An assessment of the diurnal variation of upper tropospheric humidity in reanalysis data sets

Authors


Corresponding author: E.-S. Chung, Rosenstiel School of Marine and Atmospheric Science, University of Miami, 4600 Rickenbacker Causeway, Miami, FL 33149, USA. (echung@rsmas.miami.edu)

Abstract

[1] Diurnal variations of upper tropospheric humidity (UTH) in five different reanalysis data sets are compared over convective land and ocean regions and evaluated using multiple satellite observations as a reference. All reanalysis data sets reproduce the day-night contrast of upper tropospheric humidity and the land-ocean contrast in the diurnal amplitude. The infrared satellite measurements indicate a slightly later diurnal minimum over land relative to most reanalyses and the microwave satellite measurements, suggesting that cloud masking of the infrared radiances may introduce a small (∼3 h) bias in the phase. One reanalysis exhibits a substantially different diurnal cycle over land which is inconsistent with both infrared and microwave satellite measurements and other reanalysis products. This product also exhibits a different covariance between vertical velocity, cloud water, and humidity than other reanalyses, suggesting that the phase bias is related to deficiencies in the parameterization of moist convective processes.

1 Introduction

[2] The absorption of outgoing longwave radiation from water vapor is a key process in amplifying both internal and external climate variations [e.g., Held and Soden, 2000; Sherwood et al., 2010]. Because the trapping of longwave radiation is proportional to the logarithm of water vapor concentration at high altitudes [Held and Soden, 2000], even small variations of upper tropospheric water vapor can have significant implications for the magnitude of water vapor feedback [e.g., Pierrehumbert and Roca, 1998; Held and Soden, 2000; Brogniez and Pierrehumbert, 2006]. As a result, it is important to assess whether reanalysis data sets and climate models can reproduce the observed distribution and variability of upper tropospheric water vapor. For example, Paltridge et al. [2009] noted that specific humidity in the National Centers for Environmental Prediction (NCEP)/National Center for Atmospheric Research (NCAR) reanalysis has decreased in the middle to upper troposphere over the past several decades, indicating the need for careful evaluation of the reanalysis.

[3] Because it is one of the basic forcing modes in the climate system, many studies have examined the diurnal cycle of upper tropospheric humidity (UTH) using radiance measurements from geostationary satellites [e.g., Udelhofen and Hartmann, 1995; Soden, 2000; Tian et al., 2004; Chung et al., 2007, 2009]. These studies have found coherent diurnal variations of upper tropospheric humidity over convectively active regions where deep convection peaks in the late afternoon over land and in the early morning over ocean. The upper tropospheric humidity tends to reach a maximum during nighttime (00:00–06:00 local time) with larger diurnal amplitude over land. Although such diurnal characteristics are good bench marks, they have not been widely used in the evaluation of reanalysis data sets or climate models.

[4] Recently, substantial efforts have been made to improve the quality of reanalysis systems by employing more sophisticated physics and through advanced data assimilation systems, which can more effectively exploit the available satellite measurements [e.g., Dee et al., 2011; Rienecker et al., 2011]. In this study, we compare the diurnal cycle of upper tropospheric humidity produced in five different reanalysis data sets to observations from both infrared and microwave satellite instruments.

2 Satellite Observations

[5] The 6.7 µm water vapor channel radiances on geostationary satellites have been the primary observational tool for the analysis of upper tropospheric humidity and its diurnal variation [Soden and Bretherton, 1993; Udelhofen and Hartmann, 1995; Soden, 2000; Tian et al., 2004; Chung et al., 2007, 2009; Thapliyal et al., 2011]. In this study, a 3-hourly archive of clear-sky brightness temperatures from the first generation Meteosat satellites with homogenization to Meteosat-5(Meteosat Water Vapor Channel Clear-sky Radiances: Brogniez et al. [2006] constructed a long-term archive of homogenized Meteosat water vapor channel clear-sky radiances with cloud clearing based on the International Satellite Cloud Climatology Project DX product. The archive covers the period July 1983 to June 2005 with a spatial resolution of 0.625º × 0.625º over Africa and adjacent Atlantic Ocean (45ºS-45ºN, 45ºW-45ºE) and with a 3-hourly time step. The data set calibrated with NOAA as reference is used in this study. The reader is referred to Brogniez et al. [2006] for complete details on this dataset.) [Brogniez et al., 2006; ftp://ftp.climserv.ipsl.polytechnique.fr/clear_sky_WV/] is analyzed for January and July for the period 1984–2004 over selected convectively active regions of Africa and the Atlantic Ocean: (land) 20ºS–Eq, 10ºE–40ºE and (ocean) Eq–10ºN, 45ºW–10ºW for January, and (land) Eq–15ºN, 10ºW–45ºE and (ocean) Eq–10ºN, 45ºW–10ºW for July (refer to Table 1). Previous studies [e.g., Tian et al., 2004; Chung et al., 2007; Schröder et al., 2009] showed that convective activities are frequent in Central Africa, the Ethiopian highlands, and West Africa in summer. In winter, the convection centers are moved southward (e.g., Zaire, Angola, and Zambia). Convective activities are also found over the northern part of the tropical Atlantic Ocean for both seasons.

Table 1. Analysis Domains Used in This Study
 LandOcean
January20ºS–Eq, 10ºE–40ºEEq–10ºN, 45ºW–10ºW
JulyEq–15ºN, 10ºW–45ºEEq–10ºN, 45ºW–10ºW

[6] Figure 1 displays the diurnal anomaly of Meteosat-5 water vapor channel brightness temperature for the entire period (thick red line) as well as individual years (blue lines). The diurnal anomaly of brightness temperature is computed by subtracting the diurnal mean from brightness temperature at a given time. The local times of brightness temperature maximum and minimum are 17:00–19:00 LT (local time) and 4:00–6:00 LT, respectively, over land for the climatological January case (thick red line in the upper left panel). The pattern of diurnal anomalies indicates that the upper troposphere tends to be more humid (colder brightness temperatures) in the nighttime compared to the hours of midafternoon to early evening. The diurnal variations for individual years are generally similar to that of the climatological mean case. The brightness temperature for July (upper right panel) exhibits similar local times of maximum and minimum. However, the range of diurnal variation is relatively smaller in July. This seasonal/regional difference may stem from differences in the intensity of convective activity or from sampling differences due to the presence of high-to-middle level clouds [e.g., Sohn et al., 2006; Sohn and Bennartz, 2008; John et al., 2011].

Figure 1.

Diurnal anomaly of Meteosat-5 water vapor channel brightness temperature over convectively active regions of Africa (upper panels; 20ºS–Eq, 10ºE–40ºE for January and Eq–15ºN, 10ºW–45ºE for July) and the Atlantic Ocean (lower panels; Eq–10ºN, 45ºW–10ºW for January and Eq–10ºN, 45ºW–10ºW for July) for the period 1984–2004. Each blue line denotes an individual year, and thick red lines represent the diurnal anomaly of brightness temperature for the entire period. Also included is the diurnal variation of 183.31 ± 1 GHz channel brightness temperature of MHS instruments onboard NOAA-18 and MetOp-A for the period 2007–2009 (green squares and lines). The ascending (descending) node equatorial crossing times are ~14:00 (2:00) LT for NOAA-18 and ~21:30 (9:30) LT for MetOp-A. Note that the axis and corresponding labels for the diurnal variation of MHS brightness temperature are placed on the right of the plots.

[7] A distinct diurnal variation of Meteosat-5 water vapor channel brightness temperature is also found over the oceanic domains for both months with the brightness temperature maximum and minimum occurring a couple of hours earlier relative to the land cases (lower panels), indicating that the day-night contrast of upper tropospheric humidity is a prevalent feature over convectively active regions. However, the interannual variation in diurnal phase is larger over the ocean, particularly for January. In addition, the diurnal amplitude of brightness temperature is much smaller over the oceanic domains consistent with previous studies [e.g., Soden, 2000; Tian et al., 2004; Chung et al., 2007, 2009].

[8] Despite their improved sampling in cloudy regions [e.g., John et al., 2011], microwave observations have rarely been used for decomposing the diurnal cycle of upper tropospheric humidity due to the temporal sampling limitations of polar-orbiting satellites. However, the sampling challenge can be mitigated by combining multiple polar-orbiting satellites with different equatorial crossing times [e.g., Eriksson et al., 2010].

[9] For example, the combination of NOAA-18 and MetOp-A which cross the equator at ~14:00 (2:00) LT and 21:30 (9:30) LT for ascending (descending) node, respectively, enables one to decompose the diurnal variation with four times of observations. The 183.31 ± 1 GHz channel brightness temperatures of Microwave Humidity Sounder (MHS) onboard NOAA-18 and MetOp-A are used for the period 2007–2009 to examine whether the day-night contrast might be a fortuitous feature due to the inherent clear-sky sampling bias of infrared measurements. It is noted that the 183.31 ± 1 GHz channel is sensitive to the nearly same atmospheric layer as the infrared counterpart [e.g., Sohn et al., 2000].

[10] The MHS brightness temperatures are averaged over the analysis domain only when all four measurements are available for a given location and for a given day (Figure 1, refer to green lines). Despite the differences in spatiotemporal sampling between geostationary and polar-orbiting satellites, the diurnal cycles from these two instruments are very similar to each other, confirming that the higher nighttime relative humidity is not a spurious feature related to clear-sky sampling.

3 Comparison to Reanalysis Data Sets

[11] The diurnal variations of upper tropospheric humidity depicted in reanalysis products are compared to observations using a profile-to-radiance approach. Atmospheric profiles of temperature and humidity are inserted with surface information into a radiative transfer model [Hocking et al., 2011] to simulate the radiance which would be observed by the satellites under those conditions. Reanalysis data sets examined in this study are NCEP/DOE Reanalysis [Kanamitsu et al., 2002], NASA's Modern-Era Retrospective Analysis for Research and Applications [Rienecker et al., 2011; MERRA], ERA-40 [Uppala et al., 2005] and ERA-Interim reanalysis [Dee et al., 2011] of ECMWF, and the NOAA/CIRES Twentieth Century Reanalysis [Compo et al., 2011]. The temporal sampling interval is 6 hours except for MERRA which produces 3-hourly meteorological parameters.

[12] Figure 2 illustrates the diurnal variation of simulated Meteosat-5 water vapor channel brightness temperature for the period 1984–2004. In agreement with the satellite observations (red lines), most of the reanalysis data sets exhibit larger diurnal amplitudes over the continental convective regions. By contrast, substantially different patterns are identified in MERRA. This feature of MERRA is not limited to specific years. The diurnal amplitude over land is significantly smaller compared to other reanalysis data sets, and the land-sea contrast of the diurnal amplitude is not reproduced in MERRA regardless of seasonal migration. Although the presence of clouds does not cause any sampling issue, reanalysis data sets tend to produce consistently smaller diurnal amplitudes. In particular, ERA-Interim shows relatively muted diurnal ranges over the oceanic convective regions compared to other reanalysis data sets as well as satellite observations. These differences in diurnal amplitude are consistent with previous studies that also found a smaller diurnal cycle of upper tropospheric humidity in reanalysis data sets [e.g., Eriksson et al., 2010].

Figure 2.

Diurnal anomaly of Meteosat-5 water vapor channel brightness temperature simulated from NCEP/DOE, 20th Century Reanalysis, ERA-40, ERA-Interim, and MERRA over the convectively active regions of Africa and the Atlantic Ocean for the period 1984–2004. The geographical domains are same as in Figure 1. The red lines denote the diurnal anomaly of observed brightness temperature for the same period.

[13] Variations in the diurnal phase of the brightness temperature between reanalysis data sets are clearly evident over the continental convective regions. Brightness temperature maximum occurs from ~8:00 LT (MERRA) to ~19:00 LT (NCEP/DOE). A wide spread is also found for the brightness temperature minimum. This implies a phase shift of up to ~10 h compared to the local times of observed brightness temperature maximum and minimum. However, day-night contrast of upper tropospheric humidity is consistent with the satellite observations except for MERRA. On top of the relatively muted diurnal amplitude, MERRA produces a brightness temperature maximum in the early morning over the convectively active regions of Africa. The peculiar characteristics of MERRA occur for both months of the analysis period. In contrast, the reasonable depiction of diurnal phase in the 20th Century Reanalysis is remarkable in that no upper air or satellite observations are assimilated into the reanalysis system [Compo et al., 2011].

[14] The diurnal phase of brightness temperature appears to be more consistent with the satellite observations over the oceanic domains. Moreover, the diurnal phase is reasonably reproduced in MERRA unlike the land case. While interannual variations of the diurnal phase are significant for NCEP/DOE and ERA-40, the newer generations of the reanalysis system tend to have less interannual variations in both phase and amplitude of the diurnal cycle (not shown).

4 Diurnal Cycles in Related Fields

[15] The previous section demonstrated that the diurnal variations of upper tropospheric humidity in reanalyses differ substantially from that observed by satellites. Furthermore, substantial discrepancies exist in the upper tropospheric water vapor between different reanalyses, especially for newer reanalysis data sets [e.g., Dee et al., 2011; Rienecker et al., 2011]. In this section, we investigate the possible causes of these discrepancies by examining the relationship of upper tropospheric humidity with convection, cloudiness, and the large-scale circulation [e.g., Pierrehumbert and Roca, 1998; Soden, 2004; Schmetz et al., 2005; Sohn et al., 2008; Chung et al., 2011]. Our analyses are focused on ERA-Interim and MERRA since they are the only reanalysis products which archived ice water content.

[16] Over land the diurnal variation of relative humidity at 300 hPa (unit: %) is compared to that of 500 hPa vertical wind (hPa/d) and ice water content at 300 hPa (unit: kg/kg) in Figure 3. It is noted that the diurnal variation of relative humidity is not sensitive to the choice of pressure level in the upper troposphere. Consistent with previous observational studies [e.g., Soden, 2000; Tian et al., 2004; Chung et al., 2007, 2009; Eriksson et al., 2010], the ERA-Interim shows higher humidity during nighttime whereas MERRA exhibits the relative humidity peak in the midafternoon. This feature of MERRA is produced for both months and for the entire analysis period.

Figure 3.

Diurnal anomalies of relative humidity at 300 hPa (red lines, unit: %), vertical velocity at 500 hPa (blue lines, unit: hPa/d), and cloud ice water content at 300 hPa (green lines, unit: kg/kg) over the convectively active regions of Africa for (upper panels) January and (lower panels) July: (left) ERA-Interim and (right) MERRA. Each line denotes an individual year of the analysis period (1984–2004), and the diurnal anomaly of 500 hPa vertical wind is multiplied by 0.2 for display purpose. The sign for Omega is inverted by multiplying with −1 so that positive –Omega means upward motion. The geographical domains are same as in Figure 1. Note that the axis and corresponding labels for the diurnal anomaly of cloud ice water content are placed on the right of the plots.

[17] The early morning peak of descending motion (in relative sense) and the midafternoon maximum of the upward motion is similar between the two reanalyses (refer to blue lines). However, the upward motion is maintained throughout evening to midnight in MERRA with a wider diurnal range. In addition, the relation of the vertical wind with relative humidity is substantially different. The diurnal cycles of relative humidity and upward vertical wind are in phase for MERRA, implying the dominant role of the large-scale vertical motion in shaping the diurnal variation of upper tropospheric humidity. On the other hand, the peak of upward motion (12:00–15:00 LT) concurs with the minimum relative humidity in ERA-Interim which is contradictory to satellite observations [e.g., Schmetz et al., 2005; Sohn et al., 2008; Liu and Zipser, 2008; Schröder et al., 2009].

[18] The ice water content at 300 hPa (green lines) exhibits a more consistent diurnal variation between reanalysis data sets. The maximum ice water content is produced in the midafternoon when the upward motion is prevalent. In addition to the midafternoon peak, a secondary maximum is produced during nighttime. The local time of ice water content minimum coincides with that of the descending motion peak, indicating the dominant role of large-scale vertical motion on the diurnal variation of ice water content. Meanwhile, the concurrence of the maximum ice water content with upward motion implies that convective parameterization schemes produce strong ascending motion from noon to midafternoon over land. On the other hand, satellite observations of precipitation systems, deep convective cloud, and the upper tropospheric divergence in convective cloud systems have shown that tropical convective systems attain their diurnal maximum in the late afternoon and evening over land [e.g., Soden, 2000; Yang and Slingo, 2001; Tian et al., 2004; Schmetz et al., 2005; Chung et al., 2007, 2009; Liu and Zipser, 2008; Eriksson et al., 2010]. The earlier maximum of ascending motion in the reanalyses may contribute to the biases in the diurnal phase of ice water content and thus upper tropospheric humidity.

[19] In the case of MERRA, the ice water content and relative humidity exhibit a similar diurnal variation except for a slight phase lag between them. In contrast, the phase relationship in ERA-Interim is represented in a substantially different way. The temporal evolution of relative humidity is approximately synchronized with that of ice water content throughout nighttime. However, the local time of minimum relative humidity (12:00–15:00 LT) coincides with the time of highest loading of ice water content. This pattern appears to be inconsistent with the conclusion of previous studies that the onset of vertical motion not only produces cloud condensates but also leads to ambient moistening through detrainment of moisture and cloud condensates [e.g., Soden, 2004; Schmetz et al., 2005; Sohn et al., 2008; Liu and Zipser, 2008; Schröder et al., 2009; Eriksson et al., 2010].

[20] The diurnal variation shows a less organized feature over the oceanic domains for all three variables with a significant interannual variability and discrepancies between reanalysis data sets (Figure 4). The seasonal difference is also noticeable for the vertical motions of ERA-Interim. As a result, the phase relationships cannot be clearly determined, and the similar diurnal variations of relative humidity for the entire period (thick red lines) and simulated Meteosat-5 water vapor channel brightness temperature (bottom panels of Figure 2) might be a fortuitous feature. Figure 4 indicates that the roles of vertical motion in the diurnal variations of ice water content and relative humidity are indistinct over the oceanic domains. In addition, in ERA-Interim, the phase relation between relative humidity and ice water content is noticeably different from the land case. Although the differences from the land case could be attributable to the disparate convection mechanisms between land and ocean [e.g., Yang and Slingo, 2001; Yang and Smith, 2006; Liu and Zipser, 2008] and the differences in convective parameterization schemes and data assimilation processes between reanalyses, the lack of coherence between even the latest reanalyses raises doubts regarding the performance of convective parameterizations and the upper tropospheric moistening processes, as well as the use of observations to effectively constrain the moisture budget of the upper troposphere.

Figure 4.

Same as in Figure 3, but for the convectively active regions of the Atlantic Ocean. Thick lines denote the diurnal anomaly for the entire period (1984–2004).

5 Summary

[21] The latest generations of reanalysis products have made substantial efforts to reduce the uncertainties in the distribution and variations of the water and energy budgets by employing advanced parameterization schemes and data assimilation systems. However, this study has found that the diurnal variation of upper tropospheric humidity depicted in reanalysis data sets are distinctly different from that determined from satellite observations with noticeable differences among reanalyses. This suggests that the physical processes governing the diurnal cycle of deep convection and the consequential moistening of the upper troposphere in reanalyses are incomplete and that observations to accurately constrain the moisture budget of the upper troposphere are not being used effectively.

Acknowledgments

[22] We would like to thank three anonymous reviewers for their constructive and valuable comments which led to an improved version of the manuscript. Meteosat clear-sky WV radiances were obtained from the IPSL CLIMSERV Website. MHS data were provided by the NOAA Comprehensive Large Array-data Stewardship System. NCEP/DOE Reanalysis data and the 20th Century Reanalysis V2 data were provided by the NOAA/OAR/ERSL PSD, Boulder, Colorado, USA, from their Website at http://www.esrl.noaa.gov. ERA-40 and ERA-Interim reanalysis data were obtained from the ECMWF data server. MERRA data were obtained from the NASA Goddard Earth Sciences Data and Information Services Center. This study was supported by a grant from the NOAA Climate Program Office. Byung-Ju Sohn was supported by the Korea Meteorological Administration Research and Development Program under grant CATER 2012–2061.