T. Rousselin – MinesParisTech – GEO212 - France

Remote sensing has been an important part of space activities right from the beginning. The first earth observation image was sent by a meteorological satellite on April 1st, 1960. The first civilian remote sensing satellite (ERTS1, better known as Landsat1) was launched in 1972.

Today, a large array of sensors is available allowing the user to choose the best images for his application.
In this short overview, we will focus on three aspects: multitemporal capabilities, multispectral processing and spatial resolution.

 

Multitemporal Capabilities

One of the main user requirements is to have the right image at the right time. In fact, it is not that simple. Kepper's laws forbid to have a very high resolution sensor stationary at low altitudes over our heads.

Hence, there are either satellites on high altitude geostationary orbits allowing to take as much images as we want, but with fairly low spatial resolution, or lower orbit satellites with higher spatial resolution, but no continuous capability to observe one particular location.

The best example of geostationary earth observation is given by meteorological satellites. GOES, GMS, Meteosat ... allow to scan constantly the earth (5 satellites provide a complete meteorological survey) from an equatorial orbit (spatial resolution is best at the equator and decreases towards the poles). MSG (Meteosat Second Generation) gives a new image in each spectral band every 15 minutes with 12 channels (visible and InfraRed). Two satellites are operated right now by EUMETSAT.

This particular image on the right was taken on October 7th, 2008 over Sudan and Egypt (the coast line on the right is the Red Sea). The small bright object in this Thermal InfraRed channel is asteroid 2008TC3 exploding in the atmosphere over NorthEastern Sudan. The fireball appeared at 02:45 UTC and had faded away by 02:50 UTC. This exemple shows the power of very high revisit capability.

Of course, all natural and man made events do not need such high revisit times.

Here we have a very good example in our area of interest with the progressive evolution of Lake Chad between 1973 (image taken by Landsat 1) and 2001 (image taken by Landsat 7). The full Nasa Case study is available here.

For environmental applications it is very important to have regular surveys (every year, every semester, every particular season, every month, every week, every day) whose frequency depends on the phenomenon you analyse.

A large array of sensors is available. It starts with daily coverage in intermediate resolutions (1 kilometer to 100 meters) with instruments like Vegetation (on Spot 4 and 5 satellites), MODIS (on Terra and Aqua satellites) or NPP Suomi (with its extraordinary night imagery). Those images are generally available in near real time (a few hours after they were acquired) on dedicated portals like WorldView for Modis or NPP or through web services on Google Earth.

Satellites offering better spatial resolutions like the new Landsat 8 cannot image every day a particular area of interest but may provide a few images every year to measure slower evolutions.

On an area like Lake Chad, it is easier to have a good image during the dry season rather than in the middle of the rainy season (when most images are cloudy). But even during the dry season, sand winds may alter the image. Hence, even if the sensor has a frequent revisit capability, the end user will have generally to settle with a more limited useful time sequence.

Engineers found a way to have very high revisit with very high resolution. The first way to do it is to have a constellation of satellites. This has existed for a long time in the military field of remote sensing but over the past few years we have seen civilian constellations launched. Each satellite cannot revisit the same area of interest everyday, but the combination of the different satellites gives users that capability. The RapidEye constellation launched in August 2008 relies on 5 satellites. The Italian Cosmo-Skymed radar constellation relies on 4 satellites, providing daily coverages at the Equator but much better revisit times towards the polar areas (every 6 hours over the Mediterranean Sea). When the French Pléiades constellation of two satellites was operational (two satellites on the same orbit phased 180° apart), giving Airbus Defence & Space a little competitive edge, Digital Globe, its main US competitor, changed in September 2013 GeoEye-1 (launched in 2008) orbit to have the exact same constellation effect with its WorldView-2 satellite … with a slightly better spatial resolution.

Another way to have a daily revisit with a single satellite is to place it on a geosynchronous orbit. It does not allow you to image the whole planet but only the same 14 orbits every day. But if you have always the same areas of interest (your country or your best friend country) it is fine.

A good example is the Formosat-2 satellite launched in 2004 by the Taiwanese Space Agency (NSPO). Originally designed to image the Chinese Sea (between Continental China and Taiwan Island) it also images western Chad every day. On the example on the right, the Chadian Doba oil fields have been imaged for 7 consecutive days. It gives the end user security because if clouds are most of the time over the area (because of the season or because you're in England), there is a good chance to have at least one clear image of the target over 7 days. It also allows (through image processing) to use the power of multitemporal sequences.

A last way to get excellent temporal coverage is to rely on the performance of state of the art very high resolution satellites in terms of agility. In this exemple on the left we see successive images taken by the WorldView-2 satellite over London during 2 minutes. In fact the first image is taken when the satellite is approaching (looking forward), the middle image is taken when the satellite is above the site, and the last image is taken looking backwards. It is possible to appreciate that when you concentrate on the shape of the buildings.

Already in a few applications, different satellites or constellations from different companies or countries can work together for a specific purpose. It has been common practice since 2000 in crisis situations like catastrophies (earthquakes, floodings, forest fires...) where the International Charter on Space and Major Disasters coordinates satellite acquisitions, allowing rescuers to get imagery products as soon as possible, no matter which particular satellite or sensor is involved. The Charter work has been examplified during the January 2010 Haiti Earthquake or the March 2011 Japanese Tsunami / Earthquake

In the future, the combination between satellite performance (agility), power of constellations, and the capability to exchange data for multiple applications, should allow more users to have access to timely data. 2014 in this regard is the turning point with the operational start of two new constellations, SkyBox Imaging (with 15 Skysat satellites scheduled) and Planet Labs (with 28 Doves satellites).

 

Multispectral Processing

Image sensors scan the earth in various spectral bands. Each object at the surface of the earth reacts differently to a spectral band (we use the term of object's spectral signature). Hence, through image processing, it is possible to combine the best spectral bands for a particular object or phenomenon. Of course to represent them visually we will have to use the three basic colors Red, Green and Blue (RGB).

This small exercice uses Landsat 7 ETM+ spectral bands (Visible, Near Infra Red and Middle Infra Red in this example):

  • You can first look at the original images for each spectral band (click on band 1, band 2, band 3 or band 5 in the left column). They will appear as black and white images coded on 256 levels (8 bits). You will notice reflectance difference between those original bands (based on objects properties).
  • Then, you can look the result of various color combinations using RGB or CMY (Cyan Magenta Yellow) representations or even a different color space IHS (Intensity, Hue, Saturation). Click on the combinations on the first line.

On those examples you will particularly look at various objects : the reflectance difference between various water bodies (e.g. : the Chari river and its affluent), the particular reflectance of the large city of Sarh, the aspects of sugar cane irrigated fields (circular patterns in the lower right part of the images).

Spatial Resolution

Spatial resolution has always been the main focus of the media and public interest ("Can you read a newspaper from space?"). In fact, it is both linked to the performances of your observation instrument and the orbit you choose. If you fly very low, spatial resolution is better, but earth attraction is very important. This simple dilemma explains why in the early days in the sixties, military earth observation sensors could reach 12 m spatial resolution, but the satellite would not stay very long in orbit. The first civilian terrestrial earth observation satellite ERTS-1 had 80 m resolution. Today commercial systems reach 40 cm (and defense systems do better).

Choosing the best spatial resolution for a particular application is not always easy because it depends first on the type of objects one has to discriminate (with different levels : detect, recognize, identify). But it is also linked to other factors. Very High Resolution instruments are very sharp but they have small field of view (e.g. a single VHR image cannot image the city of London). And sometimes, spectral and / or temporal resolution have greater value than pure "pixel size".

Multisensor Capabilities

Creation of Digital Elevation Models with TanDEM-X (1 mn)
DLR TanDEM-X Presentation (4 mn)
Spot Image satellite constellation presentation (3mn)

As we saw over the previous topics, many sensors are available allowing a continuous array of observation tools with different resolutions, fields of views, spectral bands, temporal revisit capabilities... The user will have to decide which image better suits his needs :

  • spatial resolution vs spectral resolution
  • spatial resolution vs field of view
  • highest geometric quality allowing to locate perfectly ground truth GPS based measurements
  • adequate revisit capability for the application
  • Image quality and metadata completion
  • ...

 

Depending on his application, his requirements and his budget, the user will rely on free imagery available through virtual globes (like Google Earth or Bing), will download open data collected for scientific applications and made available to the general public (like the Glovis or GENESI DR sites), or buy archive imagery or task image acquisition from commercial operators (like Digital Globe in the US or Airbus Defence & Space, e-Geos or DMCII in Europe).

 

New capabilities

Over the past year, the remote sensing capabilities were greatly improved, especially with the emerging of innovative small satellites like Skysat-1 which is able to provide a multispectral image at a resolution of 2m and a 90 seconds video at approximately 1m resolution, Urthecast, assembled on the ISS which can provide a 5m image and a 60 seconds video or the Flock-1 constellation formed by 28 dove satellites launched from the ISS. 

 

FURTHER READINGS

 

As a complement to the documents available on our course site,other useful links are proposed (links have been verified on March 4th, 2013).

 

Remote sensing courses:

 

Space Agencies:

 

Image Providers:

 

 

Last Updated ( Saturday, 15 March 2014 )