This protocol outlines how to obtain measures of spectral reflectance at many scales, for further information, see https://specnet.info/
Reflectance is a form of optical sampling that characterizes surface reflectance (also called reflectivity or albedo, the fraction of irradiance that is reflected from a surface). Spectralreflectance measures reflectance as a function of wavelength (figure 1). When broad spectral bands are used (e.g. visible or short-wave radiation), reflectance is often referred to as albedo, a measure of the overall reflectivity covering a large region of the solar spectrum, and a critical component of surface energy balance. Hyperspectral reflectance refers to spectral reflectance where many narrow spectral bands are used, allowing the discrimination of absorption features diagnostic of particular substances of interest (e.g. pigments, water, or particular chemical composition). Often, the purpose of measuring reflectance is to obtain a proxy measure that can be used (often in a model) to estimate something else that is harder to measure at that particular scale. For example, spectral reflectance is often used to estimate chlorophyll levels, photosynthetic rate or biospheric carbon gain (gross primary production or net primary production) from satellite. Reflectance has applications in many science disciplines. The discussion here will be limited to applications involving vegetation, ecophysiology, and ecosystems analysis.
(Awaiting approval to display figure – see Gamon et al. 2004)
Figure 1 – Spectral reflectance for several dominant tree species in the Canadian boreal forest in two seasons (spring and summer). Different reflectance “signatures” can be used to identify vegetation species (figure 4), and the changing patterns over time can indicate seasonal change and evolving physiological state associated with vegetation health or photosynthetic rate (figure 4). Reflectance spectra obtained with the AVIRIS sensor during the BOREAS field campaign (Gamon et al. 2004)
Unlike many other measurements, reflectance is easily applied at many spatial scales (figure 2), allowing multi-scale analyses. For example, leaf reflectance can be calibrated against pigment contentor photosynthetic rate, which can be the basis for analyzing vegetation health and productivity at larger spatial scales using aircraft or satellite remote sensing. Airborne and satellite sensors are often used as platforms for measuring reflectance over large areas, which can be used for measuring vegetation cover, ocean color, ecosystem health, species composition, diversity, and productivity (see figures 3-5).
(Awaiting approval to display figure – see Gamon et al. 2006)
Figure 2– Spectral reflectance methods and instruments operate at many spatial scales, from satellite or airborne remote sensing (top) to field or lab sampling (bottom). By linking reflectance patterns to other measurements, reflectance can be calibrated to provide proxy measurements of other variables of interest. For example, eddy covariance tower on left) or chamber gas exchange (bottom) can be used to derive estimates of photosynthesis or primary production from reflectance measurements. Because reflectance can be sampled at multiple scales, it provides a useful way to extrapolate from simple point measurements to larger regions. From Gamon et al. 2006.
Figure 3 – Image of global Net Primary Productivity (NPP) derived from reflected radiation sampled by the MODIS satellite. Image released April 2003 (courtesy NASA Earth Observatory).
(Awaiting approval to display figure – see Gamon et al. 2004)
Figure 4: Comparison of land-cover types (A) with midday gross carbon uptake rates (2-h averages near solar noon, September 16, 1994) derived from AVIRIS imagery for a portion of the BOREAS Southern Study Area (Manitoba, Canada). The land-cover types were derived from pigment and water absorption features using a combination of spectral mixture analysis and a maximum likelihood classification (Fuentes et al., 2001). The CO2flux image (B) was derived from spectral reflectance using a light-use efficiency model. (Rahman et al., 2001). From Gamon et al. (2004).
(Awaiting approval to display figure)
Figure 5: Image of a the Canadian boreal forest near Hinton, Alberta, in May 2010. The red stripe indicates a false-color infrared image sampled from a helicopter flying in a northeast direction, and red shapes indicate individual evergreen tree canopies. Each pixel in this image represents approximately ¼ meter on the ground. This image is overlaid on a Google Earth image showing the same forest from a coarser resolution satellite image. The detailed false-color infrared image was derived from an imaging spectrometer (MicroHyperspec, Headwall Photonics, Fitchburg, Massachusetts, USA), flown on a helicopter by VeriMap Plus, Calgary. Unpublished data of J. Gamon, University of Alberta.
Reflectometers (instruments for measuring reflectance) can be imaging or non-imaging, and there is a wide array of instruments available for sampling reflectance. Non-imaging devices are typically called albedometers or radiometers (if broad-band); or spectrometers, spectroradiometers, or hyperspectral spectrometers (if narrow-band). Imaging devices can be as simple as a digital camera, film camera, video camera, or web camera, sometimes supplemented by color filters, or split into color coordinates (e.g. red, blue, and green, or RGB) in subsequent processing. For example, most color cameras can be expressed as composite, color images (typically RGB or CMYK) and simple algorithms (“band math”) can be applied to create ratios or reflectance indices from different wavebands for specific purposes. Similarly, color bands of camera images can be reassigned in image processing to highlight key features; a common example in “false-color infrared” imagery assigns near-infrared, red, and green bands to red, blue, and green colors to highlight healthy green vegetation in red (figure 5). Complex imaging devices measure several wavebands simultaneously or sequentially using filter wheels, allowing collection of specific wavebands. When high spectral resolution (adjacent narrow waveband) measurements are used, instruments are often called hyperspectral imaging spectrometers. Some people object to the use of the term hyperspectralin this context, arguing that it is redundant, and that imaging spectrometeris sufficient to define an imaging device with many narrow spectral bands. The term hyperspectralis commonly used when emphasizing any instrument measuring in many narrow, adjacent spectral bands.
Historically, hyperspectral spectrometers were made with a moving monochromator in front of a detector; modern instruments typically use gratings in front of photodiode arrays, simultaneously sampling all wavelengths simultaneously. In imaging spectrometers, a common design is to focus the field of view on a slit (defining the spatial dimension and band width) positioned in front of a grating (diffracting the radiation into many wavebands). In this case the diffracted radiation is projected onto one dimension of a 2-dimensional detector array. The slit is projected onto the second dimension of the detector array. To create a 2-dimensional image with this kind of “pushbroom” imaging spectrometer, forward motion is required (e.g. pan and tilt platform, moving vehicle, aircraft, or satellite) to “build” the image one line a time. Consequently, imaging spectrometers generate 3-dimensional “image cubes,” with two data dimensions creating a spatial image, and the third data dimension depicting wavebands in the spectral domain (figure 6).
Figure 6:Data cube derived from reflected radiation for a forest nursery in Smoky Lake, Alberta, Canada. In the bottom left is a true-color image of the ground surface showing nursery plots ranging from bare ground to densely vegetated. Each pixel in this image covers approximately one meter on the ground. The third dimension extending backwards from the surface image represent the spectral dimension, and different colors indicate the intensity of reflected radiation in each band. This image was derived from an imaging spectrometer (MicroHyperSpec, Headwall Photonics, Fitchburg, Massachusetts, USA), flown on an airplane (data collected by VeriMap Plus, Calgary, Alberta). Unpublished data of J. Gamon, University of Alberta.
Units, terms, definitions
Reflectance is usually indicated by the Greek letter rho ( ). Since reflectance is usually expressed as a ratio (reflected radiation divided by irradiance), it is typically unitless, and expressed as a decimal value. For example, 0.5 indicated a surface that is 50% reflective. Some users prefer to multiply the decimal value by 100 to express a reflectance ratio as a percentage, sometimes called a “reflectance factor.” In the case of a surface with a reflectance value of 0.5, the reflectance factor would be 50%. When using image data, it is common to multiply a decimal percentage by a “scaling factor” (e.g. 1000) to facilitate visualization. In this case, a surface with a reflectance of 0.5 would have the value 500. Since it is often difficult to calculate reflectance accurately, reflected radiation is sometimes expressed as raw reflectance using digital numbers or machine units, but this makes it difficult to compare with between instruments or measurement campaigns. Similarly, if the instrument is calibrated radiometrically, reflected radiation can be expressed in absolute energy units (e.g. watts m-2nm-1 steradian-1).
Strictly speaking, reflectance should consider the wavelength (or wavelength range) being measured. For example, if reflectance in is measured in the 400-700 nm region (the region of photosynthetically active radiation region, or PAR region), then it would be appropriate to specify the metric as PAR reflectance, PAR albedoor PAR If a particular wavelength (waveband) is used, then that wavelength should be specified. For example, reflectance in the red spectral region (approximately 600-700 nm) might be called red reflectanceor RED. When reflectance is measured in a specific, narrow waveband, then the wavelength (in nm or μm) should be specified. For example, reflectance at 660 nm could be called 660. In this case, it would also be appropriate to state the bandwidth, usually specified as the full-width-half-maximum(FWHM).
An often-overlooked aspect of measuring spectral reflectance is the need to specify the geometry of sampling. Usually, this can be classified according to the direction of irradiance (the illuminating radiation), the direction of sampling (view angle), and the field-of-view of the sensor. For example, if the irradiance comes from a single point source (e.g. sunlight on a bright day, or a lamp), and if the reflected radiation is measured with a spectrometer fitted with a narrow field-of-view, then the term “bidirectional reflectance” can be used. Usually, the angle of the sensor view (view angle) should also be specified. For example, a sensor looking straight down at a target would be called a “nadir” view (elevation angle 90 degrees, zenith angle 0 degrees).
The terminology for illumination and sensor view angle can get quite specific and complex see LI-COR document or other published sources for examples. In some cases, reflectance is expressed for many view angles for a given sun angle as a function (the Bidirectional Reflectance Distribution Function, or BRDF). http://en.wikipedia.org/wiki/Bidirectional_reflectance_distribution_function
Procedures vary greatly depending upon the exact purpose, the particular instrument used, and the scale of sampling (figure 1).
Possible examples include: leaf reflectance, canopy reflectance, airborne imaging spectrometery
Links to resources and suppliers
Simple radiometers used for broadband reflectance (albedo):
Apogee Instruments (Logan, Utah, USA) – http://www.apogeeinstruments.com/
Decagon Devices (Pullman Washington, USA) – http://www.decagon.com/
Kipp & Zonen (Delft, Netherlands) – http://www.kippzonen.com/
LI-COR (Lincoln, Nebraska, USA) – http://www.licor.com/env/
Onset Computer Corporation (Bourne, Massachusetts, USA) – http://www.onsetcomp.com/
Skye Instruments (Llandridod Wells, UK) – http://www.skyeinstruments.com/
Analytical Spectral Devices (Boulder, Colorado, USA) – http://www.asdi.com/
Ocean Optics (Dunedin Florida, USA) – http://www.asdi.com/
PP Systems (Amesbury, MA) – http://www.ppsystems.com/
Spectral Evolution (North Andover, Massachusetts) – http://www.spectralevolution.com/
Stellarnet (Tampa, Florida, USA) – http://www.stellarnet-inc.com/
Hyperspec and Micro Hyperspec (Headwall Photonics, Fitchburg, Massachusetts, USA)
Pika (Resonon Inc., Bozeman, Montana, USA)
Fuentes DA, Gamon JA, Qiu H-L, Sims DA, Roberts DA (2001) Mapping Canadian boreal forest vegetation using pigment and water absorption features derived from the AVIRIS sensor. Journal of Geophysical Research106(D24):33, 565-33,577.
Gamon JA, Huemmrich KF, Peddle DR, Chen J, Fuentes D, Hall FG, Kimball JS, Goetz S, Gu J, McDonald KC, Miller JR, Moghaddam M, Rahman AF, Roujean J-L, Smith EA, Walthall CL, Zarco-Tejada P, Hu B, Fernandes R, Cihlar J (2004) Remote sensing in BOREAS: Lessons learned. Remote Sensing of Environment89: 139-162.
Gamon JA, Rahman AF, Dungan JL, Schildhauer M, Huemmrich KF (2006) Spectral Network (SpecNet): what is it and why do we need it Remote Sensing of Environment103: 227-235.
Rahman AF, Gamon JA, Fuentes DA, Roberts DA, Prentiss D (2001) Modeling spatially distributed ecosystem flux of boreal forests using hyperspectral indices from AVIRIS imagery Journal of Geophysical Research 106(D24): 33,579-33,591.
LI-COR documentation on irradiance and reflectance measurements http://www.licor.com/