Direct solar spectral irradiance and transmittance measurements from 350 to 2500 nm

Download Direct solar spectral irradiance and transmittance measurements from 350 to 2500 nm

Post on 02-Oct-2016




0 download

Embed Size (px)


<ul><li><p>ce50</p><p>Go</p><p>rciale to msolais 3</p><p>0% ovtter ohat cll asexpe</p><p>solact sole maarisof t 5</p><p>days. 2001 Optical Society of AmericaOCIS codes: 010.1320, 010.1110, 120.5630, 300.6190.1. Introduction</p><p>Traditionally, the measurement of extinction of solarradiation by the various constituents of the Earthsatmosphere by sunphotometry or, more properly, so-lar radiometry has been undertaken with filter-basedinstruments. These instruments typically containrelatively narrow-bandpass ~,10-nm! filters that areevenly distributed through the shorter wavelengths~300 to ;1000 nm! of the solar spectrum, placed care-fully to avoid regions of strong molecular absorption.They may also contain a channel placed inside awater-vapor absorption region ~e.g., 940 nm! to inferwater-vapor amount. The measurements of aerosoloptical depth and water-vapor amount are used toconstrain atmospheric radiative transfer models usedin, for example, radiation budget studies, remote</p><p>sensing ground reflectance retrievals, and satellitesensor calibration studies.</p><p>More recently, instruments based on prisms or dif-fraction gratings in conjunction with photodiode ar-rays are being utilized to make spectrally continuoustransmittance and irradiance measurements.13 Wepresent the use of a commercially available spectrora-diometer with a custom-designed telescope that signif-icantly extends solar spectral extinction andirradiance measurements to 2500 nm. The wave-length range of this instrument, 3502500 nm, encom-passes 1273 Wm22 or 92% of the Suns radiant energybefore interaction with the terrestrial atmosphere.In addition to the parameters usually determined bysolar radiometry ~e.g., aerosol optical depth, ozone andwater-vapor amount, aerosol size distribution!, thesemeasurements will permit the direct determination ofin-band spectral transmittance for virtually all thecurrent Earth-observing instruments with coarserspectral resolution. These include the moderate-resolution imaging spectroradiometer4 ~MODIS!, theMultiangle Imaging Spectroradiometer4 ~MISR!, andthe advanced spaceborne thermal emission and reflec-tion radiometer4 ~ASTER! on the Earth ObservingPlatform AM-1, the Landsat Thematic Mapper,5 theimaging spectrometer HYPERION6 on the EO-1 plat-form, as well as the airborne visible infrared imaging</p><p>All the authors are with the University of Colorado at Boulder,Campus Box 216, Boulder, Colorado 80309-0216. B. C.!, Z. Qu, A. F. H. Goetz are with theCenter for the Study of Earth from Space, Cooperative Institute forResearch in Environmental Sciences. A. F. H. Goetz is also withthe Department of Geological Sciences.</p><p>Received 28 November 2000; revised manuscript received 11April 2001.</p><p>0003-6935y01y213483-12$15.00y0 2001 Optical Society of America</p><p>20 July 2001 y Vol. 40, No. 21 y APPLIED OPTICS 3483Direct solar spectral irradianmeasurements from 350 to 2</p><p>Bruce C. Kindel, Zheng Qu, and Alexander F. H.</p><p>A radiometrically stable, commesimple, custom-designed telescopmittance and directly transmitted3502500 nm and the resolutionment to be stable to better than 1.portable, can be set up in a maabsolute radiometric calibration tin valid Langley channels as wealtitude Langley plot calibrationcurrent uncertainties in the TOAsured and MODTRAN-modeled direwith measurements over the largcases shown. Side-by-side compwith a mean absolute difference oand transmittance0 nm</p><p>etz</p><p>ly available spectroradiometer was used in conjunction with aake spectrally continuous measurements of solar spectral trans-</p><p>r spectral irradiance. The wavelength range of the instrument is11.7 nm. Laboratory radiometric calibrations show the instru-er a nine-month period. The instrument and telescope are highlyf minutes, and can be operated by one person. A method of</p><p>an be tied to published top-of-the-atmosphere ~TOA! solar spectraregions of strong molecular absorption is also presented. High-riments indicate that this technique is limited ultimately by ther spectra, approximately 23%. Example comparisons of mea-ar irradiance show that the model can be parameterized to agreejority of the wavelength range to the 3% level for the two examplens with a filter-based solar radiometer are in excellent agreement,</p><p>0.0036 for eight overlapping wavelengths over three experiment</p></li><li><p>spectrometer ~AVIRIS!7 instrument on NASAs ER-2.In the case of imaging spectrometers, in addition to</p><p>F</p><p>cdasp3sVfptrtaeTmplsoSictpnlttEwasrw</p><p>from approximately 7 to 3 nm was determined whenthe instrument was scanned with a 0.5-m monochro-mT1andatwcct</p><p>oitatnscictiamastrt</p><p>3better parameterization of the atmosphere ~ozone,aerosol, and water-vapor amount!, it may also be pos-sible to apply a residual correction directly to thetransmission term in the surface reflectance retrieval.In this paper we describe the design of the spectrora-diometer and telescope, its radiometric stability, and amethod of absolute radiometric calibration for bothspectral regions that obey a type of Beers law trans-mission ~a valid Langley channel! and spectral regionsthat contain strong molecular absorption and thuscannot be calibrated with the traditional Langleymethod. Finally we present the results of a side-by-side experiment with a filter-based solar radiometerand some example comparisons with the atmosphericradiative transfer code MODTRAN8 with the measureddirect solar spectral irradiance.</p><p>2. Methods</p><p>A. Instrument Design</p><p>The Analytical Spectral Devices Full Range ~ASD-R! is a commercially available, portable spectrome-</p><p>ter designed originally and primarily for themeasurement of surface spectral reflectance.9 Re-ent upgrades to the temperature stabilization andark current control have made the instrument suit-ble as a spectroradiometer. The instrument con-ists of three separate spectrometers: a siliconhotodiode array in the visible near infrared ~VNIR,501000 nm! and two scanning spectrometers in thehortwave infrared ~SWIR, 10002500 nm!. In theNIR a fixed concave holographic diffraction grating</p><p>ocuses light onto a temperature-stabilized siliconhotodiode array. An order-sorting filter placed onop of the array prevents higher-order light fromeaching the detectors. The two SWIR spectrome-ers designated SWIR1 for the 10001770-nm regionnd SWIR2 for the 17702500-nm region are single-lement, thermoelectrically cooled InGaAs detectors.he SWIR1 and SWIR2 gratings are attached to aotor that scans back and forth every 0.1 s. The</p><p>osition of the grating shaft, and thus the wave-ength, is determined with an optical encoder. In-trumental dark current is measured and subtractedut from each spectrum prior to its recording in theWIR1 and SWIR2. In the VNIR, the dark current</p><p>s measured at the beginning of the measurementycle, and masked detectors in the array thereafterrack dark current fluctuations. The spectral sam-ling is approximately 1.4 nm in the VNIR and 2.0m in the SWIR. The spectrum is linearly interpo-</p><p>ated to every nanometer prior to being recorded onhe personal computer that controls the spectrome-er. For relatively smooth varying spectra fromarth surface cover materials, the errors associatedith this interpolation are small. The errors gener-ted from this interpolation for atmospheric mea-urements are addressed in Subsection 3.A. Theesolution of the ASD-FR as measured by the fullidth at half-maximum ~FWHM! varies in the VNIR</p><p>484 APPLIED OPTICS y Vol. 40, No. 21 y 20 July 2001ator every 0.1 nm over the entire wavelength range.he FWHM measured in the SWIR1 is 11.4 and is1.7 nm in the SWIR2; the resolution does not varyppreciably across these wavelengths. It should beoted that, although time-consuming and rather te-ious to measure, it is critical to know the resolutionccurately for atmospheric measurements that con-ain sharp, strong absorption features. Simulationsith a radiative transfer code have shown that a</p><p>hange in FWHM of only 0.5 nm, at 1-nm sampling,an result in nearly a 10% change in irradiance overhe oxygen A band.</p><p>The spectrometers are fed light from a 1.0-m fiber-ptic cable that also eliminates any polarization ofncoming light. A laptop computer mounted on theop of the ASD-FR serves to control the spectrometernd record the spectra. The spectrometer measureshe entire wavelength range in 0.1 s, and a largeumber of spectra can be averaged to achieve a highignal-to-noise ratio ~SNR!. The spectra are en-oded to a 16-bit resolution, and real-time display ofrradiance spectra is possible with the inclusion ofalibration coefficients. Either ac or batteries powerhe spectroradiometer. It is highly portable ~weigh-ng only 6.6 kg!, can be set up in a matter of minutes,nd is easily operated by one person. With someodification to the ASD-FR control software as well</p><p>s weatherproofing of the spectroradiometer and tele-cope system, this system could be utilized as a long-erm monitoring system. Currently, however, itequires an operator to periodically check the opera-ion of the ASD-FR.</p><p>B. Radiometric Stability</p><p>Solar radiometers depend on a radiometrically stablesystem of detectors and electronics. Once the cali-bration of the instrument is established, a stable ra-diometer can be deployed to make instantaneousmeasurements of transmittance and irradiance. Todetermine the long-term radiometric stability of anASD-FR, we undertook a series of laboratory calibra-tions. The laboratory calibration uses a commer-cially available quartz-halogen irradiance standard~Optronic Laboratories10! 1000-W lamp in conjunc-tion with a Spectralon diffusing panel. Spectralon isa commercially available, National Institute of Stan-dards and Technology-traceable, diffuse reflectancestandard.11 The lamp is powered by a highly stablecurrent source, the output of which is crossed checkedwith a National Institute of Standards andTechnology-traceable current shunt and digital volt-meter. Because we were determining only the long-term stability of the radiometer and not its absolutecalibration, seasoned uncalibrated lamps were used.Forty-two calibrations taken from January to Sep-tember 1999 are plotted on the left-hand side of Fig.1. The coefficient of variation ~standard deviationdivided by the mean! is plotted on the right-hand sideof Fig. 1. In addition to the statistics for all 42 cal-ibrations, a second set of statistics was calculated for</p></li><li><p>rAdthe calibrations made when the instrument was innear constant use during the summer field season,either making solar radiometric measurements ormeasurements of surface reflectance. The instru-ment shows excellent radiometric stability for thisnine-month period, better than 1% for virtually theentire wavelength region and better than 0.5% forwavelengths beyond 1000 nm. The higher devia-tions centered on 1400 and 1900 nm are the result ofwater-vapor absorption bands present in the atmo-sphere surrounding the laboratory calibration setup.This demonstrates the strong sensitivity of thesebands to even small amounts of water vapor over ashort path length ~,70 cm! in a relatively dry envi-onment. These measurements indicate that theSD-FR is suitable for use as a precision spectrora-iometer.</p><p>C. Telescope Design</p><p>To measure the direct normal solar spectral irradi-ance, the field of view ~FOV! of the radiometer mustbe restricted to only that of the solar disk. This isaccomplished either by use of a telescope with a FOVslightly larger than the angle subtended by the Sun,approximately 0.5, or by a combination of a diffuserto measure total downwelling irradiance and a shad-ing device to determine the diffuse component. Bysubtracting the diffuse from the total, we can deter-mine the direct component. Finding a diffuser withexcellent cosine response over such a wide wave-length range proved difficult. For this reason, wedesigned a telescope-type instrument. The tele-scope contains front and rear apertures that definethe FOV and does not contain any optical elements~see Fig. 2!. The apertures are removable and canbe replaced easily with apertures of various sizes toincrease or decrease the FOV. Attached to the rearof the telescope is a 2-in. ~5-cm! Spectralon integrat-ing sphere. The integrating sphere serves to pro-vide a highly spatially uniform source of light for thefiber optic to view as well as to minimize the effects ofpointing errors as long as the 2.0 FOV of the tele-scope encompasses the solar disk. In addition, Spec-tralon has been shown to be resistant to degradationby ultraviolet radiation after treatment of baking invacuo at 90 C for several hours.12 Thus this design</p><p>Fig. 1. Digital number for all 42 calibrations are plotted on theleft. Coefficient of variation is plotted on the right for all 42calibrations ~solid curve! and for the summer months when theinstrument was in near constant use ~dotted curve!.should be free of the temporal degradation thatplagues filter-based solar radiometers. On the top ofthe telescope a pinhole and target allow for precisealignment to the Sun. The telescope was designedfor use with commercially available solar trackers.We made all the measurements presented in thispaper by manually tracking the Sun using a simple,commercially available, elevation-azimuth mount.An example ASD-FR optical depth spectrum is plot-ted on the left-hand side of Fig. 3. On the right-handside of Fig. 3 is an ASD-FR solar irradiance spectrum.</p><p>D. Radiometric Calibration</p><p>Careful radiometric calibration of a solar radiometeris essential to obtain meaningful results. The twocommon methods of calibration are the Langley plotmethod and the standard lamp method. In the Lan-gley plot method, the directly transmitted solar irra-diance Fl can be described by the BeerBouguerLambert law13:</p><p>Fl 5 F0l R22Tl , (1)</p><p>Tl 5 exp~mtl! , (2)</p><p>where F0l is the top-of-the-atmosphere ~TOA! solarspectral irradiance, R is the EarthSun distance in</p><p>Fig. 2. Simple schematic drawing, in cross section ~top view!, ofthe telescope. Front and rear apertures are exchangeable tochange the FOV. The telescope is 570 cm in length and weight 2kg. It can be attached to a solar tracking device or track the Sunmanually. The pinhole and target are on top of the telescope andare not shown in this view.</p><p>Fig. 3. Example optical depth spectrum is plotted on the left, andthe solar irradiance spectrum is plotted on the right. The opticaldepth spectrum, from Mauna Loa, is the result of a Langley plot;thus the optical depths are not correct for regions of strong absorp-tion. The example surface solar irradiance spectrum is from thecoast of Hawaii.</p><p>20 July 2001 y Vol. 40, No. 21 y APPLIED OPTICS 3485</p></li><li><p>astronomical units, m is the relative air mass, and tis the total optical depth at wavelength l. The air</p><p>pttr</p><p>mean of the ratio of the Langley-determined radio-metric coefficients CL~l! to the lamp coefficients CB~l!</p><p>oat</p><p>cptioRoamsw</p><p>m</p><p>3mass m is commonly determined from an air-masstable that accounts for spherical geometry and theeffects of atmospheric refraction.</p><p>For a series of measurements taken over a range ofair mass m under stable conditions of constant opticaldepth t, a plot of ln~Fl! versus m prod...</p></li></ul>