Anisotropia del campo della radiazione riflessa sulle superfici naturali

Sommario
viii
di due superfici (un campo di Colza e una superficie composta da diverse specie erbacee). I risultati
dell’inversione sono stati confrontati con misure goniometriche effettuate in situ durante i due
sorvoli: i coefficienti di correlazione tra i valori d’albedo simulati e quelli misurati sono risultati
maggiori di 0.9 sia per i dati MIVIS sopra la copertura di Colza, sia per il DAIS sopra l’altra
superficie vegetata.
Successivamente è stata analizzata l’entità degli effetti direzionali in sensori ad ampio
abbracciamento (swath) utilizzando una serie multitemporale di dati SPOT-VGT con risoluzione di
1 km. Le osservazioni multiangolari sono state ricavate scegliendo le misure spettrali in
corrispondenza di pixel appartenenti alla medesima classe d’area bruciata ipotizzando che,
nell’intervallo di tempo considerato, fossero trascurabili i cambiamenti delle caratteristiche spettrali
delle superfici. Nel medesimo intervallo temporale cambiava invece la posizione del sole e questo
ha permesso di costruire un set di dati col quale invertire il modello AMBRALS. Il medesimo
modello è stato quindi applicato in modo diretto per simulare la risposta spettrale delle aree
bruciate in condizioni d’illuminazione differenti. Il modello mostra che, quando il sole è alto
sull’orizzonte, la BRDF delle aree bruciate è abbastanza indipendente dagli angoli d’osservazione;
al contrario, quando il sole si abbassa, la BRDF delle aree bruciate di cambia sensibilmente (anche
di un fattore 10) con l’angolo di visione.
In conclusione, la ricerca presentata in questa Tesi mostra che la dipendenza della riflettività dalle
geometrie d’illuminazione e di visione può migliorare l’interpretazione dei dati acquisti dai sistemi
d’osservazione della Terra. In particolare, la misura e la modellistica della BRDF permettono di
determinare valori accurati dell’albedo e dei parametri biogeofisici delle superfici naturali. Appare
quindi evidente che questo potrà essere di grande beneficio per le applicazioni del telerilevamento
ed anticipa quello che sarà possibile nei prossimi mesi quando saranno disponibili i dati
multiangolari della nuova generazione di spettroradiometri ad immagine.
Abstract
ix
Abstract
The general aim of the study presented here is the assessment of non-Lambertian properties of the
reflected radiation over natural surfaces. This anisotropic behaviour may be demonstrated by the
Bidirectional Reflectance Distribution Function (BRDF) that describes the dependence of the
reflectivity by the illumination and viewing geometries. This study is of interest at least for two
reasons. First, it provides the means by which we can accurately estimate the albedo of Earth
surface materials. Albedo constitutes the boundary condition for any radiative transfer problem in
the atmosphere and hence is of relevance for climate modelling and energy budget investigations.
Second, the BRDF is a function of the physical and chemical parameters of the reflecting surface
and the retrieval of these parameters is possible by solving the inverse problem.
In this context a portable goniometer was originally designed and used outdoors to measure the
BRDF over natural surfaces. To the knowledge of the author, it is the first multiangular device
designed in Italy for these purposes. This device was tested over several natural surfaces and 66
Bidirectional Reflectance Factors (BRFs) were collected for each target, under stable
environmental conditions. These data allowed to analyse the anisotropic behaviour of each surface,
besides the actual albedo retrieval. Among the surfaces here investigated, white snow surfaces
showed larger anisotropy than vegetated targets. The differences between the actual albedos and
the albedos estimated under the hypothesis of Lambertian surfaces varied from 7.1% for the snow
to 3.3% for an Alpine pratum. Moreover, it was observed that the reflectance anisotropy of
vegetated surfaces still persisted in the spectral indices such as the Normalised Difference
Vegetation Index (NDVI). Empirical and semi-empirical approaches were also adopted to
parameterise the BRDFs and to model the surface albedo under different illumination conditions.
Two applications concerning with the acquisition of multiangular observations from remote sensors
have been also analysed. The AMBRALS model was used to estimate the bidirectional reflectance
effects in imaging sensors that, due to a field of view (FOV) larger than ±20 degrees, add a
brightness gradient to the intrinsic variation within the image. In June 2001, two cross-shape
patterns of airborne DAIS and MIVIS spectroradiometers were acquired on a vegetated area
(Ticino river valley) and multidirectional set of atmospherically corrected reflectances were used to
invert the AMBRALS over two natural surfaces (i.e., a Colza crop and a mixture of herbaceous
species). Radiometric measurements collected at the ground over the same surfaces at the time of
Abstract
x
the aerial surveys using a spectroradiometer mounted on the portable goniometer were then
compared to the results given by the model inversion and acted as a reference for quality control:
correlation coefficients between simulated and measured albedos were higher than 0.9 both for
MIVIS over the Colza field and for DAIS over the other vegetated target.
The magnitude of BRDF effects was finally investigated for large swath sensor such as the 1-km
SPOT-VGT that allows a daily repeat cycle. The multiangular observations were obtained over
different regions of the same surface, using a time series of data during a time period over which
surfaces do not change significantly but Sun zenith angle does. The AMBRALS model was used to
model the BRDFs over burned areas. Results from this simulation show that when the Sun is high
above the horizon, the BRDF is quite independent from the viewing zenith angles. On the contrary,
when the Sun zenith increases, the BRDF of burned area changes with the viewing observation
geometries 10 times more than in the previous case.
In conclusion, the research presented here shows that the dependence of the reflectivity by the
illumination and viewing geometries may improve the interpretation of current Earth observation
data. In particular, measurements and modelling of BRDF allow the accurate evaluation of albedo
values besides the retrieval of surface biogeophysical parameters. It is obvious that this would be of
great benefit for remote sensing applications, and it prefigures what may become possible in next
months when data from the new generation of spaceborne multiangle imaging spectroradiometers
will be delivered.
List of tables
xi
List of tables
Table 1: Current and forthcoming satellite instruments capable of multi-viewing angle sampling of the surface BRDF,
modified from Asner et al. (1998). .......................................................................................................................... 37
Table 2: Results of the propagation of variance to the NDVI. The “% of variation" represents the change of vegetation
indices due to variation of solar zenith angles (from 50° to 25° degrees). ............................................................... 48
Table 3: Summary of non-Lambertian behaviours over several surfaces and estimated with directional data acquired
during three surveys accomplished in northern Italy, between March and June 2000. ............................................ 61
Table 4: The distances (d) between the elliptical contour and the iron azimuthal ring are showed as a function of sensor
FOVs and vertical arm lengths, for the maximum view angle (i.e., 75°) and for the basement-pivot of the vertical
arm 6 cm above the ground. Both major and minor elliptical axis are reported, for four vertical arm lengths
-including the minimum (min) and the maximum (Max) chances- and for four lens of sensors available at the
Remote Sensing Dept.. d values lower than zero are critical, because include the iron ring in the BRF sampling, as
described in the text. ................................................................................................................................................ 66
Table 5: Summary of main features of the goniometric device that allow the directional sampling of the optical properties
of surfaces, under natural illumination conditions. .................................................................................................. 67
Table 6: Statistics of modeling provided by AMBRALS using ground observations processed according to DAIS and
MIVIS FWHMs. The “RMSE (inversion)” are derived by inversion of AMBRALS, the “RMSE (prediction)” are
obtained comparing modelled BRFs and measured BRFs (not used in the inversion), “R2 (simulation)” is the
regression coefficient between this data-set............................................................................................................. 81
Table 7: Headings of flight lines defining the cross-shape pattern permitting to define the MVA data-set. ...................... 81
Table 8: Illumination and viewing conditions (in degrees) under which were seen the pixels over the Colza crop and over
the mixture of herbaceous species, from DAIS........................................................................................................ 81
Table 9: Statistics of modeling provided by inversion of AMBRALS using aerial observations from DAIS and MIVIS
cross-shaped over two vegetated areas..................................................................................................................... 81
Table 10: Values of the three-parameters of the Isotropic-RossThick-LiSparseRModis model, derived for each kernel and
band. The RMSE is the root of the sum of the absolute errors in each band, absolute errors being the squared
deviations of the modelled from the observed reflectance. ...................................................................................... 81
Table 11: Summary of non-Lambertian behaviours observed over the natural surfaces sampled in this study. ................ 81
Table 12: Description of the BRF sampling over the snowfield. ....................................................................................... 81
Table 13: Description of the BRF sampling over the sandy target..................................................................................... 81
Table 14: Description of the BRF sampling over the Alpine pratum................................................................................. 81
Table 15: Description of the BRF sampling over the compact snow. ................................................................................ 81
List of tables
xii
Table 16: Description of the BRF sampling over the Alpine pratum. ................................................................................ 81
Table 17: Description of the BRF sampling over a target composed by a mixture of herbaceous species......................... 81
Table 18: Description of the BRF sampling over the Colza field. ...................................................................................... 81
Table 19: Specifications of spectroradiomters used in this study to collect radiometric data over the natural surfaces. .... 81
List of figures
xiii
List of figures
Figure 1: The colours of natural surfaces change as a function of the position of the observer with respect to the Sun.
Each pair of photos shows, on the left the photo taken when the Sun is behind observer (back-scattering), and on
the right the photo taken when the Sun is opposite the observer (forward-scattering). Clockwise from upper left:
black spruce forest, a soybean field, a barren field with a rough surface (Don Deering's photos, with the courtesy of
Wolfgang Lucht and Crystal Schaaf), an the Alpine pratum sampled in this study (cf., Appendix, part a, Table 16).
................................................................................................................................................................................. 19
Figure 2: Basic concepts of FW− and BW−scattering, hot− and dark−spot, and definition of the principal and orthogonal
planes. ...................................................................................................................................................................... 30
Figure 3: Polar coordinate system used for presenting BRDF (from Sandmeier, 2000). ................................................... 31
Figure 4: Three or more sources which illuminate the target in the field (from Curtiss and Goetz, 2000). ....................... 32
Figure 5: Examples of multiangular devices used to collect BRFs in the field. Starting from left to right: PARABOLA,
FIGOS, the goniometer used by the MISR-ground team, the device used by the Geological Survey of Japan, and
the apparatus used by Institute for Marine and Atmospheric Research of Utrecht University (The Netherlands). .. 33
Figure 6: This diagram illustrates how CHRIS can hold a target in view by using PROBA's pitch control. The target can
from the centre of an Earth image by pitching the telescope bore sight at a constant rate relative to the satellite-
target line of sight. The 4 images collected are at different viewing angles: 60°,45°,0°, and -45°........................... 36
Figure 7: Basic illustration of direct and inverse modes of a physically-based approach. ................................................. 40
Figure 8: Average values and standard deviations, of diffuse and global irradiances from 380 and 780 nm, measured
during the morning, at Passo del Pordoi (Italy)........................................................................................................ 46
Figure 9: Mean and standard deviation of grass reflectance factor, measured in the PP at view zenith angle of 0°, and
solar zenith angles ranging between 50° and 25°..................................................................................................... 47
Figure 10: Sampled directions used for the construction of BRDF of snow. The radial co-ordinate represents the view
zenith angle, the azimuthal co-ordinate the relative azimuth direction in degrees. .................................................. 48
Figure 11: BRDFs measured with the ASD-FieldSpec-FR spectroradiometer over compacted snow, in simulated TM
bands 2, 4 and 7. The Sun zenith angle is 54°. In the plots the radial direction represent the view zenith angle
(maximum 60°), and the azimuth direction the relative azimuth angle (i.e., Sun azimuth minus view azimuth)..... 49
Figure 12: Variation of broadband albedo (BRFbb) of snow surface, supposed isotropic, as a function of view zenith
angles, observed in the PP and at Sun zenith of 54°. Negative values of view zenith angles refer to forward
scattering (the Sun is in front of you), while positive values refer to backward scattering (the Sun is behind you). 50
Figure 13: Acquisition scheme of directional data from MIVIS sensor over sandy targets. The aircraft surveyed the study
area almost in the PP, therefore the scanning was orthogonal to this plane. The Sun zenith angle was equal to 45°.
The sandy targets were viewed at the three zenith angles of 7°, 10° and 15° degrees, respectively. ....................... 51
List of figures
xiv
Figure 14: ANIFs over a sandy target, as a function of wavelengths, computed in conformity with the three directions of
observation from atmospherically corrected MIVIS data (VAA=90°; SZA=45°, SAA=176°)................................ 52
Figure 15: Diagram of the solar radiation interactions in the atmosphere. E0 is the top of atmosphere solar irradiance, that
is split in the diffuse Ediffuse and direct Edirect components when interacts with the atmosphere. Ls refers to the
reflected radiance coming from the target surface, Ld to the atmospheric path radiance and L0 to the radiance
measured at the sensor. tz and tv are, respectively, downward and upward atmospheric transmittance, and qz and qv
solar zenith angle and sensor viewing zenith angle.................................................................................................. 54
Figure 16: Sketch representing the viewing observation conditions that were simulated using BRF data measured at the
ground. ..................................................................................................................................................................... 55
Figure 17: Measurement of the “error” (∆) due to the atmospheric path and committed if slanted acquisitions
(VZA=±37°, VAA=90°) are examined as they would be nadiral (VZA=0°, VAA=90°), for SZA=45° and
SAA=176° and two flight altitudes. The vertical line separates the first spectrometer (20 channels) from the second
(8 channels).............................................................................................................................................................. 57
Figure 18: Measurement of the average “error” (∆) of the firsts two MIVIS spectrometers, due to the atmospheric path
and committed if off-nadir acquisitions (VZA=±37°, VAA=90°) are analysed as they would be nadiral (VZA=0°,
VAA=90°). Results for the actual illumination conditions (SZA=45°), and winter (SZA=69°) and summer
(VZA=23°) simulations are plotted.......................................................................................................................... 58
Figure 19: Calibration coefficients from 350 to 2500 nm, determined for the reference-reflectance of the optical grade
Spectralon® panel..................................................................................................................................................... 60
Figure 20: Mean values of bidirectional reflectance factors at nadir, of two reference panels and standard deviations,
measured for Sun zenith angles between 50° and 25°. ............................................................................................. 60
Figure 21: The multiangular device at the carpenter’s shop during an intermediate assembling step: on left an overview of
the entire system, on right the particular of the slide, that permit the azimuthal sampling. The strong structure
constrains the vertical arm to be apeak, even if heavy instrument are mounted on.................................................. 64
Figure 22: Graph showing how the risk to perform an iron−corrupted directional sampling can run: when the basement-
pivot of the vertical arm, tilted by the view zenith angle (b), is at the ground (path 1, point A), the centre of the
elliptical footprint (grey) coincides with the pivot (point A). Due to technical constrains, the pivot is h cm above
the ground (path 2, point B), and the centre of the new ellipse (dotted) shifts s cm (from point A to point C), so that
the sensor FOV includes the metal of the azimuthal ring, besides the sampled surface. .......................................... 65
Figure 23: On the left: testing the capability of the device to bring a 6 kg block inside the shop where the device was
designed and assembled. On the right: the azimuthal arc dived in two parts. .......................................................... 67
Figure 24: Sampled directions used for the ordinarily construction of surface BRDF. ...................................................... 68
Figure 25: Configuration of the goniometer to measure bidirectional reflectance factors of grass nadir-viewing. ............ 69
Figure 26: ANIFs (i.e., nadir-normalised reflectance values) from 400 to 1000 nm in the PP, of the Alpine pratum. Sun
zenith is 46°. ............................................................................................................................................................ 70
Figure 27: Spectral response functions of Landsat-7 ETM+ bands 3 and 4, plotted together with BRF of the Alpine
pratum measured at nadir......................................................................................................................................... 71
List of figures
xv
Figure 28: BRDFs of grass: on the left the behaviour in the TM band 3, on the right in the TM band 4, the SZA is 46°.
Both plot exhibit the hot-spot................................................................................................................................... 71
Figure 29: Differences from albedo values computed using the whole data-set (in the graph called Albedo=66), and values
estimated using different sets of viewing angles (in the graph called Albedo<66) for the Alpine pratum. The
VZA=75 was not shown because out of range. The difference between actual albedos and values derived under
Lambertian hypothesis and nadir observation, is also shown. The Sun zenith is 46°. ............................................. 72
Figure 30: Reflectance anisotropy for TM3 and TM4 bands, as a function of viewing zenith angles, in the PP (Sun zenith
is 46°)....................................................................................................................................................................... 73
Figure 31: Anisotropy of NDVI calculated from goniometric data, as a function of viewing zenith angles, in the PP (Sun
zenith is 46°). ........................................................................................................................................................... 74
Figure 32: Configuration of the goniometer collecting the backward scattering of snow in the principal plane, with a
viewing zenith angle of 75°. .................................................................................................................................... 75
Figure 33: FW scattering of snow for 6 viewing geometries in the PP, the SZA is 57°..................................................... 76
Figure 34: Anisotropy factors of snow for 5 viewing geometries in the PP, the SZA is 57°. ............................................ 77
Figure 35: Spectral response function of the Landsat-7 ETM+ band 4, and the BRF of snow, measured with the optic
nadir viewing. .......................................................................................................................................................... 77
Figure 36: On the left: albedo values estimated by empirical and semi-empirical modelling of BRFs of snow, along with
the actual albedo derived from goniometric observations. On the right: differences, expressed in percentage, of
goniometric-based and modelled albedo values, models. Each value is refereed to the TM band 4 band and SZA of
57°. .......................................................................................................................................................................... 78
Figure 37: Simulations of albedo values of bright snow using the AMBRALS model, in the TM band 4, during the
daylight in an Alpine environment........................................................................................................................... 79
Figure 38: Differences from albedo values computed using the whole data-set (in the graph called Albedo=66), and values
estimated using different sets of viewing angles (in the graph called Albedo<66) for the snowy surface. The VZA of
75° and 60° were not shown because out of range. The difference between actual albedos and values derived under
Lambertian hypothesis and nadir observation, is also shown. The Sun zenith is 57°. ............................................. 80
Figure 39: On the left: the selection of the site where perform the goniometric measurements to be used as ground
reference of DAIS data. On the right: measuring the BRFs over the Colza crop during the MIVIS survey. ........... 82
Figure 40: Scatter-plots of BRFs modelled by AMBRALS vs. BRFs measured with the goniometer. On the left: the
comparison using ASD-FR data integrated over the 72 DAIS bands (72 per 20 samples). On the right: the
comparison using ASD-FR data integrated over the 92 MIVIS bands (92 per 20 samples). ................................... 83
Figure 41: Rough scheme of the flight-plane for the DAIS sensor (similar to MIVIS): (a) shows the planimetrical view,
(b) shows the frontal view where the flight-track RUN4 is not delineated. RUN1, RUN2 and RUN3 collected the
BW and FW scattering in the PP. ............................................................................................................................ 84
Figure 42: BRFs over the mixture of herbaceous sp. (on left) and over the Colza crop (on right) observed from the four
runs acquired by DAIS............................................................................................................................................. 86
List of figures
xvi
Figure 43: Spectral DAIS-based albedos over the mixture of herbaceous species measured with the goniometer and
modelled by AMBRALS, according to the actual atmospheric conditions. ............................................................. 87
Figure 44: Spectral MIVIS-based albedos over the Colza crop measured with the goniometer and modelled by
AMBRALS, according to the actual atmospheric conditions................................................................................... 87
Figure 45: Differences between the Lambertian-based albedos derived from MIVIS (over Colza) and DAIS (over mixture
of herbaceous sp. target) and the actual albedos computed from goniometric data and considering the anisotropic
properties of investigated surface. The first pair of column shows the difference between the anisotropic-based
albedo derived from the MVA data-set acquired by aerial sensors and the ground reference. ................................. 88
Figure 46: Spectral profiles of a savanna environment before (not burnt) and after (burnt) a fire event............................ 89
Figure 47: Daily reflectance values in the 4 spectral SPOT-VGT channels from 1st to 18th December 1999. ................... 90
Figure 48: On the left the RMSE values between observed and modelled reflectances in the four VGT channels. On the
right the scatter-plot between observed and modelled reflectance (R2 is the determination coefficient). ................ 92
Figure 49: Bidirectional reflectance of burned areas measured by SPOT-VGT in its azimuthal plane (i.e., 90°-180°) and
Sun zenith 13°, as a function of viewing zenith angles. ........................................................................................... 92
Figure 50: Bidirectional reflectance of burned areas measured by SPOT-VGT in its azimuthal plane (i.e., 90°-180°) and
Sun zenith 49°, as a function of viewing zenith angles. ........................................................................................... 93
Figure 51: ANIFs (i.e., nadir-normalised reflectance values) from 400 to 1000 nm in the PP. Clockwise from upper left:
snow, Alpine pratum, Colza and mixed herbaceous species. ................................................................................... 96
Figure 52: Broadband values of the bidirectional reflectance factors BRFbb over the investigated surfaces. Clockwise
from upper left: snow, Alpine pratum, Colza and the mixed herbaceous species. ................................................... 98
Figure 53: Comparison of broadband albedo (percentage values) computed under the Lambertian hypothesis or
considering the anisotropic properties of the reflected radiation field of investigated targets.................................. 98
Figure 54: Performances of the Landsat-7 spectral response functions, in the reflective bands. The spectral bandwidths
are determined by the combined response of all optical path mirrors (i.e., primary, secondary, scan line corrector,
scanning), the spectral filters, and the individual detectors. ................................................................................... 112
List of acronyms
xvii
List of acronyms
3D Three Dimensional
6S Second Simulation of the Satellite Signal in the Solar Spectrum
AMBRALS Algorithm for MODIS Bidirectional Reflectance Anisotropies of the Land
Surface
ANIF ANIsotropy Factor
ANIX ANIsotropy indeX
ASAS Advanced Solid-state Array Spectroradiomters
ASD-FieldSpec-FR Analytical Spectral Device- FieldSpec-Full Range (radiometer)
ATCOR ATmospheric CORrection
AVHRR Advanced Very High Resolution Radiometer
BGC Bio-Geo-Chemical cycles
BRDF Bidirectional Reflectance Distribution Function
BRF Bidirectional Reflectance Factor
BW BackWard
CHRIS Compact High Resolution Imaging Spectrometer
CNR−IREA Consiglio Nazionale delle Ricerche − Istituto per il Rilevamento
Elettromagnetico dell’Ambiente (National Research Council − Institute for
Electromagnetic Sensing of the Environment)
DAIS Digital Airborne Imaging Spectrometer
DIIAR Dipartimento di Ingegneria Idraulica, Ambientale e del Rilevamento
(Department of Environmental, Hydraulic and Survey Engineering)
DLR Deutschen Zentrum für Luft und Raumfahrt
EGO European GOniometric facility
EOS Earth Observation System
ESA European Space Agency
ETM+ Enhanced Thematic Mapper Plus
EU European Union
FAPAR Fraction of Absorbed Photosynthetically Active Radiation
FIGOS FIeld GOniometer System
FOV Field Of View
FW ForWard
FWHM Full Width at Half Maximum
HYMAP HYperspectral MAPper
List of acronyms
xviii
IDL Interactive Data Language
IGBP International Geosphere-Biosphere Programme
LADF Leaf Angle Distribution Function
LAI Leaf Area Index
LUT Look Up Table
MERIS Medium Resolution Imaging Spectrometer
MIR Mid-InfraRed
MISR Multiangle Imaging SpectroRadiometer
MIVIS Multispectral Infrared and Visible Imaging Spectrometer
MODIS Moderate Resolution Imaging Spectrometer
MVA MultiView Angle
NIR Near-InfraRed
PARABOLA Portable Apparatus for Rapid Acquisition of Bidirectional Observations of
the Land and Atmosphere
POLDER POlarization and Directionality of the Earth’ Reflectance
PP Principal Plane
PROBA PRoject for On-Board Autonomy
RCR Remote Cosine Receptor
RMSE Root Mean Square Error
RS Remote Sensing
SAA Sun Azimuth Angle
SFG Sandmeier Field Goniometer
SPECTRA Surface Process and Ecosystem Changes Through Response Analysis
SPOT Satellite Pour l’Observation de la Terre
SS-PR SpectraScan-PhotoResearch (radiometer)
SVAT Soil Vegetation Atmosphere Transfer
SWIR ShortWave- InfraRed
SZA Sun Zenith Angle
TIR Thermal-InfraRed
UTC Universal Time Coordinate
VAA View Azimuth Angle
VGT VeGeTation
VIS VISible
VZA View Zenith Angle
1. Introduction
19
Chapter 1
Introduction
Everyday outdoor scenes give ample opportunity to understand why bidirectional reflectance is an
important issue for remote sensing research and/or applications. Better than words, the following
images (Figure 1) help to explain the concept of bidirectional reflectance: the colours (i.e., the
visible part of electromagnetic spectrum reflected by targets) of the objects depend on the position
of the observer with respect to the Sun.
Figure 1: The colours of natural surfaces change as a function of the position of the observer with respect to
the Sun. Each pair of photos shows, on the left the photo taken when the Sun is behind observer (back-
scattering), and on the right the photo taken when the Sun is opposite the observer (forward-scattering).
Clockwise from upper left: black spruce forest, a soybean field, a barren field with a rough surface (Don
Deering's photos, with the courtesy of Wolfgang Lucht and Crystal Schaaf), an the Alpine pratum sampled in
this study (cf., Appendix, part a, Table 16).
1. Introduction
20
Assuming that other sensors could take the place of our eyes, the same effects also happen with the
amount of invisible part of electromagnetic radiation interacting with the surface. Therefore, to
estimate what is actually reflected by a surface towards a close-range, airborne or spaceborne
sensor, the Sun-target and the target-observer directions should be kept in mind, hence the term
“bidirectional” reflectance.
1.1 Rationale
Observations of angular variations of radiance values, reflected or emitted from a surface, are very
important to interpret and analyse remotely sensed data. Basically, it is useful to identify two
principal research fields according to the physical processes that mainly govern the interaction
between the electromagnetic radiation and the surface. In the visible (VIS) and solar-infrared
portion of the electromagnetic spectrum, between 0.4 and 5.0 µm (Rott, 2000), the solar radiation
reaching the ground is mainly reflected, and/or absorbed by the surfaces according to their nature,
whereas at the same wavelengths the electromagnetic grey-body emission of the surfaces
themselves is very low at environmental temperatures. In the terrestrial-infrared part of the
spectrum, between 5.0 and 20 µm (Rott, 2000) the radiometry of the ground surfaces is mainly
characterised by the emission of electromagnetic radiation according to the Wien Law, which is
strictly related to their brightness temperatures. Even thought thermal emission may dominate, the
measured radiance is a combination of the thermal emission from the surface and the reflection of
incident, hemispherical flux from the sky and surrounding objects into the direction of the
radiometer (Norman and Becker, 1995). In particular, the reflection factor as well as the brightness
temperatures show angular variations as a consequence of the direction (characterised by the zenith
and azimuth of the instrument). Again, both for the reflection factor and for the radiative (or skin)
temperature the angular change is also due to the direction of the illuminating source (above all the
Sun, with its zenith and azimuth) and this double dependence is commonly known as bidirectional.
This introduction is therefore divided into two sections in order to give an overview of
achievements and needs related to the angular variations of reflectance (Section I) and brightness
temperature (Section II). Both sections focus on directional measurements in the field.
1. Introduction
21
Section I-Directionality in the visible and solar-infrared portion of the
electromagnetic spectrum
The remote sensing community is becoming more and more interested in the anisotropy of the
reflected radiation field over natural surfaces due to its influence on remote sensing data. The
anisotropy of the reflected radiation field is described by the Bidirectional Reflectance Distribution
Functions (BRDFs) which are equations describing how the reflectance of a surface varies with the
solar-view geometry.
Several papers have been published concerning the algorithms for the BRDF modelling (Chopping
2000a; Chopping 2000b; Lucht et al., 2000; Duchemin, 1999; Rahman et al., 1999; Verstraete et
al., 1990; Pinty et al., 1989; Hu et al., 1997), and about methods and instrumentation for measuring
the Bidirectional Reflectance Factor (BRF) (Jensen and Schill, 2000; Sandmeier, 2000; Deering et
al., 1999; Greuell and de Ruyter de Wildt, 1999; Sandmeier and Itten, 1999; Tsuchida et al., 1999).
The modelling approach has been well studied in order to normalise the off-nadir acquisition of
sensors having large swath (e.g., NOAA-AVHRR, SPOT-VEGETATION, low altitude
spectrometry imageries) or specifically designed off-nadir acquisitions (e.g., MISR, IKONOS), and
to correct the geometry of nadir-viewing sensors, such as the Landsat series, in mountainous terrain
(Greuell and de Ruyter de Wildt, 1999; Schaaf et al., 1994). The directional features of remote
sensors imply the need to add the “directional reflectance” resolution (Myneni et al., 1995) to the
basic four resolutions (i.e., spatial, spectral, radiometric and temporal) characterising the Earth
observation systems. The off-nadir normalisation allows reflectance values, band ratios, indexes
(e.g., normalised vegetation indices) and all the others parameters related to remote sensing
measurements of surfaces to be compared independently of whether they are positioned in the
centre or at the boundaries of an image. Moreover, after having modelled the BRDFs over natural
surfaces it is possible to estimate albedo as the ratio between the hemispheric fluxes, or irradiances,
of the reflected and incoming radiation. Accurate estimation of surface albedo is of critical
importance in understanding circulation models and climate dynamics
BRDF modelling is commonly based on semi-empirical models, which need an accurate
parameterisation, based on the comparison between modeled and measured values of the BRFs for
several Sun-target-sensor geometries. In order to acquire a sufficient set of BRFs it is necessary to
observe the surface from several angles of view in a short time avoiding the changes of the surface
properties. Ground bidirectional reflectance data will be important not only for parameterisation of
existing and new BRDF models but also to validate satellite related measurements acquired from
multiangular spaceborne sensors, such as MISR, MODIS and POLDER, as well as to investigate
the relationship between biogeophysical parameters and BRDF effects (Sandmeier, 2000).