r^WV. Vr£-' .*;'■••■•> ;.,.-. •;,-:■ NOAA Technical Report ERL 378-AOML 23 An Experiment to Evaluate SKYLAB Earth Resources Sensors for Detection of the Gulf Stream George A. Maul Howard R. Gordon Stephen R. Baig Michael McCasI in Roger DeVivo August 1976 U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration Environmental Research Laboratories Digitized by the Internet Archive in 2013 http://archive.org/details/experimenttoevalOOmaul ''■"li^rs ^ NOAA Technical Report ERL 378-AOML 23 N i )V 1 7 1976 An Experiment to Evaluate SKYLAB Earth Resources Sensors for Detection of the Gulf Stream o ■D O George A. Maul Howard R. Gordon Stephen R. Baig Michael McCaslin Roger DeVivo Atlantic Oceanographic and Meteorological Laboratories Miami, Florida August 1976 U.S. DEPARTMENT OF COMMERCE Elliot Richardson, Secretary National Oceanic and Atmospheric Administration Robert M. White, Administrator Environmental Research Laboratories Wilmot Hess, Director ,0^ T 'Oa, Boulder, Colorado o ■o NOTICE The Environmental Research Laboratories do not approve, recommend, or endorse any proprietary product or proprietary material mentioned in this publication. No reference shall be made to the Environmental Research Laboratories or to this publication furnished by the Environmental Research Labora- tories in any advertising or sales promotion which would in- dicate or imply that the Environmental Research Laboratories approve, recommend, or endorse any proprietary product or proprietary material mentioned herein, or which has as its purpose an intent to cause directly or indirectly the adver- tised product to be used or purchased because of this Envi- ronmental Research Laboratories publication. 11 Page CONTENTS ABSTRACT 1. INTRODUCTION 1 1.1 Background £ Purpose 2 1.2 Test Site 2 1.3 Prior Investigations 3 2. SURFACE TRUTH DATA 3 2.1 Cruise Report 4 2.2 Trackline Profile Data 4 2 . 3 Spectrometer Data 8 3. PHOTOGRAPHIC EXPERIMENT 11 3.1 Measurements 11 3.2 Data Analysis 12 3.3 Discussion 16 4. SPECTROMETER EXPERIMENT 17 4.1 Tracking Data 17 4.2 Infrared Radiance 18 4.2.1 Theoretical calculations 19 4.2.2 Comparisons of S-191 and models 21 4.3 Visible Radiance 23 4.3.1 Theoretical calculations 2 3 4.3.2 Technique for atmospheric correction 35 4.3.3 Recovery of R(A) from the S-191 data 36 in Page 5. MULTISPECTRAL SCANNER EXPERIMENT 37 5.1 S-192 Data 38 5.2 Computer Enhancement 3 9 5.3 Discussion ^ 6. SUMMARY 46 7 . ACKNOWLEDGMENTS 4 8 8. REFERENCES 48 IV AN EXPERIMENT TO EVALUATE SKYLAB EARTH RESOURCES SENSORS FOR DETECTION OF THE GULF STREAM George A. Maul Howard R. Gordon Stephen R. Baig Michael McCaslin Roger DeVivo An experiment to evaluate the SKYLAB Earth Resources Package for observing ocean currents was performed in the Straits of Florida in January 19 74. Data from the S-190 photographic facility, S-191 spectroradiometer , and the S-19 2 multispectral scanner were compared with surface observations made simultaneously by the R/V VIRGINIA KEY and the NASA C-13 aircraft. The anticyclonic edge of the Gulf Stream could be Identified in the SKYLAB S-190 A and B photographs, but the cyclonic edge was obscured by clouds. The aircraft photographs were judged not useful for spectral analysis because vig- netting caused the blue/green ratios of selected areas to be dependent on their position in the photograph. The spectral measurement technique could not identify the anticyclonic front, but a mass of Florida Bay water, which was in the process of flowing into the Straits could be identified and classified. No calibration was available for the S-191 infrared detector, so the goal of comparing the measurements with theoretical calculations was not accomplished. Monte Carlo simu- lations of the visible spectrum showed that the aerosol concentration could be estimated and a correction technique was devised. The S-19 2 scanner was not useful for detecting the anticyclonic front because the radiance resolution was inadequate. An objective cloud discrimination technique was developed; the results were applied to the several useful oceanographic channels to specify the radiance ranges required for an ocean tuned visible multispectral scanner. 1. INTRODUCTION An important problem in physical oceanography is deter- mining the boundaries of surface currents. Many techniques have been proposed to study such boundaries from space, but the actual process of extracting the correct information from satellite data is in the early stages of development. SKYLAB, with its several types of sensors (NASA, 1975), afforded the means of testing three techniques simultaneously: photography, spectroscopy, and multispectral imagery. 1.1 Background and Purpose Major ocean currents are known to have several observable surface features that make them distinguishable from the surround- ing waters . The Gulf Stream system is used as an example to typify these changes because it is one of the most important ocean currents, and because understanding of its features can be applied to the study of other current systems. Because of its subtropical origin, the Gulf Stream is typically warmer than surrounding waters and thus has a surface thermal signature that often can be detected in infrared (IR) imagery. The waters of the current are also much lower in biological productivity and hence there are fewer particles and biological pigment molecules in the Stream; this translates to a deep blue color of water. Conversely, the juxtaposed water masses are frequently higher in biological productivity, and this can cause that water to be greener. Another feature of the current that makes it visibly distinguishable is caused by the large horizontal velocity shear. Frequently the faster moving water in the current has a different sea state than surrounding waters. Just as common are the many slick lines associated with the shear. Finally, modifications of the at- mosphere above the Gulf Stream, under certain conditions, can also give an indication of the current's location. Several other features of the Gulf Stream, potentially detectable by satellite altimetry and other microwave techniques, will not be discussed in this report. The approach here is confined to visible and infrared wavelengths. The goal of this experiment was to contribute to the determination of the loca- tion of the Gulf Stream by visible and infrared measurements of radiance . 1.2 Test Site The site chosen for the experiment was in the Straits of Florida along a suborbital track in the vicinity of Key West, Florida. In this channel, the Gulf Stream runs approximately perpendicular to the satellite ground track. This track would maximize the changes in oceanic variables while minimizing the impact on the data acquisition facility onboard SKYLAB . Further- more, the logistics of obtaining the surface-truth data from a 65-foot vessel in January weather made the choice of a semi- protected body of water mandatory. Hydrographic conditions in the test site are controlled by the location and intensity of the Gulf Stream (also called Florida Current in this vicinity). The cyclonic edge, defined as the left hand edge facing downstream, has horizontal excursions of approximately 50 km; that is, at some times of the year the current's edge may be found 2 km south of Key West and at other times 70 km to the south (Maul, 19 75). The location of the cyclonic front determines the location of the major hydrographic features of the Straits. Materials from Florida Bay are also known to flow into the Straits and at times become entrained in the current. Occurrences involving mixing of Gulf Stream and Florida Bay waters are of fundamental importance to the understanding of the dispersal of natural and man- introduced materials . 1.3 Prior Investigations A general review of remote sensing of ocean color was given by Hanson (1972); Maul (1975) discussed the application of visible spectroscopy to locating ocean current boundaries. Gordon, in a series of papers (e.g. Gordon, 1973; Gordon and Brown, 1973; Maul and Gordon, 19 75) discussed the spectra of upwelling irradi- ance as a function of the optical properties of the water as calculated by Monte Carlo simulations; those studies are directly related to the current boundary location problem because the spectrum of light changes from Gulf Stream water to coastal water. Techniques for determining ocean chlorophyll (e.g. Baig and Yentsch, 1969; Mueller, 1973; Duntley et_al. , 1974) are also related to current boundary determination because pigment-forming molecules, along with suspended materials, affect the light spectrum. Remote sensing of ocean currents in the infrared region of the electromagnetic spectrum has been attempted for many years (e.g.: Warnecke, e_t a_l. , 1971; Hanson, 1972; Richardson, e_t al. , 19 73). However, there have been questions concerning the radia- tive transfer model dependency of the atmospheric correction (Maul and Sidran, 1972; Anding and Kauth, 19 72) that have awaited SKYLAB to be addressed. Adequate atmospheric correction tech- niques are required for ocean current boundary determination using once-or twice-daily observations because compositing of images is required in order to fill in the areas covered by clouds; composites must be based on a common measurement, that of the sea surface temperature itself. SKYLAB provided the first opportunity to evaluate photo- graphic, spectrometric, and multispectral imagery in a specific experiment designed for current boundary location^ It will be seen that each instrument has unique advantages, disadvantages, and limitations. It is the intent of this report to objectively evaluate each technique and to provide recommendations for future equipment and measurements . 2. SURFACE-TRUTH DATA This section gives the details of how the ocean surface data were obtained, calibrated, and analyzed. In many cases surface optical measurements are useful indicators of the pro- perties of the water that need to be measured. This is because the theory is well ahead of the measurements, and adequate 3 instruments are not yet designed or built. In the case of spectrometer measurements, the ideal observations are in fact physically impossible. 2 . 1 Cruise Report The at-sea observations were designed to provide simultan- eous measurements of the ocean while the aircraft and satellite transited the area. Since the speeds of the three vehicles are mismatched, the assumption must be made that the oceanic con- ditions are a steady-state for 12 hours or so. While it is recognized that this is not strictly true, it is a necessary assumption in view of resources available. Underway operations on the Virginia Key included gathering data on ocean salinity, chlorophyll-a concentration, surface nutrients, seawater scattering properties, sea surface tempera- ture by bucket and by a continuous radiometric profile, and ocean temperature down to 450 m with expendable bathythermographs; ship was hove-to for these spectrometry observations. Collection of data started at 24° 39'. 1 N, 81° 08.1 W at 1253 GMT, 8 January 19 74. This point is about 7 km SSE of Marathon in the Florida Keys. The track was directed SW and ended at 23° 33'. 2 N, 81° 55'. 5 W at 0150 GMT 9 January. This was 41 km off the north coast of Cuba on the evening of the same day. Weather conditions for the experiment were not ideal, with partly cloudy skies and moderate seas. Wind was from 045° at 4 ms - l and remained steady most of the day. Air temperature ranged from 2 3.8° to 26.1°C. Wet bulb values were 2 3.0°C most of the day. Visibility was 20 km except in a rain shower at 1700 GMT when it dropped to less than 5 km. Barometer was steady at 10 2 5 mb until 19 GMT when it abruptly dropped to 10 2 3 mb and remained so thereafter. Wave height was one-fourth meter until 210 GMT when it abruptly increased to one-half meter; wave period remained 4 seconds throughout the day. The Virginia Key traveled at 8 knots while on track. The boat stopped only for spectrometry stations; all the trackline profile data were collected while making speed. 2.2 Trackline Profile Data Figure 2.1 is a plot of the trackline data, after reduction. Fhe stippled profile below the 22°C isotherm shows the bottom profile along the track. The arrow shows where the 2 2°C isotherm crosses a depth of 10 0m, a point interpreted to be approximately 15 km south of the boundary of the Gulf Stream (Maul, 1975). Fhe B(45) curves show the volume scattering function at 45°, and are in units of meter"-'- steradian _1 (m"l sr~l) . A detailed discussion of the data collection, reduction, and interpretation follows . .34 .30H 24° 39.1 'N 81°08.l'W 33.2' N 55.5' W Figure 2. 1 Surface truth profiles across the Straits of Florida on 8 January 1974. The pro file s 3 from top to bottom are: Continuous chlorophyll-a (mg m~3; discrete salinity (°/oo) : discrete volume scattering function at 45° (m~lsr~l) : discrete thermometric tempera- ture (°C); continuous radiometric temperature in 10.5 - 12.5 \im band (°C) discrete depth of 22°C isotherm. a) Chlorophyll-a CCL-a) Cl-a concentrations were obtained continuously by measuring the fluorescence when Cl-a was exposed to blue light. The data are reported as If all the pigment-forming molecules (in- cluding pheophytins) were chlorophylls. The continuous record was obtained by using a Turner fluorometer, Model 111, which measured the fluorescence of surface water drawn through a continuous-flow intake system. This method is as described in Strickland and Parsons (1968), with the addition of a bubble trap. In order to calibrate the continuous record, three dis- crete samples were obtained by filtration and measured against a known standard after the cruise. This also is as described by Strickland and Parsons (1968), and uses the SCOR/UNESCO equation. Table A.l in Appendix A shows the times and positions of the three samples . The large gap in the Cl-a curve on Fig. 2.1 is due to a combination of drift on a particularly long spectrometry ob- servation and a delay in turning on the fluorometer after leaving the station. The three other breaks in the record represent a change in fluorescence during stops for short spectrometry observations. The degree of variability in Cl-a concentration over the short distances indicated in the record shows the desirability of a continuous record instead of discrete samples as a source of the profile. The high values at the northern (left hand) end of the line occur over the reefs of the Florida Keys . In addition to the three discrete surface samples, discrete samples at various depths were obtained during stops at two of the spectrometer stations. This was achieved by acquiring water at the various depths for filtration and measurement later with the surface samples. Times, depths, and positions are given in Table A.l. b) Salinity (S o/oo) Ten salinity samples were obtained on the trackline. These were surface water samples which were bottled for measurement after the cruise. The times and positions of the salinity samples are given in Table A. 2. The salinity profile in Fig. 2.1, which starts at 1200 GMT, is a straight-line plot of the ten values obtained. c) Volume scattering (3) The volume scattering function is a measure of the amount of scattering at various angles by a sample of seawater irradiated by a beam of light. In this case a single angle of 45° was measured, for a beam with a blue filter (436 nm) and a beam with a green filter (546 nm) , with a Brice-Phoenix light-scattering photometer. $(.4 5) was calculated by using: B(45°) = a TD D(45°) x sin 45° tt h D(OO) where a is the ratio of the working standard diffuser to the reference standard diffuser, TD is the transmittance of the reference standard diffuser, h is the dimension of the irradiat- ed element, D is the deflection of the galvonometer , and x is the transmittance of the neutral density filters. Measurements were obtained by collecting water samples with PVC sampler bottles. Thirty surface samples were measured, and at five stations samples were collected at various depths. Values, times, and positions appear in Table A. 3. The curve in Fig. 2.1, which starts at 1200 GMT, is a straight-line plot of the surface values. d) Bucket Temperatures (T ) As is customary, bucket temperatures were acquired at each XBT cast. Additional bucket temperatures were acquired at spectrometry stations and at samplings for scattering measure- ments. The curve in fig. 2.1, which starts at 1215 GMT, is a straight-line plot of the 28 total temperatures obtained. See Table A. 4 for values, times, and positions. e) Radiometric Temperature A continuous sea-surface temperature profile was obtained by using a Barnes, Model PRT-5, precision radiometric thermo- meter. This radiometer has a special 10.5-12.5 ym filter that approximates those in the SKYLAB multispectral scanner. The instrument's voltage output was converted to temperature based on a calibration performed in March 19 74. The profile in Fig. 2.1 begins at 1215 GMT. The large gap in the temperature profile matches the gap in the Cl-a profile and exists for the same reasons . f) Expendable Bathythermograph (XBT) The 23 XBT casts used the Sippican XBT system with 450-m probes. These casts provided the information to plot the depth of the 22°C isotherm. The point where the 22°C isotherm crosses the 10 0-m depth (arrow in Fig. 2.1) is taken to mark the zone of maximum horizontal velocity shear of the Gulf Stream. This crossing happened at 23° 40.9' N, 81° 51.0' W, about 57 km north of the coast of Cuba. See Table A. 4 for times and positions of casts. Position SoUar Zenith angle 24° 38.9 N 81° 08.0 W 64° 24° 30.7 M 81° 16.3 w 51° 24° 19.1 N 81° 27.3 w 47° 24° 08.1 N 81° 34.6 w 56° 24° 04.0 N 81° 37.0 w 65° 23° 58.8 N 81° 40.4 w 76° c , O M o CO rd rd Mh U Mh Ph o +J o +-> o O • CD • CD O ft O ft S CO S to 3 1 3 2 3 2 4 1 2 3 1 2 . 3 SPECTROMETER DATA The spectral signature of the ocean and the sky at various points along the trackline was measured in the visible range of light. The vessel was stopped for the spectrometry observations; six observations were made to obtain a total of 25 spectra. Table 2.1 shows the times and positions of these stations. A detailed discussion of the instrument, data reduction and wave- length calibration follows . TABLE 2.1 Spectrometry Observations Time (GMT) 1406 1603 1815 1956 2100 2210 a) Instrument The instrument was a Gamma Scientific, Model 2400 SR, spectroradiometer in a special water tight case. The Model 2400 SR scans in a wavelength range of 350 to 750 nm by rotating a high efficiency diffraction grating that faces a narrow aperture slit. It has a wavelength accuracy of +_2 . 5 nm. For this experiment the instrument was set to scan from 3 70 to 725 nm with a Wratten 2b filter installed over the entry slit. This filter effectively cuts transmission below 400 nm, insuring that the results will not contain secondary diffraction return. An opal glass diffuser plate was used as a cosine (Lambertian) collector; all measurements are irradiances. The data were recorded on a dual-channel strip chart recorder. b) Data Reduction Fig. 2.2 is a typical spectral scan. It is a scan of ocean upwelling light from the station at 2 210 GMT (see Table 2.1). IThe dashed curve with the many peaks is the original unsmoothed data. The peaks are due to changes in the angle of the water surface relative to the instrument as waves pass underneath. These changes impose a fairly regular periodic variation over the general trend of the spectral return; all of the spectra acquired on the cruise showed this variation to some extent. It is necessary to remove the peaks if the data are to be useful. Digital low-pass time series filtering techniques were used to produce the smooth curve in Fig. 2.2. These techniques , _ to be explained in some detail below, permit an objective filtering of the unwanted periodicities with a minimum loss of significant data trends. The filtering was performed on a UNIVAC 110 8 using a FESTSA (Herman and Jacobson, 19 75) software system at the Atlantic Oceanographic and Meterological Laboratories. T 300 - UJ _i < o c/) > < b 200- 00 on < UJ o ■z. < < or 100 - 400 450 500 550 600 650 700 WAVELENGTH (nm) Figure 2.2 Example of upwelling spectral irradiance before (broken line) and after (solid line) filtering to eliminate the effect of ocean surface glitter variations due to surface waves. The spectra were digitized off the strip chart at intervals of 20 points per inch; the points were sufficiently close to retain the shape of the original traces. The trace in Fig. 2.2, provided 248 data points. As different wavelength drive speeds were used on different stations, this number varied from scan to scan . In order to choose a filter that successfully removes high frequency energy with a minimum loss of significant trends, it is important to identify the periods of the high frequency signals. Fig. 2.3 contains a plot (light, broken line) of the relative strengths of various periodicities of the spectrometry scan shown in Fig. 2.2. This is a power spectrum of the spectro- metry scan using Tukey ' s method (see Herman and Jacobson, 1975). One can see strong periodicities at approximately 6,7,9,20, and 30 data points per cycle. These are the dominant high frequency signals that give the original scan in Fig. 2.2 its sawtooth appearance . 2 >- or 0.99 at low nadir aggies; hence p(=l-e) is very small and equation (4.1) may be written N(e) = (j>(A) L (Ts,A) x(0,X)dA 00 Po ) is zero outside the interval discussed above. The radiance may be converted to equivalent blackbody temperature by inverting the Planck equation (L) which has been integrated over the same 10 . 5<_<_12 . 5 interval. The calculations were carried out on the AOML computer. A special radiosonde was released by the Key West office of the 19 National Weather Service at the time of SKYLAB transit (see Fig. 4.1). Eefore the ^radiative transfer from a radiosonde is computed the data must be inspected to insure that no clouds are in the path of ascent, in order to compute a cloud-free radiance. Clouds are _ readily identified by their characteristically high relative humidity and _ isothermal temperature. There is evidence of clouds in the data in Fig. 4.1, so calculations were made to test the effect of clouds. RELATIVE HUMIDITY (%) 20 40 60 80 100 < 1000 ♦20 +40 AIR TEMPERATURE (°C) Figure 4. 1 Vertical "profiles of atmospheric; pressure and rela- tive humidity taken at the times of SKYLAB transit. The dotted lines on the relative humidity profile are the oloud-free estimate of atmospheric moisture. Two cloud layers are in evidence, one centered at 744 mb and one centered at 6 71 mb . Clouds are characterized by a sudden increase in relative humidity and a small (near zero) lapse rate. An equivalent clear sky estimate is made by assuming the clouds are absent; the estimated relative humidity profile in the clouds region is given by the dotted curve. The calculated equivalent blackbody temperatures for T s 2 9 8.15°1C are: Wavelength Observed (Appendix B) Cloud-free Equivalent llym 12. 5ym 293.22° K 290.46° K 293.85° 291.28° K K The differences in this case are small, 0.6 3°K at llym, and 0.8 2°K integrated over the 10.5-12.5ym region where the S-192, NOAA-4/5, and SMS- 1/2 observe. Other experience with this type of cloud- free equivalence has been as high as 5°K over the Gulf of Mexico. 20 4.2.2 Comparison of S-191 and models As stated in section 1.3, the wavelengths chosen for the two-channel technique, CAnding and Kauth, 1970) of atmospheric correction depend on the radiative transfer model. SKYLAB was to be used to study that question but since the calibration of the S-191 infrared detector is unknown, the problem cannot be investigated . The mean sea surface temperature along the trackline was 2 5.0°C. This value has been used in the calculations shown in Fig. 4.2. The Davis and Viezee (1964) model does not include absorption due to the ozone molecules which show up as a maximum at 9.6ym in the S-191 observation. The comparison shows that ozone does not affect the 10 . 5-12 . 5um window and hence is not a factor in the S-192 infrared scanner data. At ll.Oym, the ap- parent difference between the observed and calculated equivalent blackbody temperature CTgg) is 3.5°C. This seems to be the approximate error estimate of other SKYLAB investigators (personal communications), but no conclusions can be drawn. T M .293I5 , K(200«C) T M -28965»K(l6yC) OBSERVED CALCULATED 70 aO 9.0 10.0 1 1.0 12.0 13.0 14.0 150 WAVELENGTH (fj.m) Figure 4.2 Spectral infrared radiance observed by the S-191 speetroradiometer (dashed line) 3 and calculated by the ozone excluding model of Davis and Viezee (fine solid line) . Heavy solid lines are blackbody curves. Since the calibration uncertainity is wavelength-dependent the data at llym were studied to determine the shape of the nadir angle dependence curve. In the lowe-p half of Fig. 4.3 is a least squares fourth-order polynomial fit to all the observed radiances as a function of time Cdashed line). Since the same ocean spot was to be tracked, radiance should be a function of nadir angle up to 16:30:45 GMT. The maximum on this curve (arrow) is at 21 16:30:19. The upper curve is the theoretical calculation using equation 4.2 for the same atmosphere in Fig. 4.1 and for T=25°C. Nadir angles were computed using a start time of 16:29:35 GMT (45°) and a stop time of 16:30:45 GMT (0°) following the discus- sion in section 4.0. The match in the curves maxima would be approximately coincident if the tracking started with a 3 7° nadir angle, which is in agreement with the voice tape transcript. NADIR ANGLE E 45* 40* 35* 30° 25*20° 15* 10* 5° 0* T 1 i i i i i i l i & 860 -CALCULATED - *"* 'E 850 tf> 830 *1 'E o 820 ■— 810 UJ o - 1 '. - 1 / MAXI IMAX V^7 Z 800 < < 790 / ' f A ' f \ cr 1 1 1 1 1 1 i i I6'29'30 40 50 l&30'00 10 20 30 40 50 I6'3I'00 GMT Figure 4. 3 Radiance at 11 \im as a function of nadir angle of the same ocean spot as that tracked on the S-191 (dots). The dashed vertical line separates channel Al and channel A6 data. Dashed curve is fourth-order polynomial fit to all data; fine solid line is fourth-order polynomial fit to A6 data only. Heavy solid line is the calculated nadir angle dependence. The S-191 spectroradiometer uses a series of detectors that cover a segment of the spectrum. In some regions these overlap and ambiguity often exists as to which detector to use. In the sections to follow, those detectors (channels) chosen were as recommended in the NASA reports on instrument performance. It is suggested that only those data that are well calibrated be reported to non-instrument engineering investigators in the future . Barnett (personal communication) cautioned against the use of radiometer channel Al in the S-191 (see again Fig. 4.3). Accordingly a second fourth-order polynomial (solid line) was fitted to the channel A6 data only. The rms spread of the ra- diance about this polynomial is 4.48yW cm - 2 sr"l. This corresponds to a noise-equivalent temperature difference (NEAT) of +0.3°K at 2 89.65°K. Since the atmospheric attenuation tends to diminish surface gradients, this results in a calculated NEAT of 22 approximately +_0.6°C in T for this model at 11pm on this day. The equivalent blackbody temperature at the maximum in the poly- nomial is 16. 5° C at the top of the atmosphere. This implies a temperature correction of 8.5°C which is not unreasonable for a tropical winter atmosphere whose precipitable water vapor is 3.6 cm (cf. Maul and Sidran, 1973). 4.3 VISIBLE RADIANCE The visible radiance experiment was designed to study the accuracy with which spectral changes across oceanic fronts can be observed and interpreted from satellite altitudes. Unfortunately, the strongest front expected in the experiment area was missed by the S-191 so this objective, as discussed before, could not be accomplished. However, during the course of this work, a theore- tical technique for recovering the "ocean color spectrum" through the atmosphere was developed. This is discussed in detail below and an attempt is made to compare the predictions of the theory with the S-191 data and the associated ground truth. 4.3.1 Theoretical calculations It is clear that the full potential of oceanic remote sens- ing from space in the visible portions of the spectrum can be realized only if the radiance that reaches the top of the atmos- phere can be related to the optical properties of the ocean. To effect this , the radiative transfer equation must be solved for the ocean-atmosphere system with collimated flux incident at the top of the atmosphere. In such calculations the optical proper- ties of the ocean that must be varied are the scattering phase function (P (6)) and the single scattering albedo (w ; defined as the ratio of the scattering coefficient to the total attenua- tion coefficient). Furthermore, unless the ocean is assumed to be homogeneous, the influence of vertical structure in these properties must be considered. To describe the cloud-free at- mosphere, the optical properties of the aerosols and their variation with wavelength and altitude as well as the ozone concentration must be known. Considering the ocean for the pres- ent to be homogeneous , the radiance at the satellite can be related to the ocean's properties by choosing an atmospheric model and solving the transfer equation for several oceanic phase functions and w 's at each wavelength of interest. The number of separate computational cases required is then the product of the number of phase functions, the number of values of to , and the number of wavelengths. Even If the multi-phase Monte Carlo method (MPMC) (Gordon and Brown, 19 75) is used, the 03 resolution of Gordon and Brown C19 73) would require a number of simulations equal to ten times the number of wavelengths for each atmospheric model considered. It is possible, however, to obtain the necessary information without modeling the ocean's optical properties in such detail. 23 The model is based on an observation evident in results of computations given by Plass and Kattawar C1969) and by Kattawar and Plass (197 2) on radiative transfer in the ocean-atmosphere system, namely, that when the solar zenith angle is small, the upwelling radiance just beneath the sea surface is approximately uniform, (i.e., not strongly dependent on viewing angle) and hence determined by the upwelling irradiance . This observation is utilized in simulations of oceanic remote sensing situations by assuming that a fraction R of the downwelling photons are absorbed. The ocean is then treated as if there is a Lambertian reflecting surface of albedo R just beneath the sea surface. In this case Gordon and Brown (19 74) have shown that any radiometric quantity Q-, can be written. Q R Q = Q , + _J (4.3) 1-rR Qq_ is the contribution to Q from photons that never penetrate the sea surface (but may be specularly reflected from the surface) . Q2 is the contribution to Q from photons that interact with the hypothetical "Lambertian surface" once for the case R=l. r is the ratio of the number of photons interacting with the "Lambertian surface" twice, to the number of photons interacting once, again for R=l. By use of equation 4.3, any radiometric quantity can then be computed as a function of R. Physically the quantity R is the ratio of upwelling to downwelling irradiance just beneath the sea surface and is known as the reflectance function [R(0,-)] in the ocean optics literature (Preisendorfer , 1961). Spectral measurements of the reflectance function R(A) have been pre- sented for various oceanic areas by Tyler and Smith (19 70). Henceforth, R(X) will be referred to as the "ocean color spectrum" A series of Monte Carlo computations have been carried out to see if an approximate simulation (AS1) , using this assumption of uniform upwelling radiance beneath the sea surface , yields results that agree with computations carried out using an exact simulation (ES) , in which the photons are accurately followed in the ocean as well as the atmosphere. The Monte Carlo codes used Gordon and Brown (19 73, 19 74) were modified by the addition of an atmosphere. The atmosphere consisted of 50 layers and includes the effects of aerosols, ozone, and Rayleigh scattering, using data taken from the work of Elterman (1968). The aerosol scat- tering phase functions were computed by Fraser (NASA-GSFC, personal communication) from Mie theory assuming an index of re- fraction of 1.5 and Deirmendjian ' s (19 6 4) "haze-C" size distribution. Also, to determine the extent to which the vertical structure of the atmosphere influences the approximate simulation, a second approximate simulation (AS2) was carried out in which the atmosphere was considered to be homogeneous; i.e., the aerosol scattering, Rayleigh scattering, and ozone absorption were inde- pendent of altitude. The oceanic phase functions in the ES 24 are based on Kullenherg's C196 8) observations in the Sargasso Sea, and are given in Table 4.1 CNote that all the phase functions in the present paper are normalized according to 2tt/ 7T PCB) sin d6 = l). o Table 4.1 The Three Ocean Scattering Phase Functions _ (deg) KA KB KC (xlO 2 ) (xlO 2 ) (xlO 2 ) 10924 10171 9521 4916 4577 4285 573.5 534.0 499.9 169.3 157.7 147.6 29.5 29 . 39 29.31 12.56 11.9 5 11.42 3.059 3.661 4.189 1.092 1.577 1.999 0.546 0.915 1.190 0.344 0.661 0.952 0.311 0.641 0.928 0. 317 0.732 1.094 0.410 0. 829 1.309 0.492 1.017 1.618 0.579 1.261 1.856 0.617 1.357 1.999 1 5 10 20 30 45 60 75 90 105 120 135 150 165 180 KA is roughly an average of Kullenberg's phase function at 632.8 nm and 655 nm, and KC is his phase function at 460 nm. KB is an average of KA and KC . These phase functions show con- siderably less scattering at very small angles (8<1°) than was observed by Petzold (19 72) in other clear-water areas; however, the exact form of the oceanic phase function is not very important, since it has been shown (Gordon, 19 73) to influence the diffuse reflectance and R(0,-) only through the back-scatterinj probability (B) J TT, = 2tt f P (6) sinede. 72 In all of the computations reported here the solar beam incident on the top of the atmosphere is from the zenith , and with unit flux. At visible wavelengths the variable atmospheric constit- uent that will most strongly influence the radiance at the top of the atmosphere is the aerosol concentration, so the computations have all been carried out as a function of the aerosol computa- tion . Table 4.2 gives a sample comparison of upward fluxes at the top of the atmosphere at 40 nm in the three simulation models (ES, AS1, and AS 2) as a function of the aerosol concentration. N, 3xN , and lOxN refer to aerosol concentrations in each layer 25 of 1, 3, and 1Q times the normal concentration given by Elterman. 400 nra is chosen because in the visible portion of the spectrum it is the wavelength at which the atmospheric effects are ex- pected to be most severe. The ES case uses u = 0.8 and phase function KC . The values of R used to effect The AC computations were taken from the EC computation of this quantity. However, if R is taken from 3 R = 0.0001 + 0.3244x + 0.1425x + 0.1308x (4.4) where x = co B/ (l-u) Q (1-B) which, according to Gordon, Brown and Jacobs (197§), reproduces the in-water reflection function for the corresponding case but with no atmosphere present, the results of the AS model computations agree with those listed to within 0.2%. The numbers in the parenthesis next to each flux value represent the statistical error in the flux based on the actual number of photons collected in each case. It is seen that ES and AS simulations generally agree to within the accuracy of the computations. Notice also the excellent agreement between the AS1 and AS2 fluxes. Table 4.2: Comparison of the flux at the top of the atmosphere for the ES , AS1 , and AS2 simulations . Aerosol Concentration ES AS1 AS2 N 0.222 (+.002) 0.224 (+.001) 0.226 (+.001) 3xN 0.274 (+.003) 0.273 (+.001) 0.275 (+.001) lOxN -.423 (+.004) -.426 (+.002) 0.425 (+.002) Fig. 4A presents a comparison between the ES , AS1, and AS2 upward radiances at the top of the atmosphere. The step-like curve in the figure is for ES , the solid circles for AS1, and the open circles for AS2 , and y Q is the cosine of the angle be- tween the nadir and the direction toward which the sensor is viewing. The radiances in Fig. 4.4 for the ES cases are accurate to about 3% in the range y=l to about 0.4, while for the AS cases the accuracy is about 1%. To within the accuracy of the computa- tions , the three simulations again agree for all the aerosol concentrations except within the range y=0 to about 0.3; i.e. viewing near the horizon. These computations appear to demons strate that the transfer of the ocean color spectrum through the atmosphere can be studied with either the AS1 or AS2 model as long as radiances close to the horizon are not of interest. Furthermore, from the reciprocity principle (Chandrasekhar , 19 60) -26 the nadir radiance, when the solar heara makes an angle 6 Q with the zenith, can be found by multiplying the radiance I Cp) in Fig. 4.4 by p where p is taken to be cos 6 . This implies that as long as the Sun is not too near the horizon, the AS1 and AS 2 methods of computation can be used to determine the nadir radiance at the tip of the atmosphere as a function of the ocean's prop- erties through equation 4.4-. The fact that the AS2 model (homogeneous atmosphere) yields accurate radiances is very im- portant in remote sensing since it implies that only the total concentration (or equivalently the total optical thickness) of the aerosol need be determined to recover the ocean color spec- trum from satellite spectral radiometric data. H Figure 4.4 Comparison between ES (step-like curve) AS1 (solid circles) and AS2 (open circles) upward radiances at the top of the atmosphere for an ocean with w ^ 0.8 and phase function KC and an atmosphere with a normal (lxN) 3 three times normal (3xN) and ten times normal (lOxN) aerosol concentration. 27 It should be noted that these results also strongly suggest that RCA) is the quantity relating to the subsurface conditions that can be determined from space, and hence, is the most natural definition of the "ocean color spectrum". Moreover, it has been shown (Gordon, Brown , and Jacobs, 1975) that R(A) is not a strong function of the solar zenith angle (the maximum variation in R(0,-) with O is of the order of 15% for 0<8 o <60°), in contrast with other definitions (Curran, 19 7 2; Mueller, 1973). 12 10 I.C/0 (xlOO) =1 3xN 7xN, J^_J- \-400 n m 1.0 0.8 0.6 0.4 02 H Figure 4.5 IjCv) as a function of y for various aerosol concentrations i wavelength of calculations is 400 nanometers. The only way spacecraft data can be used to obtain informa- tion concerning subsurface conditions (such as concentrations of chlorophyll, suspended sediments, etc.) is through determination of R(A).Elt is assumed here that the relationship between R(X) and the ocean constituents is well known, whereas in fact much work still remains to be carried out before such a relationship can be established^ . This can be effected by applying equation (4.3) to radiance I(u) at the top of the atmosphere with the Sun at the zenith, which yields, R ICu) ICy) = I Cy) + 1 - rR I-,Cy) and Io Cy) are presented in Figs. 4.5 and 4.6 for the three aerosol models discussed above as well as for an aerosol free model (OxN) and a model with seven times the normal aerosol concentration (7xNj. 12 i i 1 1 1 OxN r s i IxN n , 10 — 1 L_ 1 1 3xN ' — " ^J = 8 |_ - 1(H) 2 (xl00)6 1 _,7xN I—, IOxN 1 , ^^-^. ^ 4 M ^j— L^ 2 1 \=400nm = 1 1 1 1 1 1 1 1.0 0.8 0.6 0.4 0.2 Figure 4.6 r^fuJ as a function of y for vaxious aerosol concentrations; wavelength of calculations is 400 nanometers. For the cases considered, R£0 . 5 , and since R is usually about 0.0 3 to 0.10 at this wavelength we can rewrite equation (.4.5) approximately as so I (y ) = I, (y ) + Rig (y ) Rgl Q-O I.. GO. (4.5 (y ) 29 Applying the reciprocity theorem to equation 4 . 6 it is found for nadir viewing that ■ 6 - nadir- ° 1 6 v I 2 CP ) (4.7) where y is the cosine of the solar zenith angle . Noting again that R is 0.10 it is seen that the difference between I nac ji r and y I-,(y ) must be small, which implies that the accuracy in R will be limited by knowledge of I-,(y ). Since I-,(y ) depends strongly on the aerosol concentration, it is absolutely necessary to be able to determine the aerosol concentration if an accurate value of R is desired. Curran (19 72) has suggested that this can be accomplished by observing the ocean (assumed free of white caps) in the near infrared where R(X) 0. In section 4.3.2 of this report, Curran ' s suggestion is utilized to find the aerosol concentration from the S-191 data. Before trying to apply the relationships developed here to the S-191 spectra, there are several important implications of the theory to be discussed. Noting that ItCvO and IoCvO depend only on the direction of the incident solar beam, the properties of the atmosphere and ocean surface, but not R (if it is assumed these latter properties remain essentially constant over horizon- tal distances large compared with those over which R changes significantly), one can directly relate changes in I(y) to changes in R. From equation 4.5 9I(y) a I 2 (y)« 9R Figure 4.6 shows that 91/ 9R is not an extremely strong function of the aerosol concentration for concentrations up to three times normal and viewing .angles up to 35° from nadir. This suggests that horizontal gradients in R can be estimated without an accurate aerosol optical thickness. When equation 4.5 is used to relate changes in radiance (Al(u)) to changes in R(AR), I(y) = [9I(y)/9R ]AR2Tl 2 (y)AR. (4.8a) Equation (4.8a) makes possible a determination of minimum radi- ance change the sensor must be able to detect for a given AR. For example, suppose that observing at ^=0.85 it is desired to detect a 5% change in R for clear ocean water at 400 nm (R~.l) through an atmosphere with three times the normal aerosol con- centration. Figure 4.6 shows that I 2 (.0 85) is about 0.11 and noting that the extraterrestrial flux 40 0nm is about 140 fa cm" 2 nm -1 , we find from equation (4.6) that AI(0.85) is 30 0.077 Wcm~2 nm~lsr~ . In a similar way radiance changes can be related to AR for a nadir-viewing sensor and any solar zenith angle. As mentioned previously from the reciprocity principle, nadir -wo where P = cos e , e Is "the solar zenith angle and I(^o) is the radiance at the top of the atmosphere seen be a sensor viewing at 6 when the Sun is at the zenith. Following through with the same arguments that led to equation (4.8a) it is found that Ai ,. =y o E3I(y )/3R)> R^y I (y n )AR. (4.8b) nadir ° u ° I ° Clearly, for a given AR, I na <3i r decreases substantially with increasing solar zenith angle because of the presence of the y factor in equation (4.8b). For example, with a three-times normal aerosol concentration, a nadir- viewing sensor would need about 2.5 times more sensitivity at 6 o = 60° as compared with 6 o ~0 to detect the same R. The above examples indicate how the theory (AS1) can be used in the design of a satellite sensor system for estimating some ocean property such as the concentration of suspended sediments or organic material. Specifically, one must first determine the effect of the property on R. Then, on the basis of the sensi- tivity desired, find AR, and finally, use equation (4.8a) or (4.8b) to find the minimum radiance change the sensor must be capable of detecting. If the sensor has a limited dynamic range, then equation (4.5) can be used with equation (4.8a) or (4.8b) to aid in the sensor performance design trade-offs. [Unfortunately at this time, relationships between R(X) and sea- water constituents are not well established.] Considering the fact that we have used only the "haze-C" aerosol phase function (which is clearly only approximately characteristic of the actual aerosol scattering) it is natural to inquire how strongly the computations of I]_(u) and l2(y) presented in Figs. 4.5 and 4.6 depend on the shape of the aerosol phase function. To effect a qualitative understanding of the influence of the aerosol phase function, computations of Ij_ and 1 2 have been carried out using the well known Henyey-Greenstein (HG) phase function P C 6) = (l-g2)/4TT HG (l+g"-2g cos e) 3 / 2 , where the asymmetry parameter g is defined according to g = 2tt / PCe) cos 6 sin 6 d 6, and 6 is the scattering angle. Since g for the haze-C phase 31 function is 0.69Q, computations haye been made with P Hp ( 6) for g values of 0.6, 0.7, and Q.8. Figure 4.7 compares these P Hf ( 8) ' s with the haze-C phase function. The HG phase function for g=0.7 clearly fits the haze-C phase function quite well in the range <_0<^140° ; however, as is well known, the HG formula is incapable 10 icr i i i — r i i i i i i i i i i r ••• "HAZE C" HENYEY- GREENSTEIN i i i i i i 20 40 J L_J ' ' 60 80 100 120 140 160 180 9 (degrees) Figure 4.7 Comparison between the "haze-C" and various Eenyey- Greenstein phase functions characterized by asymmetry parameters 0.6 3 0.7 3 and 0.8 3 as a function of scattering angle ( 6 ) . of reproducing phase functions computed from Mie theory in the extreme forward and backward directions. The HG phase functions with asymmetry parameter 0.6 and 0.8 are seen to be substantially different from the haze-C distribution at nearly all scattering angles. On the basis of Fig. 4.7 it should be expected that I]_ and 1 2 computed with P^gC e <* will be in close agreement with the haze-C computations only for g close to 0.7. Figures 4.8 and 4.9, which compare the results of computations of I]_ and 1 2 for P C e) for the normal aerosol concentration, show that this is indeed the case. It is seen that except for apparent statistical fluctuations, the HG phase function for g=0.7 yields values of 32 II 10 1 "1 1 1 1 1 1 ••• "HAZE C HENYEY - l 1 ■I GREEf 1 1 OSTEIN _• X = 400nm 9 — r - 8 • — i 1 1 1 r t i i • 7 g=o.6 • O g=oj Q 6 * U- - - X • 1 y=u.o Hh" 5 1 1 4 3 2 1 n 1 1 1 i ' 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Figure 4.8 Comparison between Ij(v) computed for the "haze-C" and Henyey-Greenstein phase functions for an atmosphere with a normal aerosol concentrations as a function of cosine SfyJ. wavelength of calculations is 400 nanometers . 1 2. and 1 2 in good agreement with the haze-C computations. This suggests that the detailed structure of the phase function is not of primary importance in determining I]_ and I2 S and it may be sufficient for remote sensing purposes to parameterize the phase function by g. To get a feeling for the importance of variations in the phase function in the remote sensing of ocean color, consider the effect of changing the aerosol phase function from an HG with g = 0.6 to one with g = 0.8 over an ocean with R = 0.1. From Figs. 4.8 and 4.9 it is found that the normalized radiance at y = 0.85 (the assumed observation angle) decreases by 4. 9x10 3. This decrease in radiance would be interpreted under the assump- tion of no atmospheric change as a decrease in R from 0.10 to 0.056. This clearly indicates then that variations in the aerosol phase function in the horizontal direction could be erroneously interpreted as horizontal variations in the optical properties of the ocean. However, it is probably unlikely that the clear 33 12 II 10 9 8 O Q 7 X 6- 5 - 4 - 3 2 ll- 1 ~l 1 I • •• " g=o.8 1 1 1 i HAZE C" HENYEY-GREENSTEIN X = 400nm ^71 . ^ — ■ — y=uv 1 i i i i i i i i i i gfo.6 • i • • • • ' ' 1 1 i 1 1 ! 1 1 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 Figure 4.9 Comparison between I^ty.) computed for the "haze-C" Eenyey-Greenstein phase functions for an atmosphere with a normal aerosol concentration, as a function of cosine 8 (y.) ; wavelength of calculations is 400 nanometers. atmospheric oceanic aerosol phase function will exhibit varia- tions as large as that considered in this example , except in extreme cases. Assuming that the aerosol concentration of the atmosphere can be determined, the uncertainity in the aerosol phase function will still of course provide a limit to the ac- curacy with which the ocean color-spectrum can be retrieved from satellite radiance measurements. In summary then, the theory CAS1) leads to the natural definition of R( *) [RC0.-)J as a function of wavelength!! as the "ocean color-spectrum". The determination of subsurface oceanic properties from space can thus be divided into two problems: 34 1) the determination of RCA) from satellite radiance measure ments , and 2) the establishment of relationships between RCA) and the desired ocean properties. Since the method of computa- tion conveniently separates the radiance into a component that interacts with the ocean CI?) and a component due to reflection from the atmosphere and sea surface CI-. ) , it is easy to relate changes in radiance to changes in R(X). It is found that for viewing angles up to 35° from nadir, h is a relatively weak function of the aerosol concentration for concentrations up to three times normal. This suggests that spatial gradients of R(A) can be determined with only a rough estimate of the aerosol concentration. It is further found that variations in the aerosol phase function can strongly influence the interpretation of the radiance at the satellite. Clearly then, it is vital to under- stand the magnitude of aerosol phase function variations. 4.3.2 Technique for Atmospheric Correction As discussed in section 4.3.1 it is necessary to know the aerosol concentration in order to recover the ocean color- spectrum R(A) from the nadir radiance spectrum observed at the satellite. In this section a method based on Curran's suggestion of using the near-infrared radiance to determine the concentra- tion is developed and applied to the S-191 data. The method involves finding a band of wavelengths in the near infrared for which the absorption by ozone and water vapor is negligible. Since the Rayleigh scattering by air is very small in the near-infrared, the greatest contributor to the optical thickness at the wavelength in question is the aerosol. It is found that at 7 80 nm ozone and water vapor do not absorb significantly, and the Rayleigh scattering contributes only about 0.023 to the total optical thickness of the atmosphere. This implies that aerosols play the dominant role in the radiative transfer here with the normal aerosol concentration yielding an optical thickness of about 0.2. Also since R(0,-) for wave- lengths greater than about 70 nm is essentially zero, the upward radiance at the top of the atmosphere at 7 80 nm simply becomes I(y) = I-,(y) F o where F Q is the solar irradiance (mW cm"2ym ) at the top of the atmosphere (Kondrat'ev 19 73). By use of the reciprocity principle, for nadir viewing and any solar zenith angle ( 8 Q ) , I na( ^i r can be written I nadir = y o F W* ItCp) and I ? Cjj) have been computed for aerosol concentrations OxN, lxN, 2xN, and 3xN at 400 nm, 500 nm, 600 nm, and 780 nm. 35 The results for the OxN and lxN computations are presented in Appendix C. Using I-^Cy) for 780 nm and noting that the S-191 nadir radiance was recorded with 6 40° the upward radiance at the top of the atmosphere for nadir viewing is found to be 0.32 3 and 0.9 38 mWcm-2 sr~lym~l for aerosol concentrations of OxN and lxN respectively. Using the S-191 radiances at 780 nm for the nadir viewing spectra taken on Jan. 8, 19 75 at 16:30:45.75 GMT (spec- trum A) and 16:30:52.2 GMT (spectrum B) which respectively were 0.64 and 0.72 mW cm - 2 sr-1, it is found that the theory suggests the aerosol concentration was 0.51xN for spectrum A and 0.6 4xN for spectrum B. In order to compute R(A) from the S-191 data the assumption is made that the variation of the aerosol extinc- tion coefficient with wavelength is exactly as given by Elterman. 4.3.3 Recovery of R(A) from the S-191 data As mentioned above, in order to recover R(A) from the S-191 data it is necessary to assume that the variation of the aerosol extinction coefficient with wavelength is identical to that given by Elterman. Also, since I-j_(y) and InCy) at 7 80 nm were derived using the haze-C phase function for the aerosols, the assumption is implicit that this phase function is correct. With these assumptions the nadir radiance at the top of the atmosphere has been computed at 400, 500, 600, and 780 nm for aerosol concentrations OxN and lxN , assuming that R(A) is zero. These radiances are presented in Table 4.3 along with those from spectra A and B. Since the actual aerosol concentration is known to be be- tween OxN and lxN , it appears that the S-191 data at 400 nm are in error. It is virtually impossible for the nadir radiance to be less than that for a OxN atmosphere. (It should be noted that the discrepancy here is great, i.e., the S-191 radiances at 400 nm appear to be too small by more than a factor of 2.) The radiances at the other wavelengths listed in Table 4.3 seem to be reasonable and were used in equation (4.7) to estimate R(a)« The results are shown in Table 4.4. It is seen that R(^) is negative except in the spectral region 500-550 nm where the values shown compare well with the Tyler and Smith Gulf Stream data for R(0,-). As discussed above, the 400 nm data are apparently in error. However, the data at other wavelengths appear realistic, so the negative R(0,-) values are probably due to the assumptions that the haze-C phase func- tion characterizes the aerosol, and that the spectral variation of the aerosol scattering coefficient is correctly described by Elterman' s data. It is clear that considerably more experimental needed to test the ability of the theory discussed in 4.3.1 ' o obtain an accurate R(A ) from the satellite radiance. 36 Table 4.3 Wavelength nadir (R=0) (nm) _ 9 _ i lmW cm ^ym mW cm" 2 ym sr ^ 400 157 500 201 600 184 780 125 OxN lxN Spect A Spect B 5.61 6.58 2. 30 2.50 2.93 4.10 4.13 4.39 1.25 2.13 1.55 1.67 0. 323 0.938 0.64 0. 72 Table 4.4 Wavelength (nm) 400 450 500 550 600 780 RCO,-) Spectrum A -0 -0 -0 274 0242 0318 0295 00715 Spectrum B -0 -0 -0 268 0235 0370 0303 00547 5. MULTI SPECTRAL SCANNER EXPERIMENT SKYLAB's multispectral scanner was a unique design that had 13 spectral channels of data spread over the visible and infra- red bands. The system used a conical scan which had the advantage of keeping the atmospheric path length the same at all times. The visible region of the spectrum (0.4 - 0.75 ym) was divided into 6 channels, each about 0.05 ym wide. Two reflected infra- red (0.75 - 1.0 ym) channels and one in the emitted infrared (10.2 - 12.5 ym) were also provided. The channels useful to Table 5.1. LANDSAT-1 has been shown to have several useful applications of visible region imagery to marine science (Maul, 1974). The much finer spectral resolution of the S-19 2 provided an opportun- ity to expand those results to ocean current boundary determination and to test if the lower wavelength (0.0 5ym) inter- vals were useful through the intervening atmosphere. 37 Table 5 . 1 Spectral Channels Useful for Oceanography BAND DESCRIPTION RANGE (ym) 1 Violet 0.41 - 0.46 : Violet-Blue 0.46 - 0.51 3 Blue-Green 0.52 - 0. 56 4 Green-Yellow 0.56 - 0.61 5 Orange-Red 0.62 - 0.67 5 Red 0.68 - 0. 76 7 Reflected Infrare d 0.78- 0.88 8 Reflected Infrare d 0.98 - 1.03 ? Thermal Infrared 10.2 - 12.5 5.1 S-192 Data S-19 2 data were collected from 16:29:22 GMT (over the open sea just north of the Cuban coastline) to 16:31:04 GMT (over the mainland Florida coast north of Florida Bay) . All channels listed in Table 5 . 2 were carefully examined in the analog format pro- vided by NASA to the principal investigator. The data in the images were compared with the S-19 0A and S-190B photographs to see if what is interpreted in section 3.2 as the anticyclonic edge of the current could be detected. This feature was not observable in the standard data product. The cyclonic edge of the stream appears to be obscured by clouds. This is often a useful means of locating the edge of the current but unfortunately made the objective of directly sensing the edge an impossibility. However, an unexpected opportunity to evaluate the S-192 developed by the photographic detection ( section 3) of a mass of water from Florida Bay flowing south into the Straits of Florida just west of Key West. This water is milky in appearance and somewhat greener in color. No ocean surface spectra were ob- served inside or outside of the plume of Florida Bay water, although it could have been easily accomplished if the SKYLAB crew had observed the feature and notified the ship of its presence. Upwelling spectral irradiance reported by Maul and Gordon (1975) probably describes the essential features of the plume and water in the straits. An intensive effort was made by Norris (NASA-JSC) , Johnson (Lockheed-JSC), and Maul (NOAA-AOML) to identify from S-192 data the plume and the anticyclonic edge, using the computer enhance- ment facilities at NASA-JSC. After approximately 10 hours of 38 machine time on both, conical and line straightened data, the feature described as the. anticyclonic edge could not be identi- fied, although it is clearly brought out in the photographic enhancements Csee Fig. 3.2). Further effort to bring out the anticyclonic edge was judged to be unwarranted and attention was turned to the plume feature which is visible in Fig. 3.3, and which preliminary computer enhancement showed to be a useful area in which to work. 15 12 10 9 8 7 6 DATA SAMPLE INTERVAL Figure 5. 1 Power spectrum of the radiance in the unfiltered S-192 conical format. Significant noise is noted every 15 a 8-9 j, and 6 data points. Before a general computer enhancement technique was develop- ed, the data were examined for periodic features in a spectrum. Figure 5.1 is a spectrum of data specially provided for this experiment that was to be high-pass filtered only; the calibrat- ion of the S-19 2 data is considered a high-pass filter. Significant periods at about 15 data sample intervals are noted in these conical data as has been reported (Schell, Philco-JSC, personal communication, 1975). The line-straightened data (see Figure 5.2) have been band-pass filtered to remove this 15-data- sample periodicity. The wavy patterns near the edge of clouds are the result of filter ringing. 5 . 2 Computer Enhancement Computer enhancement of S-19 2 data was an objective of the experiment. The technique described below is a step toward automatic detection of clouds in multispectral data. The goal is to use a near infrared channel Cchannel 8 in this case) to specify where cloud-free areas are, for analysis of sea surface temperature or ocean color. Channel 8' CO. 9 8 -1.0 3 jim) is selected as the cloud discri- mination channel because there is a maximum in the atmospheric transmissivity at this wavelength, and a maximum in the 39 Figure 5.2 S-192 Line- straightened, filtered, scanner data over the Straits of lorida near the western Florida Keys. The appropriate S-192 channel number is at the top of each panel. 40 absorption coefficient of water. The high absorption coefficient of water at 1 urn causes the ocean surface to have a very low radiance when compared with land or clouds. Thus there should be two modes in the frequency distribution of radiance: one mode for the clear ocean and another mode for land and/or clouds. An example of such a bimodal distribution is given by the histogram in Fig. 5.3. N - 2.15 /iW cm-«8r-' a = ± 4.00 /x.W cm" 1 si-" 1 CLASS INTERVAL ' 0.5 1 23456789 10 II RADIANCE (/iW cm^sr" 1 ) Figure 5. 3 Histogram (normalized to unity) of the radiance over the area shown for channel 8 in figure 5.2. The primary peak at the left is clear ocean; the broad peak centered at 7 \im crrr^sr'^ is due to clouds and. land.. In this figure, the low ocean radiances are clustered at the mode centered at N = 0.2 yW cm" 2 sr*"- L . The other mode, centered at N = 5.7 yW cm" 2 sr~l is a contribution of the clouds. (There is no land in this example.) If these modes can be identified and separated, a statistical identification of cloud- free ocean pixels can be made. Cox and Munk (19 54) observed that the radiance reflected from the ocean is essentially Gaussian in character. The problem then is to fit a curve of the form y = ni exp E-CN - N) 2 /sa ] (5.1) 41 to the data at the lower valued mode. In this equation, the normal frequency curve Cy) is a function of the total number of observations Cn) , the class interval Ci) , and the standard de- viation OJ ; the overbar on the dependent variable CN) denotes ensemble average. Fitting equation (5.1) to the data is done in an iteration scheme that uses the lower 1 valued mode as a first estimate of N. (Only the values M +_ 2c from the original ensemble are used in this first iteration; this eliminates many of the cloud contaminated data. ) After the first fit using a predescribed N, the scheme is_to iterate the data using only +_2a of each new fit. When a (or N) changes less than 0.1% between iterations, the fit is considered acceptable and the cloud- free pixels are defined as those between 0_N + ica C = M KN+ko) - N] for (N-Ka) 24035 T N 788 011.2 72 A 81°42'W 744 009.9 66 740 008.3 72 722 008.5 30 700 007.8 31 671 005.2 55 640 003. 3 42 621 003.1 22 530 -06. 3 21 500 -09.4 13 470 -12.4 10 384 -24.9 14 300 -39.1 14 250 -48.9 224 -50. 3 212 -48.9 200 -50. 3 150 -62.6 100 -77.1 070 -75.7 066 -75.2 061 -71.5 058 -72.7 054 -67.6 050 -67.8 045 -64.7 043 -58.1 030 -51.2 023 -46. 3 020 -47.4 017 -48.0 r , : f . -48.2 13.5 -43.9 (,f; APPENDIX C - Monte Carlo Simulations This appendix lists the Monte Carlo simulations of radiances I-, and Io as described in Section 4.3.2. Wave^ lengths at which the OxN and lxN atmospheric aerosol concentrations were computed are 400, 500, 600, and 780 nm. The cosines of ten zenith angles (m) were the in- dependent variables . 1^ and 1 2 are normalized to unit solar flux on a surface normal to the solar beam. 61 Wavelength = 400 nm 0.00 0.20 0.30 3.40 0.50 0.60 0.70 .80 .90 .95 1.00 IiCv) .10089+00 .88850-01 .78011-01 .67433-01 .57839-01 .52975-01 .48109-01 .46607-01 .44244-01 .42872-01 .75049-01 Aerosol = OxN I 2 ( v) .53255-01 .72697-01 .87278-01 .98883-01 .10661+00 .11159+00 .11406+00 .11634+00 .11615+00 .11701+00 .11782+00 F,2 Wavelength = 400 run 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 .90 .95 1.00 I x (y) .93753-01 .90653-01 .83232-01 .76449-01 .67122-01 .61046-01 .57118-01 .54662-01 .52434-01 .52278-01 .71513-01 Aerosol = lxN . 51310-01 .63140-01 .77826-01 .87640-01 .96479-01 .10118+00 .10457+00 .10757+00 .10919+00 .10916+00 .10906+00 63 Wavelength =500 nm Aerosol = OxN y It ( v) I 2 ( v) 0.00 0.10 0.20 0.30 0.40 0.50 0-60 0.70 0.80 0.90 0.95 1.00 63253-01 .59700-01 52971-01 .91621-01 37429-01 .10902+00 29710-01 .12474+00 24246-01 .13138+00 22112-01 .13641+00 19925-01 .13898+00 19015-01 .14171+00 18002-01 .14225+00 17044-01 .14326+00 67308-01 .14203+00 64 Wavelength = 500 nm Aerosol = lxN y I-l(p) I 2 (v) 0.00 0.10 0.20 0.30 0.40 .50 0.60 0.70 .80 .90 .95 1.0 .62318-01 .56038-01 .59056-01 .75227-01 .49220-01 .96058-01 .41716-01 .11022+00 .35218-01 .11781+00 .31928-01 .12576+00 .27678-01 .12875+00 .26638-01 .13063+00 .25573-01 .13228+00 .26419-01 .13315+00 .57385-01 .13155+00 65 Wavelength = 6 00 nm Aerosol = lxN v I^v) I 2 (p) 0.00 0.10 0.20 0.30 0.40 .50 3 .60 .70 .80 .90 .95 1.00 .27632-01 .36991-01 .36531-01 .67189-01 .29628-01 .92357-01 .25526-01 .10823+00 .21423-01 .11883+00 .17841-01 .12614+00 .16190-01 .13059+00 .15107-01 .13303+00 .14509-01 .13448+00 .16199-01 .13524+00 .52310-01 .13560+00 (,f, Wavelength = 600 ran V 0.00 O.OiO 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 .95 1.00 I-lC y) 30110-01 24508-01 17292-01 13224-01 10725-01 98865-02 90777-02 88360-02 80934-02 77006-02 61865-01 Aerosol = OxN I 2 (y) .42002-01 .80947-01 .10497+00 .12254+00 .13037+00 .13672+00 .14052+00 .14304+00 .14358+00 .14664+00 .14694+00 67 Wa ve length = 780 nm y 0.00 0.10 0.20 0.30 .40 .50 .60 .70 .80 .90 .95 1.00 i-lCh) .27838-01 .13026-01 .76052-02 .51505-02 .42167-02 . 37095-02 .34865-02 .33889-02 .30901-02 .30811-02 .65909-01 Aerosol = OxN l 2 (u) .69923-01 .10854+00 .13086+00 .14849+00 .15502+00 .16096+00 .16342+00 .16269+00 .16464+00 .16649+00 .16607+00 f/6 Wavelength = 780 run Aerosol = lxN 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 0.95 1.00 .40506-01 .67910-01 .33748-01 .95346-01 .24905-01 .11832+00 .18224-01 .13459+00 .14512-01 .14180+00 .11931-01 .14856+00 .10491-01 .15236+00 .98166-02 .15399+00 .98476-02 .15395+00 .11295-01 .15613+00 .57149-01 .15511+00 69 "fr U. S. GOVERNMENT PRINTING OFFICE: 1976— 677-347/1293 REGION NO 8 r LABOR AT ORIES The mission of the Environmental Research Laboratories (ERL) is to conduct an integrated program of fundamental research, related technology development, and services to improve understanding and prediction of the geophysical environment comprising the oceans and inland waters, the lower and upper atmosphere, the space environment, and the Earth. The following participate in the ERL missions: MESA Marine EcoSystems Analysis Program. Plans, directs, and coordinates the regional projects of NOAA and other federal agencies to assess the effect of ocean dumping, municipal and industrial waste discharge, deep ocean mining, and similar activities on marine ecosystems. OCSEA Outer Continental Shelf Environmental Assessment Program. Plans, directs, and coordinates research of federal, state, and private institutions to assess the primary environmental impact of developing petroleum and other energy resources along the outer continental shelf of the United States. WM Weather Modification Program Office. Plans, directs, and coordinates research within ERL relating to precipitation enhancement and mitigation of severe storms. Its National Hurricane and Experimental Meteorology Laboratory (NHEML) studies hurricane and tropical cumulus systems to experiment with methods for their beneficial modification and to develop techniques for better forecasting of tropical weather. The Research Facilities Center (RFC) maintains and operates aircraft and aircraft instrumentation for research programs of ERL and other govern- ment agencies. AOML Atlantic Oceanographic and Meteorological Laboratories. Studies the physical, chemical, and geological characteristics and processes of the ocean waters, the sea floor, and the atmosphere above the ocean. PMEL Pacific Marine Environmental Laboratory. Monitors and predicts the physical and biological effects of man's activities on Pacific Coast estuarine, coastal, deep-ocean, and near-shore marine environments. GLERL Great Lakes Environmental Research Labora- tory. Studies hydrology, waves, currents, lake levels, biological and chemical processes, and lake-air interaction in the Great Lakes and their watersheds; forecasts lake ice conditions. GFDL Geophysical Fluid Dynamics Laboratory. Studies the dynamics of geophysical fluid systems (the atmosphere, the hydrosphere, and the cryosphere) through theoretical analysis and numerical simulation using power- ful, high-speed digital computers. APCL Atmospheric Physics and Chemistry Labora- tory. Studies cloud and precipitation physics, chemical and particulate composition of the atmosphere, atmospheric electricity, and atmospheric heat transfer, with focus on developing methods of beneficial weather modification. NSSL National Severe Storms Laboratory. Studies severe-storm circulation and dynamics, and develops techniques to detect and predict tornadoes, thunderstorms, and squall lines. WPL Wave Propagation Laboratory. Studies the propagation of sound waves and electro- magnetic waves at millimeter, infrared, and optical frequencies to develop new methods for remote measuring of the geophysical environment. ARL Air Resources Laboratories. Studies the diffusion, transport, and dissipation of atmos- pheric pollutants; develops methods of predicting and controlling atmospheric pollu- tion; monitors the global physical environment to detect climatic change. AL Aeronomy Laboratory. Studies the physical and chemical processes of the stratosphere, ionosphere, and exosphere of the Earth and other planets, and their effect on high-altitude meteorological phenomena. SEL Space Environment Laboratory. Studies solar-terrestrial physics (interplanetary, mag- netospheric, and ionospheric); develops tech- niques for forecasting solar disturbances; provides real-time monitoring and forecasting of the space environment. U.S. DEPARTMENT OF COMMERCE National Oceanic and Atmospheric Administration BOULDER, COLORADO 80302