpur University in India. During a research internship at IIT Bombay in 2018, he worked on developing a technology for in situ arsenic detection in water. Like Bessette, he noted the impact of this prior research experience on his interest in global water challenges, along with his personal experience growing up in an area in India where access to safe drinking water was not guaranteed. More

    • 125 Shares119 Views

      in Environment

      Using deep learning to image the Earth’s planetary boundary layer

      by

      Although the troposphere is often thought of as the closest layer of the atmosphere to the Earth’s surface, the planetary boundary layer (PBL) — the lowest layer of the troposphere — is actually the part that most significantly influences weather near the surface. In the 2018 planetary science decadal survey, the PBL was raised as an important scientific issue that has the potential to enhance storm forecasting and improve climate projections.  

      “The PBL is where the surface interacts with the atmosphere, including exchanges of moisture and heat that help lead to severe weather and a changing climate,” says Adam Milstein, a technical staff member in Lincoln Laboratory’s Applied Space Systems Group. “The PBL is also where humans live, and the turbulent movement of aerosols throughout the PBL is important for air quality that influences human health.” 

      Although vital for studying weather and climate, important features of the PBL, such as its height, are difficult to resolve with current technology. In the past four years, Lincoln Laboratory staff have been studying the PBL, focusing on two different tasks: using machine learning to make 3D-scanned profiles of the atmosphere, and resolving the vertical structure of the atmosphere more clearly in order to better predict droughts.  

      This PBL-focused research effort builds on more than a decade of related work on fast, operational neural network algorithms developed by Lincoln Laboratory for NASA missions. These missions include the Time-Resolved Observations of Precipitation structure and storm Intensity with a Constellation of Smallsats (TROPICS) mission as well as Aqua, a satellite that collects data about Earth’s water cycle and observes variables such as ocean temperature, precipitation, and water vapor in the atmosphere. These algorithms retrieve temperature and humidity from the satellite instrument data and have been shown to significantly improve the accuracy and usable global coverage of the observations over previous approaches. For TROPICS, the algorithms help retrieve data that are used to characterize a storm’s rapidly evolving structures in near-real time, and for Aqua, it has helped increase forecasting models, drought monitoring, and fire prediction. 

      These operational algorithms for TROPICS and Aqua are based on classic “shallow” neural networks to maximize speed and simplicity, creating a one-dimensional vertical profile for each spectral measurement collected by the instrument over each location. While this approach has improved observations of the atmosphere down to the surface overall, including the PBL, laboratory staff determined that newer “deep” learning techniques that treat the atmosphere over a region of interest as a three-dimensional image are needed to improve PBL details further.

      “We hypothesized that deep learning and artificial intelligence (AI) techniques could improve on current approaches by incorporating a better statistical representation of 3D temperature and humidity imagery of the atmosphere into the solutions,” Milstein says. “But it took a while to figure out how to create the best dataset — a mix of real and simulated data; we needed to prepare to train these techniques.”

      The team collaborated with Joseph Santanello of the NASA Goddard Space Flight Center and William Blackwell, also of the Applied Space Systems Group, in a recent NASA-funded effort showing that these retrieval algorithms can improve PBL detail, including more accurate determination of the PBL height than the previous state of the art. 

      While improved knowledge of the PBL is broadly useful for increasing understanding of climate and weather, one key application is prediction of droughts. According to a Global Drought Snapshot report released last year, droughts are a pressing planetary issue that the global community needs to address. Lack of humidity near the surface, specifically at the level of the PBL, is the leading indicator of drought. While previous studies using remote-sensing techniques have examined the humidity of soil to determine drought risk, studying the atmosphere can help predict when droughts will happen.  

      In an effort funded by Lincoln Laboratory’s Climate Change Initiative, Milstein, along with laboratory staff member Michael Pieper, are working with scientists at NASA’s Jet Propulsion Laboratory (JPL) to use neural network techniques to improve drought prediction over the continental United States. While the work builds off of existing operational work JPL has done incorporating (in part) the laboratory’s operational “shallow” neural network approach for Aqua, the team believes that this work and the PBL-focused deep learning research work can be combined to further improve the accuracy of drought prediction. 

      “Lincoln Laboratory has been working with NASA for more than a decade on neural network algorithms for estimating temperature and humidity in the atmosphere from space-borne infrared and microwave instruments, including those on the Aqua spacecraft,” Milstein says. “Over that time, we have learned a lot about this problem by working with the science community, including learning about what scientific challenges remain. Our long experience working on this type of remote sensing with NASA scientists, as well as our experience with using neural network techniques, gave us a unique perspective.”

      According to Milstein, the next step for this project is to compare the deep learning results to datasets from the National Oceanic and Atmospheric Administration, NASA, and the Department of Energy collected directly in the PBL using radiosondes, a type of instrument flown on a weather balloon. “These direct measurements can be considered a kind of ‘ground truth’ to quantify the accuracy of the techniques we have developed,” Milstein says.

      This improved neural network approach holds promise to demonstrate drought prediction that can exceed the capabilities of existing indicators, Milstein says, and to be a tool that scientists can rely on for decades to come. More

    • 113 Shares189 Views

      in Environment

      MIT-derived algorithm helps forecast the frequency of extreme weather

      by

      To assess a community’s risk of extreme weather, policymakers rely first on global climate models that can be run decades, and even centuries, forward in time, but only at a coarse resolution. These models might be used to gauge, for instance, future climate conditions for the northeastern U.S., but not specifically for Boston.

      To estimate Boston’s future risk of extreme weather such as flooding, policymakers can combine a coarse model’s large-scale predictions with a finer-resolution model, tuned to estimate how often Boston is likely to experience damaging floods as the climate warms. But this risk analysis is only as accurate as the predictions from that first, coarser climate model.

      “If you get those wrong for large-scale environments, then you miss everything in terms of what extreme events will look like at smaller scales, such as over individual cities,” says Themistoklis Sapsis, the William I. Koch Professor and director of the Center for Ocean Engineering in MIT’s Department of Mechanical Engineering.

      Sapsis and his colleagues have now developed a method to “correct” the predictions from coarse climate models. By combining machine learning with dynamical systems theory, the team’s approach “nudges” a climate model’s simulations into more realistic patterns over large scales. When paired with smaller-scale models to predict specific weather events such as tropical cyclones or floods, the team’s approach produced more accurate predictions for how often specific locations will experience those events over the next few decades, compared to predictions made without the correction scheme.

      Play video

      This animation shows the evolution of storms around the northern hemisphere, as a result of a high-resolution storm model, combined with the MIT team’s corrected global climate model. The simulation improves the modeling of extreme values for wind, temperature, and humidity, which typically have significant errors in coarse scale models. Credit: Courtesy of Ruby Leung and Shixuan Zhang, PNNL

      Sapsis says the new correction scheme is general in form and can be applied to any global climate model. Once corrected, the models can help to determine where and how often extreme weather will strike as global temperatures rise over the coming years. 

      “Climate change will have an effect on every aspect of human life, and every type of life on the planet, from biodiversity to food security to the economy,” Sapsis says. “If we have capabilities to know accurately how extreme weather will change, especially over specific locations, it can make a lot of difference in terms of preparation and doing the right engineering to come up with solutions. This is the method that can open the way to do that.”

      The team’s results appear today in the Journal of Advances in Modeling Earth Systems. The study’s MIT co-authors include postdoc Benedikt Barthel Sorensen and Alexis-Tzianni Charalampopoulos SM ’19, PhD ’23, with Shixuan Zhang, Bryce Harrop, and Ruby Leung of the Pacific Northwest National Laboratory in Washington state.

      Over the hood

      Today’s large-scale climate models simulate weather features such as the average temperature, humidity, and precipitation around the world, on a grid-by-grid basis. Running simulations of these models takes enormous computing power, and in order to simulate how weather features will interact and evolve over periods of decades or longer, models average out features every 100 kilometers or so.

      “It’s a very heavy computation requiring supercomputers,” Sapsis notes. “But these models still do not resolve very important processes like clouds or storms, which occur over smaller scales of a kilometer or less.”

      To improve the resolution of these coarse climate models, scientists typically have gone under the hood to try and fix a model’s underlying dynamical equations, which describe how phenomena in the atmosphere and oceans should physically interact.

      “People have tried to dissect into climate model codes that have been developed over the last 20 to 30 years, which is a nightmare, because you can lose a lot of stability in your simulation,” Sapsis explains. “What we’re doing is a completely different approach, in that we’re not trying to correct the equations but instead correct the model’s output.”

      The team’s new approach takes a model’s output, or simulation, and overlays an algorithm that nudges the simulation toward something that more closely represents real-world conditions. The algorithm is based on a machine-learning scheme that takes in data, such as past information for temperature and humidity around the world, and learns associations within the data that represent fundamental dynamics among weather features. The algorithm then uses these learned associations to correct a model’s predictions.

      “What we’re doing is trying to correct dynamics, as in how an extreme weather feature, such as the windspeeds during a Hurricane Sandy event, will look like in the coarse model, versus in reality,” Sapsis says. “The method learns dynamics, and dynamics are universal. Having the correct dynamics eventually leads to correct statistics, for example, frequency of rare extreme events.”

      Climate correction

      As a first test of their new approach, the team used the machine-learning scheme to correct simulations produced by the Energy Exascale Earth System Model (E3SM), a climate model run by the U.S. Department of Energy, that simulates climate patterns around the world at a resolution of 110 kilometers. The researchers used eight years of past data for temperature, humidity, and wind speed to train their new algorithm, which learned dynamical associations between the measured weather features and the E3SM model. They then ran the climate model forward in time for about 36 years and applied the trained algorithm to the model’s simulations. They found that the corrected version produced climate patterns that more closely matched real-world observations from the last 36 years, not used for training.

      “We’re not talking about huge differences in absolute terms,” Sapsis says. “An extreme event in the uncorrected simulation might be 105 degrees Fahrenheit, versus 115 degrees with our corrections. But for humans experiencing this, that is a big difference.”

      When the team then paired the corrected coarse model with a specific, finer-resolution model of tropical cyclones, they found the approach accurately reproduced the frequency of extreme storms in specific locations around the world.

      “We now have a coarse model that can get you the right frequency of events, for the present climate. It’s much more improved,” Sapsis says. “Once we correct the dynamics, this is a relevant correction, even when you have a different average global temperature, and it can be used for understanding how forest fires, flooding events, and heat waves will look in a future climate. Our ongoing work is focusing on analyzing future climate scenarios.”

      “The results are particularly impressive as the method shows promising results on E3SM, a state-of-the-art climate model,” says Pedram Hassanzadeh, an associate professor who leads the Climate Extremes Theory and Data group at the University of Chicago and was not involved with the study. “It would be interesting to see what climate change projections this framework yields once future greenhouse-gas emission scenarios are incorporated.”

      This work was supported, in part, by the U.S. Defense Advanced Research Projects Agency. More

    • 75 Shares149 Views

      in Energy

      Optimizing nuclear fuels for next-generation reactors

      by

      In 2010, when Ericmoore Jossou was attending college in northern Nigeria, the lights would flicker in and out all day, sometimes lasting only for a couple of hours at a time. The frustrating experience reaffirmed Jossou’s realization that the country’s sporadic energy supply was a problem. It was the beginning of his path toward nuclear engineering.

      Because of the energy crisis, “I told myself I was going to find myself in a career that allows me to develop energy technologies that can easily be scaled to meet the energy needs of the world, including my own country,” says Jossou, an assistant professor in a shared position between the departments of Nuclear Science and Engineering (NSE), where is the John Clark Hardwick (1986) Professor, and of Electrical Engineering and Computer Science.

      Today, Jossou uses computer simulations for rational materials design, AI-aided purposeful development of cladding materials and fuels for next-generation nuclear reactors. As one of the shared faculty hires between the MIT Schwarzman College of Computing and departments across MIT, his appointment recognizes his commitment to computing for climate and the environment.

      A well-rounded education in Nigeria

      Growing up in Lagos, Jossou knew education was about more than just bookish knowledge, so he was eager to travel and experience other cultures. He would start in his own backyard by traveling across the Niger river and enrolling in Ahmadu Bello University in northern Nigeria. Moving from the south was a cultural education with a different language and different foods. It was here that Jossou got to try and love tuwo shinkafa, a northern Nigerian rice-based specialty, for the first time.

      After his undergraduate studies, armed with a bachelor’s degree in chemistry, Jossou was among a small cohort selected for a specialty master’s training program funded by the World Bank Institute and African Development Bank. The program at the African University of Science and Technology in Abuja, Nigeria, is a pan-African venture dedicated to nurturing homegrown science talent on the continent. Visiting professors from around the world taught intensive three-week courses, an experience which felt like drinking from a fire hose. The program widened Jossou’s views and he set his sights on a doctoral program with an emphasis on clean energy systems.

      A pivot to nuclear science

      While in Nigeria, Jossou learned of Professor Jerzy Szpunar at the University of Saskatchewan in Canada, who was looking for a student researcher to explore fuels and alloys for nuclear reactors. Before then, Jossou was lukewarm on nuclear energy, but the research sounded fascinating. The Fukushima, Japan, incident was recently in the rearview mirror and Jossou remembered his early determination to address his own country’s energy crisis. He was sold on the idea and graduated with a doctoral degree from the University of Saskatchewan on an international dean’s scholarship.

      Jossou’s postdoctoral work registered a brief stint at Brookhaven National Laboratory as staff scientist. He leaped at the opportunity to join MIT NSE as a way of realizing his research interest and teaching future engineers. “I would really like to conduct cutting-edge research in nuclear materials design and to pass on my knowledge to the next generation of scientists and engineers and there’s no better place to do that than at MIT,” Jossou says.

      Merging material science and computational modeling

      Jossou’s doctoral work on designing nuclear fuels for next-generation reactors forms the basis of research his lab is pursuing at MIT NSE. Nuclear reactors that were built in the 1950s and ’60s are getting a makeover in terms of improved accident tolerance. Reactors are not confined to one kind, either: We have micro reactors and are now considering ones using metallic nuclear fuels, Jossou points out. The diversity of options is enough to keep researchers busy testing materials fit for cladding, the lining that prevents corrosion of the fuel and release of radioactive fission products into the surrounding reactor coolant.

      The team is also investigating fuels that improve burn-up efficiencies, so they can last longer in the reactor. An intriguing approach has been to immobilize the gas bubbles that arise from the fission process, so they don’t grow and degrade the fuel.

      Since joining MIT in July 2023, Jossou is setting up a lab that optimizes the composition of accident-tolerant nuclear fuels. He is leaning on his materials science background and looping computer simulations and artificial intelligence in the mix.

      Computer simulations allow the researchers to narrow down the potential field of candidates, optimized for specific parameters, so they can synthesize only the most promising candidates in the lab. And AI’s predictive capabilities guide researchers on which materials composition to consider next. “We no longer depend on serendipity to choose our materials, our lab is based on rational materials design,” Jossou says, “we can rapidly design advanced nuclear fuels.”

      Advancing energy causes in Africa

      Now that he is at MIT, Jossou admits the view from the outside is different. He now harbors a different perspective on what Africa needs to address some of its challenges. “The starting point to solve our problems is not money; it needs to start with ideas,” he says, “we need to find highly skilled people who can actually solve problems.” That job involves adding economic value to the rich arrays of raw materials that the continent is blessed with. It frustrates Jossou that Niger, a country rich in raw material for uranium, has no nuclear reactors of its own. It ships most of its ore to France. “The path forward is to find a way to refine these materials in Africa and to be able to power the industries on that continent as well,” Jossou says.

      Jossou is determined to do his part to eliminate these roadblocks.

      Anchored in mentorship, Jossou’s solution aims to train talent from Africa in his own lab. He has applied for a MIT Global Experiences MISTI grant to facilitate travel and research studies for Ghanaian scientists. “The goal is to conduct research in our facility and perhaps add value to indigenous materials,” Jossou says.

      Adding value has been a consistent theme of Jossou’s career. He remembers wanting to become a neurosurgeon after reading “Gifted Hands,” moved by the personal story of the author, Ben Carson. As Jossou grew older, however, he realized that becoming a doctor wasn’t necessarily what he wanted. Instead, he was looking to add value. “What I wanted was really to take on a career that allows me to solve a societal problem.” The societal problem of clean and safe energy for all is precisely what Jossou is working on today. More

    • 100 Shares169 Views

      in Energy

      Cutting carbon emissions on the US power grid

      by

      To help curb climate change, the United States is working to reduce carbon emissions from all sectors of the energy economy. Much of the current effort involves electrification — switching to electric cars for transportation, electric heat pumps for home heating, and so on. But in the United States, the electric power sector already generates about a quarter of all carbon emissions. “Unless we decarbonize our electric power grids, we’ll just be shifting carbon emissions from one source to another,” says Amanda Farnsworth, a PhD candidate in chemical engineering and research assistant at the MIT Energy Initiative (MITEI).

      But decarbonizing the nation’s electric power grids will be challenging. The availability of renewable energy resources such as solar and wind varies in different regions of the country. Likewise, patterns of energy demand differ from region to region. As a result, the least-cost pathway to a decarbonized grid will differ from one region to another.

      Over the past two years, Farnsworth and Emre Gençer, a principal research scientist at MITEI, developed a power system model that would allow them to investigate the importance of regional differences — and would enable experts and laypeople alike to explore their own regions and make informed decisions about the best way to decarbonize. “With this modeling capability you can really understand regional resources and patterns of demand, and use them to do a ‘bespoke’ analysis of the least-cost approach to decarbonizing the grid in your particular region,” says Gençer.

      To demonstrate the model’s capabilities, Gençer and Farnsworth performed a series of case studies. Their analyses confirmed that strategies must be designed for specific regions and that all the costs and carbon emissions associated with manufacturing and installing solar and wind generators must be included for accurate accounting. But the analyses also yielded some unexpected insights, including a correlation between a region’s wind energy and the ease of decarbonizing, and the important role of nuclear power in decarbonizing the California grid.

      A novel model

      For many decades, researchers have been developing “capacity expansion models” to help electric utility planners tackle the problem of designing power grids that are efficient, reliable, and low-cost. More recently, many of those models also factor in the goal of reducing or eliminating carbon emissions. While those models can provide interesting insights relating to decarbonization, Gençer and Farnsworth believe they leave some gaps that need to be addressed.

      For example, most focus on conditions and needs in a single U.S. region without highlighting the unique peculiarities of their chosen area of focus. Hardly any consider the carbon emitted in fabricating and installing such “zero-carbon” technologies as wind turbines and solar panels. And finally, most of the models are challenging to use. Even experts in the field must search out and assemble various complex datasets in order to perform a study of interest.

      Gençer and Farnsworth’s capacity expansion model — called Ideal Grid, or IG — addresses those and other shortcomings. IG is built within the framework of MITEI’s Sustainable Energy System Analysis Modeling Environment (SESAME), an energy system modeling platform that Gençer and his colleagues at MITEI have been developing since 2017. SESAME models the levels of greenhouse gas emissions from multiple, interacting energy sectors in future scenarios.

      Importantly, SESAME includes both techno-economic analyses and life-cycle assessments of various electricity generation and storage technologies. It thus considers costs and emissions incurred at each stage of the life cycle (manufacture, installation, operation, and retirement) for all generators. Most capacity expansion models only account for emissions from operation of fossil fuel-powered generators. As Farnsworth notes, “While this is a good approximation for our current grid, emissions from the full life cycle of all generating technologies become non-negligible as we transition to a highly renewable grid.”

      Through its connection with SESAME, the IG model has access to data on costs and emissions associated with many technologies critical to power grid operation. To explore regional differences in the cost-optimized decarbonization strategies, the IG model also includes conditions within each region, notably details on demand profiles and resource availability.

      In one recent study, Gençer and Farnsworth selected nine of the standard North American Electric Reliability Corporation (NERC) regions. For each region, they incorporated hourly electricity demand into the IG model. Farnsworth also gathered meteorological data for the nine U.S. regions for seven years — 2007 to 2013 — and calculated hourly power output profiles for the renewable energy sources, including solar and wind, taking into account the geography-limited maximum capacity of each technology.

      The availability of wind and solar resources differs widely from region to region. To permit a quick comparison, the researchers use a measure called “annual capacity factor,” which is the ratio between the electricity produced by a generating unit in a year and the electricity that could have been produced if that unit operated continuously at full power for that year. Values for the capacity factors in the nine U.S. regions vary between 20 percent and 30 percent for solar power and for between 25 percent and 45 percent for wind.

      Calculating optimized grids for different regions

      For their first case study, Gençer and Farnsworth used the IG model to calculate cost-optimized regional grids to meet defined caps on carbon dioxide (CO2) emissions. The analyses were based on cost and emissions data for 10 technologies: nuclear, wind, solar, three types of natural gas, three types of coal, and energy storage using lithium-ion batteries. Hydroelectric was not considered in this study because there was no comprehensive study outlining potential expansion sites with their respective costs and expected power output levels.

      To make region-to-region comparisons easy, the researchers used several simplifying assumptions. Their focus was on electricity generation, so the model calculations assume the same transmission and distribution costs and efficiencies for all regions. Also, the calculations did not consider the generator fleet currently in place. The goal was to investigate what happens if each region were to start from scratch and generate an “ideal” grid.

      To begin, Gençer and Farnsworth calculated the most economic combination of technologies for each region if it limits its total carbon emissions to 100, 50, and 25 grams of CO2 per kilowatt-hour (kWh) generated. For context, the current U.S. average emissions intensity is 386 grams of CO2 emissions per kWh.

      Given the wide variation in regional demand, the researchers needed to use a new metric to normalize their results and permit a one-to-one comparison between regions. Accordingly, the model calculates the required generating capacity divided by the average demand for each region. The required capacity accounts for both the variation in demand and the inability of generating systems — particularly solar and wind — to operate at full capacity all of the time.

      The analysis was based on regional demand data for 2021 — the most recent data available. And for each region, the model calculated the cost-optimized power grid seven times, using weather data from seven years. This discussion focuses on mean values for cost and total capacity installed and also total values for coal and for natural gas, although the analysis considered three separate technologies for each fuel.

      The results of the analyses confirm that there’s a wide variation in the cost-optimized system from one region to another. Most notable is that some regions require a lot of energy storage while others don’t require any at all. The availability of wind resources turns out to play an important role, while the use of nuclear is limited: the carbon intensity of nuclear (including uranium mining and transportation) is lower than that of either solar or wind, but nuclear is the most expensive technology option, so it’s added only when necessary. Finally, the change in the CO2 emissions cap brings some interesting responses.

      Under the most lenient limit on emissions — 100 grams of CO2 per kWh — there’s no coal in the mix anywhere. It’s the first to go, in general being replaced by the lower-carbon-emitting natural gas. Texas, Central, and North Central — the regions with the most wind — don’t need energy storage, while the other six regions do. The regions with the least wind — California and the Southwest — have the highest energy storage requirements. Unlike the other regions modeled, California begins installing nuclear, even at the most lenient limit.

      As the model plays out, under the moderate cap — 50 grams of CO2 per kWh — most regions bring in nuclear power. California and the Southeast — regions with low wind capacity factors — rely on nuclear the most. In contrast, wind-rich Texas, Central, and North Central don’t incorporate nuclear yet but instead add energy storage — a less-expensive option — to their mix. There’s still a bit of natural gas everywhere, in spite of its CO2 emissions.

      Under the most restrictive cap — 25 grams of CO2 per kWh — nuclear is in the mix everywhere. The highest use of nuclear is again correlated with low wind capacity factor. Central and North Central depend on nuclear the least. All regions continue to rely on a little natural gas to keep prices from skyrocketing due to the necessary but costly nuclear component. With nuclear in the mix, the need for storage declines in most regions.

      Results of the cost analysis are also interesting. Texas, Central, and North Central all have abundant wind resources, and they can delay incorporating the costly nuclear option, so the cost of their optimized system tends to be lower than costs for the other regions. In addition, their total capacity deployment — including all sources — tends to be lower than for the other regions. California and the Southwest both rely heavily on solar, and in both regions, costs and total deployment are relatively high.

      Lessons learned

      One unexpected result is the benefit of combining solar and wind resources. The problem with relying on solar alone is obvious: “Solar energy is available only five or six hours a day, so you need to build a lot of other generating sources and abundant storage capacity,” says Gençer. But an analysis of unit-by-unit operations at an hourly resolution yielded a less-intuitive trend: While solar installations only produce power in the midday hours, wind turbines generate the most power in the nighttime hours. As a result, solar and wind power are complementary. Having both resources available is far more valuable than having either one or the other. And having both impacts the need for storage, says Gençer: “Storage really plays a role either when you’re targeting a very low carbon intensity or where your resources are mostly solar and they’re not complemented by wind.”

      Gençer notes that the target for the U.S. electricity grid is to reach net zero by 2035. But the analysis showed that reaching just 100 grams of CO2 per kWh would require at least 50 percent of system capacity to be wind and solar. “And we’re nowhere near that yet,” he says.

      Indeed, Gençer and Farnsworth’s analysis doesn’t even include a zero emissions case. Why not? As Gençer says, “We cannot reach zero.” Wind and solar are usually considered to be net zero, but that’s not true. Wind, solar, and even storage have embedded carbon emissions due to materials, manufacturing, and so on. “To go to true net zero, you’d need negative emission technologies,” explains Gençer, referring to techniques that remove carbon from the air or ocean. That observation confirms the importance of performing life-cycle assessments.

      Farnsworth voices another concern: Coal quickly disappears in all regions because natural gas is an easy substitute for coal and has lower carbon emissions. “People say they’ve decreased their carbon emissions by a lot, but most have done it by transitioning from coal to natural gas power plants,” says Farnsworth. “But with that pathway for decarbonization, you hit a wall. Once you’ve transitioned from coal to natural gas, you’ve got to do something else. You need a new strategy — a new trajectory to actually reach your decarbonization target, which most likely will involve replacing the newly installed natural gas plants.”

      Gençer makes one final point: The availability of cheap nuclear — whether fission or fusion — would completely change the picture. When the tighter caps require the use of nuclear, the cost of electricity goes up. “The impact is quite significant,” says Gençer. “When we go from 100 grams down to 25 grams of CO2 per kWh, we see a 20 percent to 30 percent increase in the cost of electricity.” If it were available, a less-expensive nuclear option would likely be included in the technology mix under more lenient caps, significantly reducing the cost of decarbonizing power grids in all regions.

      The special case of California

      In another analysis, Gençer and Farnsworth took a closer look at California. In California, about 10 percent of total demand is now met with nuclear power. Yet current power plants are scheduled for retirement very soon, and a 1976 law forbids the construction of new nuclear plants. (The state recently extended the lifetime of one nuclear plant to prevent the grid from becoming unstable.) “California is very motivated to decarbonize their grid,” says Farnsworth. “So how difficult will that be without nuclear power?”

      To find out, the researchers performed a series of analyses to investigate the challenge of decarbonizing in California with nuclear power versus without it. At 200 grams of CO2 per kWh — about a 50 percent reduction — the optimized mix and cost look the same with and without nuclear. Nuclear doesn’t appear due to its high cost. At 100 grams of CO2 per kWh — about a 75 percent reduction — nuclear does appear in the cost-optimized system, reducing the total system capacity while having little impact on the cost.

      But at 50 grams of CO2 per kWh, the ban on nuclear makes a significant difference. “Without nuclear, there’s about a 45 percent increase in total system size, which is really quite substantial,” says Farnsworth. “It’s a vastly different system, and it’s more expensive.” Indeed, the cost of electricity would increase by 7 percent.

      Going one step further, the researchers performed an analysis to determine the most decarbonized system possible in California. Without nuclear, the state could reach 40 grams of CO2 per kWh. “But when you allow for nuclear, you can get all the way down to 16 grams of CO2 per kWh,” says Farnsworth. “We found that California needs nuclear more than any other region due to its poor wind resources.”

      Impacts of a carbon tax

      One more case study examined a policy approach to incentivizing decarbonization. Instead of imposing a ceiling on carbon emissions, this strategy would tax every ton of carbon that’s emitted. Proposed taxes range from zero to $100 per ton.

      To investigate the effectiveness of different levels of carbon tax, Farnsworth and Gençer used the IG model to calculate the minimum-cost system for each region, assuming a certain cost for emitting each ton of carbon. The analyses show that a low carbon tax — just $10 per ton — significantly reduces emissions in all regions by phasing out all coal generation. In the Northwest region, for example, a carbon tax of $10 per ton decreases system emissions by 65 percent while increasing system cost by just 2.8 percent (relative to an untaxed system).

      After coal has been phased out of all regions, every increase in the carbon tax brings a slow but steady linear decrease in emissions and a linear increase in cost. But the rates of those changes vary from region to region. For example, the rate of decrease in emissions for each added tax dollar is far lower in the Central region than in the Northwest, largely due to the Central region’s already low emissions intensity without a carbon tax. Indeed, the Central region without a carbon tax has a lower emissions intensity than the Northwest region with a tax of $100 per ton.

      As Farnsworth summarizes, “A low carbon tax — just $10 per ton — is very effective in quickly incentivizing the replacement of coal with natural gas. After that, it really just incentivizes the replacement of natural gas technologies with more renewables and more energy storage.” She concludes, “If you’re looking to get rid of coal, I would recommend a carbon tax.”

      Future extensions of IG

      The researchers have already added hydroelectric to the generating options in the IG model, and they are now planning further extensions. For example, they will include additional regions for analysis, add other long-term energy storage options, and make changes that allow analyses to take into account the generating infrastructure that already exists. Also, they will use the model to examine the cost and value of interregional transmission to take advantage of the diversity of available renewable resources.

      Farnsworth emphasizes that the analyses reported here are just samples of what’s possible using the IG model. The model is a web-based tool that includes embedded data covering the whole United States, and the output from an analysis includes an easy-to-understand display of the required installations, hourly operation, and overall techno-economic analysis and life-cycle assessment results. “The user is able to go in and explore a vast number of scenarios with no data collection or pre-processing,” she says. “There’s no barrier to begin using the tool. You can just hop on and start exploring your options so you can make an informed decision about the best path forward.”

      This work was supported by the International Energy Agency Gas and Oil Technology Collaboration Program and the MIT Energy Initiative Low-Carbon Energy Centers.

      This article appears in the Winter 2024 issue of Energy Futures, the magazine of the MIT Energy Initiative. More

    • 63 Shares199 Views

      in Energy

      Moving past the Iron Age

      by

      MIT graduate student Sydney Rose Johnson has never seen the steel mills in central India. She’s never toured the American Midwest’s hulking steel plants or the mini mills dotting the Mississippi River. But in the past year, she’s become more familiar with steel production than she ever imagined.

      A fourth-year dual degree MBA and PhD candidate in chemical engineering and a graduate research assistant with the MIT Energy Initiative (MITEI) as well as a 2022-23 Shell Energy Fellow, Johnson looks at ways to reduce carbon dioxide (CO2) emissions generated by industrial processes in hard-to-abate industries. Those include steel.

      Almost every aspect of infrastructure and transportation — buildings, bridges, cars, trains, mass transit — contains steel. The manufacture of steel hasn’t changed much since the Iron Age, with some steel plants in the United States and India operating almost continually for more than a century, their massive blast furnaces re-lined periodically with carbon and graphite to keep them going.

      According to the World Economic Forum, steel demand is projected to increase 30 percent by 2050, spurred in part by population growth and economic development in China, India, Africa, and Southeast Asia.

      The steel industry is among the three biggest producers of CO2 worldwide. Every ton of steel produced in 2020 emitted, on average, 1.89 tons of CO2 into the atmosphere — around 8 percent of global CO2 emissions, according to the World Steel Association.

      A combination of technical strategies and financial investments, Johnson notes, will be needed to wrestle that 8 percent figure down to something more planet-friendly.

      Johnson’s thesis focuses on modeling and analyzing ways to decarbonize steel. Using data mined from academic and industry sources, she builds models to calculate emissions, costs, and energy consumption for plant-level production.

      “I optimize steel production pathways using emission goals, industry commitments, and cost,” she says. Based on the projected growth of India’s steel industry, she applies this approach to case studies that predict outcomes for some of the country’s thousand-plus factories, which together have a production capacity of 154 million metric tons of steel. For the United States, she looks at the effect of Inflation Reduction Act (IRA) credits. The 2022 IRA provides incentives that could accelerate the steel industry’s efforts to minimize its carbon emissions.

      Johnson compares emissions and costs across different production pathways, asking questions such as: “If we start today, what would a cost-optimal production scenario look like years from now? How would it change if we added in credits? What would have to happen to cut 2005 levels of emissions in half by 2030?”

      “My goal is to gain an understanding of how current and emerging decarbonization strategies will be integrated into the industry,” Johnson says.

      Grappling with industrial problems

      Growing up in Marietta, Georgia, outside Atlanta, the closest she ever came to a plant of any kind was through her father, a chemical engineer working in logistics and procuring steel for an aerospace company, and during high school, when she spent a semester working alongside chemical engineers tweaking the pH of an anti-foaming agent.

      At Kennesaw Mountain High School, a STEM magnet program in Cobb County, students devote an entire semester of their senior year to an internship and research project.

      Johnson chose to work at Kemira Chemicals, which develops chemical solutions for water-intensive industries with a focus on pulp and paper, water treatment, and energy systems.

      “My goal was to understand why a polymer product was falling out of suspension — essentially, why it was less stable,” she recalls. She learned how to formulate a lab-scale version of the product and conduct tests to measure its viscosity and acidity. Comparing the lab-scale and regular product results revealed that acidity was an important factor. “Through conversations with my mentor, I learned this was connected with the holding conditions, which led to the product being oxidized,” she says. With the anti-foaming agent’s problem identified, steps could be taken to fix it.

      “I learned how to apply problem-solving. I got to learn more about working in an industrial environment by connecting with the team in quality control as well as with R&D and chemical engineers at the plant site,” Johnson says. “This experience confirmed I wanted to pursue engineering in college.”

      As an undergraduate at Stanford University, she learned about the different fields — biotechnology, environmental science, electrochemistry, and energy, among others — open to chemical engineers. “It seemed like a very diverse field and application range,” she says. “I was just so intrigued by the different things I saw people doing and all these different sets of issues.”

      Turning up the heat

      At MIT, she turned her attention to how certain industries can offset their detrimental effects on climate.

      “I’m interested in the impact of technology on global communities, the environment, and policy. Energy applications affect every field. My goal as a chemical engineer is to have a broad perspective on problem-solving and to find solutions that benefit as many people, especially those under-resourced, as possible,” says Johnson, who has served on the MIT Chemical Engineering Graduate Student Advisory Board, the MIT Energy and Climate Club, and is involved with diversity and inclusion initiatives.

      The steel industry, Johnson acknowledges, is not what she first imagined when she saw herself working toward mitigating climate change.

      “But now, understanding the role the material has in infrastructure development, combined with its heavy use of coal, has illuminated how the sector, along with other hard-to-abate industries, is important in the climate change conversation,” Johnson says.

      Despite the advanced age of many steel mills, some are quite energy-efficient, she notes. Yet these operations, which produce heat upwards of 3,000 degrees Fahrenheit, are still emission-intensive.

      Steel is made from iron ore, a mixture of iron, oxygen, and other minerals found on virtually every continent, with Brazil and Australia alone exporting millions of metric tons per year. Commonly based on a process dating back to the 19th century, iron is extracted from the ore through smelting — heating the ore with blast furnaces until the metal becomes spongy and its chemical components begin to break down.

      A reducing agent is needed to release the oxygen trapped in the ore, transforming it from its raw form to pure iron. That’s where most emissions come from, Johnson notes.

      “We want to reduce emissions, and we want to make a cleaner and safer environment for everyone,” she says. “It’s not just the CO2 emissions. It’s also sometimes NOx and SOx [nitrogen oxides and sulfur oxides] and air pollution particulate matter at some of these production facilities that can affect people as well.”

      In 2020, the International Energy Agency released a roadmap exploring potential technologies and strategies that would make the iron and steel sector more compatible with the agency’s vision of increased sustainability. Emission reductions can be accomplished with more modern technology, the agency suggests, or by substituting the fuels producing the immense heat needed to process ore. Traditionally, the fuels used for iron reduction have been coal and natural gas. Alternative fuels include clean hydrogen, electricity, and biomass.

      Using the MITEI Sustainable Energy System Analysis Modeling Environment (SESAME), Johnson analyzes various decarbonization strategies. She considers options such as switching fuel for furnaces to hydrogen with a little bit of natural gas or adding carbon-capture devices. The models demonstrate how effective these tactics are likely to be. The answers aren’t always encouraging.

      “Upstream emissions can determine how effective the strategies are,” Johnson says. Charcoal derived from forestry biomass seemed to be a promising alternative fuel, but her models showed that processing the charcoal for use in the blast furnace limited its effectiveness in negating emissions.

      Despite the challenges, “there are definitely ways of moving forward,” Johnson says. “It’s been an intriguing journey in terms of understanding where the industry is at. There’s still a long way to go, but it’s doable.”

      Johnson is heartened by the steel industry’s efforts to recycle scrap into new steel products and incorporate more emission-friendly technologies and practices, some of which result in significantly lower CO2 emissions than conventional production.

      A major issue is that low-carbon steel can be more than 50 percent more costly than conventionally produced steel. “There are costs associated with making the transition, but in the context of the environmental implications, I think it’s well worth it to adopt these technologies,” she says.

      After graduation, Johnson plans to continue to work in the energy field. “I definitely want to use a combination of engineering knowledge and business knowledge to work toward mitigating climate change, potentially in the startup space with clean technology or even in a policy context,” she says. “I’m interested in connecting the private and public sectors to implement measures for improving our environment and benefiting as many people as possible.” More

    • 75 Shares169 Views

      in Energy

      Generative AI for smart grid modeling

      by

      MIT’s Laboratory for Information and Decision Systems (LIDS) has been awarded $1,365,000 in funding from the Appalachian Regional Commission (ARC) to support its involvement with an innovative project, “Forming the Smart Grid Deployment Consortium (SGDC) and Expanding the HILLTOP+ Platform.”

      The grant was made available through ARC’s Appalachian Regional Initiative for Stronger Economies, which fosters regional economic transformation through multi-state collaboration.

      Led by Kalyan Veeramachaneni, research scientist and principal investigator at LIDS’ Data to AI Group, the project will focus on creating AI-driven generative models for customer load data. Veeramachaneni and colleagues will work alongside a team of universities and organizations led by Tennessee Tech University, including collaborators across Ohio, Pennsylvania, West Virginia, and Tennessee, to develop and deploy smart grid modeling services through the SGDC project.

      These generative models have far-reaching applications, including grid modeling and training algorithms for energy tech startups. When the models are trained on existing data, they create additional, realistic data that can augment limited datasets or stand in for sensitive ones. Stakeholders can then use these models to understand and plan for specific what-if scenarios far beyond what could be achieved with existing data alone. For example, generated data can predict the potential load on the grid if an additional 1,000 households were to adopt solar technologies, how that load might change throughout the day, and similar contingencies vital to future planning.

      The generative AI models developed by Veeramachaneni and his team will provide inputs to modeling services based on the HILLTOP+ microgrid simulation platform, originally prototyped by MIT Lincoln Laboratory. HILLTOP+ will be used to model and test new smart grid technologies in a virtual “safe space,” providing rural electric utilities with increased confidence in deploying smart grid technologies, including utility-scale battery storage. Energy tech startups will also benefit from HILLTOP+ grid modeling services, enabling them to develop and virtually test their smart grid hardware and software products for scalability and interoperability.

      The project aims to assist rural electric utilities and energy tech startups in mitigating the risks associated with deploying these new technologies. “This project is a powerful example of how generative AI can transform a sector — in this case, the energy sector,” says Veeramachaneni. “In order to be useful, generative AI technologies and their development have to be closely integrated with domain expertise. I am thrilled to be collaborating with experts in grid modeling, and working alongside them to integrate the latest and greatest from my research group and push the boundaries of these technologies.”

      “This project is testament to the power of collaboration and innovation, and we look forward to working with our collaborators to drive positive change in the energy sector,” says Satish Mahajan, principal investigator for the project at Tennessee Tech and a professor of electrical and computer engineering. Tennessee Tech’s Center for Rural Innovation director, Michael Aikens, adds, “Together, we are taking significant steps towards a more sustainable and resilient future for the Appalachian region.” More

    • 100 Shares129 Views

      in Environment

      MIT researchers remotely map crops, field by field

      by

      Crop maps help scientists and policymakers track global food supplies and estimate how they might shift with climate change and growing populations. But getting accurate maps of the types of crops that are grown from farm to farm often requires on-the-ground surveys that only a handful of countries have the resources to maintain.

      Now, MIT engineers have developed a method to quickly and accurately label and map crop types without requiring in-person assessments of every single farm. The team’s method uses a combination of Google Street View images, machine learning, and satellite data to automatically determine the crops grown throughout a region, from one fraction of an acre to the next. 

      The researchers used the technique to automatically generate the first nationwide crop map of Thailand — a smallholder country where small, independent farms make up the predominant form of agriculture. The team created a border-to-border map of Thailand’s four major crops — rice, cassava, sugarcane, and maize — and determined which of the four types was grown, at every 10 meters, and without gaps, across the entire country. The resulting map achieved an accuracy of 93 percent, which the researchers say is comparable to on-the-ground mapping efforts in high-income, big-farm countries.

      The team is applying their mapping technique to other countries such as India, where small farms sustain most of the population but the type of crops grown from farm to farm has historically been poorly recorded.

      “It’s a longstanding gap in knowledge about what is grown around the world,” says Sherrie Wang, the d’Arbeloff Career Development Assistant Professor in MIT’s Department of Mechanical Engineering, and the Institute for Data, Systems, and Society (IDSS). “The final goal is to understand agricultural outcomes like yield, and how to farm more sustainably. One of the key preliminary steps is to map what is even being grown — the more granularly you can map, the more questions you can answer.”

      Wang, along with MIT graduate student Jordi Laguarta Soler and Thomas Friedel of the agtech company PEAT GmbH, will present a paper detailing their mapping method later this month at the AAAI Conference on Artificial Intelligence.

      Ground truth

      Smallholder farms are often run by a single family or farmer, who subsist on the crops and livestock that they raise. It’s estimated that smallholder farms support two-thirds of the world’s rural population and produce 80 percent of the world’s food. Keeping tabs on what is grown and where is essential to tracking and forecasting food supplies around the world. But the majority of these small farms are in low to middle-income countries, where few resources are devoted to keeping track of individual farms’ crop types and yields.

      Crop mapping efforts are mainly carried out in high-income regions such as the United States and Europe, where government agricultural agencies oversee crop surveys and send assessors to farms to label crops from field to field. These “ground truth” labels are then fed into machine-learning models that make connections between the ground labels of actual crops and satellite signals of the same fields. They then label and map wider swaths of farmland that assessors don’t cover but that satellites automatically do.

      “What’s lacking in low- and middle-income countries is this ground label that we can associate with satellite signals,” Laguarta Soler says. “Getting these ground truths to train a model in the first place has been limited in most of the world.”

      The team realized that, while many developing countries do not have the resources to maintain crop surveys, they could potentially use another source of ground data: roadside imagery, captured by services such as Google Street View and Mapillary, which send cars throughout a region to take continuous 360-degree images with dashcams and rooftop cameras.

      In recent years, such services have been able to access low- and middle-income countries. While the goal of these services is not specifically to capture images of crops, the MIT team saw that they could search the roadside images to identify crops.

      Cropped image

      In their new study, the researchers worked with Google Street View (GSV) images taken throughout Thailand — a country that the service has recently imaged fairly thoroughly, and which consists predominantly of smallholder farms.

      Starting with over 200,000 GSV images randomly sampled across Thailand, the team filtered out images that depicted buildings, trees, and general vegetation. About 81,000 images were crop-related. They set aside 2,000 of these, which they sent to an agronomist, who determined and labeled each crop type by eye. They then trained a convolutional neural network to automatically generate crop labels for the other 79,000 images, using various training methods, including iNaturalist — a web-based crowdsourced  biodiversity database, and GPT-4V, a “multimodal large language model” that enables a user to input an image and ask the model to identify what the image is depicting. For each of the 81,000 images, the model generated a label of one of four crops that the image was likely depicting — rice, maize, sugarcane, or cassava.

      The researchers then paired each labeled image with the corresponding satellite data taken of the same location throughout a single growing season. These satellite data include measurements across multiple wavelengths, such as a location’s greenness and its reflectivity (which can be a sign of water). 

      “Each type of crop has a certain signature across these different bands, which changes throughout a growing season,” Laguarta Soler notes.

      The team trained a second model to make associations between a location’s satellite data and its corresponding crop label. They then used this model to process satellite data taken of the rest of the country, where crop labels were not generated or available. From the associations that the model learned, it then assigned crop labels across Thailand, generating a country-wide map of crop types, at a resolution of 10 square meters.

      This first-of-its-kind crop map included locations corresponding to the 2,000 GSV images that the researchers originally set aside, that were labeled by arborists. These human-labeled images were used to validate the map’s labels, and when the team looked to see whether the map’s labels matched the expert, “gold standard” labels, it did so 93 percent of the time.

      “In the U.S., we’re also looking at over 90 percent accuracy, whereas with previous work in India, we’ve only seen 75 percent because ground labels are limited,” Wang says. “Now we can create these labels in a cheap and automated way.”

      The researchers are moving to map crops across India, where roadside images via Google Street View and other services have recently become available.

      “There are over 150 million smallholder farmers in India,” Wang says. “India is covered in agriculture, almost wall-to-wall farms, but very small farms, and historically it’s been very difficult to create maps of India because there are very sparse ground labels.”

      The team is working to generate crop maps in India, which could be used to inform policies having to do with assessing and bolstering yields, as global temperatures and populations rise.

      “What would be interesting would be to create these maps over time,” Wang says. “Then you could start to see trends, and we can try to relate those things to anything like changes in climate and policies.” More