Global Remote Sensing
of Water Vapor and the Hydrological Cycle
Background
The water vapor
remote sensing challenge
What GPS RO offers
Research Goals
GPS RO moisture accuracy and information content
Global water vapor climatology
GPS and NWP global moisture comparison: Means and Biases
Variability of water vapor and accuracy of global water analyses
The tropical water vapor distribution and water vapor feedback
GPS-derived Water Vapor and Clouds
Background
My interests lie largely in atmospheric water, the hydrological cycle and how water concentrations are controlled particularly the climatic feedbacks related to water. Much of my work focuses on determining the global water vapor distribution and its variations.
Water is the most important greenhouse gas and small amounts of water in the form of clouds strongly affect both short wave and long wave radiation. Buoyancy created by changes in the phase of water in large part drives the vertical motion in the atmosphere. Precipitation of water largely defines the extent and type of the continental biosphere. Furthermore all of these are likely changing in response to changing greenhouse gas concentrations (GHG) concentrations. It is therefore crucial that we understand and can correctly model the hydrological cycle and the distribution of atmospheric water and its phase changes and their evolution in a changing climate.
The complex water-related interactions and feedbacks in our weather and climate system make such a task daunting but do not diminish its necessity. Reducing the present uncertainty in the future state of our climate requires the understanding and modeling skill necessary to make predictions. This in turn requires observations of sufficient quality to discern the true state of the atmosphere at its important temporal and spatial scales of variability that capture trends in behavior and in particular reveal tell-tale signatures of processes at work, signatures critical for evaluating and improving the realism and predictive skill of models.
The water vapor remote sensing challenge
Despite water vapor's fundamental importance, it is very difficult to observe the water vapor variations over the enormous range of spatial and temporal scales and the 5 orders of magnitude variation in concentration over which it varies from the tropical surface into the lower stratosphere. Furthermore, a small amount of water condensed into cloud droplets strongly affects the propagation of light at IR and shorter wavelengths which limits the ability to remotely sense the portion of the atmosphere within and below clouds.
Characterization of atmospheric water vapor by passive nadir-viewing sounders is limited inherently by vertical weighting functions that are coarse relative to the vertical scales at which water varies. NASA's Atmospheric IR Soundet (AIRS) profiles water vapor ~2 km vertical resolution and 10 to 15% accuracy in clear skies (Chahine et al. 2001). Such information is available only in a small percentage of the globe because ~95% of the AIRS pixels are cloud contaminated [J. Joiner, pers. comm.]. Success in applying AIRS data has been achieved via selectively choosing cloud free pixels that has allowed AIRS to have a significant positive impact on NWP forecasts (Lemaster et al., 2005). The climatology of an AIRS or ARS equivalent set of observations are inherently biased to clear conditions leaving the water vapor in, around and below clouds largely to the realm of models.
Microwave observations can penetrate clouds far better particularly at lower microwave frequencies providing a less biased set of observations. Nadir-viewing, passive microwave observations provide still coarser vertical resolution than IR and are limited largely to marine environments where the surface emissivity varies relatively little. The Microwave Limb Sounder (MLS) provides very powerful observations of water vapor in the upper troposphere and stratosphere in cloud free regions.
Surface lidars can provide very powerful information on water vapor, temperature and winds in clear air below clouds but they are very expensive and fairly delicate instruments of limited deployment and therefore coverage.
What GPS RO offers
GPS RO certainly has yet another set of features and limitations. Its features are largely unique and therefore complementary to those of other water vapor sensors. When combined with independent temperature analyses, GPS RO provides
á Profiles of specific and relative humidity
á With high precision
1. Specific humidity: s = 0.2 to 0.5 g/kg , bias ~0.1 g/kg (Kursinski et al., 1995;Kursinski and Hajj, 2001)
2. Relative humidity see below)
á And high vertical resolution (~200 m)
á Spanning the lower troposphere into the upper troposphere to temperatures of approximately –25oC
á in virtually any weather conditions [Kursinski et al., 1995, 1997; Kursinski and Hajj, 2001].
Because GPS RO measurements are made in a limb viewing geometry, they have relatively coarse along-track resoltuion of approximately 300 km, similar to many present GCMs which limits it to larger horizontal scales of variability. The ~20 cm wavelength GPS observations are essentially independent of cloudiness (clouds affect GPS RO water vapor estimates by 1% or less [Kursinski et al., 1995]). GPS RO therefore provide
á a global, high vertical resolution water vapor dataset that is the closest to an unbiased global dataset that exists at present (until ATOMMS reaches orbit).
I have focused much effort on understanding, assessing and developing the unique capabilities and information content of GPS RO for characterizing the hydrological cycle.
Research Goals:
Some of my water vapor research goals include
a. Quantifying the water vapor information content of GPS RO
b. Developing an unbiased global, high vertical resolution global climatology of water vapor including its diurnal cycle,
c. Assessing the realism and accuracy of the water distribution in NWP analyses and free-running models,
d. Determining the impact of GPS on the weather forecasting and climate moisture (re)analyses
e. Establishing the relation between relative humidity and clouds observationally
f. Refine and improve weather and climate model hydrological cycle realism and forecasting skill.
GPS RO moisture accuracy and information content
Understanding the information and value of the GPS water vapor requires knowledge of both the quality of other water vapor observations and analyses and that of the GPS water vapor estimates. I have derived estimates of the GPS RO accuracy of water vapor partial pressure (Kursinski et al., 1995, 1997) and specific humidity (Kursinski and Hajj, 2001). Because of our recent focus in modeling of tropical moisture (see below and Sherwood et al., 2006), I have been developing a relative humidity error analysis that includes estimates of the variation of the GPS refractivity errors in as a function of the concentration of water in the atmosphere. As the figure to the right shows, results indicate relative humidity errors will be approximately 15% in very wet regions and < 2% in areas of very low relative humidity. These results are very important for assimilating GPS RO profiles into NWP models as well as for climatological accuracy. A manuscript describing this error model is in progress.
Global Water Vapor Climatology
We are developing a water vapor climatology from GPS RO in terms of monthly 1st and 2nd moments (means and variances) as well as the entire probability density function and clustering. Clusters and their geographical patterns reveal new features of the character of the moisture distribution captured for the first time with the high vertical resolution GPS global database. We will show examples of each of these aspects of the climatology.
Coverage: The water vapor results derived from the CHAMP and SAC-C GPS RO data since 2001 have enabled us to begin regional, seasonal and year-to-year characterization of variability. The sampling density will increase by a factor of 10 to 20 together with complete diurnal sampling with the COSMIC GPS occultation profiles. COSMIC launched in April 2006.
GPS and NWP global moisture comparison: Means and Biases
GPS-ECMWF
moisture biases: Evaluating model and analysis
quality is to assess biases relative to independent data sets. Hagemann et al. [2003] found the ECMWF analyses generally
capture the column water quite well in comparison with ground-based GPS
estimates of column water, largely reflecting accurate estimates of moisture in
the lowermost troposphere. The
small column water errors are a bit misleading as Figure 5 reveals systematic
problems in the vertical distribution ECWMF assigns to the column
moisture.
Using GPS RO we
have extended the ECMWF bias evaluation into the free troposphere. Figure 8 shows the July 2002 GPS-ECWMF
moisture bias in two 500 m thick slices centered at 5.25 and 6.75 km altitude
respectively revealing substantial free tropospheric differences. The
differences are significant with large regions of 16 to 32% and larger biases
and a bias pattern that changes substantially over the 1.5 km altitude interval
separating the two panels. So
substantial discrepancies exist in the free troposphere (where the water vapor
feedback is most important) revealing systematic problems likely related to the
vertical resolution of the ECMWF moisture analyses and ECMWF model used in the
data assimilation process. Such differences must be understood before we can
assess how our climate is changing.
Figure 8: Mean GPS minus ECMWF specific humidity difference in % for
the month of July 2002 for two 500 m thick layers centered at 5.25 and 6.75 km
altitude respectively. Contours
are at 0, +8, +16, +32 and +64%. Green
indicates GPS moisture is greater than ECMWF estimates. Red
indicates areas where ECMWF estimates are greater.
Variability of water vapor and accuracy of global water analyses
We have used the GPS RO moisture observations to
determine errors in the global moisture analyses and estimate the impact the
GPS observations should have on the global analyses. Large discrepancies often exist between individual GPS and
ECMWF moisture estimates. To sort
out what was happening, we decomposed each specific humidity (
q) estimate from GPS and ECMWF into truth plus error
and each of these into a mean plus variable component:
(1)
where qG and qE are q from GPS and ECMWF respectively, eE and eG are the errors in the ECMWF and GPS
estimates of q
respectively and overbars and primes refer to the zonal mean and variable
components of each term. From the
observations we can directly calculate the variances of qG, qE and qG - qE which by (1) are combinations
of the variance of the true moisture variations (sq2),
the GPS errors (seG2), and the ECMWF errors (seE2)
and the crosscorrelations: <eG'eE'>, <q'eG'> and <q'eE'>. By applying our understanding of the GPS errors, we extracted
the latitude versus height dependence of the ECMWF moisture analysis errors in
June-July 1995 (Kursinski et al.,
2003) and more recently the July 2002 analyses shown in Figure 9. Fractional ECMWF errors in the summer
Northern Hemisphere are similar to moisture error covariances used in the ECMWF
NWP system. Winter Southern
Hemisphere errors are significantly worse but have also improved significantly
between 1995 and 2002. Examination
of behavior in other months have revealed the errors are worst in the winter
subtropical free troposphere regardless of hemisphere.
Figure 9. Estimated ECWMF moisture errors. a. June 22-July 4 1995. b. July 2002. Error is in fraction of zonal mean water vapor. (0.5 = 50%)
Contours are 25, 50,
100, 140 and 200%.
Using our estimates of GPS and the ECWMF moisture errors, we derived a simple theoretical, least squares improvement in the July 1995 and 2002 global moisture analyses if the GPS moisture information were assimilated. As shown in Figure 10, the improvement is expressed as the ratio of the analysis moisture error with GPS assimilation divided by the analysis error without GPS. The impact is substantial with errors reduced by factors of 2 to 3 over much of the troposphere. With our new 2D to 1D mapping occultation data assimilation scheme, forward modeling errors should decrease in some cases perhaps by as much as factor of 5-10 in the lower troposphere increasing the weight given to GPS and its impact in the lower troposphere [Syndergaard et al,. 2005].
Figure 10: Estimated improvement in ECMWF moisture
analyses with the assimilation of GPS refractivity for July 1995 and 2002
conditions. Contours show the
moisture error after
assimilation of the GPS refractivity divided by the moisture error prior to
assimilation of the GPS results
(in %). Contours are 30, 50, 70 and 90%.
The tropical water vapor distribution and
water vapor feedback
Most estimates of the water vapor feedback indicate the feedback will approximately double the surface temperature increase due to CO2 increases alone. In the tropics this is less clear because climate models may not be controlling the tropical water vapor distribution consistent with reality. Many researchers expect the free tropospheric water vapor distribution in the tropics is controlled by detrainment from deep convective towers and subsequent sinking or subsidence where the specific humidity is largely unchanged on the way down. Models on the other hand largely control the free tropospheric water distribution via vertical transport of water upwards into the tropical free troposphere from the moist boundary layer. The figure below shows a group of very dry water vapor profiles from GPS RO.
I co-authored a paper with Steve Sherwood and Bill Read (Sherwood et al., 2006) that developed a very simple model of the probability distribution of relative humidity and then compared it with GPS and MLS relative humidity estimates.
A key point is that in studying the accuracy of GPS RO, I have found that GPS RO is able to capture an accurate reconstruction of the complete probability distribution of relative humidity in the tropical (30oS to 30oN) free troposphere up to about 9 km altitude. A manuscript is in progress describing the accuracy.
I am continue to work on this model, understanding its seasonal dependence because there is such a dependence clearly evident in the GPS RO data.
GPS-derived Water Vapor and Clouds
Parameterization of clouds in models often relies on a parameterization involving relative humidity. Such a relation is not well established observationally, particularly at the vertical scales of clouds. Our ability to measure relative humidity in and around clouds at the important vertical and horizontal scales of clouds is quite limited. Spaceborne passive IR and visible instruments observe cloud tops and can penetrate somewhat into the clouds where optical depths are low with vertical resolution much coarser than the vertical scales at which clouds are known to vary. Cloudsat will provide a great deal of information about the vertical structure of clouds within a narrow swath. One thing missing are commensurate observations of water vapor within and below the clouds. GPS provides high vertical resolution but coarse horizontal resolution similar to typical climate models. We have begun a statistical characterization of the GPS relative humidity profiles and concurrent clouds from ISCCP, passive sensors and eventually Cloudsat to determine whether there is a strong relation between cloudiness and relative humidity at the 300 km horizontal scales that GPS can resolve and at which climate models are typically run.
References
Chahine et al. [2001]
Hagemann et al. [2003]
Kursinski et al., 1995,
Kursinski et al., 1997;
Kursinski and Hajj, 2001
Kursinski et al., 2003
Lemaster et al. 2005
Sherwood et al., 2006
Syndergaard et al,. 2005