The concept of a "spectral signature", another name for a plot of the variations of reflected (or absorbed) EM radiation as function of wavelengths, gives rise to the widely used approach to identifying and separating different materials or objects using multispectral data obtained by remote sensors.
Electromagnetic Spectrum: Spectral Signatures
For any given material, the amount of solar radiation that it reflects, absorbs, transmits, or emits varies with wavelength. When that amount (usually intensity, as a percent of maximum) coming from the material is plotted over a range of wavelengths, the connected points produce a curve called the material's spectral signature (spectral response curve). Here is a general example of a reflectance plot for some (unspecified) vegetation type (bio-organic material), with the dominating factor influencing each interval of the curve so indicated:
This important property of matter makes it possible to identify different substances or classes and to separate them by their individual spectral signatures, as shown in the figure below. *
For example, at some wavelengths, sand reflects more energy than green vegetation but at other wavelengths it absorbs more (reflects less) than does the vegetation. In principle, we can recognize various kinds of surface materials and distinguish them from each other by these differences in reflectance. Of course, there must be some suitable method for measuring these differences as a function of wavelength and intensity (as a fraction [normally in percent] of the amount of irradiating radiation). Using reflectance differences, we may be able to distinguish the four common surface materials in the above signatures (GL = grasslands; PW = pinewoods; RS = red sand; SW = silty water) simply by plotting the reflectances of each material at two wavelengths, commonly a few tens (or more) of micrometers apart. Note the positions of points for each plot as a reflectance percentage for just two wavelengths:
In this instance, the points are sufficiently separated to confirm that just these two wavelengths (properly selected) permit notably different materials to be distinguished by their spectral properties. When we use more than two wavelengths, the plots in multi-dimensional space (3 can be visualized; more than 3 best handled mathematically) tend to show more separability among the materials. This improved distinction among materials due to extra wavelengths is the basis for multispectral remote sensing (discussed on page I-6).
I-11: Referring to the above spectral plots, which region of the spectrum (stated in wavelength interval) shows the greatest reflectance for a) grasslands; b) pinewoods; c) red sand; d) silty water. At 0.6 micrometers, are these four classes distinguishable? ANSWER
I-12: Which material in these plots is brightest at 0.6 micrometers; which at 1.2 micrometers? ANSWER
I-13 Using these curves, estimate the approximate values of % Reflectance for rock (sand), water, and vegetation (choose grasslands) at two wavelengths: 0.5 and 1.1 micrometers, putting their values in the table provided below. Then plot them as instructed on the lower diagram. Which class is the point at X in this diagram most likely to belong? (Note: you may find it easier to make a copy of the diagram on tracing paper.) ANSWER
I-14: Presume that two unknown surface features in an image or photo, which actually are a forest and a field crop with the plants close-spaced, are measured for their spectral values, and both display quite similar reflectances at three chosen wavelengths. How might these be separated and perhaps even identified? (Hint: think spatially.) ANSWER
Spectral signatures for individual materials or classes can be determined best under laboratory conditions, where the sensor is placed very close to the target. This results in a "pure" spectral signature. But what happens if the sensor is well above the target, as when a satellite remote sensing device looks down at Earth. At such heights the telescope that examines the scene may cover a large surface area at any moment. Individual objects smaller than the field of view are not resolved (this is akin to spatial resolution limitations). Each object contributes its own spectral signature input. In other words, for lower resolution conditions several different materials/classes each send (unresolved) radiation back to the sensor. The resulting spectral signature is a compound of all components in the scene. Analytical techniques (e.g., Fourier analysis) can extract individual signatures under some circumstances. But the sampled area (corresponding to the pixel concept introduced on the next page) is usually assigned a label equivalent to its dominant class). This integration of several signatures is inherent to the "mixed pixel" concept examined in the bottom half of page 13-2.
* The principles of spectroscopy in general, as well as a survey of imaging spectroscopy and hyperspectral remote sensing, are explored in greater detail in Section 13 (pages 13-5 through 13-10). Also treated in that part of Section 13 is a brief review of the concept of "spectral resolution".