Lisbon - Portugal

Summary
For the last 41 years, Landsat satellites have been looking at the Earth!
The objective of this 2 day workshop is to introduce participants to the understanding and analysis of Landsat imagery products (Landsat 5 - TM, through Landsat 8 - OLI sensors) and their use in a context of landscape ecology – mapping of ground units.
We will use the latest version of the IDRISI software (IDRISI Selva vs. 17.02) to progress from imagery selection through image classification using both hard (automatic and supervised) and soft classifiers, to the production of basic thematic maps and their evaluation in terms of accuracy assessment. Time changes and balances in land cover will also be assessed.
Language and concepts presented will be kept at a basic level to ensure that first-time users will benefit from the workshop and gain competences that will allow them to further explore and progress in the field of remote sensing of environment in their professional context.
 

Syllabus
The course will be organized in 4 sessions: 9h30-13h00 and 14h00 to 17h30 on each day.
 

Session 1:
An introduction to Remote Sensing of the Earth surface:
• Digital images – what is it all about? Pixels and DN matrixes; image perception; from an object to its image;
• From balloons, kites and planes to spacecraft – what phenomena and scales, in space and time can we tackle;
• Orbital platforms and instruments/sensors – looking at Earth phenomena from space; land and ocean, two worlds apart;
• Different space, time and spectral resolutions – their applicability and corresponding issues in the field of (landscape) ecology;
• Availability and cost of some of the most common imagery products;
• The Landsat program – background and present-day products; comparing, retrieving and downloading Landsat imagery.

Session 2:
The IDRISI Selva software (vs. 17.02)
• Importing digital images into IDRISI Selva – imagery suppliers and formats; spectral bands recognition and selection; instruments comparison throughout the Landsat program: LS5/TM, LS7/ETM+, LS8/OLI + TIRS;
• Visualizing digital images - general commands; using the composer to control the display of layers and to overlay images; querying an image or an image group; using, creating and editing palettes; radiometric resolution of the image and color resolution/rendition – how much radiometric resolution can we display and analyze?
• DNs distribution in a digital image – image histograms and reflectance classes; how do reflectance classes translate into the terrain; masking submerged vs. non-submerged areas; different approaches and corresponding results; superimposing and comparing results;

Session 3:
Image classification – from spectral information to ecological/terrain significance
• Spectral signatures – spectral resolution in a digital image; spectral ranges and corresponding terrain information; from DNs to reflectance/radiance;
• Developing interpreted maps from remotely sensed imagery – comparing spectral signatures and grouping pixels; measuring spectral similarity between pixels; reflectance patterns vs. landcover units; hard classifications;
• Unsupervised classification – segmenting an image based on its pixels spectral signatures; distinctive reflectance patterns and how much do different approaches influence the resulting landcover patterns;
• Supervised classification – comparing pixels to pre-assumed/defined reflectance patterns; probabilistic vs. raw approaches; understanding and managing non-classified areas.

Session 4
How good is a classification – measuring its quality/accuracy
• Understanding and constructing an error matrix – the producer and the user perspectives; overall and class/unit accuracies; user and producer accuracy evaluation and perception;
• Ground truthing points – going to the field? When? How to position field observations?
• Fuzzy classifications – how likely is a pixel to belong in any given landcover class; managing and comparing results;
• Wrap-up of the sworkshop.