Using Landsat imagery a practical How-To guide Intended audience: This one isn't intended to be a guide for GIS and/or remote sensing professionals: they already know all this stuff, and may easily find this document absolutely superficial, inaccurate and too much simplistic and approximate. This short guide is intended for a generic and absolutely prophane audience: I'm thinking of lots of peoples genuinely interested for one reason or another in using satellite imagery for their practical purposes, but lacking any specific academic or practical background. i.e. someone moving his first steps on this complex ground, but curious and willing to learn. Google Earth (... Google Maps and other similar web services ...) had the big merit to put a strong and widespread attention on geography, maps, satellite imagery and so on. A new generation of webmasters, software engineers and other professionals interested on GIS related technologies for several different scopes is quickly growing: I hope this short guide can help all them to get a better comprehension and a deepest knowledge ... … and to get started on their own activity in the quickest time as possible.

Intro: NASA started the Landsat program in 1972: since then these satellites year after year have accumulated an impressive amount of high-quality Earth images. And that's not all: the USGS EROS project make the whole Landsat dataset to be freely available for everyone. You are simply required to fill a sign-in form in order to get your own username and password, and then you can immediately start any download, at absolutely no cost. And you are absolutely free to use, redistribute or rearrange Landsat images at your will. As USGS states: There are no restrictions on Landsat data downloaded from USGS EROS, and it can be used or redistributed as desired. A statement of the data source when citing, copying, or reprinting USGS Landsat data or images is requested; details are on the EROS Data Citation page. Accordingly to the above statement you are absolutely free to produce by your own an high-quality Landsat-derived map of your Country / State / Region then using this map for e.g.: • if you are a webmaster, you can implement a Web site supporting map based contenents • if you are a teacher, you can use the map for didactic and you can redistribute the map to your students (e.g. copying the images on a CD on pen drive) • if you are an industry, you can freely incorporate the map into your products A short history: The most recent (and advanced) Landsat satellite is Landsat 7, and was launched on 15 April, 1999. Unhappily in May 2003 an unrecoverable fault occurred, severely impairing the satellite operations: NASA and USGS still continue to offer Landsat 7 imagery (so called SLC-off mode), but using such kind of imagery is strongly discouraged. So we are forced for practical reasons to use the full-quality Landsat 7 dataset (SLC-on mode), which is only available for the period April 1999 – May 2003: this is a little bit annoying, but is not at all catastrophic.

A few technical details: A Landsat 7 image (or scene) isn't at all an ordinary picture, as you can get by yourself using a digital camera. Landsat 7 adopts a most complex layout, so you cannot directly use any downloaded Landsat 7 image. Some kind of post-processing is absolutely required in order to get an usable image: so at least a moderate degree of technical understanding is required. 1. Each Landsat 7 scene you'll get by download physically consist in a single .tar.gz file • if you are a Linux user, no explanation is required (at least, I hope well …) • if you are a Windows user some further explanation is probably required: • this is a compressed file: you can use several tools in order to decompress it: my personal suggestion is to use the open source 7-Zip [http://www.7-zip.org/] • after decompressing the .tar.gz file, you get an archive file .tar • and finally you can extract a directory (aka folder) from the archive file: you can use 7-Zip for this task as well. 2. NASA and USGS adopts a quite complex (and obscure) naming convention to identify each scene ID: basically the orbit position and the date-time are encoded within the directory and the file names. Anyway they are strictly related, so after a short time you'll become familiar with this odd naming schema. 3. I strongly suggest you to keep each single scene within its own directory (aka folder), because this helps a lot in maintaining a proper order when you have to handle lots and lots of different scenes (a very common case). 4. Several distinct GeoTIFF images are contained within the same scene directory: each one of them represents a determinate band (i.e. the output of a single spectral sensor). ● ● ● ● ● ● ● ●

Band #1 (blue) identified by suffix _B10.tif Band #2 (green) identified by suffix _B20.tif Band #3 (red) identified by suffix _B30.tif Band #4 (near infrared) identified by suffix _B40.tif Band #5 (mid infrared) identified by suffix _B50.tif Band #6 (thermal infrared) identified by suffixes _B61.tif and _B62.tif Band #7 (mid infrared) identified by suffix _B70.tif Band #8 (panchromatic) identified by suffix _B80.tif

Directory LE71930312002168EDC00 L71193031_03120020617_B10.TIF L71193031_03120020617_B20.TIF L71193031_03120020617_B30.TIF L71193031_03120020617_B40.TIF L71193031_03120020617_B50.TIF L71193031_03120020617_B61.TIF L71193031_03120020617_GCP.txt L71193031_03120020617_MTL.txt L72193031_03120020617_B62.TIF L72193031_03120020617_B70.TIF L72193031_03120020617_B80.TIF README.GTF

the standard Landsat 7 directory/scene layout

● For bands #1,#2,#3,#4,#5 and #7 a single pixel corresponds to a square cell of 30m x 30m on the terrain ● Band #6 adopts an half-size resolution, so a single pixel corresponds to a square cell of 60m x 60m on the terrain ● Band #8 adopts an enhanced resolution, so a single pixel corresponds to a square cell of 15m x 15m on the terrain We'll ignore at all any infrared band: they are really useful for scientific purposes, but are quite useless in order to produce realistic (picture like) images. But on the other side we must take note that panchromatic band #8 offers an very interesting high resolution: yes, it's B&W (not so glamorous as true-color is), but anyway this can be very useful in many cases. Technical hilights: A GeoTIFF image simply is an ordinary TIFF image containing some extra tags used to encode a Spatial Reference System and related coordinates. So a GeoTIFF image is self-conscious about spatial (geographic) positioning. You can directly visualize a GeoTIFF using quite any ordinary picture tool, but you must aware that when attempting to edit a GeoTIFF image using i.e. Photshop or Gimp you'll unrecoverably lost any related spatial infos. Don't be surprised if any Landsat 7 image will be shown as grayscale: they actually are grayscale (i.e. each pixel is represented by a single byte expressing an intensity value ranging from 0 to 255). Any subsequent color interpretation is merely conceptual, not factual. And don't be surprised if Landsat 7 images looks slant and surrounded by large black areas: • Landsat 7 follows a circumpolar orbit, so the satellite path isn't exactly meridian aligned, mainly because the Earth rotates while the satellite goes along its own orbit. • any black pixel (i.e. zero) identifies a so called NoDataValue, i.e. a zone falling outside the sensor coverage. Finally, you must not assume that simply gluing all together the Red, Green and Blue band you can get an useful RGB image: that's not so simple. Some kind of post-processing is absolutely required (i.e. color normalization and recalibration) in order to get a good quality image.

The Landsat 7 orbit

A typical unprocessed Landsat single band scene: not so exciting, that's true ?

Things doesn't go much better even after reassembling an RGB image, that appears to be very dark. And colors aren't at all brilliant as you can expect: this one surely represents the kind of image none of us is willing to use.

Finally this one is a properly post-processed Landsat 7 scene. Wow, that's a completely different thing, really … but in order to achieve this final result you must continue to patiently read this tutorial.

Please note: all the above examples represented reduced scale images (a whole scene thumbnail). But the intrinsic resolution supported by Landsat imagery (30m per pixel) actually is the one exemplified by this full sized detail.

And that's not all: this full sized detail exemplifies what you can actually expect to get when using the high-resolution (15m per pixel) Panchromatic sensor (B&W, band #8).

Key Points: • Processing Landsat imagery by your own isn't a too complex task. A basic computing background is enough to start. No specific skill or knowledge is strictly required. • Anyway, this specific activity requires a good dose of time, care and attention. • Satellite imagery is huge: accordingly to this, download times are quite lenghty, and a big amount of free disk space is surely required. • In order to get optimal results, a patient approach by trial and error is usually required. So, if you are a guy of the fire and forget kind ... is probably better you'll waste your time in some different activity. • A sound familiarity in using shell commands surely helps a lot (no GUI here, please). This can sound a bit strange speaking about imagery ... but processing huge amounts of image data is a task strictly related to data processing, and has very little to do with photofinishing and desktop graphics as usually intended.

Getting started: How-To download Landsat scenes EROS sign-in: This is the first step: you can freely download any Landsat imagery from USGS-EROS, but you are requested to sign-in using your own username and password. The registration process is quite simple and easy: just answer to any required question, and then submit your registration form. Take careful note of your username / password for any further use. Bulk download: Usually you'll need several Landsat scenes in order to compose your own map: in other words, it's very alike your map would require to create a mosaic of several overlapping scenes. How many scenes are requested obviously depends on the extension of the Country / State / Region you are interested in. Just as an example, covering Italy requires several tens distinct scenes (about thirty), for a total download size of several GBs. The bulk download tool is really useful because using it you can submit more request in a single time: http://landsat.usgs.gov/Landsat_Search_and_Download.php Anyway, submitting a Bulk Download request isn't a so simple process: there are thousands and thousands of different Landsat scenes, so we must first check out the ones we really intend to download. Using GloVis to select Landsat scenes: The USGS offers a good tool [GloVis] you can use to browse the Landsat archive. GloVis simply is a Java applet, so you can use this application on any modern web browser without any further complication. http://glovis.usgs.gov/

First of all you must select Landsat 7 SLC-on (1999-2003) [Glovis supports other satellite datasets as well]. Please note well: don't use the default selected Landat 7 SLC-off option: these scenes where taken after Landsat 7 suffered the infamous SLC fault, and are quite useless for our purposes.

If you are interested on checking out the whole coverage for your Country activating the Country Boundaries option is surely advisable: but if you are interested on a more limited State / Region coverage you can activate the Admin Boundaries option instead. Using GloVis is quite intuitive and immediate. Just a few hints: • use the left / right / up / bottom arrows to locate a scene on the space. • then use the Prev. Scene / Next Scene buttons to check the available dates: i.e. you can now locate the current scene at a given time. • please note well: not all scenes offers the same real quality. ◦ avoid at all to select scenes taken on bad weather, because clouds will obviously hide any terrain detail ◦ don't be afraid when you'll see scenes showing odd colors on preview: they simply need a thorough post-processing. ◦ if you are attempting to build a mosaic of several different scenes, you must carefully check to select all scenes taken in the same month / season. Otherwise you'll get awful looking sudden color changes when changing from a scene to the adjacent one. ▪ in other worlds, it's safest to overlap a scene taken on June 2000 and a scene taken on June 2002 (same vegetation stage). ▪ on the other side overlapping a scene taken on February 2002 and another one taken on August 2002 will surely produce a very poor result. ▪ just as a suggestion, May-June is the most advisable period of the year (northern emisphere, of course), because the weather conditions are usually clear, and the vegetation shows brightest colors.

GloVis supports an useful tool: Scene List When this tool is activated a small popup panel shows the list of any scene you've selected since now, and allows for easy reviewing the whole selection. You can also consult the metadata for any selected scene, you can display a preview of each scene, you can locate a single scene on the map, and so on … A little detail: some scene displays a bag-like symbol following the ID: this means that this scene is available for immediate download. i.e. someone else has already requested the same scene in the past. Scenes doesn't showing this symbol are available as well, but they had to be prepared first, and this introduces some (little) more complexity in the process. But we'll explain better this step in the next tutorial section.

All right: once you've selected any interesting scene (and carefully revised your selections), you can export your scene list. This actually simply means that GloVis will create somewhere (at your choice) a plain text file, enumerating any scene ID you've selected since now. GloVis Scene List sensor=L7 SLC-on (1999-2003) LE71920281999281EDC00 LE71920282000044EDC00 LE71920282000316AGS00 LE71920282001046EDC01 LE71920282001238EDC00 LE71920282001286EDC00 LE71920282001334EDC00 LE71920282002017EDC00 LE71920282002113EDC00 LE71920282002353EDC00 LE71920282003052EDC00

Such text file exactly corresponds to a request you can send to the Bulk Download tool (from where we started). You simply have now to open this file (you can use any trivial text editor), deleting the first two lines (i.e. the title), and then this request file will be ready to be submitted to Bulk Download.

Bulk download (really): All right, this time we can actually upload a valid request file (the one produced in the above steps).

Do you remember that using GloVis a curious bag-like symbol marked some scene ID ? All right, here is the full explanation. • The scenes in the green panel are immediately available to be downloaded. Someone else already requested the same scene in the past. • The scenes in the red panel aren't available to be downloaded. You are the first people on the World interested in these specific scenes. (you can be really proud of this …) ◦ technically speaking, USGS has yet to perform some processing on this scenes before they can be offered for download. ◦ so you have to submit a formal request to activate this process (don't be afraid, it's completely cost-free). ◦ go to Shopping Basket, fill any requested form and submit your request. ◦ when the datasets will be ready for download you'll get an e-mail notification. Although USGS fixes a (very prudential) three or four days deadline, usually you can expect that your request will be satisfied just in one or two hours, but sometimes you must be patient and wait one day or two … it depends. Using Landsat 5 imagery: There is another good source for Landsat imagery: the venerable Landsat 5 (launched on 1984) is fully operative and still continues producing images you can actually request and download using GloVis. Landsat 5 is not affected by the infamous SLC-off failure plaguing Landsat 7, so Landat 5 imagery is the only full-quality dataset available for years 2003-2010. Landsat 5 and 7 are really different spacecrafts (using different generation technologies): anyway they are quite similar in overall layout. You can safely mix scenes coming from both sources. Anyway a paramount difference exists: Landsat 5 doesn't supports Band #8 (panchromatic). Only RGB bands (#3, #2, #1) are supported, so the unique available resolution is 30m.