Spectral resolution is defined as the width of the region of the EM spectrum recorded as one 'band'. Typically this band width is related to the number of wavelength bands recorded by the sensor. Modern sensors can be divided into three broad classes based on their spectral resolution:

• panchromatic (one very wide band, typically covering the visible spectrum)
• multispectral (several coarse bands, typically limited to 15 or fewer bands)
• hyperspectral (many discrete, fine bands, typically far more bands than a multispectral system)

In today's lecture I'll provide an overview of spectral resolution and provide examples of the three broad classes. Recall that in Lecture 3 I reviewed the names for each of the different spectral regions.

There are several informative web sites that should help you understand how spectral resolution varies in different satellites, and why satellites are designed with different spectral resolutions.

• Overlaying spectral curves on sensor bands (and a description of the MODIS bands).
• Definitions: Signal-to-noise ratio (SNR), radiance, noise-equivalent temperature (NET)
• A library of spectral reflectance curves.
• False colours (i.e., band combinations).
• An explanation by NASA of why satellite images are different from photographs (goes into false colours and band combinations).
• A description of the Gambit KH-8 satellite (aka "Key Hole")--a US photographic surveillance system.

Text: Chapter 2.5.1: Sensing properties