<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[ImpactObservatoryInc - Medium]]></title>
        <description><![CDATA[Impact Observatory’s team of Earth Observation and AI experts produce cutting-edge, decision-support tools powered by deep-learning algorithms and automated metrics optimized for modern cloud infrastructure. - Medium]]></description>
        <link>https://medium.com/impactobservatoryinc?source=rss----4fe1f8128369---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>ImpactObservatoryInc - Medium</title>
            <link>https://medium.com/impactobservatoryinc?source=rss----4fe1f8128369---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 12:57:45 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/impactobservatoryinc" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Mapping the world with unmatched frequency]]></title>
            <link>https://medium.com/impactobservatoryinc/mapping-the-world-with-unmatched-frequency-54a83527f138?source=rss----4fe1f8128369---4</link>
            <guid isPermaLink="false">https://medium.com/p/54a83527f138</guid>
            <category><![CDATA[open-data]]></category>
            <category><![CDATA[land-use-land-cover]]></category>
            <category><![CDATA[satellite-imagery]]></category>
            <category><![CDATA[maps]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Mark Hannel]]></dc:creator>
            <pubDate>Thu, 10 Mar 2022 15:18:58 GMT</pubDate>
            <atom:updated>2022-03-10T17:35:01.514Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*51S07scdZvIi7KXoN7F1XQ.png" /><figcaption>Figure 1–2021 Land-use and Land-cover classifications over Southern Asia.</figcaption></figure><p><a href="https://www.impactobservatory.com/">Impact Observatory</a> (IO) is excited to announce the release of a time-series of annual, global land use and land cover (LULC) maps covering the years 2017 through 2021. IO developed these maps in collaboration with the geospatial experts at<a href="https://livingatlas.arcgis.com/landcover/"> ESRI</a> and the incredible cloud computing resources of <a href="https://www.microsoft.com/en-us/ai/ai-for-earth">Microsoft</a>, using satellite imagery from the European <a href="https://www.copernicus.eu/en">Copernicus</a> Sentinel constellation. These maps of changing land use and land cover provide leaders in governments, industry, and finance with a new AI-powered capability for timely, actionable, geospatial insights on demand.</p><h3>Automated mapping with unprecedented accuracy, speed, and scale</h3><p>Images of Earth collected by the Sentinel-2 satellite constellation observe the world as millions of pixels, each showing a patch of ground 10 meters across. IO turns millions of these satellite images collected each year into maps that label every pixel into one of nine discrete categories. Since the release of our global <a href="https://www.arcgis.com/home/item.html?id=d6642f8a4f6d4685a24ae2dc0c73d4ac">2020 LULC map</a> in June 2021, we have been thrilled to receive hundreds of thousands of visitors accessing and downloading our map via Esri Living Atlas, Microsoft Planetary Computer, and the UN Biodiversity Lab. Based on the feedback we have received, we have been working hard to improve our map-making abilities. Notably, due to limitations of observations at 10m resolution, we have merged the “shrub/scrub” and “grass” classes into a single class “rangeland”.</p><p>We have also improved the overall accuracy of our annual maps by increasing the number of scenes going into each annual result. At the same time, our engineering team has significantly improved our <a href="https://customers.microsoft.com/en-us/story/1459632165188903952-impact-observatory-professional-services-azure-hpc">processing time per scene</a>: we are now processing four times as many scenes with only marginal increases in our computation time, allowing us to combine many more observations per pixel into our final maps.</p><h3>Changing Landscapes</h3><p>Many of the most useful applications of LULC maps require the ability to measure changes in land use and land cover over time. With a time-series of LULC maps, monitoring of deforestation, urban expansion, agricultural land conversion, and surface water scarcity all become possible.</p><p>Classifications transitioning to built area can reveal new construction due to industrialization and population growth. Reductions in built area can signify urban decline and human migration that can be triggered by climate change, natural disasters, resource depletion, and human conflict. Losses in tree cover can now be understood by tracking what the forest turns into: tree-covered regions turning bare may be the result of deforestation or forest fire while tree-covered regions turning to agriculture may be the result of slash-and-burn agriculture. Changes in surface water and wetlands can inform drought assessments, floodplain management, hydropower monitoring, and the accessibility of outlying areas to vehicle passage. A few specific examples of these changes are explored below.</p><p><strong>Catastrophic fires and drought: </strong>Below, you can see our map outside Oroville, California overlaid on Sentinel-2 satellite imagery of the scene in 2020 and 2021. The extent of a 2020 catastrophic forest fire dominates the scene in 2021 as a large portion of land shifts from trees to rangeland. Additionally, the drier conditions of 2021 appear as lower water lines in the surrounding lakes and are highlighted by the transition of water to rangeland.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hBV9ZvDf9ShpFpDReLtUjg.gif" /><figcaption>Figure 2 — Forest fires in 2020 ravaged Butte County, California, resulting in significant tree loss in the 2021 land-cover.</figcaption></figure><p><strong>Rapid growth of cities in the developing world: </strong>The population of Bujumbura, Burundi is rapidly growing. In fact, the World Bank projects that Bujumbura will be one of the ten fastest growing cities in the world through 2025 growing at an annual rate of 5.75%. Our LULC map below highlights urban expansion between 2017 and 2021, showing the expanding perimeter of the African city.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7ybcufHJHfSQQ9mN_ZGL8w.gif" /><figcaption>Figure 3 — Eastward urban expansion between 2017 and 2021 in Bujumbura, Burundi. Built area classifications are highlighted in red.</figcaption></figure><p><strong>Human modification of the landscape: </strong>The Aswan High Dam was built to control annual flooding of the Nile. During periods of excess flooding, several lakes known as the Toshka Lakes appear northwest of the artificially-created Lake Nasser. The Toshka lakes periodically swell and evaporate subject to the climate and the dam’s high levels. Below our LULC maps in 2017 and 2021 readily demonstrate the differences in average waters levels during the two calendar years.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dSmreb7dXjJtGL9uHiAnYA.gif" /><figcaption>Figure 4 — The lower average water levels of 2017 are juxtaposed against the higher average water levels of 2021.</figcaption></figure><h3>Delving deeper — Landscape changes within each year</h3><p>Our annual maps utilize a year’s worth of Sentinel-2 satellite images. In addition to these publicly available maps, <a href="https://www.esri.com/en-us/arcgis-marketplace/listing/products/43c1d2612d134cc19b7f95deef5469db">Impact Observatory is now offering sub-annual time periods</a>, which reveal a trove of information on cyclical changes in landscapes and how these cycles change from year to year.</p><p>Melting mountain snow is a critical source of irrigation for surrounding regions. However, changing precipitation patterns along with higher average temperatures have limited the amount of snow cover stored on snow capped mountains. Because snow cover varies greatly throughout the calendar year, the average snow cover shown in our annual maps is sometimes significantly less than the total amount of snow cover in the winter season. By limiting our LULC analysis to the winter months in the Andes, we are able to see the changing snow cover between the winter of 2020 and the winter of 2021.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iW77Vu8n5cv6uQdLUu6Bmw.gif" /><figcaption>Figure 5 — Winter 2020 to Winter 2021 comparison of the Andes Mountain snow extent.</figcaption></figure><h3>Conclusion</h3><p>With the public release of our 2017–2021 LULC time-series, governments, environmental organizations, and researchers alike will have access to an unprecedented series of annual, global land use and land cover maps. We are positive that the above examples will encourage users of our maps to discover many more exciting and practical use cases.</p><p>Authored by Mark Hannel and Steven Brumby.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*X3KnzoXxY_F1dgO51WFmNw.png" /></figure><p><strong><em>About Impact Observatory</em></strong></p><p><a href="https://www.impactobservatory.com/"><em>Impact Observatory</em></a><em> develops AI-powered, on-demand, geospatial data that map changing land use and land cover at unprecedented scale and speed. Impact Observatory provides world leaders in governments, industry, and finance with timely, actionable, data-driven insights for sustainability and environmental risk analysis. Impact Observatory is a US tech company based in Washington DC.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=54a83527f138" width="1" height="1" alt=""><hr><p><a href="https://medium.com/impactobservatoryinc/mapping-the-world-with-unmatched-frequency-54a83527f138">Mapping the world with unmatched frequency</a> was originally published in <a href="https://medium.com/impactobservatoryinc">ImpactObservatoryInc</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mapping the world in unprecedented detail]]></title>
            <link>https://medium.com/impactobservatoryinc/mapping-the-world-in-unprecedented-detail-7c0513205b90?source=rss----4fe1f8128369---4</link>
            <guid isPermaLink="false">https://medium.com/p/7c0513205b90</guid>
            <category><![CDATA[land-use-land-cover]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[remote-sensing]]></category>
            <dc:creator><![CDATA[Caitlin Kontgis]]></dc:creator>
            <pubDate>Wed, 26 Jan 2022 16:46:08 GMT</pubDate>
            <atom:updated>2021-06-25T04:53:38.241Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*g1hX54Ph8HNzTzi2uWrjQg.jpeg" /><figcaption>The Esri 2020 Land Cover map, created in partnership with Impact Observatory and Microsoft.</figcaption></figure><p>Impact Observatory (IO) is thrilled to share a 2020 land use / land cover (LULC) map of the world, developed with <a href="https://livingatlas.arcgis.com/landcover/">Esri</a> and in partnership with <a href="https://www.microsoft.com/en-us/ai/ai-for-earth">Microsoft AI for Earth</a>. Built using the highest resolution, publicly available satellite data from the European Space Agency, this artificial intelligence powered map provides unprecedented insight into the world’s landscapes. We used a deep learning model to classify hundreds of thousands of satellite images into ten discrete LULC classes, creating an open source dataset that provides global scale understanding of global scale challenges, including conservation planning, resource management, and climate change mitigation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9MXSnrTf1Dlz28kPRGTx-Q.jpeg" /><figcaption>The Mekong River Delta in southern Vietnam, seen in the 2020 10-m LULC map. Here, you can see dense lines of built area, which are villages built along roads and irrigation canals. The central and northern areas are dominated by rice paddies, while the water further south represents aquaculture ponds.</figcaption></figure><p>The IO team evolved out of the Geographic Visualization Lab at National Geographic Society. During that time, our founders led the creation of the largest ever training dataset for medium-resolution Earth observation satellite imagery, generating over 5 billion hand-labeled pixels. Using these labels, we trained a UNet, a convolutional neural network specifically tailored for image segmentation. To get a truly representative look at 2020, we composited model outputs from multiple scenes across the entire year. We ran our model on Sentinel-2 imagery provided by the Microsoft Planetary Computer to generate a map at 10-m resolution, making this the highest resolution global map to ever be publicly released. At 10-m, the 2020 map reveals features previously unseen in global land cover maps; meandering streams, rural villages, and pockets of permanent snowfields all jump to life.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*vtChlvdlhqT1FMzNV95F2Q.jpeg" /><figcaption>The greater Seattle area, as seen in the 2020 10-m LULC map, is surrounded by the snow capped Cascades. This region is dominated by dense forests, which give way to scrub/shrub in the rain shadow of the mountains, which can be seen in the eastern portion of the image.</figcaption></figure><p>Thanks to our partners at Esri and Microsoft, we developed this map in just a few months. Using over a million core hours on Azure Batch, we were able to deploy our final model and produce the global map in a single week. Now, rather than waiting months or even years for an updated global LULC map, we can process the entire world in near real-time.</p><p>We’re proud to share this map with the world. Get acquainted with the data on the beautiful <a href="https://storymaps.arcgis.com/stories/486cd2ae2016454f951c97f802f125b3">story map</a> Esri created, and explore the map itself on <a href="https://www.arcgis.com/home/item.html?id=d6642f8a4f6d4685a24ae2dc0c73d4ac">Esri’s Living Atlas</a> and <a href="https://aka.ms/ai4edata-io-lulc">Microsoft’s Planetary Computer</a>. We can’t wait to see what you’ll build with it.</p><p>Learn more at: <a href="https://learn-about.impactobservatory.com/maps">https://learn-about.impactobservatory.com/maps</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*GSfBEKHgyN11opU9EUY3Kw.jpeg" /><figcaption>Melbourne and Eastern Victoria from the 2020 10-m LULC map. The city on Port Phillip Bay is surrounded by crops and grass to the northwest and southeast, with Yarra Ranges National Park to the northeast, which is home to kangaroos, wallabies, and over 100 species of native birds.</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*EUEhpae_PGgdmnJIjzFpFA.png" /></figure><p><strong><em>About Impact Observatory</em></strong></p><p><a href="https://www.impactobservatory.com/"><em>Impact Observatory</em></a><em> is a mission-driven technology company bringing artificial intelligence powered algorithms and on-demand data to sustainability and environmental risk analysis for governments, non-profit organizations, companies, and markets. Founded in 2020 in Washington, DC, Impact Observatory’s goal is to empower global decision-makers to be planetary superheroes, with the timely, actionable science-based insights they need to succeed.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7c0513205b90" width="1" height="1" alt=""><hr><p><a href="https://medium.com/impactobservatoryinc/mapping-the-world-in-unprecedented-detail-7c0513205b90">Mapping the world in unprecedented detail</a> was originally published in <a href="https://medium.com/impactobservatoryinc">ImpactObservatoryInc</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>