Using lidar to assess destruction in Puerto Rico

In the wake of a disaster, responding agencies need to assess damage quickly in order to figure out where to focus their efforts and where debris might block rescue crews.

Adam Norige, associate leader of MIT Lincoln Laboratory's Humanitarian Assistance and Disaster Relief Systems Group, says currently these assessments are conducted by “literally driving around or flying a small aircraft and taking digital camera pictures to document the damage,” 

But manually monitoring debris is a slow process. So Lincoln Laboratory is undertaking a multidivisional effort to revolutionize this task in disaster response.

“We are trying to show, from a research and development perspective, that we can automate the debris quantification process with Lincoln Laboratory optics and specialized algorithms,” Norige says. By automating the analysis of debris data, laboratory researchers hope to better prepare the Federal Emergency Management Agency (FEMA) for future disasters and reduce the time and cost of planning tasks to repair or clear damages.

A team in Puerto Rico has used the Airborne Optical Systems Testbed (AOSTB) to develop a baseline lidar map of the entire island, showing the latest topographical conditions and debris resulting from Hurricane Irma and Hurricane Maria in 2017. If another hurricane hits the island in the future, FEMA can track the damage that occurs by comparing subsequent lidar scans to the baseline data. 

The AOSTB utilizes single-photon-sensitive, time-of-flight imaging technology to collect information about the surface characteristics of the land below. This advanced lidar system, developed by the Active Optical Systems Group, is 10 to 100 times more capable than any commercial system available and can collect wide-area, high-resolution, 3-D datasets very rapidly.

Jalal Khan, leader of the Active Optical Systems Group, says: “Data is great, but what people really want are answers to specific questions … Where can I drive? Where can I position relief supplies? Where can I pitch tents? Where are there downed power lines? Where can I land a helicopter?” 

The maps generated by the AOSTB will help FEMA personnel assess damages, quantify debris, inspect infrastructure, and monitor erosion and reconstruction.

Since the first mission was flown on May 31, laboratory staff, assisted by engineers from 3DEO (a small business located in Massachusetts), have now mapped the entire island of Puerto Rico and the Puerto Rican islands of Vieques and Culebra, crisscrossing over the land during nightly sorties on a BT-67 (a remanufactured and modified DC-3) aircraft. They plan to complete two additional flight campaigns over the next nine months, or as the hurricane season demands.

“One of our goals in this effort was to increase our daily lidar area collection rate,” says John Aldridge, the assistant leader of the Humanitarian Assistance and Disaster Relief Systems Group. “This aircraft is a key enabler, in that it is a long-endurance aircraft capable of eight-hour missions with plenty of space on board for flight crew and support equipment.”

Back on the ground, Anthony Lapadula and Matthew Daggett of the Humanitarian Assistance and Disaster Relief Systems Group led the data management and analysis efforts, while Luke Skelly and Alexandru Vasile of the Active Optical Systems Group led the development of advanced data exploitation algorithms.

Khan says the foundational data collected in Puerto Rico “will form the basis of all analysis and will also be used in the development of automated algorithms that will find points of interests — buildings, roads, powerlines.”

“The map will also form the baseline data against which to compare future maps,” Khan says. “If another hurricane hits, we will be able to see the damage. That is really powerful.”

Since completing the first flight campaign, the team is now working with FEMA to understand how the lidar data can best be translated for operational uses. The FEMA Transportation Sector can immediately use the data to identify sections of roads that were damaged or washed away by Hurricane Maria. Data such as those used to model flood plains are valuable for planning new infrastructure.

Over the next few months, three staff members will be stationed at the FEMA Joint Recovery Office in Puerto Rico to help facilitate this work.

"There are so many questions that can be asked and answered with the data and we are only just now getting started," Aldridge says.

Lincoln Laboratory previously used the AOSTB to assess damage in Houston, Texas, after Hurricane Harvey.

Eventually, Lincoln Laboratory staff plan to add additional sensors to the test bed and make maps of other areas throughout the United States that are susceptible to disasters.

“We hope [the AOSTB] opens the door to all the other sensors that could be used on this platform,” Norige says. “For example, if you used infrared sensors after a disaster, you could find people who are stuck on rooftops or under rubble. There are quite a few modalities we envision coming together on an aircraft.”

de MIT News https://ift.tt/2NABfrI

Q&A: Luisa Kenausis ’17 — A passion for policy

While still an undergraduate at MIT, Luisa Kenausis ’17 co-founded MIT Students for Nuclear Arms Control. The organization’s goal: to raise awareness of nuclear arms control issues. As a Herbert Scoville Jr. Peace Fellowship at the Center for Arms Control and Non-Proliferation this spring, Kenausis continued her work to raise public awareness of these issues.

Kenausis grew up in Bethel, Connecticut, a small town a couple of hours from New York City, and attended the town’s public high school, Bethel High School, where she played the saxophone in the school’s award-winning marching band and developed a love for mathematics. “I ran out of math classes to take when I was a sophomore,” Kenausis recalls, “so during my junior year I started taking math classes at a community college.” At MIT, she double-majored in political science and nuclear science and engineering, working with professors Scott Kemp and Vipin Narang. Her senior thesis, “North Korea’s Nuclear Weapons: Understanding the Nuclear Tests and the Current Trajectory of the Weapons Program,” explored both technical and political aspects of North Korea’s nuclear weapons program. Kenausis has just accepted a job at the Stanley Foundation in Iowa where she will continue work in nuclear policy.

Kenausis spoke with the Department of Nuclear Science and Engineering about her fellowship experience and next steps in her career.

Q: What did you hope to achieve this spring though the Herbert Scoville Jr. Peace Fellowship?

A: I had a couple of goals. I wanted to meet and connect with people in the field — the fellowship has really helped a lot with that. We had small meetings of the Scoville Fellows with leaders who were expert in a particular field. We met senior level staff from non-governmental organizations and former government officials though these sessions. Those were super helpful, not only because we got to speak with that person but for practicing how to speak with someone in that position and thinking about how to formulate questions around issues. My other goal was to expand my issue areas, expose myself to new questions, and new areas of research that I might not have been aware of when I was a student. As a result I have become interested in defense spending policy — the way that money is authorized and allocated to the military, for instance. It’s an area that is really fascinating but not really well understood.

Q: What was the most rewarding part of this fellowship experience?

A: A couple of individual experiences stand out. We prepared questions to send to the congressional offices before Secretary of State Mike Pompeo testified in front of the Senate Foreign Relations Committee. It was just super exciting to be working on questions that might be asked in a congressional hearing in a directly meaningful way. Another was a little unexpected and rewarding — I was able to contribute and participate as part of a team straight out of college. Being a Scoville Fellow gave me credibility with people I met for the first time and opened doors to experiences I would never have had otherwise.

Q: What’s your favorite memory from MIT?

A: It’s from my senior year. My group in the senior nuclear design class, taught by Professor Mike Short, won the final competition for the best project. As the winners we got a trip to Singapore to participate in a hackathon. We not only competed in the hackathon but also got to explore the city. It was just such an amazing and fun experience. We had an awesome time.

Q: How did you become interested in nuclear policy?

A: I started out in nuclear science, but I was not really in love with just the technical side. I thought it was interesting and challenging, and I’ve always loved to be challenged and it was one of the things I really liked about NSE, but it didn’t really ignite my passion. Then when I took my first class on nuclear weapons proliferation, with nuclear weapons historian Professor Frank Gavin, during my junior fall semester. That’s when I felt “wait, this is why I think nuclear science is so important”, this is so insane and such a huge issue, and it just seems like people aren’t thinking about it. So after I took that one class I added political science as my second major and built up my own little nuclear weapons-focused curriculum. I got into nuclear policy because I just didn’t understand why people weren’t really concerned about this.

Q: What do you do for fun outside of research and work?

A: I like to work out, and I really like dancing, and I really love cooking. In the last couple of years, I’ve really gotten into cooking. Now I spend a lot of my weekends just cooking for the week. It’s really fun.

de MIT News https://ift.tt/2Ny1i2V

MIT researchers 3-D print colloidal crystals

MIT engineers have united the principles of self-assembly and 3-D printing using a new technique, which they highlight today in the journal Advanced Materials.

By their direct-write colloidal assembly process, the researchers can build centimeter-high crystals, each made from billions of individual colloids, defined as particles that are between 1 nanometer and 1 micrometer across.

“If you blew up each particle to the size of a soccer ball, it would be like stacking a whole lot of soccer balls to make something as tall as a skyscraper,” says study co-author Alvin Tan, a graduate student in MIT’s Department of Materials Science and Engineering. “That’s what we’re doing at the nanoscale.”

The researchers found a way to print colloids such as polymer nanoparticles in highly ordered arrangements, similar to the atomic structures in crystals. They printed various structures, such as tiny towers and helices, that interact with light in specific ways depending on the size of the individual particles within each structure.

Nanoparticles dispensed from a needle onto a rotating stage, creating a helical crystal containing billions of nanoparticles. (Credit: Alvin Tan)

The team sees the 3-D printing technique as a new way to build self-asssembled materials that leverage the novel properties of nanocrystals, at larger scales, such as optical sensors, color displays, and light-guided electronics.

“If you could 3-D print a circuit that manipulates photons instead of electrons, that could pave the way for future applications in light-based computing, that manipulate light instead of electricity so that devices can be faster and more energy efficient,” Tan says.

Tan’s co-authors are graduate student Justin Beroz, assistant professor of mechanical engineering Mathias Kolle, and associate professor of mechanical engineering A. John Hart.

Out of the fog

Colloids are any large molecules or small particles, typically measuring between 1 nanometer and 1 micrometer in diameter, that are suspended in a liquid or gas. Common examples of colloids are fog, which is made up of soot and other ultrafine particles dispersed in air, and whipped cream, which is a suspension of air bubbles in heavy cream. The particles in these everyday colloids are completely random in their size and the ways in which they are dispersed through the solution.

If uniformly sized colloidal particles are driven together via evaporation of their liquid solvent, causing them to assemble into ordered crystals, it is possible to create structures that, as a whole, exhibit unique optical, chemical, and mechanical properties. These crystals can exhibit properties similar to interesting structures in nature, such as the iridescent cells in butterfly wings, and the microscopic, skeletal fibers in sea sponges.

So far, scientists have developed techniques to evaporate and assemble colloidal particles into thin films to form displays that filter light and create colors based on the size and arrangement of the individual particles. But until now, such colloidal assemblies have been limited to thin films and other planar structures.

“For the first time, we’ve shown that it’s possible to build macroscale self-assembled colloidal materials, and we expect this technique can build any 3-D shape, and be applied to an incredible variety of materials,” says Hart, the senior author of the paper.

Building a particle bridge

The researchers created tiny three-dimensional towers of colloidal particles using a custom-built 3-D-printing apparatus consisting of a glass syringe and needle, mounted above two heated aluminum plates. The needle passes through a hole in the top plate and dispenses a colloid solution onto a substrate attached to the bottom plate.

The team evenly heats both aluminum plates so that as the needle dispenses the colloid solution, the liquid slowly evaporates, leaving only the particles. The bottom plate can be rotated and moved up and down to manipulate the shape of the overall structure, similar to how you might move a bowl under a soft ice cream dispenser to create twists or swirls.

Beroz says that as the colloid solution is pushed through the needle, the liquid acts as a bridge, or mold, for the particles in the solution. The particles “rain down” through the liquid, forming a structure in the shape of the liquid stream. After the liquid evaporates, surface tension between the particles holds them in place, in an ordered configuration.

As a first demonstration of their colloid printing technique, the team worked with solutions of polystyrene particles in water, and created centimeter-high towers and helices. Each of these structures contains 3 billion particles. In subsequent trials, they tested solutions containing different sizes of polystyrene particles and were able to print towers that reflected specific colors, depending on the individual particles’ size.

“By changing the size of these particles, you drastically change the color of the structure,” Beroz says. “It’s due to the way the particles are assembled, in this periodic, ordered way, and the interference of light as it interacts with particles at this scale. We’re essentially 3-D-printing crystals.”

The team also experimented with more exotic colloidal particles, namely silica and gold nanoparticles, which can exhibit unique optical and electronic properties. They printed millimeter-tall towers made from 200-nanometer diameter silica nanoparticles, and 80-nanometer gold nanoparticles, each of which reflected light in different ways.

“There are a lot of things you can do with different kinds of particles ranging from conductive metal particles to semiconducting quantum dots, which we are looking into,” Tan says. “Combining them into different crystal structures and forming them into different geometries for novel device architectures, I think that would be very effective in fields including sensing, energy storage, and photonics.”

This work was supported, in part, by the National Science Foundation, the Singapore Defense Science Organization Postgraduate Fellowship, and the National Defense Science and Engineering Graduate Fellowship Program.

de MIT News https://ift.tt/2Nu6wg1

Model can more naturally detect depression in conversations

To diagnose depression, clinicians interview patients, asking specific questions — about, say, past mental illnesses, lifestyle, and mood — and identify the condition based on the patient’s responses.

In recent years, machine learning has been championed as a useful aid for diagnostics. Machine-learning models, for instance, have been developed that can detect words and intonations of speech that may indicate depression. But these models tend to predict that a person is depressed or not, based on the person’s specific answers to specific questions. These methods are accurate, but their reliance on the type of question being asked limits how and where they can be used.

In a paper being presented at the Interspeech conference, MIT researchers detail a neural-network model that can be unleashed on raw text and audio data from interviews to discover speech patterns indicative of depression. Given a new subject, it can accurately predict if the individual is depressed, without needing any other information about the questions and answers.

The researchers hope this method can be used to develop tools to detect signs of depression in natural conversation. In the future, the model could, for instance, power mobile apps that monitor a user’s text and voice for mental distress and send alerts. This could be especially useful for those who can’t get to a clinician for an initial diagnosis, due to distance, cost, or a lack of awareness that something may be wrong.

“The first hints we have that a person is happy, excited, sad, or has some serious cognitive condition, such as depression, is through their speech,” says first author Tuka Alhanai, a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “If you want to deploy [depression-detection] models in scalable way … you want to minimize the amount of constraints you have on the data you’re using. You want to deploy it in any regular conversation and have the model pick up, from the natural interaction, the state of the individual.”

The technology could still, of course, be used for identifying mental distress in casual conversations in clinical offices, adds co-author James Glass, a senior research scientist in CSAIL. “Every patient will talk differently, and if the model sees changes maybe it will be a flag to the doctors,” he says. “This is a step forward in seeing if we can do something assistive to help clinicians.”

The other co-author on the paper is Mohammad Ghassemi, a member of the Institute for Medical Engineering and Science (IMES).

Context-free modeling

The key innovation of the model lies in its ability to detect patterns indicative of depression, and then map those patterns to new individuals, with no additional information. “We call it ‘context-free,’ because you’re not putting any constraints into the types of questions you’re looking for and the type of responses to those questions,” Alhanai says.

Other models are provided with a specific set of questions, and then given examples of how a person without depression responds and examples of how a person with depression responds — for example, the straightforward inquiry, “Do you have a history of depression?” It uses those exact responses to then determine if a new individual is depressed when asked the exact same question. “But that’s not how natural conversations work,” Alhanai says.   

The researchers, on the other hand, used a technique called sequence modeling, often used for speech processing. With this technique, they fed the model sequences of text and audio data from questions and answers, from both depressed and non-depressed individuals, one by one. As the sequences accumulated, the model extracted speech patterns that emerged for people with or without depression. Words such as, say, “sad,” “low,” or “down,” may be paired with audio signals that are flatter and more monotone. Individuals with depression may also speak slower and use longer pauses between words. These text and audio identifiers for mental distress have been explored in previous research. It was ultimately up to the model to determine if any patterns were predictive of depression or not.

“The model sees sequences of words or speaking style, and determines that these patterns are more likely to be seen in people who are depressed or not depressed,” Alhanai says. “Then, if it sees the same sequences in new subjects, it can predict if they’re depressed too.”

This sequencing technique also helps the model look at the conversation as a whole and note differences between how people with and without depression speak over time.

Detecting depression

The researchers trained and tested their model on a dataset of 142 interactions from the Distress Analysis Interview Corpus that contains audio, text, and video interviews of patients with mental-health issues and virtual agents controlled by humans. Each subject is rated in terms of depression on a scale between 0 to 27, using the Personal Health Questionnaire. Scores above a cutoff between moderate (10 to 14) and moderately severe (15 to 19) are considered depressed, while all others below that threshold are considered not depressed. Out of all the subjects in the dataset, 28 (20 percent) are labeled as depressed.

In experiments, the model was evaluated using metrics of precision and recall. Precision measures which of the depressed subjects identified by the model were diagnosed as depressed. Recall measures the accuracy of the model in detecting all subjects who were diagnosed as depressed in the entire dataset. In precision, the model scored 71 percent and, on recall, scored 83 percent. The averaged combined score for those metrics, considering any errors, was 77 percent. In the majority of tests, the researchers’ model outperformed nearly all other models.

One key insight from the research, Alhanai notes, is that, during experiments, the model needed much more data to predict depression from audio than text. With text, the model can accurately detect depression using an average of seven question-answer sequences. With audio, the model needed around 30 sequences. “That implies that the patterns in words people use that are predictive of depression happen in shorter time span in text than in audio,” Alhanai says. Such insights could help the MIT researchers, and others, further refine their models.

This work represents a “very encouraging” pilot, Glass says. But now the researchers seek to discover what specific patterns the model identifies across scores of raw data. “Right now it’s a bit of a black box,” Glass says. “These systems, however, are more believable when you have an explanation of what they’re picking up. … The next challenge is finding out what data it’s seized upon.”

The researchers also aim to test these methods on additional data from many more subjects with other cognitive conditions, such as dementia. “It’s not so much detecting depression, but it’s a similar concept of evaluating, from an everyday signal in speech, if someone has cognitive impairment or not,” Alhanai says.

de MIT News https://ift.tt/2wqp760

Model improves prediction of mortality risk in ICU patients

In intensive care units, where patients come in with a wide range of health conditions, triaging relies heavily on clinical judgment. ICU staff run numerous physiological tests, such as bloodwork and checking vital signs, to determine if patients are at immediate risk of dying if not treated aggressively.

Enter: machine learning. Numerous models have been developed in recent years to help predict patient mortality in the ICU, based on various health factors during their stay. These models, however, have performance drawbacks. One common type of “global” model is trained on a single large patient population. These might work well on average, but poorly on some patient subpopulations. On the other hand, another type of model analyzes different subpopulations — for instance, those grouped by similar conditions, patient ages, or hospital departments — but often have limited data for training and testing.

In a paper recently presented at the Proceedings of Knowledge Discovery and Data Mining conference, MIT researchers describe a machine-learning model that functions as the best of both worlds: It trains specifically on patient subpopulations, but also shares data across all subpopulations to get better predictions. In doing so, the model can better predict a patient’s risk of mortality during their first two days in the ICU, compared to strictly global and other models.

The model first crunches physiological data in electronic health records of previously admitted ICU patients, some who had died during their stay. In doing so, it learns high predictors of mortality, such as low heart rate, high blood pressure, and various lab test results — high glucose levels and white blood cell count, among others — over the first few days and breaks the patients into subpopulations based on their health status. Given a new patient, the model can look at that patient’s physiological data from the first 24 hours and, using what it’s learned through analyzing those patient subpopulations, better estimate the likelihood that the new patient will also die in the following 48 hours.

Moreover, the researchers found that evaluating (testing and validating) the model by specific subpopulations also highlights performance disparities of global models in predicting mortality across patient subpopulations. This is important information for developing models that can more accurately work with specific patients.

“ICUs are very high-bandwidth, with a lot of patients,” says first author Harini Suresh, a graduate student in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “It’s important to figure out well ahead of time which patients are actually at risk and in more need of immediate attention.”

Co-authors on the paper are CSAIL graduate student Jen Gong, and John Guttag, the Dugald C. Jackson Professor in Electrical Engineering.

Multitasking and patient subpopulations

A key innovation of the work is that, during training, the model separates patients into distinct subpopulations, which captures aspects of a patient’s overall state of health and mortality risks. It does so by calculating a combination of physiological data, broken down by the hour. Physiological data include, for example, levels of glucose, potassium, and nitrogen, as well as heart rate, blood pH, oxygen saturation, and respiratory rate. Increases in blood pressure and potassium levels — a sign of a heart failure — may indicate health problems over other subpopulations.

Next, the model employs a multitasking method of learning to build predictive models. When the patients are broken into subpopulations, differently tuned models are assigned to each subpopulation. Each variant model can then more accurately make predictions for its personalized group of patients. This approach also allows the model to share data across all subpopulations when it’s making predictions. When given a new patient, it will match the patient’s physiological data to all subpopulations, find the best fit, and then better estimate the mortality risk from there.

“We’re using all the patient data and sharing information across populations where it’s relevant,” Suresh says. “In this way, we’re able to … not suffer from data scarcity problems, while taking into account the differences between the different patient subpopulations.”

“Patients admitted to the ICU often differ in why they’re there and what their health status is like. Because of this, they’ll be treated very differently,” Gong adds. Clinical decision-making aids “should account for the heterogeneity of these patient populations … and make sure there is enough data for accurate predictions.”

A key insight from this method, Gong says, came from using a multitasking approach to also evaluate a model’s performance on specific subpopulations. Global models are often evaluated in overall performance, across entire patient populations. But the researchers’ experiments showed these models actually underperform on subpopulations. The global model tested in the paper predicted mortality fairly accurately overall, but dropped several percentage points in accuracy when tested on individual subpopulations.

Such performance disparities are difficult to measure without evaluating by subpopulations, Gong says: “We want to evaluate how well our model does, not just on a whole cohort of patients, but also when we break it down for each cohort with different medical characteristics. That can help researchers in better predictive model training and evaluation.”

Getting results

The researchers tested their model using data from the MIMIC Critical Care Database, which contains scores of data on heterogeneous patient populations. Of around 32,000 patients in the dataset, more than 2,200 died in the hospital. They used 80 percent of the dataset to train, and 20 percent to test the model.

In using data from the first 24 hours, the model clustered the patients into subpopulations with important clinical differences. Two subpopulations, for instance, contained patients with elevated blood pressure over the first several hours — but one decreased over time, while the other maintained the elevation throughout the day. This subpopulation had the highest mortality rate.

Using those subpopulations, the model predicted the mortality of the patients over the following 48 hours with high specificity and sensitivity, and various other metrics. The multitasking model significantly outperformed a global model by several percentage points.

Next, the researchers aim to use more data from electronic health records, such as treatments the patients are receiving. They also hope, in the future, to train the model to extract keywords from digitized clinical notes and other information.

The work was supported by the National Institutes of Health.

de MIT News https://ift.tt/2BYIJDv

MindHandHeart plus art at Paint Nite

The skies were varying shades of blue, and the trees were drawn in different shapes and colors. Some of the paintings were serene, while others were abstract, but all of them reflected the creativity of the MIT community.

The Office of the Chancellor and MindHandHeart sponsored a Paint Nite for MIT graduate students on Aug. 22 at The Thirsty Ear Pub. Chancellor Cynthia Barnhart and MindHandHeart Executive Administrator Maryanne Kirkbride provided opening remarks, informing students of their offices’ programs and services, such as the Accessing Resources Coalition, the MIT Student Support Hub, and the MindHandHeart Innovation Fund.

While the students enjoyed hors d'oeuvres and refreshments, a local art instructor led them in painting a tree against a sky background. The sold-out event was organized by the Graduate Student Council Activities Committee, and spearheaded by graduate students Shaiyan Keshvari, Mukund Gupta, and Xueying Zhao.

The student organizers were motivated to establish Paint Nites to provide their peers with an opportunity to unwind in a community setting.

“We love the space at The Thirsty and want to use it as much as possible,” Keshvari said. “I thought about what activities we could do that are fun, easy to run, and absolutely stress-free for participants. I’d heard about Paint Nites from a friend, and I thought we should try it out here.”

The first in a series of Paint Nites, the event appeared to be a success. “Some people are absorbed in their painting, some people are socializing, and it seems like everyone is having a good time, which was the goal,” said Keshvari.

Zhao added: “I think most MIT grad students are scientists and engineers, and this will help them to realize that they can also be artists. Everyone’s life needs some art and painting can be de-stressing. It’s not a competition and no one is judging your work — it’s just enjoyable.”

Graduate student Nicole Moody said of her painting: “It has sort-of evolved into a willow tree. When I was making the leaves some of the paint dripped down, so it looks like a fall scene. I’m here with four of my friends and we plan to hang our paintings along the staircase of our residence, so we can see our nice trees and remember this fun activity.”

Upcoming Paint Nites will be sponsored by the Office of the Vice Chancellor, the Office of Graduate Education, Community Wellness at MIT Medical, the International Students Office, and the Atlas Service Center.

de MIT News https://ift.tt/2PKWV62

Air pollution can put a dent in solar power

Ian Marius Peters, now an MIT research scientist, was working on solar energy research in Singapore in 2013 when he encountered an extraordinary cloud of pollution. The city was suddenly engulfed in a foul-smelling cloud of haze so thick that from one side of a street you couldn’t see the buildings on the other side, and the air had the acrid smell of burning. The event, triggered by forest fires in Indonesia and concentrated by unusual wind patterns, lasted two weeks, quickly causing stores to run out of face masks as citizens snapped them up to aid their breathing.

While others were addressing the public health issues of the thick air pollution, Peters’ co-worker Andre Nobre from Cleantech Energy Corp., whose field is also solar energy, wondered about what impact such hazes might have on the output of solar panels in the area. That led to a years-long project to try to quantify just how urban-based solar installations are affected by hazes, which tend to be concentrated in dense cities.

Now, the results of that research have just been published in the journal Energy & Environmental Science, and the findings show that these effects are indeed substantial. In some cases it can mean the difference between a successful solar power installation and one that ends up failing to meet expected production levels — and possibly operates at a loss.

After initially collecting data on both the amount of solar radiation reaching the ground, and the amount of particulate matter in the air as measured by other instruments, Peters worked with MIT associate professor of mechanical engineering Tonio Buonassisi and three others to find a way to calculate the amount of sunlight that was being absorbed or scattered by haze before reaching the solar panels. Finding the necessary data to determine that level of absorption proved to be surprisingly difficult.

Eventually, they were able to collect data in Delhi, India, providing measures of insolation and of pollution over a two-year period — and confirmed significant reductions in the solar-panel output. But unlike Singapore, what they found was that “in Delhi it’s constant. There’s never a day without pollution,” Peters says. There, they found the annual average level of attenuation of the solar panel output was about 12 percent.

While that might not sound like such a large amount, Peters points out that it is larger than the profit margins for some solar installations, and thus could literally be enough to make the difference between a successful project and one that fails — not only impacting that project, but also potentially causing a ripple effect by deterring others from investing in solar projects. If the size of an installation is based on expected levels of sunlight reaching the ground in that area, without considering the effects of haze, it will instead fall short of meeting its intended output and its expected revenues.

“When you’re doing project planning, if you haven’t considered air pollution, you’re going to undersize, and get a wrong estimate of your return on investment,” Peters says

After their detailed Delhi study, the team examined preliminary data from 16 other cities around the world, and found impacts ranging from 2 percent for Singapore to over 9 percent for Beijing, Dakha, Ulan Bator, and Kolkata. In addition, they looked at how the different types of solar cells — gallium arsenide, cadmium telluride, and perovskite — are affected by the hazes, because of their different spectral responses. All of them were affected even more strongly than the standard silicon panels they initially studied, with perovskite, a highly promising newer solar cell material, being affected the most (with over 17 percent attenuation in Delhi).

Many countries around the world have been moving toward greater installation of urban solar panels, with India aiming for 40 gigawatts (GW) of rooftop solar installations, while China already has 22 GW of them. Most of these are in urban areas. So the impact of these reductions in output could be quite severe, the researchers say.

In Delhi alone, the lost revenue from power generation could amount to as much as $20 million annually; for Kolkata about $16 million; and for Beijing and Shanghai it’s about $10 million annually each, the team estimates. Planned installations in Los Angeles could lose between $6 million and $9 million.

Overall, they project, the potential losses “could easily amount to hundreds of millions, if not billions of dollars annually.” And if systems are under-designed because of a failure to take hazes into account, that could also affect overall system reliability, they say.

Peters says that the major health benefits related to reducing levels of air pollution should be motivation enough for nations to take strong measures, but this study “hopefully is another small piece of showing that we really should improve air quality in cities, and showing that it really matters.”

The research team also included S. Karthik of Cleantech Energy Corp. in Singapore, and Haohui L. of the National University of Singapore. The work was supported by Singapore’s National Research Foundation through the Singapore-MIT Alliance for Research and Technology and by the U.S. Department of Energy and National Science Foundation.

de MIT News https://ift.tt/2LzUhgf