Directional effects

Directional effects are a function of many factors, such as the sensor view-angle, altitude, the angle between sun azimuth and sensor scan plane as well as the scatter patterning of land use types (Kennedy et al., 1997; Roberts et al., 1997).

Sensors with a large FOV are characterised by a changing view angle for every pixel in a line. The view angle increases the quantity of atmosphere through which the sensor views, which can increase the path radiance and, consequently, the overall brightness (Royer et al., 1985; Irons et al., 1987). Each change in view angle also alters the sun- sensor-object geometry of the observed surface objects. This effect is influenced by the bi-directional reflectance distribution function (BRDF, see section 5.4.1.4) of surface and atmosphere. For a given wavelength and surface type, BRDF can cause either increased or decreased observed radiance at higher view angles, depending on sun-sensor-object geometry. Cross track brightness gradients are considered to be a combination of these two view angle dependent effects (Royer et al., 1985; Roberts et al., 1997). Kennedy et al. (1997) identified the BRDF as the primary determinant of view angle effects in AVIRIS data. Currently the correction of this effect is not possible.

Most of the above studies dealt with satellite or high altitude airborne measurements where cross track brightness gradients in the AVIS data are not visible to the naked eye, when slight differences are difficult to obtain in the highly heterogeneous landscape of the test area. Therefore the AVIS data must be searched for occurring brightness gradients. Influences, which depend on the sun azimuth and the altitude, are normally minimised by the atmospheric correction. Both parameters are included in the atmospheric correction leading to at-sensor radiance. The quality of the correction largely depends on the quality of the input parameters. The comparison of reflectances measured at different altitudes during one overflight is one possibility for examining this influence. Therefore the reflectances of AVIS measurements were compared for the maize canopies on 19th and 31st _{July 1999, where two altitudes were flown at almost the same daytime to provide} similar sun elevation.

Table 4-1: Differences of maize reflectances due to different altitudes for two AVIS measurement dates (sun elevation is calculated for sea level at 47°57’N 11°17’E)

Date Aquisition time

[CET] Sun elevation [°] Deviation [% reflectance] 4000ft asl 4000 - 10000ft asl Mean 0.79 1.39 19th _July 1999 10:30-11:45 ca. 54.9-61.9 _{Maximum 2.94 5.78} Mean 1.51 3.35 31st _July 1999 10:00-11:30 ca. 48.8-58.6 _{Maximum 5.73 6.99} Mean 1.15 2.72 Total Maximum 5.73 6.99

Table 4-1 presents the deviations for all maize fields that were covered on these two

dates. Figure 4-13 shows one example of spectra flown on the 19th July. Figure 4-13 also

illustrates that the differences between the two altitudes are wavelength dependent. The maximum difference between the spectra flown at 4000ft is 5.73% reflectance whereas the difference between 4000 and 1000ft spectra is 6.99% indicating that the maximum difference between the two altitudes is 1.26% reflection. The mean difference is 1.15% reflection for spectra derived at the same altitude and 2.72% for different flight levels. This small difference allows the comparison of spectra derived at the two altitudes, whereas the differences may also be due to the BRDF of the vegetation.

Figure 4-13: Field average spectra of maize field No.260 at 4000 and 1000ft asl, flown on 19th _{July 1999}

The course of the difference spectrum is similar to vegetation indicating an incompletely correction of the adjacency effect. The adjacency effect depends on the altitude flown due to scattering in the atmosphere and a different IFOV. The spectra of a target acquired in an altitude of 1000ft resulting in a small IFOV, are more affected by a superimposition of adjacent targets than a 4000ft spectrum with larger pixel spacing. When the adjacency effect is not completely eliminated for the 10000ft flight level a vegetation-shaped difference spectrum remains after subtracting.

A simple approach for the investigation of brightness gradients occurring across track, which are due to the large FOV of the sensor, is the comparison of reflectance spectra of the highway, which crosses the N-S oriented flight stripes transversally. The left graph in

Figure 4-14 shows a sequence of highway spectra within one AVIS scene recorded on 9th

May 2000 and shows the difficulty in obtaining isolated asphalt spectra, which is mainly due to the perturbing effect of the centre strip of the highway. Therefore some of the spectra show slight signs of a red edge between 700 and 740nm as well as water absorption at 820nm. This also accounts for the higher deviation from the mean value in the NIR, which is shown in the right graph of Figure 4-14. The mean deviation ranges between 0.6 and 1.18% reflectance in the NIR, while in the VIS the values are located

between 0.2 and 0.8%. On the 9th May, the maximum deviation from the mean

reflectance is 1.4% while the value for the comparison of the three acquisition dates ranges at 2.6%. This result indicates that no radiometric distortion due to the FOV of the sensor occurs. 15 20 25 30 550 600 650 700 750 800 850 Wavelength [nm] Ref lect ance [ % ] Z 1348 S 38 Z 1331 S 150 Z 1329 S 215 Z 1327 S 263 Z 1323 S 331 Z 1323 S 366 0 0.2 0.4 0.6 0.8 1 1.2 550 600 650 700 750 800 850 Wavelength [nm] M

ean deviation [% reflectance]

9 May 2000 20 June 2000 1 August 2000