Aug 11, 2009

Remote Sensing with Eight Spectral Bands

Estimated Article Reading Time: 5 min.

The launch of a new satellite is an exciting event, as it represents the culmination of several months or years of hard work and implementation of new technology. Due to launch October 6, 2009, DigitalGlobe’s WorldView-2 promises to be no exception.

In a webinar last week, Dr. Kumar Navulur, principal scientist at DigitalGlobe, discussed the satellite’s 8 new spectral bands and how they will impact different industries and sciences. WorldView-2 is said to be the first high resolution 8-band multispectral commercial satellite, giving it the ability to produce better feature classification and identification, improved bathymetry applications and change detection.

Navulur said that typically anything less than 10 bands is considered multispectral. WorldView-2 will have 8 bands and a panchromatic band. Superspectral satellites have 10-20 bands, and hyperspectral sensors have greater than 20 bands.

QuickBird satellites have multispectral plus panchromatic, WorldView-1 has just panchromatic.

For the hyperspectral bands, typically the very narrow bands are designed for very specific applications that can be as small as 5 to 10 nanometers in spectral width. “So if you want 120 spectral band, probably you’re looking at a total resolution of around 3 to 4 meters, whereas if you need ½ meter resolution then you probably need 20-30 spectral bands,” he said.

WorldView-2 has four additional spectral bands (coastal blue, yellow edge, red edge and near infrared 2) which facilitate a broader range of classification. More features can be extracted and tracking coastal changes and infractions is easier. The near infrared 2 does not have as much atmospheric noise as near infrared. The panchromatic band of WorldView-2 only goes through a visible area, it doesn’t penetrate the infrared area. The primary reason is the vegetation in black and white is more representative of the vegetation signature and does not include the high reflectance from a near infrared band.

The actual wavelength of the different spectral bands is as follows: QuickBird’s panchromatic band goes from 450 nanometers to 900 nanometers and so does that of WorldView-1. The spectrum range of WorldView-2 is 450-800 nanometers but is still in the visible region; the near infrared band 2 goes from 770 to 895. The additional bands in WorldView-2, especially the red edge, start at 705 and 745. This band has been specifically designed to capture the red edge position in vegetation.

The yellow band, also a key band for mapping vegetation, starts at 585 nanometers and ends at 625 nanometers. The coastal band before the blue band starts at 400 nanometers and goes through 450. The near infrared 2 band starts at 860 and goes through to 1040; the near infrared 2 band is comparable to one of the Landsat’s near infrared bands.

In remote sensing there are four dimensions, said Navulur, typically spectral, spatial, radiometric and temporal and the combination of these dimensions. The spectral filters on the WorldView-2 satellites are very accurate, and thus the information you can extract in this spectral band is very accurate. WorldView-2 can collect imagery at 46 centimeters “but we have license to sell only 50 centimeter imagery so I’ll refer to 50 centimeters as our spatial resolution,” Navulur explained. “We still keep the 11 radiometric resolution similar to QuickBird. If you compare this to 8-bit satellites like Landsat you see more radiometric information preserved in pixels as compared to the 8-bit bands. WorldView-2 improves a temporal dimension of the satellite.”

Spectral information content can be derived using the 8-multispectral band. WorldView-2’s coastal band, yellow band, red edge as well as near infrared band 2 allow them to extract more information content. In addition, imaging spectroscopy of the hyperspectral imaging can extract a lot more information, but are cost prohibitive for collecting large areas.

Five components characterize a high resolution image: built up or manmade features, vegetation, bare earth or soil, water and shadows. There is a very high spectral correlation between red, blue and green, which doesn’t allow enough dimension to identify these components independently. With the additional spectral bands offered in WorldView-2, the components will be able to be identified within an image, so that users will be able to do feature extraction and decompose the image accurately.

The infrared 2 band is placed where there is maximum separation between the individual components. The coastal band in combination with the blue band is very useful in trying to separate water and shadows. This will enhance bathymetry studies of sea floors, coastal plans and show the shallow ocean floor more definitively. Coastal remote sensing applications will likely benefit from this increased accuracy. With spectral filters, the response is very accurate. The red edge and yellow edge bands are ideal for various vegetation based applications, mapping vegetation and health, including the age and various stresses in modern vegetation.

With the promised increased spatial resolution, if looking at mixed or planted forest, not only can users identify whether it’s a coniferous forest or just forest, thye might be able to identify different tree species like oak, ash, etc. The combination of the radiometric, spectral and spatial resolution of WorldView-2 will allow users to extract more level of detail than was previously available in QuickBird.

WorldView-2 can collect up to 7700 square kilometers in one pass. Anywhere on the globe one can expect less than 2 days revisit time. Combined with the DigitalGlobe constellation, the company expects to get intra-day revisit on anywhere on the globe.

“Because we are collecting large areas in one path, we are reducing uncertainties caused by changes in atmosphere,” said Navulur. “The look angle is different as well as the bi-directional reflection distribution function. Those who work with creating orthomosaics can realize that by reducing the seasonal changes. By collecting all the data very quickly, you’ll be able to create more visually pleasing orthomosaics.”

WorldView-2 can also collect stereos in global color stereo

Blue band has been part of Landsat and QuickBird, and was designed for studying water, soil and vegetation mapping. The green band measures reflectance of vegetation and can be also used for culture feature identification. The red band is sensitive to chlorophyll, and agencies like NDVI take advantage of the red band to identify vegetation. The infrared band is useful for vegetation and for mapping soil moisture.

Potential areas where WorldView-2’s 8 bands may be useful: for global warming, to discover with temporal coverages the changes going on in an urban area or map where ice is melting. For pollution, to map natural resources, and map urban areas for growth and development, agricultural applications, and environmental applications. Plant species identification is another area where the satellite may be useful. Compared to farmland which has a uniform texture, wild plant populations can be very complex. Invasive species can be identified for possible future use as biofuels.

Susan Smith, Managing Editor