Abstract

The combination of multiple satellite instruments on a pixel-by-pixel basis is a difficult task, even for instruments collocated in space and time, such as the Moderate Resolution Imaging Spectroradiometer (MODIS) and Atmospheric Infrared Sounder (AIRS) on board the Earth Observing System (EOS) Aqua. Toward the goal of an improved collocation methodology, the channel- and scan angle–dependent spatial response functions of AIRS that were obtained from prelaunch measurements and calculated impacts from scan geometry are shown within the context of radiance comparisons. The AIRS spatial response functions are used to improve the averaging of MODIS radiances to the AIRS footprint, and the variability of brightness temperature differences (ΔTb) between MODIS and AIRS are quantified on a channel-by-channel basis. To test possible connections between ΔTb and the derived level 2 (L2) datasets, cloud characteristics derived from MODIS are used to highlight correlations between these quantities and ΔTb, especially for ice clouds in H2O and CO2 bands. Furthermore, correlations are quantified for temperature lapse rate (dT/dp) and the magnitude of water vapor mixing ratio (q) obtained from AIRS L2 retrievals. Larger values of dT/dp and q correlate well to larger values of ΔTb in the H2O and CO2 bands. These correlations were largely eliminated or reduced after the MODIS spectral response functions were shifted by recommended values. While this investigation shows that the AIRS spatial response functions are necessary to reduce the variability and skewness of ΔTb within heterogeneous scenes, improved knowledge about MODIS spectral response functions is necessary to reduce biases in ΔTb.

1. Introduction

A large number of atmospheric sounding and imaging instruments, located on a variety of satellite platforms and with widely varying orbital characteristics, have obtained a rich array of information, either individually or in synergy, about the earth’s atmosphere in recent decades. The A-Train (Stephens et al. 2002), a coordinated earth-viewing satellite constellation, has provided unprecedented information about the atmosphere, ocean, and cryosphere. Although the assorted A-Train platforms are closely collocated in time and space, combining them has proven to be nontrivial (Nagle and Holz 2009) with many instrument-specific challenges. However, a sufficiently precise combination may either increase the information content of existing geophysical parameters (e.g., L’Ecuyer et al. 2006) or provide entirely new ones (e.g., Baran and Francis 2004). Furthermore, improved collocation is also informative toward the development of future mission concepts, for instance, those based on combinations of high spatial and spectral resolution sounding and imaging observations (Pagano et al. 2006).

A state-of-the-art high-spatial-resolution imager and a high-spectral-resolution sounder, the Moderate Resolution Imaging Spectroradiometer (MODIS; Barnes et al. 1998) and the Atmospheric Infrared Sounder (AIRS; Aumann et al. 2003), respectively, are both residing on the National Aeronautics and Space Administration (NASA) satellite platform Aqua. MODIS has high spatial resolution (∼1 km) and several broadband channels in the visible, near-infrared, and midinfrared spectral range. AIRS has much higher spectral resolution within the midinfrared, but the spatial resolution (∼13.5 km) is an order of magnitude coarser than that of MODIS and does not have the same panspectral coverage of MODIS. The combination of MODIS and AIRS on Aqua provides an opportunity to quantify the trade-offs between spectral and spatial coverage and, furthermore, to quantify the added value of atmospheric (and surface) retrievals from a combined multisensor, multiwavelength retrieval approach.

Previous investigations have used combinations of MODIS and AIRS, either together or with other instruments, for the intercalibration of radiances (Tobin et al. 2006, hereafter T06), to characterize AIRS subpixel variability in atmospheric and surface properties and retrievals of cloud parameters (Li et al. 2004a,b), to investigate optimal cloud-clearing approaches that exploit the high spatial resolution of MODIS (Li et al. 2005), to compare individual instrument-derived infrared-based cloud products and assess their radiative consistency (Kahn et al. 2007, hereafter K07), and to intercalibrate AIRS and the Infrared Atmospheric Sounding Interferometer (IASI; see, e.g., Wang et al. 2010). However, these studies used estimates of the AIRS field of view (FOV) that assume 1) it is approximately circular at nadir view, 2) the contribution of the total radiance is uniformly distributed within the FOV, and 3) there is no variation of the spatial response between the different AIRS channels. In this study, the channel-dependent spatial response functions FAIRS are used to spatially collocate MODIS and AIRS radiances. The spatial response functions of AIRS are based on prelaunch characterization (Overoye et al. 1999). Measurements of FAIRS allow one to obtain a more “realistic” weighting of the MODIS pixels distributed within an AIRS FOV and also facilitate the quantification of the inherent scene heterogeneity in the radiances and derived geophysical fields.

In principle, improved radiance agreement can be obtained in the presence of heterogeneous clouds (e.g., K07) and variable surface types if a collocation approach does not assume 1–3 above, but rather uses values of FAIRS (e.g., Elliott et al. 2006). Furthermore, the impact of clouds, aerosols, temperature, and water vapor profiles on the radiance comparisons can be quantified. While MODIS provides information about surface, aerosol, and cloud properties, AIRS observes vertical profiles of atmospheric constituents, including temperature and water vapor. These additional retrieval products are used in this work to gain insight on physical causes for radiance differences between MODIS and AIRS.

Section 2 introduces the relevant MODIS and AIRS instrument specifications and aspects of FAIRS. Section 3 investigates the radiance comparisons using different types of FAIRS related to instrument characteristics and scan geometry (truncation, rotation, and smearing). In section 4, FAIRS is used together with derived geophysical products from MODIS and AIRS to investigate atmospheric influences on the radiance comparisons. Section 5 summarizes the main conclusions of this work.

2. Description of MODIS and AIRS

Launched in 2002, the MODIS and AIRS instruments are located onboard the Earth Observing System (EOS) Aqua, a platform in a polar sun-synchronous orbit with an equatorial local crossing time of 0130 (descending) and 1330 (ascending) local crossing time. AIRS is a grating spectrometer with a spectral resolution of νν ≈ 1200. It has 2378 channels in the range of 3.7–15.4 μm with a few spectral gaps, and the FOV is approximately 1.1°, which results in a footprint size of 13.5 km at nadir view. There are 90 cross-track scan angles, with the highest at ±48.95°, yielding a swath width of approximately 1650 km (Aumann et al. 2003). AIRS is coregistered with the Advanced Microwave Sounding Unit (AMSU; Lambrigtsen and Lee 2003) and with the Humidity Sounder for Brazil (HSB), which was operational for only a brief period prior to February 2003. The combination of AIRS–AMSU is used to “cloud clear” AIRS radiances to retrieve temperature, water vapor, and numerous other surface and atmospheric geophysical parameters in the presence of clouds (Susskind et al. 2003).

The MODIS instrument is a spectrometer based on a spectral filter aperture. There are 36 channels in the range of 0.4–14 μm, with bandwidths of 10–500 nm, depending on the channel, and scans as high as ±55° off of nadir. Furthermore, the pixel size depends on the channel and varies from 0.25 to 1.0 km at nadir view. The emphasis of this study is on the infrared channels, all of which have 1.0-km resolution at nadir view, yielding 1354 cross-track pixels resulting in a swath width of approximately 2330 km (Barnes et al. 1998). MODIS uses a radiance lookup table approach to retrieve a wide variety of land, ocean, and ice surface characteristics, and atmospheric and oceanic geophysical properties, including information about clouds and aerosols (Platnick et al. 2003; Remer et al. 2005). In this work, the cloud mask (Ackerman et al. 1998, 2008; Frey et al. 2008) and additional aerosol and cloud parameters are used to sort the AIRS–MODIS radiance comparisons; see section 4 for more detail.

a. MODIS spectral response functions within the AIRS spectrum

Although AIRS has 2378 channels between 3.7 and 15.4 μm, there are small spectral gaps in certain regions. The spectral extent of the AIRS channels is indicated as the discontinuous black line in Fig. 1. Below these lines are the corresponding spectral response functions of the MODIS channels (SRFMODIS; downloaded from http://mcst.gsfc.nasa.gov/index.php?section=32). In principle, MODIS channels 20–25 and 27–36 fall within the spectral range of AIRS. However, as shown by Fig. 1, some SRFMODIS extend across the AIRS spectral gaps, specifically affecting channels 20, 23, 27, 29, and 31. Additionally, AIRS spectra contain channels with excessive noise that must be removed from the comparison, leaving additional yet narrow gaps. The spectral extent of bad channels from AIRS within a given SRFMODIS is small and the cumulative “loss” in spectral coverage ranges from 0% to 15%, depending on the MODIS channel. Furthermore, the MODIS instrument contains an array of 10 detectors (or 20 and 40, for 500- and 250-m resolution, respectively), and each one has slight differences in the shape of SRFMODIS. This is shown with the overlapping functions in Fig. 1. Because the different SRFMODIS within a given channel are very similar, a mean SRFMODIS is calculated per MODIS channel and is used throughout the remainder of this work.

Fig. 1.

The spectral extent of the AIRS channels (discontinuous black line) and the corresponding spectral response functions of the MODIS channels within the AIRS spectral range (M20–M25 and M27–M36, colored curves). Scaling of the x axis is enhanced from 3.5 to 4.7 μm.

Fig. 1.

The spectral extent of the AIRS channels (discontinuous black line) and the corresponding spectral response functions of the MODIS channels within the AIRS spectral range (M20–M25 and M27–M36, colored curves). Scaling of the x axis is enhanced from 3.5 to 4.7 μm.

Some previous studies applied SRFMODIS to radiance comparisons with AIRS, and modifications to them are suggested. For example, T06 showed that a spectral shift of some MODIS channels reduced the bias between MODIS and AIRS for a case study. The suggested spectral “shift” of the response functions for MODIS channels 27, 28, 34, 35, and 36 will be applied to the comparison in sections 3 and 4 to demonstrate improvement resulting from the combination of accurate information about spatial and spectral characteristics of MODIS and AIRS. The reported spectral shifts are −22.9 nm (channel 27), 10.7 nm (channel 28), 15.0 nm (channel 34), −15.5 nm (channel 35), and −20.2 nm (channel 36).

T06 and Gunshor et al. (2009) present methods that “fill” the gaps in the AIRS spectra. Although quantifying the relative importance of spectral gaps on radiance comparisons is another important step toward synergetic applications of MODIS and AIRS, this investigation emphasizes the significant gains obtained with additional spatial information. Therefore, the previous approaches in spectral gap filling are not applied simultaneously with the spatial study presented herein.

b. AIRS spatial response functions

AIRS has a FOV of 1.1° and MODIS of 0.08° at nadir. If AIRS is assumed to have a circular FOV, then approximately 190 MODIS pixels are collocated within AIRS at nadir. Therefore, it is necessary to know the geometrical characteristics of the “larger” AIRS FOV, namely, the spatial distribution of the contributions to the total radiance (Cracknell 1998). Depending on FAIRS, the MODIS pixels have to be weighted differently depending on the positioning of each pixel within the AIRS FOV. Additionally, with increasing scan angle, the number of collocated pixels and their respective weighting within the FOV will also vary, resulting in scan angle–dependent FAIRS (Sun et al. 2006).

Additional complexities arise because FAIRS has a different spatial response for each channel (Elliott et al. 2006). Each AIRS channel’s spatial response was measured before launch by using a spatial collimator system with a resolution of 0.04° (Overoye et al. 1999). These measurements were used to estimate the heterogeneity of the radiances from the perspective of the instrument. Additionally, the instrument scanning characteristics in flight were also taken into account. Because AIRS has 90 different scan angles from +49.5° to −49.5°, a total of 90 values of FAIRS for each of the 2378 channels were obtained to provide a function for every scan angle.

The steps from the measured FAIRS to the final truncated, rotated, and smeared FOVs are illustrated in Fig. 2. As examples, both channels 1 (649.6 cm−1) and 2265 (2554.9 cm−1) are shown. Channel 1 is an example that contains a nearly uniform spatial response, while channel 2265 is an example of an asymmetric function because it is located at the end of the detector array and is partially obscured by detector shadowing. In Fig. 2a, channel 1 approximately resembles the circular and homogeneous view that is assumed in earlier studies (e.g., Li et al. 2004a,b; K07). In the case of channel 2265, the substantial variability of the weighting exists throughout the FOV, which results from the nonuniform detector response. In Fig. 2b, the same response functions are shown as in Fig. 2a, except they are trimmed on both sides by blinds attached to the AIRS FOV shortly before launch to avoid overlapping effects between neighboring footprints. With the resulting turn of the mirror, the spatial response is also rotating. Figure 2c shows the FOV after rotation of Fig. 2b by ∼49°. The final step is to consider the scanning motion and the time integration of the radiances that result in a “smearing” of the response function. This movement can be accounted for with a movement-weighting function (Elliott et al. 2006). The final values of FAIRS are illustrated in Fig. 2d.

Fig. 2.

Examples of the AIRS spatial response functions for (top) channel 1 and (bottom) channel 2265. (a) The prelaunch measurement of the spatial response functions (Overoye et al. 1999); (b) the truncated functions; (c) the rotated functions for a scan angle of 48°; and (d) the final truncated, rotated, and smeared functions (Elliott et al. 2006).

Fig. 2.

Examples of the AIRS spatial response functions for (top) channel 1 and (bottom) channel 2265. (a) The prelaunch measurement of the spatial response functions (Overoye et al. 1999); (b) the truncated functions; (c) the rotated functions for a scan angle of 48°; and (d) the final truncated, rotated, and smeared functions (Elliott et al. 2006).

Apart from the channel-dependent variations in FAIRS, each channel centroid is slightly shifted relative to the others. This so-called Cij shift was quantified for each channel during prelaunch calibration and is generally very small (10 mdeg or less). To test the impact of Cij on FAIRS, the Cij shift was included in the comparisons.

The four types of FAIRS illustrated in Fig. 2 are described by the following terminology: circular at nadir “ROUND” (Fig. 2a); truncated (“TRUNC”; Fig. 2b); rotated (“ROTATED”; Fig. 2c), and smeared (“SMEAR”; Fig. 2d). In the following section, the MODIS–AIRS radiance comparisons for each function type are contrasted against each other, and the added value of truncation, rotation, and smearing on the final spatial response is demonstrated and compared to those obtained from spectral shifts (e.g., T06).

3. Comparison of brightness temperature differences using different spatial response functions

The brightness temperature differences ΔTbTb,MODISTb,AIRS, are quantified for four types of FAIRS over several MODIS channels by comparing derived statistical properties of ΔTb. First, the AIRS channels and their radiances are matched to the averaged MODIS spectral response function (SRFMODIS) and then are weighted by their magnitudes. In this calculation, the wings of SRFMODIS are truncated when the magnitude of SRFMODIS was less than 1%. Second, the MODIS channel radiances are weighted within FAIRS according to their relative positioning within the four types described in Fig. 2. The spectrally and spatially smoothed collocated radiances are then obtained. If i is the index for the spatial position and j is the index for the wavelength, the comparison can be quantitatively described by the following:

 
formula
 
formula

Third, Tb for both AIRS and MODIS is calculated from the central wavelength of the MODIS channels. Fourth, ΔTb was calculated globally for several days. Because all of the days showed similar results, we highlight only one day in this analysis (1 January 2005).

a. ΔTb comparisons with different spatial weighting

The ΔTb for the four weighting functions applied to MODIS channels 32 and 36 are shown in Fig. 3, which represent a “good” and a “bad” example, respectively. On the one hand, there is an improvement for channel 32 between the ROUND and SMEAR functions, namely, that the histogram contracts around ΔTb ∼ 0 K. On the other hand, only a slight improvement is seen for channel 36; however, both maxima are still observed in the histogram and the bias in ΔTb remains. The main improvement for channel 36 results from application of the spectral shift of −20.2 nm (T06) and when combined with the smeared FAIRS, is indicated in Fig. 3 as SHIFT.

Fig. 3.

Distribution of ΔTb (≡Tb,MODISTb,AIRS) for the equivalent of MODIS channels (left) 32 and (right) 36 using the AIRS spatial response functions in Fig. 2.

Fig. 3.

Distribution of ΔTb (≡Tb,MODISTb,AIRS) for the equivalent of MODIS channels (left) 32 and (right) 36 using the AIRS spatial response functions in Fig. 2.

Figure 4 shows the mean, standard deviation, and skewness of ΔTb for all collocation occurrences. The panels on the left (right) show the data for 1 January 2005 (1 April 2005). Additional days were tested and are within a similar range as given by these examples (not shown). As seen in the upper panel, the mean value of ΔTb is marginally reduced at best between the ROUND and SMEAR functions. Therefore, the overall effect is small relative to the spectral shift proposed by T06 for channels 27, 28, 34, 35, and 36. The spectral shifts show a significant improvement in the mean ΔTb for channel 27 (from −1.85 to −0.32 K), channel 28 (from +0.32 to +0.07 K), channel 35 (from +0.61 to +0.01 K), and channel 36 (from +0.90 to +0.05 K). An exception is channel 34, where the spectral shift appears to be overestimated. The Cij shift is negligible and changes ΔTb by 0.01 K or less (not shown). Thus, the effects of Cij on the radiance comparisons do not play an important role. The dashed lines indicate the noise estimates for the different MODIS channels, and, if it is assumed to be the main source of error, both instruments should agree within this level of difference. However, some channels, like channel 27 or 36, show a higher ΔTb between the two instruments. However, a spectral shift, as indicated by the red squares, is able to reduce the difference often within the levels of allowed noise. Also, in Fig. 4b, the standard deviation of ΔTb is reduced substantially when using SMEAR FAIRS, especially for channels 29–33. For instance, channel 33 shows a reduction from 0.8 to 0.5 K and also for channels at smaller wavelengths, like channel 23, where the standard deviation is reduced from 1.2 to 0.7 K.

Fig. 4.

The (a) mean bias, (b) standard deviation, and (c) skewness of ΔTb for MODIS channels 20–36 using the different types of AIRS spatial response functions shown in Figs. 2 and 3. The noise requirements for MODIS are indicated (dashed line): (left) 1 Jan 2005 and (right) 1 Apr 2005.

Fig. 4.

The (a) mean bias, (b) standard deviation, and (c) skewness of ΔTb for MODIS channels 20–36 using the different types of AIRS spatial response functions shown in Figs. 2 and 3. The noise requirements for MODIS are indicated (dashed line): (left) 1 Jan 2005 and (right) 1 Apr 2005.

This result suggests that the SMEAR FAIRS is most important within heterogeneous cloud fields or surface types. For channels 27, 28, 34, 35, and 36, the combination of smearing and shifting FAIRS reduces the standard deviation of ΔTb the most. The skewness of ΔTb in Fig. 4c, which is a proxy for asymmetric behavior in ΔTb, shows a significant reduction using the SMEAR function for most channels. Channel 35 shows a further reduction in skewness resulting from the recommended shift by T06, whereas the spectral shift in channel 36 appears to lead to a slightly stronger asymmetry. Figure 4 demonstrates that transformed FAIRS do not necessarily reduce ΔTb, but demonstrate particular skill in reducing variability of ΔTb in heterogeneous scenes. Further improvements are observed when recommended shifts in SRFMODIS are applied (T06).

b. Cloudy and clear-sky ΔTb and variations with scan angle

A problem seen in previous works that assume a circular FOV is the increasing magnitude of ΔTb variability with scan angle, especially in the presence of clouds (e.g., K07). This effect is quantified with the scan angle–dependent FAIRS. The left column of Fig. 5 shows the scan angle dependence of ΔTb for channel 32 in clear and cloudy skies, using cloud parameters from MYD06_L2 (of collection 5) to determine the presence of cloud coverage within the AIRS footprint. The ΔTb comparison shows that a slightly larger variation for cloudy scenes exists. However, in both cloudy and clear skies, ΔTb is nearly independent of the scan angle, with only a slight increase in the scatter as the scan angle increases from 0° to 49°. This very weak dependence of ΔTb on the scan angle is directly a result of accounting for the AIRS FOV rotation with scan angle (see Fig. 2c). An important aspect of the MODIS and AIRS instruments is the large areal coverage obtained from scanning at large off-nadir scan angles. Because ΔTb is shown to depend very weakly on scan angle for channel 32, this shows that both instruments may be used synergistically within cloudy scenes, even at high scan angles.

Fig. 5.

The variation of ΔTb with scan angle: (left) channels 32 and (right) 36 for (top) clear and (bottom) cloudy sky. Each panel is normalized to its respective maximum value.

Fig. 5.

The variation of ΔTb with scan angle: (left) channels 32 and (right) 36 for (top) clear and (bottom) cloudy sky. Each panel is normalized to its respective maximum value.

The right column of Fig. 5 shows the scan angle dependence of ΔTb for channel 36. A slight dependence on scan angle is observed for angles >±35°, which is also shown in T06. In contrast to channel 32, this variation seems to be systematic and instrument related because ΔTb has opposite tendencies on either side of nadir view and does not depend on the use of FAIRS, which is included in Fig. 5 (right). Furthermore, the “double peak” in ΔTb, described in Fig. 3, is also visible in Fig. 5 (right). The double-peak structure is very weak in cloudy skies as the variation of ΔTb increases and effectively smears the two peaks together.

4. Relationships of ΔTb to clouds and atmospheric gradients

By applying channel- and scan angle–dependent FAIRS to the radiance comparisons, the variability of ΔTb was reduced and skill was shown in limiting the scan angle dependence. However, discrepancies remain in some channels, especially the occurrence of two peaks of ΔTb in channels 21, 22, 27, 30, 35, and 36, which cannot be resolved with FAIRS. The result is that these remaining problems may hinder certain synergistic applications of AIRS and MODIS that require these channels. However, Figs. 4 and 5 also suggest some influence from geophysical quantities like clouds or atmospheric profiles.

Both instruments were designed to obtain atmospheric and surface geophysical parameters and are reported as part of the suite of level 2 (L2) products separately. The combination of the instruments can help to improve the L2 products as long as the variation of ΔTb with these parameters is weak. Furthermore, the L2 products are informative in combination with the radiances, and additional information can be obtained as long as no correlation exists between ΔTb and L2 products. In this section, atmospheric properties obtained from both instruments are used to sort differences in ΔTb. MODIS was designed for spatial observation of clouds, surface, and aerosol parameters (e.g., Platnick et al. 2003; Remer et al. 2005; Menzel et al. 2006), while AIRS was designed to retrieve atmospheric profiles of temperature and water vapor, and several minor gas species (Aumann et al. 2003; Chahine et al. 2006). As shown in Fig. 5, larger differences in ΔTb arise within cloud coverage. Thus, ΔTb and its variability may be correlated to the derived L2 geophysical retrievals. These correlations are quantified below.

a. Correlation of ΔTb to heterogeneous clouds

The parameters in the MODIS MYD06_L2 product (Platnick et al. 2003) are used to quantify ΔTb with cloud coverage (named “cloud mask cloudiness”) within the MYD06_L2 product, cloud phase (“cloud phase infrared”), and the occurrence of thin cirrus at 1-km resolution (“thin cirrus”; Gao and Kaufman 1995) within the AIRS FOV. Additionally, the “heavy aerosol” flag is used to derive aerosol amount within the FOV. Each of these flags is weighted using the smeared values of FAIRS. In other words, MODIS pixels on the edge of the AIRS FOV are weighted less than those near the center. Only AIRS FOVs containing heterogeneous or homogeneous coverage of a single cloud type or aerosol were considered (e.g., no mixtures of clouds or aerosols).

The correlation coefficient r is used to illustrate the relationships between ΔTb and the fraction of coverage in an AIRS FOV for clouds/aerosols and is shown in the upper panel of Fig. 6. The most interesting result is the strong dependence of ΔTb for ice clouds within the absorbing H2O and CO2 channels. In other words, ΔTb increases as the coverage of thick ice clouds increases within the AIRS FOV. As the weighting functions of these channels peak in the midtroposphere, contributions from mid- or high-level clouds is stronger than that from low clouds within these channels.

Fig. 6.

The correlation coefficient r between ΔTb and the coverage of different parameters (aerosol, thin cirrus, water clouds, ice clouds, and all cloud types together) within an AIRS FOV for each MODIS channel. (top) All data points from 90°N and 90°S with polar regions, and (bottom) only data points between 60°N and 60°S to exclude polar regions.

Fig. 6.

The correlation coefficient r between ΔTb and the coverage of different parameters (aerosol, thin cirrus, water clouds, ice clouds, and all cloud types together) within an AIRS FOV for each MODIS channel. (top) All data points from 90°N and 90°S with polar regions, and (bottom) only data points between 60°N and 60°S to exclude polar regions.

The presence of water and thin cirrus clouds seem to have little to no effect on ΔTb, which is consistent with relatively weak radiance signatures in the infrared by these cloud types. The correlation r of ΔTb to the coverage of thicker ice clouds reaches 0.3–0.4. One could argue that problems with MODIS cloud detection over snow or ice in polar regions, or the presence of temperature inversions, could explain the correlations. Thus, the same data are shown in the lower panel of Fig. 6 but are restricted to ±60° latitude. As seen in the lower panel, r is similar, but in the case of ice clouds in the CO2 slicing channels, r increases slightly. A correlation of ΔTb within cloud coverage results from the gaseous absorption present within these channels, while the window channels are relatively unaffected. In the following subsection, the temperature and water vapor profile information from AIRS is used to sort variability in ΔTb because both temperature and water vapor will affect the emission of infrared radiation in channels with gaseous absorption and may couple to spatial and spectral mismatches.

b. Correlation of ΔTb to temperature and water vapor profiles

In Fig. 6, a channel-dependent weak-to-moderate correlation of ΔTb with ice cloud coverage is observed. Now, vertical profiles of temperature and water vapor from AIRS are used to quantify the correlations as in Fig. 6. The following two quantities from AIRS are considered: temperature lapse rate dT/dp and the magnitude of the water vapor mixing ratio q, both of which may cause Tb mismatches depending on their relative magnitudes. The values of dT/dp and q were determined at the altitude of the highest sensitivity for each MODIS channel’s weighting function based on the values from Seemann et al. (2003). The values of dT/dp and q were divided into five bins each that split the more-or-less Gaussian-distributed values into bins of 20% (Tables 1 and 2).

Table 1.

Classification (A–E) of the temperature lapse rate for the different MODIS channels as it was used for Fig. 7.

Classification (A–E) of the temperature lapse rate for the different MODIS channels as it was used for Fig. 7.
Classification (A–E) of the temperature lapse rate for the different MODIS channels as it was used for Fig. 7.
Table 2.

Classification (A–E) of the temperature lapse rate for the different MODIS channels as it was used for Fig. 8.

Classification (A–E) of the temperature lapse rate for the different MODIS channels as it was used for Fig. 8.
Classification (A–E) of the temperature lapse rate for the different MODIS channels as it was used for Fig. 8.

The upper panel in Fig. 7 presents values of r for ice cloud coverage only as in Fig. 6, but the results are sorted into the five bins of dT/dp. First, the strongest negative correlations are seen in channels 35 and 36, which also show that an increase in the correlation corresponds with an increase in dT/dp (Fig. 7, upper panel). Second, channels 24 and 27 show similarly strong correlations, except they increase in the positive direction as dT/dp decreases. Interestingly, for channel 25, the correlation is negligible for all lapse rates, except for the bin of dT/dp closest to an isothermal lapse rate.

Fig. 7.

The correlation coefficient r between ΔTb and ice cloud coverage separated into five-temperature lapse-rate bins (see Table 1 and accompanying text) centered on the vertical level of the weighting function peak. (top) The AIRS spectral response function information only and (bottom) the MODIS spectral response function corrections of channels 27, 28, 34, 35, and 36 (T06).

Fig. 7.

The correlation coefficient r between ΔTb and ice cloud coverage separated into five-temperature lapse-rate bins (see Table 1 and accompanying text) centered on the vertical level of the weighting function peak. (top) The AIRS spectral response function information only and (bottom) the MODIS spectral response function corrections of channels 27, 28, 34, 35, and 36 (T06).

However, the correlations become much weaker when the spectral correction of T06 is applied to channels 27, 28, 34, 35, and 36 (Fig. 7, lower panel). In particular, the correlation between ice cloud coverage and ΔTb drops to nearly zero for channels 35 and 36. This is additional strong evidence for the existence of a spectral shift on the order of the magnitude shown by T06 and shows that a careful determination of the shift in SRFMODIS is necessary for synergistic research applications of these two instruments that require a variance of 1 K (or less) in consistency. The ΔTb for channel 34 reveals that the shift in SRFMODIS seems to be too strong and results in a higher correlation, which additionally depends on the water vapor amount. Because the correlation is near 0 for the upper panel, the use of the shift is not recommended for channel 34. These results imply that the double-peak structure observed in Figs. 3 and 5 for channel 36 is largely a manifestation of variable dT/dp that is coupled to the erroneously shifted SRFMODIS.

The upper panel in Fig. 8 shows r for different magnitudes of q. Once again, channels 35 and 36 show a strong negative correlation of ΔTb with cloud coverage in the presence of optically thick ice clouds when water vapor amounts are relatively high; likewise, the same is true for channel 27. In general, all channels have a lower r when water vapor amounts are low. As in the lower panel of Fig. 7, the corrections of T06 show a strong reduction in r, shown in the lower panel of Fig. 8, especially for channels 27, 35, and 36. As with dT/dp, this is an indication that a precise estimate of the shift in SRFMODIS is important for the reduction of the biases in ΔTb, which are in general more strongly manifested in moist atmospheres and steeper lapse rates. However, again, there is an increased correlation by using the shift for channel 34 and neglecting the shift seems to be a better option in this case.

Fig. 8.

As in Fig. 7, but for AIRS-derived water vapor mixing ratio vertically centered at the peak of the MODIS weighting functions.

Fig. 8.

As in Fig. 7, but for AIRS-derived water vapor mixing ratio vertically centered at the peak of the MODIS weighting functions.

This analysis shows that by applying a combination of FAIRS and SRFMODIS to collocated MODIS and AIRS radiance data, the bias, standard deviation, and skewness in ΔTb are substantially reduced over more simple representations of the AIRS FOV (e.g., K07). In particular, the biases of ΔTb are primarily reduced by applying shifts in the SRFMODIS recommended by T06, except for channel 34. The standard deviation and skewness are largely reduced by applying channel- and scan angle–dependent values of FAIRS to the collocated radiances. Furthermore, correlations between ΔTb with cloud and aerosol amount, dT/dp, and q, within the strongly absorbing channels (e.g., H2O and CO2), are substantially reduced after applying the recommended shift in SRFMODIS. These results show that larger values of ΔTb—for example, the second peak observed in some channels—occur in the presence of thicker ice clouds, moist atmospheres, and steep lapse rates; and can be corrected for with more precise estimates of SRFMODIS.

5. Summary

Radiance comparisons of the Moderate Resolution Imaging Spectroradiometer (MODIS; Barnes et al. 1998) and Atmospheric Infrared Sounder (AIRS; Aumann et al. 2003) on the Aqua satellite platform are presented for a global set of collocated observations on 1 January 2005. The channel- and scan angle–dependent spatial response functions of AIRS that are obtained from prelaunch measurements and calculated effects of the scan geometry are described (Overoye et al. 1999; Elliott et al. 2006). Next, the spatial response functions are truncated, rotated, and smeared to calculate a realistic on-orbit representation of the AIRS FOV. After using them to more precisely smooth the MODIS radiances to the AIRS footprint, it is shown that the standard deviation and skewness of brightness temperature differences (ΔTb) between MODIS and AIRS are substantially reduced and quantified on a channel-by-channel basis. The scan angle dependence is also reduced in most channels, but in a few channels multiple peaks of ΔTb remain. Combining the retrieved Cij shift with the calculated “smeared” spatial response did not show any improvement in ΔTb.

Because the differences in Tb between MODIS and AIRS may be related to atmospheric or surface properties, correlations of ΔTb to surface type, aerosol and cloud characteristics derived from MODIS are used to highlight correlations between these quantities and ΔTb. They are shown to be especially significant for ice clouds in H2O and CO2 bands (e.g., MODIS channels 24, 27, 35, and 36), whereas in the presence of aerosols, thin cirrus, and water clouds, the corresponding correlations were shown to be weak. The correlation of ΔTb and ice cloud coverage was investigated further by relating these relationships to different magnitudes of temperature lapse rate (dT/dp) and the magnitude of water vapor mixing ratio (q) obtained from AIRS retrievals at the altitudes of peak sensitivity for each MODIS channel’s weighting function. Larger values of dT/dp and q correlate well to larger values of ΔTb in the H2O and CO2 bands. After the MODIS spectral response functions are shifted by values recommended in T06, these correlations are largely eliminated or reduced, especially for MODIS channels 28, 35, and 36. Interestingly, channel 34 showed no improvement using the proposed shift of the response function, and already agreed within the specified noise levels. Thus, the application and analysis of the spatial and spectral shifts is an important factor for improved L2 geophysical retrievals or additional atmospheric information when obtained from combinations of MODIS and AIRS radiances, and we strongly recommend the use of the spectral shift for MODIS channels 35 and 36.

This investigation shows that the AIRS spatial response functions are necessary to reduce the variability and skewness of ΔTb within heterogeneous scenes; because, importantly, precise estimates of spectral shifts in the MODIS spectral response functions are necessary to reduce biases in ΔTb. Both of these instrumental characteristics can couple to scene heterogeneity (especially broken clouds) and inherent variability in vertical profiles of temperature and water vapor to produce significant biases, variability, and skewness in ΔTb on the order of 1 K or more. These physical factors must be accounted for in the channels that respond most strongly to these geophysical influences if they are to be used in a synergistic manner together, for instance, in multi-instrument and multiwavelength retrievals of cloud properties that cannot be obtained by either instrument alone (e.g., Baran and Francis 2004; L’Ecuyer et al. 2006).

Acknowledgments

Funding for this project was provided by NASA Award NNX08AI09G. The authors thank the AIRS project at the Jet Propulsion Laboratory and Hal Maring at the NASA Radiation Science program for support. MODIS data were obtained through the level 1 and Atmosphere Archive and Distribution System (LAADS; http://ladsweb.nascom.nasa.gov/). AIRS data were obtained through the Goddard Earth Sciences Data and Information Services Center (http://daac.gsfc.nasa.gov). A portion of this work was performed within the Joint Institute for Regional Earth System Science and Engineering (JIFRESSE) of the University of California, Los Angeles (UCLA) and at the Jet Propulsion Laboratory, California Institute of Technology, under contract with NASA. The authors thank two anonymous reviewers for their comments and suggestions.

REFERENCES

REFERENCES
Ackerman
,
S. A.
,
K. I.
Strabala
,
W. P.
Menzel
,
R. A.
Frey
,
C. C.
Moeller
, and
L. E.
Gumley
,
1998
:
Discriminating clear sky from clouds with MODIS.
J. Geophys. Res.
,
103
,
32141
32157
.
Ackerman
,
S. A.
,
R. E.
Holz
,
R.
Frey
,
E. W.
Eloranta
,
B. C.
Maddux
, and
M.
McGill
,
2008
:
Cloud detection with MODIS. Part II: Validation.
J. Atmos. Oceanic Technol.
,
25
,
1073
1086
.
Aumann
,
H. H.
, and
Coauthors
,
2003
:
AIRS/AMSU/HSB on the aqua mission: Design, science objectives, data products, and processing systems.
IEEE Trans. Geosci. Remote Sens.
,
41
,
253
264
.
Baran
,
A. J.
, and
P. N.
Francis
,
2004
:
On the radiative properties of cirrus cloud at solar and thermal wavelengths: A test of model consistency using high-resolution airborne radiance measurements.
Quart. J. Roy. Meteor. Soc.
,
130
,
763
778
.
Barnes
,
W. L.
,
T. S.
Pagano
, and
V. V.
Salomonson
,
1998
:
Prelaunch characteristics of the Moderate Resolution Imaging Spectroradiometer (MODIS) on EOS-AM1.
IEEE Trans. Geosci. Remote Sens.
,
36
,
1088
1100
.
Chahine
,
M. T.
, and
Coauthors
,
2006
:
AIRS: Improving weather forecasting and providing new data on greenhouse gases.
Bull. Amer. Meteor. Soc.
,
87
,
911
926
.
Cracknell
,
A. P.
,
1998
:
Synergy in remote sensing—What’s in a pixel?
Int. J. Remote Sens.
,
19
,
2025
2047
.
Elliott
,
D. A.
,
T. S.
Pagano
, and
H. H.
Aumann
,
2006
:
The impact of the AIRS spatial response on channel-to-channel and multi-instrument data analyses.
Earth Observing Systems XI, J. J. Butler, Ed., International Society for Optical Engineering (SPIE Proceedings, Vol. 6296), doi:10.117/12.679542
.
Frey
,
R. A.
,
S. A.
Ackerman
,
Y.
Liu
,
K. I.
Strabala
,
H.
Zhang
,
J. R.
Key
, and
X.
Wang
,
2008
:
Cloud detection with MODIS. Part I: Improvements in the MODIS cloud mask for Collection 5.
J. Atmos. Oceanic Technol.
,
25
,
1057
1072
.
Gao
,
B-C.
, and
Y. J.
Kaufman
,
1995
:
Selection of the 1.375-μm MODIS channel for remote sensing of cirrus clouds and stratospheric aerosols from space.
J. Atmos. Sci.
,
52
,
4231
4237
.
Gunshor
,
M. M.
,
T. J.
Schmit
,
W. P.
Menzel
, and
D. C.
Tobin
,
2009
:
Intercalibration of broadband geostationary imagers using AIRS.
J. Atmos. Oceanic Technol.
,
26
,
746
758
.
Kahn
,
B. H.
,
E.
Fishbein
,
S. L.
Nasiri
,
A.
Eldering
,
E. J.
Fetzer
,
M. J.
Garay
, and
S-Y.
Lee
,
2007
:
The radiative consistency of Atmospheric Infrared Sounder and Moderate Resolution Imaging Spectroradiometer cloud retrievals.
J. Geophys. Res.
,
112
,
D09201
.
doi:10.1029/2006JD007486
.
Lambrigtsen
,
B. H.
, and
S-Y.
Lee
,
2003
:
Coalignment and synchronization of the AIRS instrument suite.
IEEE Trans. Geosci. Remote Sens.
,
41
,
343
351
.
L’Ecuyer
,
T. S.
,
P.
Gabriel
,
K.
Leesman
,
S. J.
Cooper
, and
G. L.
Stephens
,
2006
:
Objective assessment of the information content of visible and infrared radiance measurements for cloud microphysical property retrievals over the global oceans. Part I: Liquid clouds.
J. Appl. Meteor. Climatol.
,
45
,
20
41
.
Li
,
J.
,
W. P.
Menzel
,
F.
Sun
,
T. J.
Schmit
, and
J.
Gurka
,
2004a
:
AIRS subpixel cloud characterization using MODIS cloud products.
J. Appl. Meteor.
,
43
,
1083
1094
.
Li
,
J.
,
W. P.
Menzel
,
W.
Zhang
,
F.
Sun
,
T. J.
Schmit
,
J. J.
Gurka
, and
E.
Weisz
,
2004b
:
Synergistic use of MODIS and AIRS in a variational retrieval of cloud parameters.
J. Appl. Meteor.
,
43
,
1619
1634
.
Li
,
J.
,
C. Y.
Liu
,
H. L.
Huang
,
T. J.
Schmit
,
X. B.
Wu
,
W. P.
Menzel
, and
J. J.
Gurka
,
2005
:
Optimal cloud-clearing for AIRS radiances using MODIS.
IEEE Trans. Geosci. Remote Sens.
,
43
,
1266
1278
.
Menzel
,
W. P.
,
R. A.
Frey
,
B. A.
Baum
, and
H.
Zhang
,
2006
:
Cloud top properties and cloud phase.
Algorithm theoretical basis document, version 7, 55 pp
.
Nagle
,
F. W.
, and
R. E.
Holz
,
2009
:
Computationally efficient methods of collocating satellite, aircraft, and ground observations.
J. Atmos. Oceanic Technol.
,
26
,
1585
1595
.
Overoye
,
K.
,
H. H.
Aumann
,
M. H.
Weiler
,
G. W.
Gigioli
,
W.
Shaw
,
E.
Frost
, and
T.
McKay
,
1999
:
Test and calibration of the AIRS instrument.
Infrared Spaceborne Remote Sensing VII, M. Strojnik and B. F. Andresen, Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 3759), 254–265
.
Pagano
,
T. S.
,
M. T.
Chahine
,
H. H.
Aumann
, and
S. E.
Broberg
,
2006
:
Advanced Remote-sensing Imaging Emission Spectrometer (ARIES): An instrument concept for next-generation imager/sounder.
Sensors, Systems, and Next-Generation Satellites XI, R. Meynart et al., Eds., International Society for Optical Engineering (SPIE Proceedings, Vol. 6744), doi:10.1117/12.738079
.
Platnick
,
S.
,
M. D.
King
,
S. A.
Ackerman
,
W. P.
Menzel
,
B. A.
Baum
,
J. C.
Riédi
, and
R. A.
Frey
,
2003
:
The MODIS cloud products: Algorithms and examples from Terra.
IEEE Trans. Geosci. Remote Sens.
,
41
,
459
473
.
Remer
,
L. A.
, and
Coauthors
,
2005
:
The MODIS aerosol algorithm, products, and validation.
J. Atmos. Sci.
,
62
,
947
973
.
Seemann
,
S. W.
,
J.
Li
,
W. P.
Menzel
, and
L. E.
Gumley
,
2003
:
Operational retrieval of atmospheric temperature, moisture, and ozone from MODIS infrared radiances.
J. Appl. Meteor.
,
42
,
1072
1091
.
Stephens
,
G. L.
, and
Coauthors
,
2002
:
The CloudSat mission and the A-train.
Bull. Amer. Meteor. Soc.
,
83
,
1771
1790
.
Sun
,
H.
,
W.
Wolf
,
T.
King
,
C.
Barnet
, and
M.
Goldberg
,
2006
:
Co-Location for satellite observations.
Preprints, 14th Conf. on Satellite Meteorology and Oceanography, Atlanta, GA, Amer. Meteor. Soc., P6.25. [Available online at http://ams.confex.com/ams/pdfpapers/104936.pdf]
.
Susskind
,
J.
,
C. D.
Barnet
, and
J. M.
Blaisdell
,
2003
:
Retrieval of atmospheric and surface parameters from AIRS/AMSU/HSB data in the presence of clouds.
IEEE Trans. Geosci. Remote Sens.
,
41
,
390
409
.
Tobin
,
D. C.
,
H. E.
Revercomb
,
C. C.
Moeller
, and
T. S.
Pagano
,
2006
:
Use of Atmospheric Infrared Sounder high–spectral resolution spectra to assess the calibration of Moderate resolution Imaging Spectroradiometer on EOS Aqua.
J. Geophys. Res.
,
111
,
D09S05
.
doi:10.1029/2005JD006095
.
Wang
,
L.
,
X.
Wu
,
Y.
Li
,
M.
Goldberg
,
S.
Sohn
, and
C.
Cao
,
2010
:
Comparison of AIRS and IASI radiances using GOES imagers as transfer radiometers toward climate data records.
J. Appl. Meteor. Climatol.
,
49
,
478
492
.

Footnotes

Corresponding author address: Dr. Mathias M. Schreier, Jet Propulsion Laboratory, Mail Stop 169–237, 4800 Oak Grove Drive, Pasadena, CA 91109. Email: mathias.schreier@jpl.nasa.gov