Abstract

This study examines the feasibility of retrieving lower-stratospheric water vapor using a nadir infrared hyperspectrometer, with the focus on the detectability of small-scale water vapor variability. The feasibility of the retrieval is examined using simulation experiments that model different instrument settings. These experiments show that the infrared spectra, measured with sufficient spectral coverage, resolution, and noise level, contain considerable information content that can be used to retrieve lower-stratospheric water vapor. Interestingly, it is found that the presence of an opaque cloud layer at the tropopause level can substantially improve the retrieval performance, as it helps remove the degeneracy in the retrieval problem. Under this condition, elevated lower-stratospheric water vapor concentration—for instance, caused by convective moistening—can be detected with an accuracy of 0.09 g m−2 using improved spaceborne hyperspectrometers. The cloud-assisted retrieval is tested using the measurements of the Atmospheric Infrared Sounder (AIRS). Validation against collocated aircraft data shows that the retrieval can detect the elevated water vapor concentration caused by convective moistening.

1. Introduction

Despite its scarcity, stratospheric water vapor is an important atmospheric composition because of its radiative and chemical effects. It radiatively cools the stratosphere but potentially warms the troposphere and surface (Dvortsov and Solomon 2001; Dessler et al. 1995; Huang et al. 2016). It also may affect total ozone loss rate through several chemical processes (Anderson et al. 2012).

However, current satellite observations have limited ability to detect stratospheric water vapor variations. Satellite nadir-view radiance measurements usually do not have sufficient sensitivity to the low concentration of stratospheric water vapor, while limb view measurements have large sampling footprints, making small-scale water vapor variability hard to detect. Current global climate models also have various deficiencies in their simulations of lower-stratospheric water vapor, mostly caused by temperature bias (e.g., Schoeberl et al. 2012) and representation of the deep convection process (e.g., Neale et al. 2010).

Aircraft in situ observations have demonstrated that lower-stratospheric water vapor is highly variable. For instance, Anderson et al. (2012) showed that in the data samples taken by summertime flights over North America, convective moistening occurs on nearly 50% of these flights, with elevated lower-stratospheric water vapor concentrations being as high as 15 ppmv. Although composite analyses suggest deep convection tends to elevate water vapor concentration in the upper troposphere and lower stratosphere in this region (e.g., Sun and Huang 2015), there has been no conclusive evidence of convective moistening resulted from individual convective events from satellite water vapor measurements (e.g., Schwartz et al. 2013). It is likely because of the large footprint size and coarse vertical resolution (which lead to significant spatial averaging) of the current instruments, such as the Microwave Limb Sounder (MLS). These previous studies highlight the need to improve the monitoring of stratospheric water vapor variability.

As advanced nadir-view satellite measurements have relatively small footprints and high spectral resolution, they are better suited for monitoring small-scale water vapor variability in the lower stratosphere with relatively high vertical resolution. However, nadir retrieval of lower-stratospheric water vapor is challenging because 1) the water vapor variability is much smaller in the lower stratosphere than in the upper troposphere right across the tropopause, which induces significant uncertainty in lower-stratospheric water vapor retrieval as a result of the smoothing effect (see the discussion in section 3); and 2) the temperature variation is not monotonic around the tropopause, which causes the retrieval to lack the thermal contrast necessary for attributing a water vapor anomaly to a particular vertical level. In theory, these limitations can be largely relieved if there is an opaque cloud layer right at the tropopause, which blocks the upwelling radiation from the troposphere. In this study we test the feasibility of this retrieval idea, that is, the sensitivity of the retrieval to stratospheric water vapor with the presence of an opaque cloud, using an evaluation framework similar to Bani Shahabadi and Huang (2014). In section 2 we first investigate the information content in such a retrieval using synthetic satellite radiance simulated from atmospheric profiles and adopting a widely used optimal estimation method (Rodgers 2000) that simultaneously retrieves the stratospheric temperature and humidity. In section 3 we present test retrieval results under various sensor specifications based on simulation experiments. In section 4 retrievals based on the Atmospheric Infrared Sounder (AIRS; Chahine et al. 2006) radiance data are attempted and compared to collocated airborne in situ measurements.

2. Retrieval method and information content

a. Retrieval method

The retrieval method is based on the optimal estimation method (Rodgers 2000). The relationship between atmospheric quantities and measurements can be linearized as follows:

 
formula

Here the state vector x contains temperature and the logarithm of specific humidity at 60 fixed pressure levels (15 of them are in the stratosphere). The term y refers to upwelling radiances measured at the top of the atmosphere, and is the residual that includes measurement error and forward model error. We use the moderate resolution atmospheric transmission, version 5.0 (MODTRAN 5; Berk et al. 2005), to compute synthetic radiance and the Jacobian (see Fig. 1 for an illustration). The quantity is the prior guess of the state vector, and is the corresponding synthetic radiance.

The goal then is to solve for x, the truth atmospheric state vector, through an inverse model. Because the retrieval problem is ill-posed, it is not appropriate to directly invert the Jacobian matrix to solve for x. We adopt a Bayesian inference-based optimal estimation method instead (Rodgers 2000). The term , the best estimate of x, can be derived as

 
formula
 
formula

where is the covariance matrix of state vector, is the covariance matrix of measurement error, and is the retrieval gain matrix.

Applying the Gaussian–Newton iteration method, the iteration formula for at each iteration time step then becomes

 
formula

where is the Jacobian matrix computed at the ith time step.

The Jacobian matrix is computed by finite differencing, , where the perturbation is 1 K for temperature and 10% for specific humidity. Logarithmic scaling is used for water vapor—that is, —because its radiative effect is logarithmically dependent on its concentration (Huang and Bani Shahabadi 2014). As shown in Fig. 1, a positive Jacobian value reveals an increase in TOA observed radiance in response to an increase in atmospheric temperature or specific humidity.

b. Prior

To form the prior knowledge on the retrieval quantities, 5051 atmospheric profiles with their tropopause at 120 hPa are obtained from a 6-hourly ECMWF interim reanalysis (ERA-Interim) dataset (Dee et al. 2011) within the region of 40°–60°N, 100°–130°W, from the year 2000 to 2013. As mentioned earlier, this dataset does not depict small-scale (subgrid) variations of stratospheric water vapor, and it tends to have a dry bias in the stratosphere (Schoeberl et al. 2012). To account for such variability, we applied artificial moistening to the lower stratosphere. Although the exact pattern of lower-stratospheric moistening is still not clear, previous studies have shown several clues. First, enhanced water vapor may enter the lower atmosphere from several sources, including horizontal transport from a tropical tropopause layer driven by large-scale air motion and vertical transport from a troposphere driven by overshooting convection (Smith 2012; Grosvenor et al. 2007; Dessler et al. 1995). Elevated water vapor mixing ratios up to 15 ppmv are observed across the lower stratosphere ranging from the tropopause to 50 hPa (Anderson et al. 2012). To represent these observed moist features, random enhancements of water vapor that linearly decrease with pressure level are added to the original profiles from 200 to 50 hPa, where the increase of water vapor at 120 hPa has a mean of 5 ppmv and a standard deviation of 5 ppmv (a negative value is set to be zero). The prior knowledge is composed of 5051 moistened profiles and 5051 original profiles from ERA-Interim. A subset of them is shown in Fig. 2.

In our retrieval the covariance matrix is useful in that its diagonal elements show the uncertainty in the prior estimate and its off-diagonal elements reflect the correlation between the state vector elements. The covariance matrix can be calculated directly from the atmosphere profiles as mentioned above (the 10 102 profiles). However, computed this way includes the correlation between different layers as a result of artificial moistening. An alternative approach is to keep the correlation of the original dataset without moistening but add variance to the lower stratosphere and upper troposphere. Following this idea, a correlation matrix [] is first calculated based on the original data from ERA-Interim. The (shown in Fig. 3) is converted back to based on the standard deviations of the enlarged datasets with artificial moistening, as shown below:

 
formula

where vectors and d are the standard deviations of the original and moistened dataset, respectively. are two diagonal matrices with 1/d0 and d in their diagonal.

c. Forward model

We first examine the retrieval feasibility with a hypothetical ideal instrument. This ideal instrument is assumed to have spectral coverage from 200 to 2000 cm−1, with 0.1 cm−1 spectral resolution and an uncorrelated uniform noise level at σ = 2.5 × 10−8 W cm(−2) sr(−1) cm (about 0.1 K in the midinfrared (MIR), using 200 K as the background temperature). Normally distributed random noise with mean 0 and standard deviation σ is generated and added to forward model–calculated radiance F(x) to mimic measurements with sensor noise, following Eq. (1). In addition, several cases (see Table 1) are designed to examine the effect of spectral coverage, spectral resolution, and noise level, testing the possibility of performing this retrieval with different instrument specifications.

A layer of opaque cloud with uniform ice particle density of 1.5 g m−3 is prescribed right below the tropopause level, at 120 hPa, which effectively blocks radiative emission from layers below. Instead of iteratively retrieving it, cloud-top temperature is computed directly from the mean brightness temperature in the window band (wavenumber from 850 to 900 cm−1), as the emissivity of the cloud is close to 1. Cloud optical parameters are prescribed according to the library of Yang et al. (2013) to simulate reflectance and transmittance throughout the spectra.

The quality of retrieval is assessed with different metrics. The root-mean-square error (RMSE) indicates the uncertainty in the retrieval at each level, and the mean error shows the systematic bias in the retrieval.

d. Information content

A retrieval is expected to reduce the uncertainty in lower-stratospheric temperature and the water vapor mixing ratio, and to distinguish moistened cases from unmoistened cases. To test the retrieval method proposed here, a test dataset consisting of 200 profiles are randomly selected. Several scenarios (Table 1) with different instrumental parameters are designed to test the feasibility of this retrieval using current or improved satellite instruments.

We first use the concept of degree of freedom for signal (Rodgers 2000) to assess the retrieval potential. The averaging kernel , derived from Eq. (4), relates retrieved quantities to their truth values,

 
formula

where . A narrowly shaped row vector of peaking at the corresponding pressure level would indicate high vertical resolution. Figures 4a and 4b presents a few selected row vectors of the averaging kernel for temperature and water vapor retrievals at 100, 50, and 10 hPa, demonstrating that the water vapor concentrations at these levels can be delineated to a certain extent.

The degree of freedom for signal (DFS), which is the trace of , is calculated here to provide an estimate of how much independent information can be obtained in the retrieval (Rodgers 2000). Table 1 shows the DFS for a variety of instrument specifications tested in this study. The DFS values show how the retrieval is impacted by the spectral coverage [inclusion of the far infrared (FIR) in particular], spectral resolution, and instrument noise.

In agreement with the DFS values, spectral signals caused by enhanced water vapor also reveal the effects of the different instrument parameters. An effective retrieval requires signals larger than the measurement noise, the prescriptions of which in our tests are shown by dashed lines in Fig. 5. Because the signal strength essentially depends on the spectral resolution, a finer resolution potentially leads to improved retrieval performance given the same noise level.

As we are most interested in the lower stratosphere from 100 to 40 hPa, cumulative DFS (CDFS) normalized by the total DFS is used to demonstrate how the information content distributes across the vertical profile (Fig. 4c). Although the DFS has similar values in the two cases (see Table 1), the CDFS in the cloudy sky is constantly higher than the clear sky down to where the cloud is located. It is illustrated in Fig. 4 that the averaging kernels of the lowermost stratospheric layers (e.g., at 100 hPa) in the cloudy case are noticeably narrower and, very importantly, inhibit the tropospheric effects from below the cloud level, where the water vapor variability (uncertainty) is an order of magnitude larger. This advantage is verified by the retrieval tests below.

3. Simulation experiments with different instrument specifications

In case 1 we present the performance of the abovementioned ideal instrument to illustrate the theoretical limits of the retrieval. As shown in Figs. 5a and 5b, this instrument specification fully enables the TOA spectrum to capture signals coming from lower-stratospheric moistening.

Current satellite missions, however, do not meet these instrumentation requirements simultaneously. The IR spectrometer carried by AIRS has a noise level of 0.75 × 10−8 W cm(−2) sr(−1) cm (0.3 K in MIR), but its resolution is coarser than 0.5 cm−1. Though the Tropospheric Emissions Spectrometer (TES; Beer et al. 2001) has a high spectral resolution, its noise level is higher. Figure 5 suggests the inclusion of FIR measurements, as designed for the Climate Absolute Radiance and Refractivity Observatory (CLARREO; Wielicki et al. 2013), may be helpful, in that signals at some FIR frequencies are of greater magnitude.

Several cases are then designed to test the effects of different instrument settings in light of these instruments. Table 1 lists detailed instrument settings in each case (with corresponding retrieval results shown in Fig. 7). Note that all of the listed values for the RMSE are computed based on levels between 100 and 40 hPa, which is the vertical range of most concern in this study. Among them, case 1 is the ideal case that combines optimal noise and spectral resolution with an extended spectral region to obtain the best representation of the true water vapor profile. Case 2 is designed to represent instrument specifications of TES, and case 3 is for AIRS. Cases 4 and 5 consider the cases if the measurement errors for AIRS and TES are reduced, which can be realized by taking multiple measurements of the same target or extending the stare time (e.g., if deployed on a geostationary satellite) or, hypothetically, by reducing detector noise. Cases 6 and 7 are designed to assess the effect of FIR coverage. Retrieval tests are performed using these parameter specifications under both cloudy-sky (opaque cloud at the tropopause) and clear-sky scenarios.

In the ideal case, the retrieval process greatly decreases the uncertainty in the state vector compared to the prior guess, as shown in Figs. 6 and 7. With a fine spectral resolution of 0.1 cm−1 and a low noise level, the retrieval result is able to detect the elevation of water vapor (WV) concentration. The retrieval process reduces uncertainty (RMSE) in temperature, water vapor, and column-integrated water vapor (CIWV) by 87%, 40%, and 69%, respectively. As inferred from the relatively low vertical resolution indicated by the averaging kernel, the retrieval cannot perfectly reproduce an individual moistening profile (i.e., place the anomalous water vapor at the right place), but it reduces systematic bias at each level dramatically with a satisfactory estimation of CIWV.

For different sensor specifications, temperature retrieval in all the cases reaches ≤1 K uncertainty in the lower stratosphere, but the quality of water vapor retrieval differs. In accordance with our expectation, cases 2 and 3 have limited reduction of uncertainty in water vapor, although it seems case 3 (AIRS) has some potential for detecting strong moistening. Cases 4 and 5 indicate that reduction in measurement error improves retrieval accuracy. Moreover, cases 6 and 7 highlight that the coverage of FIR with good radiometric accuracy would significantly reduce the uncertainty, achieving an accuracy of ppmv for the lower-stratospheric mean water vapor mixing ratio. Among all six nonideal sensor specifications, case 6 provides the best retrieval quality, which affords 0.38-K accuracy for temperature and 0.21 g m−2 for column-integrated water vapor in the lower stratosphere.

It is interesting to note that although the information contents for water vapor retrievals are comparable in the clear-sky case, the retrieval tests show that the retrieval performance is worse in the clear-sky case. The reduction in uncertainty is generally only half of that accomplished in the cloudy case. As reasoned above, this is because of the obscuration of the upper-tropospheric water vapor variability. As indicated by the averaging kernel (Fig. 4b), there are substantial contributions across the tropopause level in the clear-sky case. The presence of a cloud layer at the tropopause effectively eliminates such complication and the degeneracy caused by the nonmonotonic change in temperature across the tropopause.

When examining the performance of this retrieval method, two different moistening patterns are investigated using ideal instrument settings. The first pattern is the case with enhanced water vapor continuously extending from cloud top to 50 hPa. It reaches 51% and 80% uncertainty reduction for the water vapor mixing ratio and CIWV, respectively. Another pattern is with isolated moistening in one 20-hPa segment. In this case, retrieval results show 21% uncertainty reduction for the water vapor mixing ratio and 42% for CIWV. The inferior performance in the latter case is because the retrieval process cannot always place the elevated water vapor at the right vertical level because of the limited vertical resolution of the retrieval technique, as indicated by the averaging kernel in Fig. 4.

This retrieval method uses known cloud properties to assist stratospheric retrieval, as uncertainty in cloud properties might affect the retrieval. The high emissivity of the cloud is essential to the retrieval process, as it provides an accurate cloud-top temperature that largely eliminates background radiance differences. The cloud-top pressure level in this study is fixed at the tropopause to simplify the retrieval process, though it is worth examining how it can be retrieved and how cloud tops that overshoot the tropopause may affect retrieval results in the future.

The forward model error might also add uncertainty to the result. First, the specification of frequency increment, spectral resolution, and slit function affect the performance of the radiative transfer model. The frequency increment is limited to 0.1 cm−1 by the minimum bin size of MODTRAN, which means that Table 1 might have underestimated the performance of TES (which has a spectral sampling size at 0.0592 cm−1). Furthermore, the retrieval considers only temperature and water vapor; other absorbers like CO2, O3, and CH4 are assumed to have constant values. Variability in these gases may affect retrieval quality, although our tests prove that 20% perturbation of these gases throughout the stratosphere does not have a significant impact on the spectral signal of water vapor.

4. Application to AIRS

a. AIRS

AIRS has provided global measurements in a sun-synchronized orbit since 2002. The AIRS Level 1B (L1B) radiance product contains 2378 infrared channels from 650 to 2665 cm−1, with accuracy varying from 0.3 to 0.5 K (against a 200-K reference temperature). It provides a Level 2 (L2) retrieval support product (version 6) for specific humidity on 100 pressure levels from 0.0161 hPa down to the surface, with a documented RMSE of 20% near the surface (Susskind et al. 2011; Kahn et al. 2014) based on a cloud clearing method. However, this retrieval algorithm does not yield accurate water vapor retrieval in the upper troposphere and lower stratosphere (UTLS) for several reasons. First, because the cloud-clearing algorithm essentially depends on valid clear-sky observations obtained from the Moderate Resolution Imaging Spectroradiometer (MODIS), it does not work well when cloud uniformly obscures 3 × 3 fields of view (FOVs). It results in a much larger RMSE with increased altitude, and dry biases over thick clouds (Wong et al. 2015). Second, the average over nine FOVs limits the ability to detect small-scale variation, although the latest single-footprint retrieval does not have this requirement anymore (Irion et al. 2018). Third, as discussed in the previous section, the AIRS retrieval scheme shares the common problem with other nadir retrieval methods of UTLS water vapor performed in the clear-sky condition, and it is insensitive to the very low water vapor concentration (<10 ppmv) in the UTLS region (Fetzer et al. 2008; Gettelman et al. 2004; Read et al. 2007).

b. Aircraft

The in situ water vapor measurements used here are from the Harvard water vapor instrument, which combines the Harvard Lyman-α photo-fragment fluorescence instrument with a tunable diode laser direct absorption instrument (Anderson et al. 2012; Weinstock et al. 2009). The maximum offset of this instrument is <0.2 ppmv. In this study we take measurements during the summer of 2005. Flights from 2 days, 17 June and 7 July, are used here in that they have the cloud condition desired by our retrieval method. Both flights made continuous measurements of the water vapor mixing ratio in the upper troposphere and lower stratosphere (see Figs. 8 and 9).

c. Retrieval test

In this section we test the cloud-assisted retrieval method explained above using the AIRS L1B radiance over regions where elevated water vapor is detected by aircraft measurements. To perform this retrieval with AIRS L1B radiance, scenes over the same area as the aircraft with at least 30 adjacent AIRS FOVs filled with deep convective clouds (DCCs) are selected. To identify DCCs, a threshold based on a positive brightness temperature difference between a water vapor absorption channel (bt1419) and an IR window channel (bt1231) is used (Aumann and Ruzmaikin 2013). A positive brightness temperature difference, bt1419 − bt1231, is induced by warmer stratospheric water vapor emission against the cold cloud top, so a large positive difference indicates a near-tropopause cloud top and possible elevated stratospheric water vapor concentration. The selection of channel bt1419 is based on its strong water vapor absorption and a relatively low noise-to-signal ratio.

The DCCs are assumed to be a blackbody. Under this assumption cloud-top temperature can be inferred from the brightness temperature in the IR window channel. Based on this, cloud-top pressure is then estimated from an a priori temperature profile (the AIRS L2 temperature retrieval), varying from 90 to 180 hPa. Only the above-cloud part is retrieved and then latterly used in the evaluation. The covariance matrix a is derived using the same training dataset described in the simulation experiments (see section 2). To start the iteration, we use the AIRS L2 temperature and water vapor retrieval at each FOV as the first guess.

Figure 8 shows the aircraft track from 1800 to 2240 UTC 17 June. The aircraft flew from south to north and then returned along the same track. It crossed the tropopause five times during the measurement period. During the flight elevated water vapor was observed with mixing ratios up to 8 ppmv at a level around 80 hPa for a roughly 10-hPa vertical interval. Around 12 h earlier, AIRS detected an area of DCCs to the east of the flight track, as shown in Fig. 8 with solid squares, among which colored squares denote a positive temperature difference. These samples are collected for later retrieval.

In Fig. 8d retrieval results for water vapor from 150 to 50 hPa are presented. We calculated the probability density function (PDF) at 2-ppmv-spaced intervals at every 20-hPa vertical level. Compared to the AIRS L2 retrieval, the water vapor retrieved here has much larger variability. Around 40% of the profiles indicate water vapor elevation at levels between 50 and 90 hPa, which agree with the aircraft measurements. The narrow vertical layer with high water vapor concentrations at around 75 hPa depicted by the aircraft data is not evident in the retrieval. This is within our expectation, as previous simulation experiments suggest limited vertical resolution in the retrieval. On the other hand, the aircraft measurements have limited spatial coverage, such that the narrowness of the layer may not be real.

On 7 July the aircraft detected very high water vapor mixing ratios, up to 18 ppmv at 117 hPa and an elevated water vapor layer over 30 hPa thick below 145 hPa. There are two concurrent AIRS scenes containing DCCs: one group of FOVs in the west, upwind of the flight track, around 6 h earlier; and another group of FOVs parallel to the flight track, about 4 h later. Although these scenes contain only 49 FOVs with DCCs, the aircraft measurements (Fig. 9c) show a clear distinction between unmoistened and moistened samples at a lower level. Interestingly, the mean of retrieved water vapor suggests a different vertical distribution of water vapor compared to the case of 17 June, which corresponds to the in situ measured enhancement up to 35 ppmv around 120 hPa very well.

Our tests here indicate that nadir-view hyperspectral radiance measurements, such as the AIRS L1B data, contain information for detecting elevated water vapor concentration in the UTLS region. Although the current retrieval methods have not fully used such information, an improved method can. In both case studies here, the cloud-assisted retrieval method detects moistening in the lower stratosphere, with the posterior mean profile showing better agreement with collocated airborne measurements than the AIRS L2 retrieval. Both cases strongly support this method regarding its ability to detect stratospheric water vapor elevation and to improve the mean climatology of a region with water vapor mixing ratios lower than the 10–25-ppmv sensitivity limit of the AIRS instrument suggested in previous studies (Fetzer et al. 2008; Gettelman et al. 2004; Read et al. 2007).

A few caveats should be noted regarding the retrieval though. First, as discussed in section 2, the optimal estimation-based retrieval algorithm depends on both radiance measurements and prior information—as a rule of thumb, the magnitudes of the covariance matrices a and e decide which dominates. Because the instrument (AIRS) noise level is relatively high, the prior information, including both the first guess and a, is expected to influence the retrieval strongly. The covariance matrix a is important in that it controls the covariance of the retrieved state vector. In the retrieval above, we have used a special way to construct the covariance matrix (see section 2). If the artificially moistened profiles are instead used to construct the off-diagonal element of , then the DFS of the retrieval will increase by 0.1 and, consequently, isolated layers with elevated water vapor concentration will be rendered by the retrieval method. In fact, when the retrieval is done this way, we find in the retrieval results narrow high water vapor concentration layers at upper levels, at around 50 hPa on 17 June and at around 50 and 120 hPa on 7 July. However, there was no aircraft measurement at the higher levels to validate such a retrieved water vapor pattern. This is a situation worth further investigating in future observational campaigns.

Second, the retrieval is limited by the present instrumentation. As the cloud-top position is derived using the AIRS L2 temperature retrieval, the uncertainty in the AIRS L2 retrieval thus affects the results. The cloud-top temperature determined in both cases is higher than the minimum temperature at the tropopause (e.g., 80 hPa) but corresponds to the temperature at 120 and 60 hPa. Because the 60-hPa level is 3 km above the tropopause and is unlikely reached by the deep convection, the cloud top is thus placed at the lower level. In this regard the uncertainty in cloud-top position can be greatly reduced if we implement cloud-top pressure data from CloudSat or Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO).

In addition, the nonmonotonic temperature structure induces ambiguity in the water vapor retrieval. Under this condition, had there been water vapor channels with narrow weighting functions peaking at different levels around the tropopause, atmospheric moistening below and above the tropopause can be distinguished in theory by their brightness temperature signals (anomalies of different signs). However, the AIRS spectral coverage and resolution are such that there is no water vapor channel with a weighting function that peaks above 100 hPa (the highest peak is around 150 hPa). As a result, the retrieved enhancement of water vapor can only be widespread in the vertical. From the simulation tests in section 2, it is expected that the retrieval could benefit from an instrument with FIR coverage (to utilize the stronger rotational lines of water vapor) or higher spectral resolution (to enhance the signals in the vibration–rotational band of water vapor).

5. Conclusions

In this study we examine the feasibility of retrieving stratospheric water vapor using nadir-view satellite infrared spectral measurements and apply this method to two case studies on the AIRS L1B radiance product. Our focus is on the detectability of small-scale water vapor variability, for instance, caused by convective moistening, leveraging on the relatively small footprint sizes of the nadir-view instruments. The feasibility of the retrieval is assessed using simulation experiments to model a variety of instrument settings (Table 1). A hypothetical case with ideal instrument settings demonstrates the theoretical limits of this retrieval, while a suite of additional cases shows the performance that can be expected from current satellite instruments, as well as improvements that can be achieved under different instrumentation conditions. Had an ideal instrument with FIR and MIR spectral coverage (200–2000 cm−1), high spectral resolution (0.1 cm−1), and low noise level (0.25 × 10−8 W / [cm2 sr cm(−1)] or W cm(−2) sr(−1) cm) been used, the measurements would have a DFS of nearly 3 for lower-stratospheric (40–100 hPa) water vapor (see case 1 in Table 1), indicating the high potential of this retrieval technique. The FIR coverage proposed for future missions, for example, the Climate Absolute Radiance and Refractivity Observatory (CLARREO, https://clarreo.larc.nasa.gov) would be especially beneficial in that the signal strength in the water vapor rotational band is an order-of-magnitude larger than in the vibration-rotational band (see Fig. 5). An advanced detector with lower noise levels would also greatly enhance the retrieval performance. In this regard, a strategy that allows increased stare time of the target (e.g., on geostationary orbits or high-altitude drones) would be beneficial.

Interestingly, we find the performance of lower-stratospheric water vapor retrievals can be considerably enhanced with the presence of a tropopause cloud layer. This is because an opaque cloud layer at this level effectively blocks the upwelling radiation from below the tropopause, which not only eliminates the uncertainty arising from the upper-tropospheric water vapor variability caused by the smoothing effect related to averaging kernel (Fig. 4) but also the degeneracy caused by the nonmonotonic vertical temperature variations about the tropopause level. With the assistance of such a cloud layer, the retrieval can substantially reduce the uncertainty in the total amount of the lower-stratospheric water vapor (see the CIWV column in Table 1). This indicates the feasibility of detecting above thick cumulonimbus or anvil clouds had the convective moistening of the lower stratosphere occurred.

We have tested the cloud-assisted retrieval technique using the infrared hyperspectral measurements of AIRS. The retrieval detects small-scale water vapor variability that is observed by in situ aircraft measurements but not shown by the current AIRS L2 retrieval. The detected lower-stratospheric moistening is in qualitatively good agreement with collocated aircraft data. Although the tests here imply that some aspects of the retrieval, such as the vertical resolution, are limited by the present instruments, the results indicate the potential of detecting highly elevated water vapor concentrations near or above the tropopause by using the nadir instruments. Given the climatic importance of atmospheric water vapor in the upper troposphere and lower stratosphere and the considerable uncertainty of its distribution in this region in the present datasets, future research is warranted to further develop and apply the retrieval technique proposed here. Besides AIRS, the technique can be applied to other nadir-view infrared hyperspectrometers, such as the Infrared Atmospheric Sounding Interferometer (IASI) and the Cross-Track Infrared Sounder (CrIS), on polar orbiters. Similar instruments on geostationary satellites, such as the Geostationary Interferometric Infrared Sounder (GIIRS) on Fengyun-4, may be especially suitable for this application because of their ability to continually monitor overshooting targets.

Acknowledgments

We thank Kevin Bowman, Louis Garand, Brian Kahn, Jessica Smith, and two anonymous reviewers for their comments. We thank Jessica Smith for providing the aircraft water vapor measurements. We acknowledge the ECMWF for the ERA-Interim data (http://apps.ecmwf.int/datasets/data/interim-full-moda/) and MODTRAN 5 used in this study. This work is supported by grants from the Discovery Program of the Natural Sciences and Engineering Council of Canada (Grant RGPIN 418305) and the Class G&C Program of the Canadian Space Agency (Grant 16SUASURDC).

REFERENCES

REFERENCES
Anderson
,
J. G.
,
D. M.
Wilmouth
,
J. B.
Smith
, and
D. S.
Sayres
,
2012
:
UV dosage levels in summer: Increased risk of ozone loss from convectively injected water vapor
.
Science
,
337
,
835
839
, https://doi.org/10.1126/science.1222978.
Aumann
,
H.
, and
A.
Ruzmaikin
,
2013
:
Frequency of deep convective clouds in the tropical zone from 10 years of AIRS data
.
Atmos. Chem. Phys.
,
13
,
10 795
10 806
, https://doi.org/10.5194/acp-13-10795-2013.
Bani Shahabadi
,
M.
, and
Y.
Huang
,
2014
:
Measuring stratospheric H2O with an airborne spectrometer
.
J. Atmos. Oceanic Technol.
,
31
,
1502
1515
, https://doi.org/10.1175/JTECH-D-13-00191.1.
Beer
,
R.
,
T. A.
Glavich
, and
D. M.
Rider
,
2001
:
Tropospheric emission spectrometer for the Earth Observing System’s Aura satellite
.
Appl. Opt.
,
40
,
2356
2367
, https://doi.org/10.1364/AO.40.002356.
Berk
,
A.
, and Coauthors
,
2005
: MODTRAN 5: A reformulated atmospheric band model with auxiliary species and practical multiple scattering options: Update. Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XI, S. S. Shen and P. E. Lewis, Eds., Society of Photo-Optical Instrumentation Engineers (SPIE Proceedings, Vol. 5806), 662–667, https://doi.org/10.1117/12.606026.
Chahine
,
M. T.
, and Coauthors
,
2006
:
AIRS: Improving weather forecasting and providing new data on greenhouse gases
.
Bull. Amer. Meteor. Soc.
,
87
,
911
926
, https://doi.org/10.1175/BAMS-87-7-911.
Dee
,
D.
, and Coauthors
,
2011
:
The ERA-Interim reanalysis: Configuration and performance of the data assimilation system
.
Quart. J. Roy. Meteor. Soc.
,
137
,
553
597
, https://doi.org/10.1002/qj.828.
Dessler
,
A.
,
E.
Hintsa
,
E.
Weinstock
,
J.
Anderson
, and
K.
Chan
,
1995
:
Mechanisms controlling water vapor in the lower stratosphere: “A tale of two stratospheres.”
J. Geophys. Res.
,
100
,
23 167
23 172
, https://doi.org/10.1029/95JD02455.
Dvortsov
,
V. L.
, and
S.
Solomon
,
2001
:
Response of the stratospheric temperatures and ozone to past and future increases in stratospheric humidity
.
J. Geophys. Res.
,
106
,
7505
7514
, https://doi.org/10.1029/2000JD900637.
Fetzer
,
E. J.
, and Coauthors
,
2008
:
Comparison of upper tropospheric water vapor observations from the Microwave Limb Sounder and Atmospheric Infrared Sounder
.
J. Geophys. Res.
,
113
,
D22110
, https://doi.org/10.1029/2008JD010000.
Gettelman
,
A.
, and Coauthors
,
2004
:
Validation of Aqua satellite data in the upper troposphere and lower stratosphere with in situ aircraft instruments
.
Geophys. Res. Lett.
,
31
,
L22107
, https://doi.org/10.1029/2004GL020730.
Grosvenor
,
D.
,
T.
Choularton
,
H.
Coe
, and
G.
Held
,
2007
:
A study of the effect of overshooting deep convection on the water content of the TTL and lower stratosphere from Cloud Resolving Model simulations
.
Atmos. Chem. Phys.
,
7
,
4977
5002
, https://doi.org/10.5194/acp-7-4977-2007.
Huang
,
Y.
, and
M.
Bani Shahabadi
,
2014
:
Why logarithmic? A note on the dependence of radiative forcing on gas concentration
.
J. Geophys. Res. Atmos.
,
119
,
13 683
13 689
, https://doi.org/10.1002/2014JD022466.
Huang
,
Y.
,
M.
Zhang
,
Y.
Xia
,
Y.
Hu
, and
S.-W.
Son
,
2016
:
Is there a stratospheric radiative feedback in global warming simulations?
Climate Dyn.
,
46
,
177
186
, https://doi.org/10.1007/s00382-015-2577-2.
Irion
,
F. W.
, and Coauthors
,
2018
:
Single-footprint retrievals of temperature, water vapor and cloud properties from AIRS
.
Atmos. Meas. Tech.
,
11
,
971
995
, https://doi.org/10.5194/amt-11-971-2018.
Kahn
,
B. H.
, and Coauthors
,
2014
:
The Atmospheric Infrared Sounder version 6 cloud products
.
Atmos. Chem. Phys.
,
14
,
399
426
, https://doi.org/10.5194/acp-14-399-2014.
Neale
,
R. B.
, and Coauthors
,
2010
: Description of the NCAR Community Atmosphere Model (CAM 5.0). NCAR Tech. Note NCAR/TN-486+STR, 268 pp., www.cesm.ucar.edu/models/cesm1.1/cam/docs/description/cam5_desc.pdf.
Read
,
W. G.
, and Coauthors
,
2007
:
Aura Microwave Limb Sounder upper tropospheric and lower stratospheric H2O and relative humidity with respect to ice validation
.
J. Geophys. Res.
,
112
,
D24S35
, https://doi.org/10.1029/2007JD008752.
Rodgers
,
C. D.
,
2000
: Inverse Methods for Atmospheric Sounding: Theory and Practice. Vol. 2, World Scientific Publishing, 238 pp.
Schoeberl
,
M.
,
A.
Dessler
, and
T.
Wang
,
2012
:
Simulation of stratospheric water vapor and trends using three reanalyses
.
Atmos. Chem. Phys.
,
12
,
6475
6487
, https://doi.org/10.5194/acp-12-6475-2012.
Schwartz
,
M. J.
,
W. G.
Read
,
M. L.
Santee
,
N. J.
Livesey
,
L.
Froidevaux
,
A.
Lambert
, and
G. L.
Manney
,
2013
:
Convectively injected water vapor in the North American summer lowermost stratosphere
.
Geophys. Res. Lett.
,
40
,
2316
2321
, https://doi.org/10.1002/grl.50421.
Smith
,
J. B.
,
2012
: The sources and significance of stratospheric water vapor: Mechanistic studies from equator to pole. Ph.D. thesis, Harvard University, 363 pp.
Sun
,
Y.
, and
Y.
Huang
,
2015
:
An examination of convective moistening of the lower stratosphere using satellite data
.
Earth Space Sci.
,
2
,
320
330
, https://doi.org/10.1002/2015EA000115.
Susskind
,
J.
,
J. M.
Blaisdell
,
L.
Iredell
, and
F.
Keita
,
2011
:
Improved temperature sounding and quality control methodology using AIRS/AMSU data: The AIRS Science Team version 5 retrieval algorithm
.
IEEE Trans. Geosci. Remote Sens.
,
49
,
883
907
, https://doi.org/10.1109/TGRS.2010.2070508.
Weinstock
,
E.
, and Coauthors
,
2009
:
Validation of the Harvard Lyman-α in situ water vapor instrument: Implications for the mechanisms that control stratospheric water vapor
.
J. Geophys. Res.
,
114
,
D23301
, https://doi.org/10.1029/2009JD012427.
Wielicki
,
B. A.
, and Coauthors
,
2013
:
Achieving climate change absolute accuracy in orbit
.
Bull. Amer. Meteor. Soc.
,
94
,
1519
1539
, https://doi.org/10.1175/BAMS-D-12-00149.1.
Wong
,
S.
,
E. J.
Fetzer
,
M.
Schreier
,
G.
Manipon
,
E. F.
Fishbein
,
B. H.
Kahn
,
Q.
Yue
, and
F. W.
Irion
,
2015
:
Cloud-induced uncertainties in AIRS and ECMWF temperature and specific humidity
.
J. Geophys. Res. Atmos.
,
120
,
1880
1901
, https://doi.org/10.1002/2014JD022440.
Yang
,
P.
,
L.
Bi
,
B. A.
Baum
,
K.-N.
Liou
,
G. W.
Kattawar
,
M. I.
Mishchenko
, and
B.
Cole
,
2013
:
Spectrally consistent scattering, absorption, and polarization properties of atmospheric ice crystals at wavelengths from 0.2 to 100 μm
.
J. Atmos. Sci.
,
70
,
330
347
, https://doi.org/10.1175/JAS-D-12-039.1.

Footnotes

© 2018 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).