Landsat Turns 50
9 min read
Landsat Turns 50
In my senior year of high school I was fortunate enough to spend one period a day working on a special project. Instead of being stuck in a classroom, my friends and I got to work on installing an antenna on the school’s roof to receive data from polar-orbiting weather satellites. Not a satellite dish, just the same simple antenna used for over-the-air TV signals. At some point during that project, I got my hands on a glossy print of a single Landsat scene — part of learning about the uses of remote sensing beyond monitoring weather.
It blew my teenage mind.
Keep in mind that in the late ’80s the only satellite imagery you’d ever see was a loop of approaching thunderstorms on the evening news; or, if you were the right kind of geek, a grainy black and white photo in the depths of a tome titled something like Modern Combat Aircraft. Getting a close up look at a high-res image of your own neighborhood was simply unheard of.
False-color (near infrared, red, green) satellite image of Washington, DC, Baltimore, MD, and the Chesapeake Bay collected on May 16, 1987. Data courtesy USGS/NASA, processing by Planet.
I don’t remember too many specifics of the image — I know it was a false-color composite because the wooded landscape of the Washington, DC suburbs was a screaming red. I could see individual buildings (RFK Stadium, but not my house) and the National Mall. And I don’t remember seeing any clouds, which is rare for a humid climate like DC. Given what I remember, maybe it was this data, a Landsat 5 scene collected on May 16, 1987.
At the time, the Landsat program was already 15-years-old, and on its 5th satellite. First launched in July 1972, the Landsat satellites provided an unprecedented view of the Earth’s land surface. The first three satellites carried two instruments — the Return Beam Vidicon (RBV) and the Multispectral Scanner System (MSS). Data from the Return Beam Vidicon proved difficult to work with, while data from the Multispectral Scanner System became a cornerstone of the emerging science of remote sensing. The Multispectral Scanner System was revolutionary in at least three ways: it offered global coverage, high resolution, and color imagery.
In the early seventies, if you wanted to study a remote corner of the Earth you pretty much had to personally go there. Expensive and time-consuming at best, impossible at worst. With the help of an onboard tape recorder, Landsat could bring data from almost anywhere into the lab. Landsat’s global coverage also helped enable the study of Earth as a system, rather than as distinct geographic regions or unrelated academic disciplines.
False-color Landsat 1 image collected on July 25, 1972 — the earliest scene available in the Landsat archive. Image courtesy NASA Earth Observatory.
Although 68 meters (220 feet) per pixel sounds modest compared to the 50-centimeters (20 inches) or even 30-centimeters (12 inches) per pixel data returned by modern ultra high resolution satellites, contemporary satellite instruments with global coverage had a maximum resolution of 1,000 meters (3280 feet) per pixel at best. Landsat provided enough detail to study urban growth, glaciers, agricultural fields, coastal wetlands, fault lines and dozens of other features of the Earth’s surface.
Another key feature of Landsat’s Multispectral Scanner System was its ability to image in several discrete spectral bands. Most satellite imagery at the time — including the Return Beam Vidicon and classified reconnaissance satellites — was black and white (or, more precisely, panchromatic). Panchromatic imagery blends many wavelengths of light into a single measurement of brightness. The Multispectral Scanner System, in contrast, split the colors of the rainbow (and two colors beyond red) into four individual bands. This unlocked the ability to determine surface properties — like vegetation health or the presence of water — by comparing the relative brightness of two or more bands.
From the beginning, Landsat was conceived as a sequence of missions to collect a continuous record of change on Earth, and two similar satellites followed Landsat 1 into orbit — Landsat 2 and Landsat 3, which collected data through early 1983. By this time a new, improved model had joined the fleet — Landsat 4. Landsat 4 retained the Multispectral Scanner System, now augmented by the Thematic Mapper (TM). This new instrument improved maximum resolution to 30 meters (98 feet) per pixel, and also added a blue band (allowing true-color imagery), more precise near infrared bands, a shortwave infrared band, and a thermal band capable of mapping temperature. These new bands expanded Landsat’s abilities to map mineral deposits, different types of vegetation, and even urban heat islands.
These images highlight the capabilities of the Thematic Mapper’s 7 spectral bands. They were collected by Landsat 5 over Dunhuang, China on September 30, 2011. Clockwise from left: red, green, and blue (so-called “true-color”, similar to what a human would see from the same vantage point ); near infrared, red, and green (this band combination highlights vegetation, which appears bright red); shortwave infrared, near infrared with a slightly longer wavelength, and green (these wavelengths help show different minerals and the presence of water), and thermal infrared (temperature — lighter areas are warmer than darker areas). Data courtesy USGS/NASA, processing by Planet.
Landsat 5 joined Landsat 4 in orbit on March 1, 1984, and became the longest-lived Earth observation satellite ever, collecting data until January 2013, more than 28 years! Landsat 5’s longevity was in part because there was no other alternative — over the span of a few months in late 1993 Landsat 4 failed and the next-generation Landsat 6 was lost shortly after launch.
Over their long lifespans, Landsat 4 and 5 were able to document many dynamic events and monitor long term change. One surprising application involved using Landsat data to study the aerosols (tiny particles like water droplets, dust and soot suspended in the atmosphere) in the smoke plumes from burning Kuwaiti oil wells , ignited at the tail end of the first Gulf War.
Smoke from hundreds of burning oil wells obscure the Kuwaiti desert in this true-color image collected by Landsat 5 on July 1, 1991. Look closely and you can see individual fires. Data courtesy USGS/NASA, processing by Planet.
Taken nearly a decade apart, this pair of Landsat images documents the impact of mine tailings on the lowland forest of New Guinea. The tailings — waste from a high-altitude gold and copper mine — filled streams flowing from the New Guinea highlands to the Arafura Sea. The choked streams eventually overflowed, covering large swaths of dark green forest with gray tailings . The images were acquired by Landsat 5 on May 27, 1988 (left), and January 31, 1998 (right). Data courtesy USGS/NASA, processing by Planet.
As a new century approached, Landsat 7 joined the veteran Landsat 5 in orbit on April 15, 1999. This new satellite carried the Enhanced Thematic Mapper Plus (yeah, NASA has a knack for names) with a 15 meters (49 feet) per pixel panchromatic band, a higher-resolution 60 meters (200 feet) per pixel thermal infrared band, and improved signal to noise. It also had features to help calibrate the instrument, ensuring the data would remain consistent over time.
Augmenting the improved capabilities of the Enhanced Thematic Mapper Plus instrument, Landsat 7 added onboard data storage and better downlink capability. This allowed more imagery to be collected than ever before, especially in areas far from a ground receiving station. More data covering a larger amount of the Earth’s surface led to the creation of several global- and continental-scale datasets, including a high resolution mosaic of Antarctica, the first-ever global map of coral reefs, and a map of the global distribution of mangroves.
The Landsat Image Mosaic Of Antarctica (LIMA) combined more than 1,000 Landsat 7 scenes to create a seamless mosaic of the southern continent. Landsat 7’s onboard data recorder and high gain antennas allowed the collection of more data over remote areas than earlier satellites in the program. Image courtesy NASA GSFC Scientific Visualization Studio.
Onboard storage and improved bandwidth allowed Landsat 7 to deliver more data of the World’s most remote locations, like the Antarctic Dry Valleys. True-color image collected on December 18, 1999. Courtesy NASA Earth Observatory.
Around the time of Landsat 7’s launch, while crafting visualizations at NASA’s Earth Observatory , I began working with the data — which was a significant challenge with contemporary desktop computers! Even opening a file took enough time to grab a cup of coffee, and offloading the data from a CD-ROM (one scene per CD) created an opportunity to go grab lunch. But my teenage enthusiasm for looking at the Earth remained (still does!), and it was worth the hassle.
It was another 14 years before the Landsat program took its next step. After several false starts Landsat 8 (initially called the Landsat Data Continuity Mission ) was launched on February 11, 2013. It carries the Operational Land Imager (OLI) which features a new “coastal blue” band, a narrower near infrared band, a band designed to detect cirrus clouds (which are translucent in visible light), and a second shortwave infrared band. In addition, Landsat 8’s panchromatic band is limited to the wavelengths observed by the red, green, and blue bands. This helps true-color data sharpened with the panchromatic band appear more realistic, without the near infrared light present in Landsat 7’s panchromatic band.
A comparison of pan-sharpened true-color imagery from Landsat 7 (left) and Landsat 8 (right) collected on the same day a few minutes apart. The infrared light in Landsat 7’s panchromatic band makes water appear darker and vegetation lighter than similarly processed data from Landsat 8. Data courtesy USGS/NASA, processing by Planet.
The Operational Land Imager aboard Landsat 8 also digitizes data at a higher bit depth than its predecessors — 12 bits (4,096 distinct levels) instead of 8 bits (256 distinct levels) — which allows small changes in brightness to be recorded. Other improvements included enhanced dynamic range, so the data doesn’t saturate over bright surfaces like ice, snow, or clouds; and higher signal to noise that improves observations over dark surfaces like ocean water or dense vegetation.
The surface of the Patagonian Ice Sheet shines brilliant white in the summer sun in this Landsat 8 scene collected on January 30, 2021. Earlier generations of Landsat sensors would have saturated over the brightest parts of the ice sheet. Data courtesy USGS/NASA, processing by Planet.
Thermal observations from Landsat 8 are made by a separate instrument, the Thermal Infrared Sensor, which dropped from 60 meters (197 feet) per pixel to 100 meters (328 feet) per pixel resolution, but added a second band.
Landsat 8 continues the trend to acquire more data than earlier satellites. Landsat is able to downlink all the data it collects over land and coastal waters, leading to a nearly complete picture of Earth every 16 days. The satellite can even collect and broadcast a continuous strip of imagery along an entire day side orbit from Siberia to South Africa.
Just over a year ago, on September 27, 2021, the newest member of the Landsat family arrived. Landsat 9 is a near-twin of Landsat 8, with improved versions of both the Operational Land Imager and the Thermal Infrared Sensor. Together, the two operational Landsats provide coverage over the Earth’s land surface once every 8 days. Carrying enough fuel to last over ten years, Landsat 9 is expected to operate well into the next decade.
Comparison of true-color (top) and a single-band coastal blue image (lower) from the eastern end of Issyk Kul, Kyrgyzstan. Light in the coastal blue band penetrates water further than longer wavelengths, showing more of the lake bottom. Landsat 9 data collected on September 1, 2022, courtesy USGS/NASA, processing by Planet.
Not long after the launch of Landsat 8, I got the opportunity to move from NASA to Planet, trading billion-dollar missions for satellites built in a “clean enough” room. This transition gave me an appreciation of how both the federal government and Silicon Valley approach science and technology. Although the New Space industry is often described as a nimble alternative to NASA’s deliberate pace, many space startups utilize government data, including the long term Landsat archive.
Over 5 decades in orbit, the Landsat program generated a dataset that now provides the foundation for Planet’s constellation of satellites. The first few generations of Doves (our medium-resolution monitoring constellation) relied extensively on Landsat data for geolocation & orthorectification — the process of aligning each pixel of data precisely to its corresponding position on the Earth’s surface. Early mosaics even used Landsat imagery as a color target, to help merge data collected at different times of year and in all sorts of conditions into a seamless composite.
Planet basemap from July through August 2017. Landsat data provided the color target used to help create a consistent appearance. © 2017, Planet Labs PBC
For Planet’s business, Landsat data proved the viability of dozens of applications for Earth observation, from urban planning and agriculture to natural resources management and disaster response.
50 years of data provides a baseline to compare against — what is “normal” for a particular area? What’s the typical range of variability from day to day, month to month, season to season, and year to year? Have permanent, long-term changes occurred?
This record is essential to placing current observations in context. Context that is essential to deriving information from raw data, and meaning from information. How have the past 50 years of changes on the Earth’s surface shaped, and been shaped by, humans? What ecosystems do we need to preserve, and where should we develop new infrastructure? And — perhaps most importantly — how can we expect the Earth to change in the future?