Report

Dynamic Processes Governing Lower-Tropospheric HDO/H2O Ratios as Observed from Space and Ground

See allHide authors and affiliations

Science  11 Sep 2009:
Vol. 325, Issue 5946, pp. 1374-1377
DOI: 10.1126/science.1173791

Abstract

The hydrological cycle and its response to environmental variability such as temperature changes is of prime importance for climate reconstruction and prediction. We retrieved deuterated water/water (HDO/H2O) abundances using spaceborne absorption spectroscopy, providing an almost global perspective on the near-surface distribution of water vapor isotopologs. We observed an unexpectedly high HDO/H2O seasonality in the inner Sahel region, pointing to a strong isotopic depletion in the subsiding branch of the Hadley circulation and its misrepresentation in general circulation models. An extension of the analysis at high latitudes using ground-based observations of δD¯ and a model study shows that dynamic processes can entirely compensate for temperature effects on the isotopic composition of precipitation.

Water vapor is the most important greenhouse gas in the atmosphere. As saturation vapor pressure increases exponentially with temperature, a positive feedback effect with respect to the current global warming trend is expected and confirmed by satellite measurements over the ocean (14). However, highly complex interactions via cloud formation and the release of latent heat, impacting convection, complicate matters and seem not to be well represented in climate models, especially in the tropics (5, 6). Land-atmosphere coupling adds further uncertainties (7, 8). An accurate knowledge of hydrological cycles and feedback mechanisms is therefore indispensable for reliable weather (7) and climate predictions (13, 8).

The isotopic fractionation of water provides a deeper insight into hydrological cycles as evaporation and condensation processes deplete heavy water in the gas phase (911). In paleoclimate applications, the isotopic composition of water in the polar ice sheets has become a key temperature proxy (10, 12, 13). The Global Network of Isotopes in Precipitation (GNIP) (14), initiated in 1961, uses isotope data in hydrological investigations for water resources inventory, planning, and development. However, it only interprets precipitation, the end product of a multitude of evaporation, condensation, and mixing processes. Isotope measurements of water vapor would give important additional information but are more difficult to make. Atmospheric applications traditionally focus on isotopes as a proxy for exchange processes in the upper troposphere and lower stratosphere (e.g., 15), but only a few in situ measurements of the isotopic composition of lower tropospheric water vapor are available (1618). Some general circulation models have incorporated water isotopologs (1921), which can, among other purposes, be used to scrutinize the linear δ18O/T relationship in ice core measurements (13). However, their sole global source of validation was, for decades, precipitation measurements (14), which partially inhibits a physical explanation for intermodel differences because large-scale vapor phase measurements were missing.

The first global measurements of atmospheric deuterated water (HDO) in the middle to lower troposphere were performed by Interferometric Monitor for Greenhouse Gases (IMG) aboard ADEOS (22) and, only recently, the Tropospheric Emission Spectrometer (TES) aboard the Aura spacecraft (23). These measurements were performed in the thermal infrared and are mostly sensitive to the 550 to 800 hPa layer, hence lacking sensitivity to the lowest troposphere where, owing to the scale height of 1 to 2 km, most water vapor resides. As this layer is of the utmost importance for an understanding of hydrological cycles, we apply a previously overlooked technique to obtain global information with high sensitivity near the ground by retrieving δD¯ from short-wave infrared spectra recorded by the SCIAMACHY (Scanning Imaging Absorption Spectrometer for Atmospheric Chartography) instrument (24) aboard the European Space Agency (ESA)’s environmental research satellite ENVISAT. Owing to the large spectral displacement of rotational-vibrational transitions of HDO, it exhibits absorption lines distinct from those of H2O. In a wavelength window ranging from 2355 to 2375 nm, this enables a simultaneous retrieval of HDO and H2O vertical column densities (equivalent to total precipitable water), denoted as VCD [see (25) for more information on retrieval procedures, instrumental details, data quality criteria, potential problems over the oceans, and precision estimates]. The HDO abundance relative to standard mean ocean water (SMOW) (Rs = 3.1152 × 10−4) in delta-notation now reads

δD¯=1000%˙(VCD(HDO)/VCD(H2O)RS1)

where δD¯ is the mass weighted column average of the relative HDO abundance expressed in per mil (‰) with almost equal sensitivity at all height layers (25), thereby allowing for a direct comparison with total column averages from atmospheric models. Because of the relatively high detector noise of SCIAMACHY in the short-wave infrared channel 8, the single measurement noise (1-sigma precision error) in δD¯ is typically 40 to 100 ‰, depending on total water column, surface albedo, and viewing geometry (25). Hence, we strongly average in space and time. Over the oceans, we mainly rely on retrievals of δD¯ above low-level clouds (25) because the ocean surface albedo is too low in the near infrared.

Figure 1 shows the global δD¯ distribution as derived from 3 years of SCIAMACHY spectra. The dominant feature in the global distribution is the strong latitudinal gradient, as also observed in precipitation (14) and by the TES instrument (23). Owing to its high sensitivity to the boundary layer, several new insights can be gained from SCIAMACHY. Over the large ocean basins, where few continuous GNIP stations are located, the high δD¯ values typical for the tropics extend farther north than over land. This is most striking in the north Atlantic, where zonal symmetry is broken and δD¯ iso-lines point northeast (see arrow). This feature can be attributed to the warm Gulf Stream and strong evaporation, associated with storm tracks efficiently transporting vapor northeastward.

Fig. 1

(A) Mean data from 2003 to 2005, shown on grids of 4° by 4° without further smoothing (only grid boxes where more than 15 measurements are available are shown). Over land, we use grids of 1° by 1° (if more than 15 measurements are available) to show small-scale features. The figure should be interpreted keeping the total number of measurements (B) and potential seasonal biases (C) in mind. See (25) for a more detailed discussion of biases toward specific seasons, potential problems over the oceans, and data quality in general. (B) Square root of the total number of measurements (on 4° by 4°) in the considered time period. Due to the very low albedo in the near infrared, only a few measurements exist over the oceans where we rely on low-level clouds for sufficient signal. (C) Maximal seasonal contribution (from four seasons: December to February, March to May, June to August, and September to November) to the long-term mean. 25% means that all seasons equally contribute, while 100% means that the mean represents one season only (see fig. S3 for the contribution of the individual seasons).

Similar to observations from the GNIP network (9, 14), the “continental effect” appears very small in tropical regions such as the Amazon basin or central Africa, where continental water vapor is least depleted. Even though we acknowledge low measurement frequency (Fig. 1B) and a potential seasonal bias (Fig. 1C and fig. S3), tropical δD¯ values are substantially higher than observed by TES (23). This points to very low depletions in the boundary layer, which can be caused by strong evapotranspiration, recycling water, and exhibiting little fractionation of the typically only slightly depleted moisture in the tropics (14). Strong continental gradients, however, are observed in North America and, to a lesser extent, in Eurasia. The smaller gradient in Eurasia can be explained by a relatively efficient transport of Atlantic water vapor into the continents, whereas in North America the Rocky Mountains and the associated precipitation obstruct the propagation of oceanic moisture (25). In addition to the general global distribution, SCIAMACHY identifies and quantifies small-scale features, for example, the altitude effect on the global scale with strong δD¯ depletions over mountain ranges such as the Andes and a strong evaporation source resulting in relatively HDO-enriched vapor amidst arid regions such as over the Red Sea (fig. S5).

Seasonal δD¯ variations provide important insights into dynamic aspects. Figure 2A shows a difference plot of the δD¯ distribution between winter and summer [δD¯ (December to February) – δD¯ (June to August)]. Similar to H2O abundances and temperature, δD is generally higher in local summer than winter, which is manifested in the clear difference between the hemispheres. However, seasonal changes in evaporation, precipitation, and the general circulation also affect variability. The highest seasonal differences at low latitudes, almost 100‰, are observed in the semiarid inner Sahel, the transition zone between the Saharan desert in the north and less arid Savannas in the south. Isotopic variations can be attributed to seasonal changes in the atmospheric circulation rather than to temperature variations: In December through February, the subsiding branch of the Hadley circulation (not accessible to precipitation measurements) transports dry and thus HDO-depleted air masses to this region, which itself, in contrast to other regions (25, Sec. 2.2), contributes little evaporation that could mix HDO-rich water vapor into the dry descending air. Figure 2B shows a time series of SCIAMACHY retrievals along with results from the general circulation model IsoGSM (21, 26) in this region. Even though the general shape of the seasonal cycle is remarkably similar, the overall amplitude and gradients in time are stronger in SCIAMACHY observations. Figure 3A shows that the integrated water vapor column (or total precipitable water, TPW) as measured by SCIAMACHY agrees well with colocated assimilated ECMWF (European Centre for Medium-Range Weather Forecasts) model fields. IsoGSM, however, overestimates the water column, especially in dry conditions (corresponding to boreal winter). We suppose that the subsiding branch of the Hadley circulation is substantially drier and hence more depleted in HDO than modeled by IsoGSM. Figure 3B underlines that most of the underestimation in the isotopic seasonality can be attributed to lower variability in the integrated water column. However, the slope of the lines, which is indicative of fractionation temperature, also differs and points to a model underestimation of fractionation occurring at colder temperatures, potentially in the convective tropical cells at higher altitudes. Considering long-term climatologies of other general circulation models (27) (note the relatively coarse resolution) over the considered region in Africa, intermodel differences in TPW and δD¯ are large, and, like IsoGSM, neither manages to simultaneously reproduce variations in observed TPW and δD¯ (fig. S12). The intensity of the Hadley circulation in general circulation models differs widely, and observational constraints are missing (28). The combination of a metric for the state (TPW) and the history (δD¯) of atmospheric water vapor from SCIAMACHY is unique and should provide hitherto missing constraints for, e.g., moist convective parameterizations or land-atmosphere coupling, both being prime uncertainties in climate simulations.

Fig. 2

(A) Seasonal differences in Embedded Image shown as December to February minus June to August averages. The latitude range is restricted because SCIAMACHY local winter retrievals are compromised by snow cover and high solar zenith angles. Retrievals over ocean are less reliable and also depend more strongly on clouds; thus, we restricted the plot to land masses only. (B) Monthly mean time series in a subregion over the Sahel [indicated as a white box in (A), 0°E to 10°E, 15°N to 23°N] for both SCIAMACHY retrievals and IsoGSM.

Fig. 3

Analysis of the Sahel region between 0°E to 10°E and 15°N to 23°N (see white box in Fig. 2A). (A) Correlation plot of SCIAMACHY and IsoGSM total precipitable water (TPW) against ECMWF assimilated model fields. (B) Rayleigh-type plot in logarithmic scale for SCIAMACHY and IsoGSM (the slopes correspond to a Rayleigh-fractionation temperature of 305±6 K and 281±9 K for IsoGSM and SCIAMACHY, respectively).

Seasonality of δD¯ at higher latitudes is expected to be more directly driven by strong temperature variations, especially in the inner North American and Eurasian land masses. This is partly reflected in SCIAMACHY observations in central North America and Mongolia. However, retrievals in winter at higher latitudes than shown in Fig. 2 are compromised by snow cover and high solar zenith angles, impeding a comprehensive seasonality analysis based on SCIAMACHY only, especially in the potentially interesting exit region of North Atlantic storm tracks. Hence we expand our analysis by retrieving δD¯ from ground-based Fourier Transform Spectrometer (FTS) (29) stations at different northern latitudes, namely in Bremen (Germany, 53.0°N 8.8°E), Kiruna (Sweden, 67.8°N 20.3°E), and Ny Ålesund (Norway, 78.9°N 11.9°E). Figure 4 shows that the monthly climatologies of IsoGSM underestimates the amplitude of the seasonality, most prominently in high arctic Ny Ålesund. If the model is sampled at the measurement times of the FTS retrievals, however, the correspondence of the model seasonal cycle with FTS retrievals is far better. Surprisingly, δD in precipitation as measured by GNIP and modeled by IsoGSM agree relatively well, both showing almost no seasonal variation in precipitation δD. This suggests that precipitation (measured by GNIP) as well as clear-sky days (measured by FTS) are linked to very specific meteorological conditions. In the case of precipitation, any correlation of δD with the seasonal cycle in local temperature is eliminated, while seasonal differences of δD¯ in vapor during clear days are stronger. A model study (25, Sec. 2.3) shows that clear-sky days, as measured by the FTS, are indeed mostly linked to polar air intrusions, whereas for precipitation, the effects of temperature variations are compensated by concurrent changes in moisture source area. We suggest that this effect is related to the contrast of sea surface temperature against air temperature in the lower troposphere, resulting in maximum ocean evaporation and more local moisture origin of precipitation in winter. The deuterium excess (d) as measured in precipitation in Ny Ålesund supports this view, because there is a clear seasonality in d-excess, with maxima in winter pointing to strong kinetic fractionation during ocean evaporation in nonequilibrium. This offers an alternative explanation for d-excess changes in ice core records, which are often attributed to moisture source region temperature [e.g., (30)]. In fact, summer precipitation in Ny Ålesund, which is more influenced by moisture evaporated from the warmer North Atlantic below 50°N (25), exhibits lower d-excess, hence an anticorrelation with source temperature. For a coastal high-arctic station, seasonal changes in the atmospheric circulation and ocean-air temperature contrast can thus compensate for temperature effects, both in δD for local temperature and d-excess for source region temperature. In the Ny Ålesund area, δD¯ seasonality is, compared with other models, best represented by IsoGSM (25). Although data are only available for a coastal station, this indicates that moisture transport largely influences high-arctic isotopic variability and that its misrepresentation in general circulation models (possibly due to differences in storm track activity between reanalysis data and general circulation models) can be critical.

Fig. 4

(Top) Modeled and measured precipitation Embedded Image and d-excess monthly averages at high arctic Ny Ålesund (Norway, 78.9°N 11.9°E). (Bottom) Modeled (climatologies as solid lines and sampled at FTS measurement times as dashed lines) and retrieved vapor Embedded Image at Bremen (Germany 53.0°N 8.8°E), Kiruna (Sweden, 67.8°N 20.3°E), and Ny Ålesund. FTS retrievals have been arbitrarily shifted to match the model as we focus on seasonality and because the different stations apply different retrieval windows.

In our study, we provide observational evidence that ice core isotope records are indeed a mixed proxy for dynamics and temperature, as suggested by model studies (13, 3133). Hence, interpretations based on simple assumptions such as pure Rayleigh fractionation processes are inaccurate, requiring the application of general circulation models. However, a combination of diverse isotope records might help simultaneously reconstruct changes in both past temperatures and dynamics. To this end and for future climate predictions, general circulation models have to be validated in their ability to correctly represent current hydrological cycles, including cloud processes, moist convection, and atmospheric transport, especially to higher latitudes. The combined information on the water vapor total column and its isotopic composition from space provides a crucial process-oriented benchmark for climate models.

Supporting Online Material

www.sciencemag.org/cgi/content/full/325/5946/1374/DC1

Materials and Methods

SOM Text

Figs. S1 to S13

References

Movie S1

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Owing to the presence of strong synoptic variations in water vapor total column and its isotopic composition, a meaningful comparison only holds for colocated measurements. A direct comparison with an atmospheric model is thus currently only possible with IsoGSM as its large-scale motions (but not water vapor, because the model has its own hydrological cycle) are nudged to NCEP2 (National Centers for Environmental Prediction) Reanalysis data, thereby excluding strong sampling biases.
  3. We thank all the scientists and engineers involved in the ESA’s ENVISAT/SCIAMACHY mission, especially J. P. Burrows and H. Bovensmann from the University of Bremen, and, apart from the ESA, the national space agencies of Germany [German Aerospace Center (DLR)], Netherlands [Netherlands Space Office (NSO)], and Belgium [Belgian User Support and Operations Center (BUSOC)]. The ground-based FTS activities at Kiruna and Izaña were supported by the Deutsche Forschungsgemeinschaft through project RISOTO. We are grateful to the Alfred Wegener Institute for providing the FTIR instrument and logistical support in Ny Ålesund. Financial support from the European Union project GEOMON for the FTS stations is acknowledged. We thank H. Sodemann for a valuable discussion on an earlier manuscript draft; R. Snel, R. van Hees, and P. van der Meer for their work on instrument calibration; and A. de Lange and P. Tol for proofreading and image creation, respectively. C.F. was supported by a VENI fellowship from the Dutch science foundation NWO.

Stay Connected to Science

Navigate This Article