What is remote sensing?
Just now you are actively engaging in the process of remote sensing. As you read this text, your eyes are collecting electromagnetic radiation emitted from your computer monitor, which your mind interprets as images. Remote sensing is thus precisely what the name implies: it is the process of sensing the environment from a distance. In the geosciences, remote sensing can be a very powerful tool. Within the discipline, remote sensing most often involves collecting data about the surface of the Earth using sensor systems located some distance above the ground, usually either mounted on an airplane or satellite.
While data about the Earth may be collected on the surface in a process known as in situ data collection (essentially “in the field”), quite often this form of data collection is infeasible over large areas. Remote sensing allows the data collector to view a large portion of the Earth’s surface at once, without the taking ground measurements at a large number of sample points. For example, if someone wanted to map the land cover within a county perhaps 200 square kilometers in area, it would be nearly impossible to take enough measurements on the ground to get an accurate picture of the entire county. Such an endeavor would simply take too long and be far too expensive. By taking remotely-sensed images from a plane or from space, and then performing a land cover classification based on the data, it is possible to identify the land cover throughout the entire county in an efficient and effective manner.
The advantages of sensing remotely
In addition to saving time, money, and other resources, remote sensing offers a number of additional advantages. First, because the data is collected at a distance, there less change that the process of data collection will interfere or disturb the environment or phenomena that is being studied. Passive remote sensing, in which the sensors are simply recording the incoming data, are particularly unobstrusive. In in situ data collection, where measurements are taken in the field, the person doing the measuring quite often has to transverse and interact with the environment, which may not only disturb the environment of interest, but also affect the data being collected.
Second, remote sensing makes it easy to collect data systematically, as most sensors collect data within specific frames or along a single line. Data collected in the field can also be done systematically along transects. This, however, is often difficult as there exist many physical and political barriers to data collection at ground-level. Remote sensing therefore automatically and effortlessly minimizes the sampling bias that might occur during in situ data collection. Third and finally, remote sensing can detect certain biological and physical data, such as exact x,y,z locations, temperature, moisture, biomass, and luminosity, that is difficult to measure in situ, especially over large areas.
One might get the impression at this point that in situ data collection is useless or unnecessary. The reality is, however, that data collected in situ plays a critical role in the remote sensing process. Data collected in situ is often used to calibrate the remotely-sensed data, and to perform accuracy assessments of final data product. In classifying land cover, for example, it may be necessary to first collect in situ data regarding the spectral reflectance of different land cover types, such as grass, trees, corn fields, roads, etc. In doing so it is possible to tell more accurately that a specific measurement represents a particular land cover type.
Once the land cover classification has been done, in situ data collection may again be performed to determine the accuracy of the classification. An accuracy assessment may be performed by choosing a certain number of random points, visiting them on the ground in situ, and determining how many of the points were accurately classified.
In remote sensing, there are four important resolutions to consider: spectral, spatial, radiometric, and temporal.
Most remote sensing instruments measure the amount and type of incoming electromagnetic radiation (or light). One defining feature of light is its wavelength, or the distance from one peak of the light wave to another. Visible light, for example, as a wavelength from about 400 nanometers (violet) to 700 nanometers (red). Wavelengths just below the visible spectrum are called ultraviolet, while wavelengths just above the visible spectrum are appropriately called infra-red. Today, there are instruments capable of detecting wavelengths of light all along the electromagnetic spectrum (below).
Spectral resolution refers to the sensitively of a remote sensing device to certain wavelengths of light. Remote sensors are generally sensitive to a particular range of wavelengths and can differentiate between a certain number of wavelength intervals known as bands or channels. Sensors that can distinguish multiple bands are known as multispectral. The Landsat Multispectral Scanners (MSS), for example, could record data in four separate bands including the visible light bands green and red, and two near-infrared bands.
There are also hyperspectral remote sensing instruments sensitive to hundreds of spectral bands. The AVIRIS sensor, for example, in sensitive to 224 different bands between the 400 and 2500 nanometer wavelengths. Instead of a standard 2d image, the AVIRIS produces what is known as a hyperspectral datacube, with the z-axis refering to the 224 data bands.
Spatial resolution is a measure of the minimal distance between any two objects that can be distinguished using a particular sensor. All other factors equal, the better the spatial resolution, the more detailed or clear the image becomes. For digital sensors, the spatial resolution is apparent in the data product’s pixel size. Pixels, short for “picture elements” are squares whose width represents the spatial resolution of the image, such as 10 x 10 meters or 1.5 x 1.5 feet. The Landsat Enhanced Thematic Mapper (ETM+), for example, has a spatial resolution of 30 x 30 meters for its six multispectral bands. Thus, when viewing these images on the computer screen, each individual pixel represents 30 meters on the ground, and 900 meters square.
While it may seem advantages to always use the highest spatial resolution possible (in order to resolve more detail), it is important to consider that higher spatial resolutions result in more data, which requires both more memory for storage and more computing power for processing. In many instances it may not be necessary or desirable to use the highest spatial resolution available. The proper spatial resolution is often dictated by the scale of study; if you need to know the exact number of trees in a given area, for example, your spatial resolution should be no more than one-half the width of the tree’s canopy, perhaps no more than 2 meters.
Radiometric resolution is a measure of a remote sensing device’s ability to finely measure quantized amounts of energy that is reflected, emitted, and back-scattered from the surface of the Earth. Sensors with a higher radiometric resolution can distinguish a wider range of energy intensities, similar to using a ruler with additional, smaller tick-marks to make more accurate measurements. Radiometric resolution is usually defined in terms of bits. Launched in 1972, the Landsat 1 Multispectral Scanner has a radiometric resolution of 6 bits. By 1984, Landsat 5 had a resolution of 8 bits. More recently, the Quickbird and IKONOS sensors each had a resolution of 11 bits. The higher the radiometric resolution of the sensor, the more likely the object or phenomena of interest will be sensed accurately.
Finally, temporal resolution refers to the amount of time between data collection. Satellites that orbit the Earth, for example, may only pass over a particular location every 16 days or more. Other imagery may only be obtained a few times a year, and therefore has relatively low temporal resolution. Temporal resolution is very important for any investigation that seeks to understand how change occurs over time. High temporal resolution imagery is used in meteorology to predict changes in the weather and track potentially dangerous storms. Satellites in geostationary orbit, like NOAA’s Geostationary Operational Environmental Satellites (GOES), can collect data every half hour.