Abstract

Spaceborne spectrometers like the Medium Resolution Imaging Spectrometer (MERIS) on board the Environmental Satellite (Envisat) are widely used for the remote sensing of atmospheric and oceanic properties and make an important contribution to the monitoring of the earth’s atmosphere system. To enable retrievals with high accuracy, the spectral and radiometric properties of the instruments have to be characterized with high precision. One of the main sources of radiometric errors is stray light caused by multiple reflections and scattering at the optical elements within the instruments. If not corrected for properly, the stray light–induced offsets of measured intensity can lead to significant errors in the derived parameters. The effect of stray light is particularly momentous in the case of measurements inside strong absorption bands like the oxygen A band at 0.76 μm or the ρστ absorption band of water vapor around 0.9 μm. For example, the retrieval of surface and cloud-top pressure from MERIS measurements in the O2 A band can be biased because of an insufficient correction of stray light in the operational processing chain.

To correct for the residual stray light influence after the operational stray light correction in the O2 A-band channel of MERIS, an empirical stray light correction of the measured radiance at 0.76 μm has been developed based on optimizing the coefficients of a simple brightness-dependent stray light model. The optimal model coefficients were found by adjusting the retrievals of surface and cloud-top pressure to accurate reference data for several selected scenes. To account for the limited accuracy of the MERIS spectral calibration, the center wavelength of the O2 A-band channel was additionally adjusted within a ±0.1-nm tolerance range. The correction was tested on a variety of clear and cloudy scenes at different locations by applying the surface and cloud-top pressure retrieval algorithms to data recorded over the whole lifetime of MERIS. The results indicate the potential to greatly improve the accuracy of the retrieved pressure values using the proposed correction factors.

1. Introduction

Satellite observations provide an indispensable contribution to the monitoring of the atmosphere, earth, and ocean. Because it is impossible to operationally perform in situ measurements of cloud properties, the spaceborne remote sensing of clouds is of special importance. The methodologies for the retrieval of cloud properties from satellite data have been constantly advanced during the past decades. In the case of the retrieval of cloud altitude, which is representing one of the most decisive parameters for the cloud radiative effect, several techniques have been developed. The most popular methodologies make use of either the thermal emission of the clouds, allowing for the determination of the cloud’s temperature and height (e.g., Smith and Platt 1978; Wielicki and Coakley 1981; Menzel et al. 1983, 2006), or stereoscopic views, exploiting the height-dependent parallax of clouds when looked at from different angles (e.g., Seiz et al. 2006). Another possibility for the detection of cloud height is the exploitation of measurements within the oxygen A absorption band at 0.76 μm (Yamamoto and Wark 1961; Wu 1985; Fischer and Grassl 1991; Preusker and Lindstrot 2009). The mass of oxygen along the path of reflected sunlight is derived from measurements inside the absorption band. Because oxygen is well mixed in the atmosphere, the traversed mass of air can directly be determined, allowing for the retrieval of cloud-top pressure. Since the launch of the Environmental Satellite (Envisat) on 1 March 2002, measurements inside the O2 A band are available from the Medium Resolution Imaging Spectrometer (MERIS; Rast et al. 1999). MERIS, primarily designed for the remote sensing of coastal waters, is operationally used for the retrieval of cloud properties like optical thickness and height. The oxygen A-band method allows for accurate retrievals of cloud-top pressure, especially in the case of low clouds, where the common techniques exploiting the thermal emission of clouds are less sensitive. In a validation campaign using airborne lidar measurements of cloud-top height, Lindstrot et al. (2006) found an accuracy of 25 hPa for low, single-layered clouds. In the case of clear sky, the MERIS measurements inside the oxygen A band can be used for a retrieval of surface pressure that can be an important tool for cloud detection in satellite imagery. Apart from the uncertainty introduced by the influence of geophysical parameters like, for example, the temperature profile, the upper limit for the achievable retrieval accuracy is fixed by instrumental constraints like sensor noise. In the case of MERIS measurements in the O2 A band, one of the largest sources of error is instrumental stray light. If properly corrected, the measurements of MERIS allow a surface pressure retrieval accuracy of approximately 10–15 hPa (Lindstrot et al. 2009).

Instrumental stray light is caused by multiple scattering and reflection at optical elements within the spectrometer like lenses or gratings. The correction of stray light is particularly important in absorption bands because weak intensities are affected strongly even by small offsets caused by stray radiation. Therefore, the O2 A-band-based algorithms for the retrieval of surface and cloud-top pressure from MERIS are susceptible to errors caused by instrumental stray light. Although there is a correction for stray radiation in the operational MERIS processing chain (MERIS ground segment, Merheim-Kealy et al. 1999), artifacts are apparent in the pressure retrievals in particular, which are likely to be caused by residual stray light. The quantification of the stray light effect on the retrieval errors is complicated by a high correlation with the effect of the spectral calibration uncertainty: a spectral shift of the oxygen A-band channel toward weaker or stronger absorption causes a similar signal as an under- or overestimation of the stray light contribution to the measured radiance. Because the MERIS swath is composed of the measurements of five identical cameras with individual characteristics (see section 2a), the errors induced by stray light and spectral calibration issues become evident particularly at the borders of the field of view of the cameras, resulting in discontinuities of the derived pressure. In the past, these errors as a whole were attributed to the spectral calibration inaccuracy, assuming that the operational stray light correction scheme worked to a satisfying degree. Therefore, an individual spectral calibration was performed for the MERIS oxygen A-band channel, deviating from the operational spectral calibration by up to 0.2 nm. In view of the contradicting results of the two calibration efforts, hinting at the presence of residual stray light in the oxygen A band and the fact that the errors in the derived pressure products could not be eliminated following this approach, the authors pursued a different strategy. Using the operational spectral calibration, the coefficients of a simple stray light model were optimized along with adjusting the retrieved pressure values to reference maps. This procedure was applied to operational level 1 data, already subject to the operational stray light correction in advance, as well as to MERIS spectral campaign data. MERIS spectral campaigns are a series of consecutive orbits dedicated to the spectral calibration of MERIS, which are conducted repeatedly during the lifetime of the mission and characterized by a modified spectral band setting, providing the maximal possible spectral resolution inside the oxygen A band. Due to the modified spectral band setting, the operational stray light correction is not applied to this data. In the case of the operationally corrected data, the results obtained with our method will not represent a realistic estimate of the stray light contribution to the oxygen A-band channel but will provide a possibility to correct the operational O2 A-band radiance to eliminate the artifacts in derived pressure products. In the case of the spectral campaign data, the retrieved stray light correction coefficients indicate the actual amount of stray light in the oxygen A band, averaging out at roughly 5% of the radiance available in the absorption-free vicinity of the absorption band (see section 4).

The instrument MERIS and the stray light model used in the frame of this work are introduced in section 2. Section 3 describes the correction strategies used for the surface and cloud-top pressure retrievals. The results of the stray light optimization and applications to independent MERIS data are shown in sections 4 and 5, respectively. Conclusions are given in section 6.

2. MERIS

a. Instrument overview

MERIS is a programmable, medium-spectral-resolution, imaging spectrometer (Rast et al. 1999). It is one of the 10 core instruments on the polar orbiter Envisat (an environmental satellite launched on 1 March 2002), flying at 800 km in a sun-synchronous orbit with an equator crossing time of 1000 local time (LT), descending node, and 98.5° inclination. MERIS consists of five identical pushbroom imaging spectrometers operating in the solar spectral range (390 to 1040 nm), arranged in a fan-shape configuration that covers a total field of view of 68.5° and spans a swath width of around 1150 km. The fields of view of the five spectrometers therefore each cover a different area on the ground while all the cameras are observing the same spectral range. The spectral dispersion is achieved by mapping the entrance slit of a grating spectrometer onto a charge-coupled device (CCD) array. The integration time, instrument optics, and CCD array resolution are adjusted such that MERIS has a spatial resolution of 260 m × 300 m, a maximum spectral resolution of 1.6 nm, and a spectral sampling of 1.25 nm. The instrument electronic data rate provides 15 channels that are programmable by ground command in spectral width and position, with each channel obtained by integrating over several detector rows in spectral direction of the CCD array. In the regular operation mode, the spatial resolution is reduced by a factor of 4 along and across track (reduced resolution mode). In the full-resolution mode, the full spatial resolution is transmitted.

The central wavelengths of the spectral channels as listed in Table 1 vary slightly across the field of view of MERIS. This “spectral smile” is caused by curvature of the image of the slit formed in the focal plane array, resulting in viewing angle–dependent central wavelengths of the spectral MERIS channels. To accurately determine the spectral smile of MERIS, spectral calibration campaigns are conducted repeatedly, using the full possible spectral resolution in the oxygen A band and solar Fraunhofer lines (Delwart et al. 2007).

Table 1.

Central wavelength and bandwidth (fwhm) of MERIS.

Central wavelength and bandwidth (fwhm) of MERIS.
Central wavelength and bandwidth (fwhm) of MERIS.

b. Stray light and spectral calibration

Stray light can, for example, refer to light of a frequency other than intended or light-following paths other than intended. It is typically caused by multiple reflection and scattering at optical elements like lenses or gratings. As a result, both the spectral as well as the spatial purity are degraded, for example, causing a virtual monochromatic point source to be imaged slightly blurred both spectrally and spatially. According to the MERIS preflight characterization, stray light within the instrument is composed of two differing contributions: a diffuse part, caused by internal light scatter, and grating ghosts. Within the MERIS ground segment (MEGS; Merheim-Kealy et al. 1999), the measured radiances are corrected for instrumental stray light. However, for computing performance reasons, the current operational stray light correction uses a rather coarse spectral sampling. As the diffuse part of the stray light is spectrally smooth, it is correctly accounted for by the operational stray light correction where the input spectra are smooth as well. This requirement is fulfilled for all MERIS bands except for channel 11 within the oxygen A band. The less pronounced ghost contributions are poorly corrected for the same reason (L. Bourg, ACRI-ST, 2009, personal communication).

As large discontinuities in the derived products can be found especially at the borders of the field of view of the MERIS cameras, the correction scheme is suspected to be insufficient. To improve the performance of the MERIS pressure retrievals, a secondary empirical stray light correction was developed in the frame of this work. A similar correction scheme has been developed for the MERIS spectral calibration orbits, characterized by a modified band setting with a maximal spectral resolution within the O2 A band (Fig. 1). These measurements are not subject to the operational stray light correction, and therefore enable a determination of the amount of stray light using the empirical correction approach.

Fig. 1.

(top) Atmospheric transmission around 0.76 μm (air mass = 1) and (bottom) MERIS spectral channel response functions for nominal and spectral campaign band setting.

Fig. 1.

(top) Atmospheric transmission around 0.76 μm (air mass = 1) and (bottom) MERIS spectral channel response functions for nominal and spectral campaign band setting.

The stray light effect is stronger in case the target is bright because more photons are available. The modeled instrumental stray light was therefore related to the brightness of the observed scene. To account for in-camera slopes of the stray light and the effects of a higher order at the camera boundaries, an additional dependency on the across-track pixel number of MERIS had to be taken into account. Stray light within and close by the O2 A band was therefore assumed to depend linearly on the brightness in window channel 10 at 754 nm, with a polynomial dependence on the normalized across-track pixel index x, with −1 ≤ x ≤ 1:

 
formula

where f is a factor relating the amount of radiance of band 10 to the corrective offset to be subtracted from the O2 A-band channel:

 
formula

Because the error caused by instrumental stray light is a radiance offset, it is hard to separate from the spectral uncertainty of MERIS band 11, which causes the channel to be slightly shifted toward weaker or stronger gaseous absorption as compared to the nominal position. The MERIS spectral calibration is operationally performed by analyzing the position of well-known solar Fraunhofer lines in the measured spectrum of MERIS. As an exception, the O2 A-band channel was spectrally calibrated by analyzing the shape of the oxygen A band, exploiting the full possible spectral resolution of 1.25 nm between 758 and 770 nm available during the spectral calibration orbits (Delwart et al. 2007). The difference between the two calibration approaches is in the region of 0.2 nm and is suspected to be caused by either residual stray light in the oxygen A band or the temperature dependence of the absorption lines. Consequently, in the frame of this study the center wavelength of MERIS band 11 was set to the values found by the Fraunhofer calibration and adjusted within a range of ±0.1 nm, represented by the fourth parameter d to be optimized:

 
formula

The coefficients a, b, c, and d were found by minimizing the retrieval errors of surface and cloud-top pressure, as detailed in section 3. Therefore, the bias of the pressure retrievals was analyzed at regularly spaced grid points of the parameter space spanned by the coefficients a, b, c, and d. The optimal set of coefficients was found by iteratively narrowing the value’s margin for each coefficient, centered around the state resulting in the smallest pressure bias in the previous iteration step. This procedure was performed separately for each MERIS camera to account for the deviating behavior of the individual modules. The pressure adjustment was conducted simultaneously for a number of selected clear-sky and cloudy scenes to ensure the global validity of the found correction coefficients.

3. Correction strategy

a. Surface pressure

The algorithm for the retrieval of land surface pressure from MERIS measurements (SPFUB) is based on the exploitation of the gaseous absorption by oxygen around 762 nm. As oxygen is well mixed both horizontally and vertically in the atmosphere, the observed amount of oxygen along the light path is directly related to the traversed air mass, allowing for the determination of the surface pressure. The amount of oxygen is derived from the transmission in the oxygen A band, approximated by the radiance ratio of the absorption channel 11 and a virtual absorption-free channel at the same spectral location, interpolated from channels 10 and 12 (see Fig. 1). The inversion of the top-of-the-atmosphere (TOA) radiance measurements is performed by an artificial neural network (ANN) trained with radiative transfer simulations. The Matrix Operator Model (MOMO; Fell and Fischer 2001) was used to build the database for the training of the ANN. An advanced k-distribution technique was used for the calculation of the oxygen absorption (Bennartz and Fischer 2000), assuming a fixed temperature profile. A detailed description of SPFUB can be found in Lindstrot et al. (2009). To use SPFUB for the optimization of surface pressure in the spectral campaign orbits, the ANN had to be adapted to the spectral campaign band setting, using three narrow channels inside the oxygen A band, instead of the nominal channel 11 (see Fig. 1).

The surface pressure retrieval algorithm is well suited for the optimization of the O2 A stray light model because

  1. the main piece of information is provided by the O2 A channel to be corrected and

  2. accurate reference data are available from terrain-corrected sea level pressure.

The European Centre for Medium-Range Weather Forecasts (ECMWF) sea level pressure is available from MERIS level 1 data files, whereas the surface elevation could be extracted from digital elevation models (DEMs). The DEM derived from measurements of the Geoscience Laser Altimeter System (GLAS) on board the National Aeronautics and Space Administration (NASA) satellite IceSat (DiMarzio et al. 2007) was used to calculate the surface pressure above Greenland; the Shuttle Radar Topography Mission (SRTM; Farr et al. 2007) DEM was used elsewhere. Both datasets are spatially highly resolving [spatial resolution: l km (GLAS/Greenland), 90 m (SRTM)], accurate height models that can be regarded as nearly free of errors for the purpose of this work. To obtain the DEM-corrected pressure SPDEM, the ECMWF sea level pressure PECMWF was converted using the terrain height h and the surface temperature t:

 
formula

with g = 9.806 65 m s−2 representing the gravitation constant; R = 287.05 J kg−1 K−1 is the specific gas constant of dry air; C = 0.11 K hPa−1 is a constant accounting for the influence of the humidity; e is the vapor pressure at the surface; and γ = 0.0065 K m−1 is the temperature lapse rate in the atmosphere. This formula, accounting for the temperature at the surface, was chosen instead of assuming an average atmosphere to be consistent with the standard procedures of reducing in situ–measured surface pressure to sea level.

b. Cloud-top pressure

Similar to the retrieval of surface pressure, the cloud-top pressure (CTP) can be derived from MERIS measurements in the oxygen A band (Preusker and Lindstrot 2009). Here, the inversion of the measured radiance ratio is further complicated by the unknown influence of multiple scattering inside the cloud layer and between the surface and the cloud. To account for these effects, the surface albedo is extracted from a global albedo database (MERIS AlbedoMap; Muller et al. 2007), whereas the in-cloud scattering is estimated from the cloud optical thickness, which is reasonably well derivable from measurements of the window channel 10 at 754 nm (Fischer and Grassl 1991). The algorithm for the retrieval of cloud-top pressure (CTPFUB) is an ANN, trained with MOMO radiative transfer simulations (Fischer et al. 1997). CTPFUB was not adapted to the spectral campaign band setting and is only used for the optimization of stray light in the nominal band setting.

Because the true height of the clouds observed by MERIS is not known with sufficient accuracy, the adjustment of CTPFUB was limited to eliminating the discontinuities in CTP at the borders of the field of view of the MERIS cameras. MERIS scenes providing extended homogeneous stratocumulus fields in the Southern Atlantic Ocean were selected for the optimization. To account for small across-track trends in cloud height, simultaneous observations of the Spinning Enhanced Visible and Infrared Imager (SEVIRI; Aminou 2002) on board Meteosat Second Generation (MSG) were analyzed—the cloud-top temperature as observed by SEVIRI at 10.8 μm was converted to cloud-top pressure using ECMWF temperature profiles. This simple method for the retrieval of CTP is susceptible to the temperature profile inaccuracy and the unknown amount of water vapor above the low stratocumulus deck, and therefore it is likely to result in a biased CTP retrieval. However, the derived cloud heights are free of any MERIS-like camera effects and therefore served as a relative reference—the SEVIRI-derived CTP was shifted to make both cloud height retrievals match on average. Then, the MERIS camera effects were eliminated by fitting CTPFUB to the shifted SEVIRI-derived CTP by optimizing the stray light coefficients, as detailed in section 2b.

c. Scene selection

1) Nominal band setting

Various MERIS scenes above the desert (from orbits 18 598, 18 755, 18 897, and 18 926) and Greenland (from orbits 17 714, 17 728, and 17 729) were selected for the surface pressure adjustment. The scenes were thoroughly checked for being free of clouds, using the MERIS cloud mask and visual inspection, and matching the temperature profiles used for the training of the SPFUB artificial neural networks on average. A tropical and a subarctic summer temperature profile were assumed for the desert and Greenland scenes, following the definition by McClatchey et al. (1972). However, the heterogeneity of the temperature profile over the selected scenes causes a small, variable bias in surface pressure, reducing the accuracy of the optimized correction factors. In addition to the clear-sky observations, several homogeneous cloud scenes above the Benguela Current region (from orbits 16 394, 18 899, and 18 956) were chosen for the optimization of the CTP retrieval. The resulting scene composite, entirely recorded between April and October 2005 and shown in Fig. 2 (left panel), covers the whole swath of MERIS to ensure that all cameras are represented in the optimization dataset.

Fig. 2.

MERIS scenes used for the optimization of (left) the stray light model for the nominal band setting and (right) the spectral campaign band setting.

Fig. 2.

MERIS scenes used for the optimization of (left) the stray light model for the nominal band setting and (right) the spectral campaign band setting.

2) Spectral campaign band setting

The stray light optimization was performed on clear-sky scenes only, using the SPFUB algorithm adapted to the spectral campaign band setting. The optimization dataset was composed of three desert scenes above the Arabian Peninsula, the Libyan Desert, and the western Sahara, recorded during a spectral campaign on 13 December 2008 (MERIS orbits 35 488–35 490). The scenes are shown in Fig. 2 (right panel).

4. Results

a. Nominal band setting

The effect of the optimization on the retrieved pressure is shown separately for scenes over desert, ice, and clouds in Fig. 3. The displayed curves are median values of the along-track columns depending on the across-track pixel number to demonstrate the camera effects visible in the across-track direction. For all regimes, the retrieval based on the Fraunhofer spectral calibration and unmodified level 1 radiances exhibits strong discontinuities at the borders and artificial slopes inside the field of view of the individual cameras. The deviation from the reference pressure exceeds 100 hPa at some locations. In contrast, the retrieval based on stray light–corrected radiances is close to the reference data for all regimes and shows hardly any camera artifacts. Apparently, the assumed simple stray light model is suitable for the removal of the strong camera effects, and the correction quality does not depend on the regime under consideration. The correction is therefore applicable over clear-sky and cloudy scenes at different brightness levels.

Fig. 3.

(top), (middle) Along-track median values of surface pressure above desert and Greenland derived from DEM (solid line), and MERIS measurements before (dashed) and after (dotted) optimization. (bottom) Same as (top) and (middle) but for cloud-top pressure, solid line represents MSG-derived CTP, shifted to match average MERIS-derived CTP.

Fig. 3.

(top), (middle) Along-track median values of surface pressure above desert and Greenland derived from DEM (solid line), and MERIS measurements before (dashed) and after (dotted) optimization. (bottom) Same as (top) and (middle) but for cloud-top pressure, solid line represents MSG-derived CTP, shifted to match average MERIS-derived CTP.

The resulting factor f for the correction of MERIS band 11 radiance and the deviation from the operational Fraunhofer calibration as found by the optimization algorithm are shown in Fig. 4. As detailed in section 2b, the nominal band setting data are operationally corrected for stray light in the MERIS ground segment; therefore, f does not represent the actual amount of stray light present within MERIS band 11, but rather it is a correction factor compensating for a too strong or too weak correction within the MERIS ground segment. Following the optimization results, the operational correction scheme underestimates the amount of stray light for cameras 1–3 and 5, whereas it is overestimated for camera 4. In addition, the Fraunhofer calibration is modified by the optimization approach, exhibiting a positive wavelength shift for cameras 1, 2, and 5 and a negative shift for cameras 3 and 4. There is a high correlation of the correction factor f and the center wavelength of band 11, presumably caused by the disability of the operational correction scheme to separate the effects of spectral shifts and stray light.

Fig. 4.

Optimization results for nominal band setting: (top) Additive stray light factor f and (bottom) central wavelength of MERIS channel 11 as resulting from optimization (gray) and Fraunhofer calibration (black), depending on MERIS detector index.

Fig. 4.

Optimization results for nominal band setting: (top) Additive stray light factor f and (bottom) central wavelength of MERIS channel 11 as resulting from optimization (gray) and Fraunhofer calibration (black), depending on MERIS detector index.

b. Spectral campaign band setting

As mentioned in section 2b, the spectral campaign data are not operationally corrected for stray light. The correction factor f, found by the pressure adjustment algorithm, therefore does not merely represent a correction factor but corresponds to the actual amount of stray light present within the oxygen A-band channels. Similar to Fig. 3 for the nominal band setting case, Fig. 5 shows the effect of the stray light optimization on the retrieved surface pressure. Here, the retrieval based on uncorrected radiance shows a large negative bias of surface pressure for all cameras, as compared to the terrain-corrected sea level pressure. The in-camera slopes of the bias are less pronounced, and so are the discontinuities at the borders of the field of view of the cameras. The mean bias of the retrieved surface pressure is about −100 hPa. The retrieval based on stray light–corrected radiances is very close to the terrain-corrected sea level pressure without any visible camera artifacts.

Fig. 5.

Along-track median values of surface pressure derived from DEM (solid lines) and MERIS spectral campaign measurements before (dashed) and after (dotted) optimization.

Fig. 5.

Along-track median values of surface pressure derived from DEM (solid lines) and MERIS spectral campaign measurements before (dashed) and after (dotted) optimization.

The correction factor f and the deviation from the spectral Fraunhofer calibration as found by the optimization approach are shown in Fig. 6. A relatively homogeneous amount of stray light was found in the oxygen A band, adding up to 4%–7% of the amount of radiance in band 10 L10. Compared to the nominal band setting data, the in-camera slope of the stray light is negligible. The dependence of f on the across-track pixel index is not correlated to the spectral slope of the center wavelength. This supports the assumption that the spectral Fraunhofer calibration is correct and a separation of the effects of spectral shifts and stray radiation is possible using the proposed empirical correction approach.

Fig. 6.

Same as in Fig. 4, but for spectral campaign band setting.

Fig. 6.

Same as in Fig. 4, but for spectral campaign band setting.

5. Application to independent MERIS data

The nominal band setting correction factors have been applied to MERIS scenes that were not part of the optimization scene composite to demonstrate the global validity of the results. Two scenes are shown, a desert scene from 15 October 2004 used for the retrieval of surface pressure and a stratocumulus cloud scene from 16 May 2007.

a. Surface pressure

The selected clear-sky scene is located above the Libyan Desert. Figure 7 shows a red–blue–green (RGB) image of the scene, the surface pressure retrieval based on both uncorrected and corrected radiances, and the terrain-corrected sea level pressure. In addition, scatterplots of the two retrievals and the terrain-corrected pressure and the along-track median of all three datasets are shown. Again, the retrieval based on unmodified radiances exhibits large camera effects with strong in-camera slopes and discontinuities at the borders of the field of view of the cameras of up to 120 hPa. The spatial structures of surface pressure are masked by the strong deviations caused by the insufficient correction for instrumental stray light. By applying the empirical correction factors, the camera effects completely disappear and the retrieved surface pressure is very close to the terrain-corrected sea level pressure. The root-mean-square error of the retrieval is reduced from 36 to 5 hPa. Apparently, the correction factors found by optimizing selected surface and cloud-top pressure scenes are suitable for the correction of independent clear-sky scenes.

Fig. 7.

Effect of empirical stray light correction on retrieval of surface pressure: (top right) RGB image of Libyan desert on 15 Oct 2004; retrieved surface pressure (SPFUB) from (top left) uncorrected and (middle left) corrected radiances; (middle right) terrain-corrected sea level pressure SPDEM; (bottom left) scatterplot of corrected/uncorrected SPFUB and SPDEM; (bottom right) along-track median of SPFUB and SPDEM, depending on across-track pixel index.

Fig. 7.

Effect of empirical stray light correction on retrieval of surface pressure: (top right) RGB image of Libyan desert on 15 Oct 2004; retrieved surface pressure (SPFUB) from (top left) uncorrected and (middle left) corrected radiances; (middle right) terrain-corrected sea level pressure SPDEM; (bottom left) scatterplot of corrected/uncorrected SPFUB and SPDEM; (bottom right) along-track median of SPFUB and SPDEM, depending on across-track pixel index.

b. Cloud-top pressure

The scene selected for the retrieval of cloud-top pressure is located above the Benguela Current region and characterized by a homogeneous stratocumulus field. Figure 8 shows an RGB image, the CTPFUB retrieval based on both uncorrected and corrected radiances, and the along-track median of both retrievals. Because the true cloud-top height is not known, the applied approach can only be evaluated with regard to what extent the artificial camera effects are reduced. Again, the uncorrected retrieval shows large camera artifacts caused by the insufficient correction of stray light. The errors are strongly reduced in case stray light–corrected radiances are used for the retrieval. However, residual camera effects are still visible.

Fig. 8.

Effect of empirical stray light correction on retrieval of cloud-top pressure: (top right) RGB image of stratocumulus clouds above Benguela Current on 16 May 2007; retrieved cloud-top pressure (CTPFUB) from (top left) uncorrected and (bottom left) corrected radiances; (bottom right) along-track median of CTPFUB for both retrievals depending on across-track pixel index.

Fig. 8.

Effect of empirical stray light correction on retrieval of cloud-top pressure: (top right) RGB image of stratocumulus clouds above Benguela Current on 16 May 2007; retrieved cloud-top pressure (CTPFUB) from (top left) uncorrected and (bottom left) corrected radiances; (bottom right) along-track median of CTPFUB for both retrievals depending on across-track pixel index.

6. Conclusions

A simple model of the instrumental stray light within the MERIS oxygen A-band channel was optimized by fitting the retrieved pressure to accurate reference data. The main results of the stray light optimization are as follows:

  1. The simple stray light model, assuming a linear dependence on the brightness of the observed scene and a polynomial dependence on the spatial across-track pixel index of each camera, is suitable for a correction of the camera artifacts apparent in the derived pressure products.

  2. A homogeneous amount of stray light was found in the spectral campaign orbits, which are not corrected for stray light in the operational processing chain. Here, 4%–7% of the window radiance L10 have to be subtracted from the oxygen A-band channels to compensate the bias in derived surface pressure (∼−100 hPa).

  3. The correction factor found for the nominal band setting data shows strong in-camera slopes that follow the shape of the center wavelengths, indicating that the operational stray light correction scheme applied within the MERIS ground segment is not able to separate the effects of stray light and wavelength shifts.

  4. The amount of stray light in the nominal channel 11 is overestimated for camera 4 and underestimated for all other cameras by the operational stray light correction scheme.

  5. The center wavelength of MERIS band 11 is slightly adjusted by the optimization algorithm. The adjustment does not exceed ±0.1 nm, which is the expected accuracy of the Fraunhofer calibration (Delwart et al. 2007).

The current operational stray light correction scheme of the MERIS ground segment is not sufficient to correct the effects of stray radiation within the MERIS instrument. The proposed empirical correction factors are suitable for a significant reduction of the camera effects, as demonstrated by applying the correction to independent data (see section 5).

A further improvement of the correction accuracy will be obtained by extending the surface pressure retrieval. The current version is based on fixed temperature profiles, causing a bias in retrieved surface pressure in case the atmosphere is warmer or colder than assumed. The enhanced version will include the actual temperature profile, enabling a more accurate correction of stray light by a reduction of the surface pressure bias.

Acknowledgments

This work was carried out in the frame of the ESA-funded project “Exploitation of the MERIS Oxygen A Band” (Contract 20693/07/I-OL). The authors thank two anonymous reviewers for their comments and L. Bourg (ACRI-ST) for valuable information about the MERIS stray light characterization and nominal correction.

REFERENCES

REFERENCES
Aminou
,
D.
,
2002
:
MSG’s SEVIRI instrument.
ESA Bull.
,
111
,
15
17
.
Bennartz
,
R.
, and
J.
Fischer
,
2000
:
A modified k-distribution approach applied to narrow-band water vapour and oxygen absorption estimates in the near infrared.
J. Quant. Spectrosc. Radiat. Transfer
,
66
,
539
553
.
Delwart
,
S.
,
R.
Preusker
,
L.
Bourg
,
R.
Santer
,
D.
Ramon
, and
J.
Fischer
,
2007
:
MERIS inflight spectral calibration.
Int. J. Remote Sens.
,
28
,
479
496
.
DiMarzio
,
J.
,
A.
Brenner
,
R.
Schutz
,
C.
Shuman
, and
H. J.
Zwally
,
2007
:
GLAS/ICESat 1 km laser altimetry digital elevation model of Greenland. NSIDC.
.
Farr
,
T. G.
, and
Coauthors
,
2007
:
The shuttle radar topography mission.
Rev. Geophys.
,
45
,
RG2004
.
doi:10.1029/2005RG000183
.
Fell
,
F.
, and
J.
Fischer
,
2001
:
Numerical simulation of the light field in the atmosphere–ocean system using the matrix-operator method.
J. Quant. Spectrosc. Radiat. Transfer
,
3
,
351
388
.
Fischer
,
J.
, and
H.
Grassl
,
1991
:
Detection of cloud-top height from backscattered radiances within the oxygen A band. Part 1: Theoretical study.
J. Appl. Meteor.
,
30
,
1245
1259
.
Fischer
,
J.
,
R.
Preusker
, and
L.
Schüller
,
1997
:
ATBD 2.3 cloud-top pressure.
European Space Agency, Algorithm Theoretical Basis Doc. PO-TN-MEL-GS, 28 pp
.
Lindstrot
,
R.
,
R.
Preusker
,
T.
Ruhtz
,
B.
Heese
,
M.
Wiegner
,
C.
Lindemann
, and
J.
Fischer
,
2006
:
Validation of MERIS cloud-top pressure using airborne lidar measurements.
J. Appl. Meteor. Climatol.
,
45
,
1612
1621
.
Lindstrot
,
R.
,
R.
Preusker
, and
J.
Fischer
,
2009
:
The retrieval of land surface pressure from MERIS measurements in the oxygen A band.
J. Atmos. Oceanic Technol.
,
26
,
1367
1377
.
McClatchey
,
R.
,
R.
Fenn
,
J.
Selby
,
F.
Volz
, and
J.
Garing
,
1972
:
Optical Properties of the Atmosphere.
3rd ed. Air Force Cambridge Research Laboratories, 108 pp
.
Menzel
,
W. P.
,
W. L.
Smith
, and
T. R.
Stewart
,
1983
:
Improved cloud motion wind vector and altitude assignment using VAS.
J. Climate Appl. Meteor.
,
22
,
377
384
.
Menzel
,
W. P.
,
R. A.
Frey
,
B. A.
Baum
, and
H.
Zhang
,
2006
:
Cloud-top properties and cloud phase algorithm theoretical basis document.
Cooperative Institute for Meteorological Satellite Studies, Algorithm Theoretical Basis Doc. MOD04, 62 pp
.
Merheim-Kealy
,
P.
,
J. P.
Huot
, and
S.
Delwart
,
1999
:
The MERIS ground segment.
Int. J. Remote Sens.
,
20
,
1703
1712
.
Muller
,
J-P.
,
R.
Preusker
,
J.
Fischer
,
M.
Zuhlke
,
C.
Brockmann
, and
P.
Regner
,
2007
:
ALBEDOMAP: MERIS land surface albedo retrieval using data fusion with MODIS BRDF and its validation using contemporaneous EO and in situ data products.
Proc. IGARSS Int. Geoscience and Remote Sensing Symp., Barcelona, Spain, Institute of Electrical and Electronics Engineers, 2404–2407
.
Preusker
,
R.
, and
R.
Lindstrot
,
2009
:
Remote sensing of cloud-top pressure using moderately resolved measurements within the oxygen A band—A sensitivity study.
J. Appl. Meteor. Climatol.
,
48
,
1562
1574
.
Rast
,
M.
,
J. L.
Bezy
, and
S.
Bruzzi
,
1999
:
The ESA Medium Resolution Imaging Spectrometer MERIS–A review of the instrument and its mission.
Int. J. Remote Sens.
,
20
,
1681
1702
.
Seiz
,
G.
,
R.
Davies
, and
A.
Gruen
,
2006
:
Stereo cloud-top height retrieval with ASTER and MISR.
Int. J. Remote Sens.
,
27
,
1839
1853
.
Smith
,
W. L.
, and
C. M. R.
Platt
,
1978
:
Intercomparison of radiosonde, ground-based laser, and satellite-deduced cloud heights.
J. Appl. Meteor.
,
17
,
1796
1802
.
Wielicki
,
B. A.
, and
J. A.
Coakley
,
1981
:
Cloud retrieval using infrared sounder data: Error analysis.
J. Appl. Meteor.
,
20
,
157
169
.
Wu
,
M-L. C.
,
1985
:
Remote sensing of cloud-top pressure using reflected solar radiation in the oxygen A band.
J. Climate Appl. Meteor.
,
24
,
539
546
.
Yamamoto
,
G.
, and
D.
Wark
,
1961
:
Discussion of the letter by R. A. Hanel: Determination of cloud altitude from a satellite.
J. Geophys. Res.
,
66
,
3596
.

Footnotes

Corresponding author address: Rasmus Lindstrot, Institut für Weltraumwissenschaften, Freie Universität Berlin, Carl-Heinrich-Becker-Weg 6-10, 12165 Berlin, Germany. Email: rasmus.lindstrot@wew.fu-berlin.de