Microsoft AI for Earth grantee gallery

Showcasing the work of Microsoft AI for Earth grant recipients

Profiles

AI for Earth supports organizations all around the world that are working on challenges in biodiversity conservation, climate change, agriculture, and water. Read more about the AI for Earth grantees and the amazing projects they’re working on in environmental sustainability.

Collapse

  • ATREE

    Building a biodiversity atlas for northeast India

    The Ashoka Trust for Research in Ecology and Environment (ATREE) plans to use machine learning and computer vision to boost its efforts to map and catalog the unique, resource-rich ecosystem of northeast India. Armed with detailed satellite images of the region and the AI for Earth grant, the team believes AI-enabled tools will help them create a comprehensive database of biodiversity to help policymakers and local communities make better-informed economic, ecological, and infrastructure-related decisions.

  • Aberystwyth University, To the Poles

    Gaining a better understanding of Earth’s melting glaciers

    Joseph Cook, a polar scientist from the United Kingdom, is applying machine learning to optical data from drones and satellites to explore the changing cryosphere of Arctic glaciers. By training the algorithms to recognize how the different surfaces reflect certain wavelengths of light — wavelengths that can be measured by satellites as well as by drones — precision study of vast areas becomes feasible. After testing the algorithms on imagery from custom-built drones, Cook’s team will then apply them to satellite remote-sensing data, enabling them to scale up to entire glaciers.

  • AdaViv

    Using AI to unleash the potential of urban agriculture

    AdaViv is developing an adaptive and efficient indoor growing system on Azure that uses sensors, actuators, and machine learning to monitor plant growth, predict yields, detect diseases, and understand precisely how nutrients, environment and light are affecting plant growth. This system will help indoor producers attain higher yields, precise quality control, and hyper-efficient production.

  • African Parks

    Creating smart, connected parks with cloud and AI

    Managing vast territories with dispersed endangered animal populations is a logistical feat, especially in regions with political instability and poaching. For two decades, the nonprofit African Parks has served as steward for parks across Africa that would otherwise lack the resources to protect vulnerable animal populations. Cloud connectivity has transformed African Parks’ management approach from ad hoc interventions by roving rangers into a centralized command center from which the park manager can coordinate operations. African Parks uses Microsoft cloud and AI tools to track elephants, detect threats to them, and coordinate an effective ranger response to protect the animals from poaching or illegal hunting. With machine learning, park managers also hope to better understand and even predict elephant behavior. As low Earth orbit satellite constellations bring internet connectivity to the most remote regions, African Parks is poised to create smart, connected parks that are true sanctuaries for vulnerable endangered species like elephants, tigers, hyenas, and Kordofan giraffes.

  • Ag-Analytics

    Improving agriculture forecasting and conservation practices

    Professor Joshua Woodard launched Ag-Analytics, a service integrated with the John Deere Operations Center that provides intelligent, easy-to-use tools to help farmers plan and monitor their crops. Ag-Analytics brings together data from farm machinery sensors with other datasets such as weather and satellite imagery to develop models for yield and crop cover forecasting. This information and more accurate forecasting can help shape policies to make it economically feasible (through insurance subsidies) for farmers to implement conservation practices.

  • Agrimetrics

    Accelerating innovation in agri-food by linking data

    The agri-food sector is one of the least digitized industries in the world, with many barriers to collecting, sharing, and using data. Agrimetrics was founded to accelerate innovation in the agri-food industry by connecting data and thus enabling advanced analytics and AI. Their goal is to help agri-food businesses produce food more efficiently and sustainably. As a long-time Microsoft Partner, Agrimetrics uses Microsoft Azure technology to power an agri-food Data Marketplace. This Data Marketplace lets data providers, from farmers to global corporations, market and manage their data, and helps data consumers, from researchers to businesses and government organizations, to find and use that data. Agrimetrics is now working with the AI for Earth program to forge new collaborations, such as with other AI for Earth grantees, and extend its capabilities to deploy innovative, sustainable, and scalable solutions to environmental and agricultural problems around the world.

  • Aquaveo

    Solving the challenge of increasing flood frequency and severity in developing countries

    Michael Souffront, a software engineer at Aquaveo, developed a high-density hydrologic model and visualization tool for forecasting global floods. The GloFAS-RAPID model produces streamflow forecasts not just from major rivers, but also medium-sized and smaller streams — helping advance flood preparedness, especially in developing countries.

  • Breeze Technologies

    Applying AI to improve the accuracy of air quality measurements

    Air pollution is the single biggest environmental health threat of our time, killing 7 million people and costing the world economy USD5 trillion per year. Data-driven decision-making around air pollution mitigation has been unfeasible, as traditional sensing equipment is expensive, stakeholders lack necessary knowledge to analyze the data, and suitable interventions are difficult to define. Breeze Technologies aims to deliver hyperlocal comprehensive and accurate air quality data from public and private data sources and low-cost sensors, as well as insights based on recent scientific studies and actionable recommendations from a growing, self-learning catalog of more than 3,500 air quality interventions.

  • Carnegie Mellon University

    Countering poaching with adaptable AI

    Poaching is one of the greatest threats to wildlife conservation and is very difficult to prevent. Rangers have had to develop their own skills and intuitions through years of field experience, and they lacked modern technological tools that could help them make better decisions. PAWS, developed by Dr. Fei Fang of Carnegie Mellon University, is an AI tool that uses machine learning and behavior modeling to help rangers plan more effective patrol routines. Through a Microsoft AI for Earth grant, Dr. Fang took the next step in developing PAWS by adding real-time interactive tools that can take new information from the rangers on patrol and offer updated strategies for tracking down poachers.

  • Chloris Geospatial

    Using spaceborne sensors and AI to measure forest biomass

    In the wake of Glasgow COP26, world leaders are aligned in their commitment to reduce atmospheric carbon and keep global warming below the critical 1.5°C threshold. Making good on this promise requires changing the way forests and landscapes are managed. Microsoft Planetary Computer partner Chloris Geospatial has developed a unique technology that goes beyond monitoring forest cover—the traditional monitoring approach—to measure directly the growth and degradation of above-ground biomass over time, providing accurate, global measurements of a crucial component of earth’s carbon stock. Chloris’ proprietary solution uses satellite data and machine learning to transform how forests and other ecosystems are monitored. At 30-meter resolution, it offers accurate insight into changing carbon stock at the global, national, regional, and even the project level.

  • City University of New York

    Furthering oceanographic study with the Microsoft cloud

    The Ocean Observatories Initiative (OOI) Cabled Array collects large quantities of data from the seafloor and overlying ocean environment of the Juan de Fuca tectonic plate in the northeast Pacific — providing a valuable opportunity for researchers and students to learn more about the ocean and seafloor processes. But currently, downloading and processing the data on local computers takes days or even weeks. With funding from the Microsoft AI for Earth program, Dr. Timothy Crone and Dr. Dax Soule are helping make this data more accessible and usable to scientists and students around the globe by building a Microsoft cloud-based system on Pangeo, an open-source platform for big data geoscience. Increasing access to technology is a passion for Dr. Soule. At the City University of New York (CUNY), he teaches students from very diverse ethnic and economic backgrounds that are often not well-represented in science-related careers. Through the partnership with Microsoft, Dr. Soule and his students now have the cloud-based tools they need to access and work with the OOI Cabled Array data, conducting important research and becoming the next generation of oceanographic scientists.

  • CoCoRaHS Network, Colorado State University

    Enhancing climate data and research with AI

    Precipitation can vary a lot over surprisingly small distances, as demonstrated by the Spring Creek flood in Fort Collins, Colorado, in 1997, when 14.5 inches of rain fell in a highly concentrated area and caused a deadly flash flood in nearby neighborhoods that had significantly less rain. From that disaster was born the Collaborative Community Rain, Snow, and Hail Network — CoCoRaHS for short — which works with thousands of volunteers to gather daily data on precipitation. CoCoRaHS provides small-scale coverage that helps weather services issue timely alerts on severe weather conditions that can save lives, and its accumulated records also help other organizations, from climatology to agriculture, engineering, and insurance, with long-term planning. Now thanks to a Microsoft AI for Earth grant, CoCoRaHS is improving the quality of its reports through AI, pulling more information out of the reports with natural language processing, and making that data more available through Azure Notebooks and Power BI.

  • Columbia University

    Keeping a close watch on our forests, for our future

    Professors Tian Zheng and Maria Uriarte at Columbia University are using ground observations of forest plots to create a machine learning pipeline that’s capable of correctly classifying the species of individual trees using aerial photographs and LiDAR data collected by NASA using remote sensing technologies, in order to better understand how storms affect a forest’s ability to store carbon and aid in climate change mitigation, and how damaged forests recover over time.

  • Conservation Metrics

    Closing the gap between field work and analysis

    Conservation Metrics is developing automated solutions using Microsoft Azure that collect, process, and analyze terabytes of wildlife data. By moving its infrastructure to Azure, Conservation Metrics hopes to give researchers more time and resources to meet their conservation goals by significantly closing the gap between field work and information and discovery.

  • Conservation Science Partners

    Protecting forest and water resources with AI

    Forests in the western United States are suffering increasing tree losses due to several causes, including droughts from climate change, wildfires, and beetle infestations. This loss of trees is a serious problem not only for maintaining carbon storage, but also for the availability of water resources, as forests play an important role in replenishing local watersheds. A regional-scale analysis of forest disturbances and their impact on water resources changes is necessary to better understand and manage these issues. With recent advances in AI, machine learning, and cloud computing, it’s now possible to combine satellite imagery at medium and high resolutions and analyze this data to see how the forest cover across the region changes from disturbance events. Additionally, that analysis can be correlated to water supply records to understand those impacts as well. Through the insights gained from this study, local communities, regional organizations, and the federal government can better manage and protect these vital resources.

  • Cornell University Center for Conservation Bioacoustics

    Monitoring insect sounds in tropical rainforests

    Led by Holger Klinck and Laurel Symes of Cornell University’s Bioacoustics Research Program, a team of researchers is looking to AI-powered acoustic monitoring of insects as a way of better understanding the dynamics of rainforest habitats. The team is focusing first on neotropical rainforest katydids, a diverse group that occupies a central position in tropical food webs. How the wide variety of katydids interacts with the rest of the forest species, both plants and animals, can provide lots of information about the overall ecosystem. Klinck aims to scale beyond insects to other species, including birds, monkeys, and other vocal animals, to help advance conservation of tropical rainforests.

  • DHI GRAS

    Improving crop water efficiency in Uganda

    At DHI GRAS, Dr. Torsten Bondo and Dr. Radoslaw Guzinski — a small Denmark-based company focused on Earth observation and satellite imaging — are using machine learning and satellite remote sensing to measure the rate of water evaporation from soil and plant surfaces into the atmosphere from fields. Their goal is to help Ugandan farmers reduce water use by knowing more precisely how much water their crops really need.

  • Digamma.ai

    Classifying land cover to improve maps of landslide susceptibility, water, and carbon storage

    Due to technological advances, scientists can now capture data on Earth and off-world at rates that greatly exceed the ability to interpret it. These massive datasets challenge the delivery of timely maps and analysis to the nation. Artificial intelligence (AI) tools, including machine learning, allow us to rapidly interpret these data sets to solve national challenges. Digamma.ai and the U.S. Geological Survey’s National Innovation Center (USGS NIC) are using machine learning to dramatically improve land cover models, with the intent of improving maps of landslide susceptibility, water, and carbon storage. In addressing this challenge, Digamma.ai used geologic mapping and USDA imagery to train machine learning models to discriminate between bare rock and exposed soil, improving land-cover maps across the Sierra Nevada in California.

  • Dr. Monique Mackenzie, University of St. Andrews

    Saving endangered vultures through AI modeling

    Vultures perform essential ecosystem services by scavenging on dead animals, which is crucial in preventing the spread of disease to other animals and humans. However, deliberately poisoned carcasses—a result of human-wildlife conflict—can result in several hundred vulture deaths at a single poisoned carcass. Dr. Monique Mackenzie, a statistician and Provost at the University of St. Andrews in the United Kingdom, is working with a team in Namibia to stop the decline of the vulture population. By analyzing the locations and activity of GSM/satellite tagged animals which locate carcasses as part of normal foraging behavior, the team can quickly locate and attend to the carcasses, preventing many deaths. Through an AI for Earth grant, Dr. Mackenzie can help the team upscale their solution and create lasting impact.

  • Duke University

    Monitoring climate change in the Antarctic with machine learning

    Climate change is disrupting the pristine ecosystem around the western Antarctic peninsula, a globally significant center of biodiversity based around the presence of krill that provide sustenance for many other species. Many of the world’s whales spend their summers here as their primary feeding grounds. By monitoring the size and health of the whales, it’s possible to gain insights on the abundance of krill and the ecosystem as a whole. Satellites and drones now enable vast amounts of image and video data on the whales to be collected — more than could ever be efficiently processed by people. The Duke University Mobile Robotics and Remote Sensing Lab is developing machine learning models on Microsoft Azure that can manage this massive data and quickly provide the statistics needed to help further research and environmental protection efforts. Making these models available as APIs on Azure will also enable other researchers to improve their work.

  • ETH Zurich

    Fighting deforestation with deep learning and smart contracts

    Deforestation is one of the significant contributors of greenhouse gases and drivers of climate change. Much of that deforestation comes from local farmers trying to make their living, which presents the possibility of combating it by offering financial incentives to preserve the trees. In many areas, such as the Amazon forest, it would be too time-consuming and difficult to determine who has the legal claim to be the caretaker for a particular section of forest. AI researcher David Dao and his team came up with an innovative alternative: help make everyone a caretaker of the forest by letting anyone put a financial stake in its well-being and earning a higher repayment when it is conserved. Through AI technologies on Microsoft Azure, including machine learning and blockchain-enabled smart contracts, Dao was able to make this concept, GainForest, a reality.

  • EcoHealth Alliance

    Tracking diseases through scientific literature with AI

    As new infectious diseases emerge and spread in different areas of the world, tracking the outbreaks is an important step in analyzing where they might emerge or spread next. Archives of scientific publications such as PubMed Central present a resource for monitoring this information, but with thousands and thousands of articles published annually without a common standard for presenting data, extracting that data is very challenging. EcoHealth Alliance, an international nonprofit organization dedicated to preventing pandemics and protecting both human lives and wildlife, is turning to AI to meet this challenge. With assistance from a Microsoft AI for Earth grant, EcoHealth Alliance is developing PubCrawler, an AI-based software project that uses natural language processing to produce high-resolution datasets of the locations where research is being done into various diseases. The tools of this project also can be applied more broadly to meet other needs in biodiversity and conservation research.

  • Farming Online

    Optimizing coffee harvesting with AI

    Coffee farming is a financially risky effort because the coffee berries ripen at varying rates even on the same tree branch, making it challenging to avoid a wasteful amount of underripe or overripe fruit. Climate change is increasing that risk by reducing yields and quality, increasing pests and diseases, and even making farmlands untenable. These changes are not merely an inconvenience for coffee drinkers around the world, but represent a serious threat to the livelihood of tens of millions of small-scale farmers in developing nations. Farming Online is working to mitigate that risk by enabling farmers to harvest a higher proportion of fully ripe coffee. Through a Microsoft AI for Earth grant, the Farming Online team is developing a machine learning model and smartphone app that will let farm workers in the field use photos of the coffee berries and current weather data to predict the best time to harvest.

  • Harrison Atelier

    Pollinators Pavilion

    Unlike the familiar honeybees which live together in hives, most bee species are solitary and therefore difficult to study. These solitary bees also play a far greater role in pollination than is commonly known, and understanding their lives is important to managing biodiversity and conservation efforts. For that purpose, Dr. Ariane Harrison and her team at Harrison Atelier created the Pollinators Pavilion, a prototype field station and educational tool that provides an artificial habitat and monitoring station for 2,000 solitary bees. Using automated cameras and machine learning analysis, the Pavilion will help researchers better study the bees, while also providing a means for the public to learn more about these important pollinators as well.

  • HighTide Intelligence

    Understanding the risk of sea level rise to populations

    Sea levels are rising and the impact on coastal communities will be far reaching. Though most communities are aware and conceptually understand what it could mean to local infrastructure, very few are focusing on addressing the inevitable challenges this will bring to individuals. HighTide Intelligence is quantifying the financial impact of climate-driven flooding with the goal of preparing people and cities to understand the risks and measures required to adapt to changing climates

  • IIIT Delhi

    Developing an intelligent tool for monitoring monkey populations

    Ankita Shukla’s and a team at IIIT Delhi are developing an intelligent tool to monitor and control rapidly growing urban monkey populations. The tool will use Microsoft cloud and AI tools to detect and identify individual monkeys from images captured by photographers and camera traps, helping researchers identify and find monkeys needing sterilization and distribute contraceptive-laden food.

  • Imazon

    Enabling equitable water distribution to residents in megacities

    The vastness of the Amazon means that identifying and understanding trends there can be a daunting task. While images were available for processing, the initial process was highly manual and time consuming. And while image processing computer models was available, the sheer volume meant that the computing power required was significant. Using Microsoft Azure, image recognition became scalable to larger geographies in order to locate and predict fires and deforestation. This prediction allows governments and NGOs to better inform their budgets and allocate resources.

  • Indian Institute of Science

    Enabling equitable water distribution to residents in megacities

    Dr. Yogesh Simmhan is part of an interdisciplinary team that is applying their experience with the Internet of Things to the challenge of water management in megacities. As part of the EqWater project, Dr. Simmhan will use data analytics and machine learning to identify inequities in water distribution and develop data-based recommendations to resolve them.

  • International Center for Tropical Agriculture (CIAT)

    Using AI to prevent malnutrition in sub-Saharan Africa

    Dr. Mercy Lung’aho and the International Center for Tropical Agriculture are tackling the issue of chronic malnutrition in sub-Saharan Africa with NEWS, a Microsoft AI-powered diagnostic model designed to predict and prevent a nutrition crisis before it occurs. NEWS will aggregate and analyze satellite imagery and traditional data, such as rainfall, temperature, and vegetation health, to help predict the nutritive value of crops. Insights from NEWS will then help inform interventions to boost nutrition in sub-Saharan Africa.

  • International Crops Research Institute for the Semi-Arid Tropics ICRISAT

    Helping solve the big challenges of small farmers

    Dr. Mamta Sharma and a team at ICRISAT are using Microsoft cloud computing and AI together with IoT sensors to help with real-time monitoring of small farms in developing countries and provide pest diagnosis and farm and market advice to farmers through an AI-supported mobile application that displays personalized prediction results and recommend actions for each farmer.

  • Ketty Adoch

    Assessing the impacts of development in Uganda

    The wilderness of the Murchison Falls National Park and nearby Lake Albert in Uganda is threatened by development for oil production. However, the potential impact is difficult to assess without knowing what changes are happening to the land. Through a Microsoft AI for Earth grant, Ketty Adoch will be applying machine learning to analyze aerial imagery of the landscape, tracking the changes in the previous and upcoming decades. These algorithms and analyses will support conservation efforts going forward.

  • Leadership Counsel for Justice and Accountability

    Forecasting regional-scale water shortages

    Residents in California’s rural Central Valley often rely on private domestic wells for drinking water, but many of these wells are vulnerable to failure when groundwater levels fall due to drought or unsustainable management. In fact, for the past century, Californians have consumed more water in any given year on average than has been naturally replenished in aquifers. Using historical groundwater level data, Leadership Counsel for Justice and Accountability and UC Davis work to predict groundwater level trends; this output is then lined up with the state’s Well Completion Report database, which shows the location, depth, and type of wells (agricultural, public supply, or domestic). Algorithms are applied to determine how vulnerable each well is to failure, based on both pump location and local groundwater levels.

  • Lion Identification Network of Collaborators

    Counting lions through AI for conservation

    To protect a threatened species, we need to know how many animals are left and where they are. That can be extremely difficult to determine for species that are wide-ranging and very similar-looking, such as lions. The Lion Identification Network of Collaborators (LINC) is working to provide a collaborative online database to help researchers overcome this problem. Through a Microsoft AI for Earth grant, LINC is developing AI techniques to identify individual lions through images with far greater accuracy than humans could manage. With that capability, lions can be more easily tracked, managed, and protected.

  • Long Live the Kings

    Using the power of Azure to save salmon in the Salish Sea

    Long Live the Kings is developing an ecosystem model on Microsoft Azure to answer critical questions facing salmon recovery and sustainable fisheries in the Salish Sea. On Azure, researchers can run up to ten simulations at a time and get results in days instead of weeks — propelling research that informs ecosystem management and policy decisions.

  • Lower Atmosphere Research Group

    Improving short-term forecasting with radar analysis

    Weather forecasting is notoriously difficult. So many factors play into predicting what is going to happen in the earth’s atmosphere. Jennifer Davison, President of the Lower Atmospheric Research Group, is taking advantage of existing Next-Generation Radar data (NEXRAD) measurements to map out the mean mesoscale, real-time vertical structure of moist, dry, and other significant layers to improve short term forecasting and our knowledge of the lower atmosphere.

  • National Audubon Society

    Protecting bird populations after hurricanes with AI-enabled monitoring

    Increasingly intense storms are causing significant erosion and land cover change along US coastlines, resulting in critical habitat loss for birds. The National Audubon Society is using Microsoft cloud and AI tools to improve bird monitoring after weather-related disasters, helping researchers quickly assess disturbance effects and act to preserve endangered coastal birds.

  • National Geographic Labs, Conservation Intelligence

    Moving conservation platforms to the cloud

    National Geographic Labs recognized that the conservation model practiced for the last several decades is failing and began exploring new ways to advance their efforts. In the conservation sector, the same model has been followed without incorporating learnings from mistakes, and there will unfortunately always be more poachers than rangers to detect them. National Geographic Labs asked itself how other industries have benefited from technology and why it’s been failing in the conservation sector, recognizing that there was a disconnect between folks in the bush and the people with technology know-how. In order to more efficiently use technology to advance conservation, the two groups needed to understand how to collaborate and communicate with each other. Using AI, technology can help conservation workers recognize events that normally require hands-on continuous observation to identify. Technology acts as a force multiplier to help the people on the ground be more efficient at their jobs and to better protect the animals and places that they’re tasked with protecting. While technology is not a silver bullet, it’s a critical component in making conservation successful moving forward.

  • National University of Ireland Galway

    Saving the bees with sustainable farming – and AI

    Honeybees are one of the most widely used pollinators, playing a vital role in maintaining the world’s food supply. However, bee populations have declined dangerously, and modern intensive agriculture is one of the main causes — including pesticide use and monoculture crop production. Agustin Garcia Pereira saw these problems firsthand, growing up in a farming community in Argentina. Now, as a software engineer and researcher with the Insight Centre for Data Analytics at the National University of Ireland Galway, he is using remote sensing data, Microsoft Geo AI Data Science Virtual Machines, and GIS mapping to develop machine learning models that can identify agricultural practices at field level across wide areas. This information will help farmers, beekeepers, and governments shift to more ecological and sustainable agriculture that also helps sustain the bees.

  • NatureServe

    Building a unique tool to map high-priority conservation areas

    NatureServe is developing an unparalleled tool for identifying the places most critical for conserving at-risk species in the contiguous United States. With support from Esri, The Nature Conservancy, and Microsoft, NatureServe and its network of state natural heritage programs are applying machine learning techniques to their comprehensive biodiversity inventory data to model habitat for more than 2,600 at-risk, taxonomically and ecologically diverse species. These spatial models will be synthesized into a map that identifies high-priority biodiversity conservation areas — a dynamic, transparent, and repeatable base layer to help guide effective conservation decision-making.

  • OceanMind

    Curbing illegal fishing with satellite data and AI

    Illegal, unreported, and unregulated fishing have significant detrimental impacts on biodiversity and exacerbate ocean impacts of climate change. Healthy and productive ocean ecosystems are necessary for human food security, livelihoods, and health, and for helping the planet be more resilient in the face of climate change. OceanMind is working to increase the sustainability of fishing by analyzing vessel movements and identifying their behavior and regulatory compliance. This helps governments enforce existing laws more effectively and helps seafood buyers make more responsible choices. Through a Microsoft AI for Earth grant, OceanMind will move its data analytics to the Microsoft Azure cloud, allowing it to analyze more data in real time, faster and more accurately. That will greatly improve OceanMind’s ability to help in the fight against illegal fishing.

  • Patagonian Institute for the Study of the Continental Ecosystems

    Automating the mapping of land use and land cover

    The Chubut watershed, located on the arid Patagonian steppe, is the main source of water for 250,000 people. Due to climate change, water yield is expected to decrease by an estimated 20 to 40 percent in the Chubut River by the end of the century. Since 2018, Dr. Ana Liberoff of the Patagonian Institute for the Study of the Continental Ecosystems and her colleagues have worked to model the impacts of human practices on water quality and quantity. An important input for modeling human impacts are land use and land cover (LULC) maps. Deep learning neural network algorithms can facilitate standardized methods for producing consistent LULC maps, allowing for more accurate tracking of changes over time. The AI for Earth Innovation grant, a partnership between the Microsoft AI for Earth program and The National Geographic Society, is helping Dr. Liberoff’s team to automate LULC map production using a transdisciplinary approach. By combining remote sensing data and vegetation indices, the team will produce maps that can then be validated by stakeholders on the ground and used to predict future changes to land and water use.

  • Peace Parks Foundation

    Connecting farmers in Africa to conservation practices

    For many years, conservation farming has been taught and implemented in marginal areas of trans-frontier conservation areas in southern Africa. The Peace Parks Foundation is extending these practices on a larger scale by offering mobile devices with a custom app that helps farmers learn the methodology and measure the process.

  • Peace Parks Foundation

    Fighting wildlife crime with intelligent poacher detection

    To combat increasing wildlife crime, the Peace Parks Foundation is developing Smart Park, an integrated set of systems and technologies on Microsoft Azure designed to significantly enhance anti-poaching methods and protection for rhinos and other endangered wildlife by providing data-driven and intelligent decision-making.

  • Peace Parks Foundation

    Perpetuating the traditional skills of animal tracking

    Traditional animal tracking skills offer value to modern needs for conservation, wildlife protection, and ecotourism. However, these skills are in danger of being lost. Peace Parks Foundation has been involved with the SA College for Tourism’s Tracker Academy for several years. The Tracker Academy seeks to restore these skills, and now bring them into the modern age with a custom app that helps students learn to track, while also teaching the general populace the value of tracking.

  • Rainforest Alliance

    Supporting sustainable livelihoods for the world’s cocoa farmers

    Climate change is threatening the livelihoods of 5 million smallholder cocoa farmers who produce roughly 90 percent of the world’s cocoa supply. The Rainforest Alliance is using its Microsoft AI for Earth grant to develop a machine learning model to predict cocoa yield and a customized digital dashboard in ArcGIS Pro that will help farmers optimize their agricultural practices and improve their incomes more sustainably.

  • Salva Rühling Cachay

    Improving deep learning models for El Niño prediction

    The El Niño Southern Oscillation (ENSO) typically happens every three to five years and affects agriculture, safety, and living conditions. Today, a deep learning model using a convolutional neural network (CNN) is used to forecast El Niño patterns; however, there are limitations with this model. Salva Rühling Cachay and partners have created a new model that overlays graph neural networks (GNN) with climate models to more accurately forecast El Niño for up to six months.

  • Scripps Institution of Oceanography, University of California, San Diego

    Improving weather forecasting for the western US with AI

    Atmospheric rivers are a weather phenomenon that cause much of the precipitation in the western coasts of continents but aren’t well represented in current forecasting models. The Center for Western Weather and Water Extremes (CW3E), a project of the Scripps Institution of Oceanography at UC San Diego, was established to better understand and predict atmospheric rivers and other phenomena affecting the western US. By applying machine learning and other AI tools to existing decades of weather data, CW3E hopes to improve the accuracy of forecasts and provide better tools and processes for managing flooding and drought conditions.

  • Scripps Institution of Oceanography, University of California, San Diego

    Monitoring mangrove deforestation in Mexico

    Mangrove forests supply many ecosystem services, from fisheries to coastal protection, carbon sequestration, and biodiversity. Mangroves are estimated to sequester up to 50 times more carbon than tropical rainforests, but existing climate policies do not include mangrove protection. And in the last 50 years, between 30 and 50 percent of the world’s mangroves have been destroyed. Researcher Octavio Aburto of Scripps Institution of Oceanography and his team, in partnership with Engineers for Exploration, are using drone technology and machine learning to identify and monitor mangrove forests in Mexico to improve measurable conservation outcomes globally. Aburto’s team is working closely with Mexican officials to assess mangrove coverage in all regions in Mexico, to define mangroves as a protected habitat and to calculate the economic value of the habitat’s ecosystem services. With machine learning and drone imagery they can evaluate mangrove ecosystems on a local and global scale, providing stakeholders and decisionmakers with the data they need for effectively management and conservation.

  • SilviaTerra

    Incentivizing small forest landowners to keep their trees

    In partnership with Microsoft, SilviaTerra is transforming how conservationists and landowners measure, monitor, and preserve forests. In 2018, AI for Earth awarded SilviaTerra a grant to develop a high-resolution national forest inventory with timber, habitat, and carbon estimates for every acre in the continental US. The first of its kind, the forest inventory enables SilviaTerra to provide landowners with recommendations to improve the value, health, and future of their forests — including providing better access to carbon markets. SilviaTerra is now working with Microsoft to demonstrate the viability and effectiveness of a data-driven, technology-enabled market for small, private landowner carbon. SilviaTerra will be sourcing and selling forest carbon from small private landowners, using an AI for Earth-enabled approach for measurement, algorithmic verification, and monitoring of carbon stocks, with Microsoft as its first corporate carbon credit purchaser.

  • SOS Mata Atlântica

    Protecting the rivers of Brazil’s Atlantic Forest

    Brazil is rich in fresh water sources, yet millions of its people lack access to clean water. Within the Atlantic Forest region along Brazil’s southeast Atlantic coast, even such major rivers as the Rio Tietê, which flows through the São Paulo metropolitan area, are unprotected from contamination by garbage and other pollutants. That leads to waterborne diseases causing millions of preventable hospitalizations and deaths. The SOS Mata Atlântica Foundation has been monitoring the water quality of the rivers in the Atlantic Forest for years, working to call attention to this plight. Through a grant from Microsoft AI for Earth and the development work of EloGroup, the Foundation now has a machine learning model that can predict water quality out to five years in advance based on historical data, and a dashboard on its website to clearly present the findings. With these tools, SOS Mata Atlântica is better positioned to advocate for changes in the laws so that the rivers will be protected and safe fresh water will be available to everyone.

  • Soundscapes to Landscapes, University of California, Merced

    Mapping biodiversity by classifying bird calls with AI

    As a global phenomenon, climate change affects biodiversity at scales that are difficult to monitor. The Soundscapes to Landscapes (S2L) project in Sonoma County, California, is developing a new methodology to measure these changes in bird communities. Citizen scientists distribute audio sensors around the county and help identify bird calls in the recordings through an online system. From these data and satellite imagery, the S2L team uses species distribution modeling to create biodiversity maps that provide a better understanding of the impact of climate change on bird diversity and help conservation planning efforts. This method generates massive amounts of sound data that requires an automated approach to analysis. Through a Microsoft AI for Earth grant, S2L team members Professor Shawn Newsam and PhD student Shrishail Baligar at the University of California, Merced, are developing and training AI deep learning models to identify and annotate the bird calls. Not only will the data be analyzed quickly, but also the process can be scaled up to large landscapes on a global level. Additionally, providing these models through an Azure API will enable other scientists to benefit from this work.

  • Stanford University Natural Capital Project

    Detecting and mapping small dams and reservoirs

    A team of researchers at the Natural Capital Project, based at Stanford University, is combining remote sensing data with machine learning to develop a model that can detect smaller dams and reservoirs. Knowing where the dams and reservoirs are located will contribute to mitigating their impact, to conserving and managing hydrological ecosystem services, and to planning development more sustainably.

  • Stony Brook University

    Tracking Antarctic penguin populations

    Dr. Heather Lynch is working at the intersection of geography, remote sensing, statistical modeling, and ecology to map the distribution and abundance of Antarctic penguins. She is using computer vision and deep learning to help train computers to identify penguin colonies and estimate penguin populations based on guano stains visible in satellite imagery.

  • SunCulture

    Improving the livelihood of African farmers with AI

    Africa offers the best potential to meet the increasing needs of food production for the world’s growing population, but the great majority of its farmers are smallholders lacking the resources to meet that demand. SunCulture was founded to help those farmers, initially by improving their productivity through solar-powered irrigation. But the company realized the farmers needed better guidance to use more efficient precision agriculture methods, and that required gathering more information on the needs of their specific farm. By working with Microsoft, first with its Airband Initiative and now with AI for Earth, SunCulture has been building an Internet of Things solution using local sensors and cloud-based machine learning models to produce detailed recommendations for farm management based on weather predictions, soil conditions, and pest levels. With access to irrigation, financing, and reliable, real-time recommendations on how best to manage their farms, the farmers can improve their lives and be better positioned to help feed the world.

  • Swedish Forest Agency

    Monitoring tree health with AI models

    Monitoring the health of trees is a big challenge in forestry. It’s important to detect signs of diseases and insect infestations in time to either provide remedies or remove affected trees before the infections spread and wipe out entire species. However, the scale, density, and difficult terrain of forests make it impractical at best to do frequent thorough surveys by humans on foot, while aerial photographic surveys can cover a lot of ground quickly but produce vast amounts of data to sort through. The Swedish Forest Agency is leading a team of partners, including Microsoft, to develop AI and machine learning models that can analyze forest photos and not only identify the trees by species but also determine their health condition and track changes over time. The Forest Agency started with the models needed for larch trees and larch casebearer moths, a relatively small problem in Sweden that served as a proof of concept that AI can do the task quickly and sufficiently accurately to be helpful. With that proof, the Forest Agency now is scaling up to bigger problems such as Dutch elm disease and spruce bark beetles which have a greater impact on forest biodiversity and the forest industry, both in Sweden and globally.

  • Symbiosis Institute of Technology

    Enabling better energy management for utilities and customers

    Smart electricity meters generate a wealth of data that can be used to improve energy management, enabling both reduced costs and reduced carbon emissions. Archana Chaudhari, JRF at the Symbiosis Institute of Technology (SIT), Symbiosis International (Deemed University) in Pune, India, with support from Dr. Preeti Mulay, is developing incremental clustering algorithms to take advantage of this wealth of data. When applied to smart meter and socioeconomic data, these algorithms will predict demand and peak loads; identify regional, seasonal, and community patterns in consumption; enable utilities to align generation with anticipated demand to reduce waste; and help consumers to plan their own electricity usage for lower demand and reduced carbon emissions.

  • Tanzania Conservation Resource Centre

    Reducing human-wildlife conflicts through machine learning

    Understanding where human and wildlife populations intersect is critical to the conservation of wildlife in Africa. The Tanzania Conservation Resource Centre (TZCRC), in conjunction with Development Seed and other partners including the Tanzania Wildlife Research Institute, is developing an AI-assisted methodology to increase the speed of counting wildlife and human activities following an aerial survey and to produce a heatmap of potential conflict areas.

  • The Freshwater Trust and Upstream Tech

    Aquifer conservation at scale for agriculture

    Groundwater plays a significant role in supplying water for drinking and agriculture, as well as supporting many ecosystems. However, the underground aquifers that naturally store this water are slow to replenish, vulnerable to overuse from wells, and difficult to manage because so many different users draw upon them. The Freshwater Trust and Upstream Tech have worked together on various projects to conserve or restore freshwater ecosystems and related areas. Now, they have collaborated to create BasinScout® Platform, a tool to help identify the best places to improve groundwater and surface water quantity and quality. The BasinScout® Platform uses satellite imagery and machine learning to provide a holistic view of the farms within an aquifer basin and compare different possible conservation actions to see which would be most effective in terms of both cost and environmental impact. This enables targeted investments to achieve specific quantifiable improvements at a scale that makes a quantifiable difference for the environment.

  • The Nature Conservancy

    Saving the oceans through tourism and AI

    The Nature Conservancy’s Mapping Ocean Wealth initiative is focused on calculating and describing the benefits the ocean provides to people, in part by using photo repositories as a data source. In collaboration with Microsoft AI for Earth and Esri, the team developed an AI-powered web application to illustrate the value of coral reefs globally as well as by country. By integrating AI and machine learning capabilities into a model that previously relied on user-input data, image recognition attains a greater level of specificity and accuracy. The Mapping Ocean Wealth application enables users to explore the tourism value in specific locations to support sustainable management goals. Mapping Ocean Wealth moves from global analysis of coast and ocean ecosystem service to looking at specific technologies around recreation and tourism in calculating the economic value of coral reef ecosystems.

  • The Ocean Cleanup

    Identifying and quantifying plastic debris removed from rivers

    The Ocean Cleanup (TOC) is a non-profit group headquartered in Rotterdam whose goal is a 90 percent reduction in ocean plastic by 2040. Knowing that millions of tons of plastic enter the oceans via rivers each year, the group focuses on both preventing plastics from reaching the ocean (via river cleanup) and on cleaning what exists in the ocean. Reducing and preventing pollution in waterways will improve environmental health, economic factors, and human wellbeing, and TOC is now using AI and other technology to improve the efficiency of waste removal and identification in the 1 percent of Earth’s rivers that contribute 80 percent of ocean-bound plastics

  • The SeaDoc Society, University of California, Davis

    Improving killer whale conservation through data in the cloud

    The SeaDoc Society (SeaDoc) and its partners proposed the creation of a killer whale health database to facilitate health evaluation of individual animals. This database would enable a variety of scientists to cross-analyze multiple well-developed datasets to better evaluate the interaction of threats and define recovery options for this endangered killer whale population. With a Microsoft AI for Earth grant, SeaDoc was able to migrate its database to Microsoft Azure, where it is readily available to many researchers for real-time data entry and analysis.

  • The Trust for Public Land

    Mapping the benefits of parks and forests nationwide

    In recent years, interest has increased in understanding the value of city parks and open spaces, not just as social and recreational areas, but also as sources of economic and health benefits. The Trust for Public Land has long worked with urban communities to create parks and protect public land to everyone’s benefit. It developed its ParkScore Index to evaluate the effectiveness of parks in the largest US cities, and recently sought to expand that work to many more communities nationwide. Through a Microsoft AI for Earth grant, the Trust for Public Land gained the powerful, scalable resources of Microsoft Azure cloud computing, enabling its ParkServe site to map this data for the 100 largest metro areas in the US.

  • Tohoku University

    Using AI to enhance disaster response and urban resilience to natural disasters

    Dr. Yanbing Bai at Tohoku University is developing an intelligent cloud-based disaster management service using Microsoft cloud and AI tools and Esri’s geospatial mapping platform to help disaster officials, stakeholders, urban engineers, and planners mitigate the effects of and increase urban resilience to natural disasters.

  • University of Alabama

    Providing early warning of harmful algal blooms

    Africa Flores, research scientist at the Earth System Science Center at the University of Alabama, and her team are using AI to conduct deep analysis of satellite image datasets and weather models to help identify the variables that could predict future algal blooms on Lake Atitlán in the Guatemalan highlands. Knowledge of what those triggers are can turn into precise preventative action, not just in the lake in Flores’s home country but also in other freshwater bodies with similar conditions in Central and South America.

  • University of Alberta

    Understanding what drives grizzly bear density in a changing landscape

    Grizzly bears only occupy a fraction of their historic range and many remaining populations now face increasing pressure due to a changing landscape. Clayton Lamb, a Vanier Scholar and PhD researcher at the University of Alberta in Canada, is using Microsoft Azure and machine learning tools to create a comprehensive analysis of the many human and environmental factors limiting grizzly bear density in British Columbia, in order to help inform collaborative conservation efforts for this iconic species.

  • University of California, Berkeley

    Predicting climate-related human migration in Africa

    Solomon Hsiang and his team at the University of California (UC) Berkeley are applying AI to see how changes in climate have affected human migrations across Africa in the past, with an eye to the future. By applying machine learning to aerial images, the team is reconstructing a chronicle of population density, urban extents, and land use across Africa over time — helping estimate migration risk across the continent in the future.

  • University of California, Santa Barbara

    Improving agricultural water use efficiency with AI

    An increase in center-pivot irrigation is straining groundwater resources and disrupting ecosystems, and too many wells already exist to effectively measure their water usage individually with sensors. Professor Kelly Caylor at the University of California, Santa Barbara devised a different method of monitoring, by applying machine learning and analysis to satellite imagery to identify and measure changes in field use, weather, and crop growth over time. With that information, water usage can also be estimated, and policies for more efficient agricultural land and water use can be implemented.

  • University of Massachusetts Boston

    An interdisciplinary approach to improving long-range forecasts for flood prediction

    An accurate early warning system for severe floods in flood-prone regions would support response teams and help build resilience among vulnerable populations. A team of computer scientists and hydrologists/meteorologists is developing a flood prediction model that uses machine learning and an advanced AI data algorithm on Microsoft Azure to identify precursors to floods in specific at-risk regions, with the aim of accurately predicting floods with up to 15 days’ lead time.

  • University of Pittsburgh

    Monitoring bird populations with AI

    Indirect observation methods, such as microphones and camera traps, offer the opportunity for scientists to gather a lot more data on threatened species. With that capability comes the need for automated intelligent tools to efficiently process and analyze the massive amounts of data. Dr. Justin Kitzes is developing machine learning models to identify bird songs in acoustic field recordings to better measure bird populations and predict biodiversity loss.

  • University of Washington

    Forecasting marine heatwaves in time and space

    Marine heatwaves cause many adverse impacts, from disruptions in the marine food chain to declines in fish stock and consequent harm to the fishing industry and rise in economic and political tensions. Early detection and prediction of marine heatwaves can inform better decisions to protect marine ecosystems and manage human activities. Using observational records, scientists have been able to study past marine heatwaves and their drivers, as well as predict future statistics based on climate change projections. However, it’s been difficult to anticipate when, where, and for how long future marine heatwaves will exist. Hillary Scannell and her advisor, LuAnne Thompson, of the University of Washington School of Oceanography will apply AI tools to learn the spatio-temporal patterns of historical marine heatwaves, then use a large climate model database to anticipate future marine heatwaves.

  • Vector Center

    Responding to the world’s water crises with AI

    Around the world, billions of people lack access to reliable and safe water sources, and an increasing number of major population centers face the threat of “day zero,” the day when an urban water supply fails. Besides shifting weather patterns and extended droughts, governmental policies, people’s beliefs and behaviors, historical context, and infrastructure conditions all play roles in whether and when a potential shortage of water quickly reaches acute danger of collapse. Historically, a lack of available, accessible data on these factors has created dangerous gaps in situational awareness that exacerbate water crises. Those gaps can quickly translate into operational failures, political and reputational problems, civil unrest, and supply chain disruptions. Now, Vector Center has created a new set of AI-powered tools and processes to provide decision makers with real-time, contextualized intelligence about water and intersecting threats. By comparing the dissonance between perception and reality, and putting data into actionable context, Vector Center empowers governments, agencies, and businesses avoid catastrophic failures like day zero and create a more water secure world.

  • WikiNet

    Providing expert recommendations for toxic site cleanup with AI

    Contamination of soil and groundwater by toxic pollutants is a serious problem worldwide, causing malnutrition, diseases, and death among many people. The degradation rate of natural environments by these contaminants reduces exploitable farmland and increases the extinction rate of fauna and flora species. As industrialization spreads globally, it brings these problems to newly developing nations that lack the experience and expertise to regulate, prevent, or clean up contaminated sites. WikiNet is developing a solution that draws upon the accumulated knowledge from past cleanup efforts to provide automated recommendations for more efficient and effective remediation methods.

  • Wildlife Protection Solutions

    Using technology to encourage conservation and protect species

    Poachers are an active threat to wildlife around the globe, particularly in parts of Africa and Southeast Asia. By using a network of sensors, cameras, and AI technology, Wildlife Protection Solutions (WPS) is automating the detection of poachers. When a human intruder is detected through machine learning algorithms, an alert is sent to staff who can dispatch anti-poaching teams much more quickly. The Microsoft AI for Earth grant allows WPS to use cloud resources to scale the storage and processing of images, with a user-friendly interface that’s easily customized. WPS also uses virtual reality to share footage of endangered wildlife in an effort to encourage more engagement in conservation. Azure Media Services will make storing and sharing this content seamless.

  • World Resources Institute

    Reducing urban heat by mapping surface reflectivity

    As the world’s populations become more urbanized and as climate continues to change, urban heat islands play a significant role in an increasing number of deaths, as well as a wide range of other problems such as worsening health conditions, increasing energy consumption, and damaging infrastructure. Adopting more reflective materials for roofs and roads in urban areas could reduce the heat island effect and the resulting mortality risk. However, up to now, city planners have lacked a tool to map, measure, and monitor the changes in urban surface reflectivity over time, which is necessary to inform policy, budget planning, and decision making. A joint team led by the World Resources Institute is using machine learning resources provided by Microsoft AI for Earth to build this tool.

  • Yellowstone Ecological Resource Center

    Powering smart ecological decisions with real-time data insights

    Our ecosystems are changing rapidly as the ripple effects of a warming planet take hold. Climate change has revealed the degree to which we are all interconnected in a vast web of life. As such, adapting to a changing climate will take coordinated action across all decision makers, from policy makers to scientists, ranchers and farmers to concerned citizens. Yellowstone Ecological Research Center (YERC) recognizes that information is at the heart of coordinating this adaptive alliance. With the 22-million-acre Greater Yellowstone Ecosystem as its laboratory, YERC strives to gather vast stores of field data into a single source of truth trusted by all stakeholders as a reliable and actionable representation of the local ecosystem. The next frontier in YERC’s ambitious project is to bring this data to life through accessible reporting, visualization, and analytics. Microsoft Azure cloud processing and machine learning are among the tools YERC is using to help land stewards to interpret signals from across the ecosystem—rivers, land, and wildlife—and to take action for a sustainable future.

  • iNaturalist

    Creating an online community of citizen scientists to help protect species

    iNaturalist is a social media platform for nature and wildlife enthusiasts that encourages engagement and stewardship while supporting species identification and the collection of biodiversity data. Azure has enabled the platform to evolve to include artificial intelligence (AI)-based species identification and better scalability to meet seasonal and ad-hoc bumps in usage.

Collapse

Published papers

The organizations supported by AI for Earth are producing groundbreaking research in diverse fields, from environmental science to artificial intelligence. Read the scientific work that our AI for Earth grantees have published based on work done on Azure, supported by AI for Earth grants.

Collapse
Collapse

Open-source code

The projects funded by AI for Earth result in innovative solutions using AI and cloud computing. Build on the work of our AI for Earth grantees by exploring their open-source repositories.

Collapse
Collapse

Data sets

AI for Earth grantees often release their training data to encourage others to build upon their work.

Collapse
Collapse

Applications, APIs, demos, and models

AI for Earth helps to democratize access to AI tools and to enable everyone, everywhere to accelerate research, innovation, and solutions to our most urgent environmental challenges. Explore the tools that our AI for Earth grantees have developed to bring their work to the environmental science community.

Collapse
Collapse