Atmospheric humidity and clouds - what can we learn from stratospheric water? Stephan Fueglistaler

Transcription

1 Atmospheric humidity and clouds - what can we learn from stratospheric water? Stephan Fueglistaler

2 The stratosphere is the driest place in the atmosphere as we know it. This workshop is about "Convection, water vapor, and climate." so why think about the region where the object of interest is virtually absent? -> Wrong session?

3 Stratospheric water vapor and the corresponding problem in the troposphere Motivation (troposphere): - Global tropospheric RH seems fairly robust; - ditto for cloud radiative forcing. These two components are the largest modifiers of the radiative budget, but we do not have a very good understanding of what controls the their global abundances. A few reasons why the stratospheric problem may be interesting:

4 (i) Stratospheric water vapor is controlled by conditions in a relatively confined region of the atmosphere (vicinity of the tropical tropopause). (ii) There is a large signal of natural variability that allows studying perturbations. (iii) The variability shows distinct patterns (importance explained later): zonally symmetric (QBO, strat. residual circulation), and asymmetric (ENSO). (iv) Only very weak feedback of resulting moisture field on conditions in controlling region. (v) Now about 20 years of remarkably accurate data (little dehydration in stratosphere also allows closure studies, tests of self-consistency of observations).

5 Coldest point in the system; regulates H 2 O "entry mixing ratios" for the stratosphere. Once in stratosphere: source: CH 4 -oxidation; sink: dehydration in polar vortices (in today's climate, weak effect from Antarctic vortex).

6 Motivation: Understanding this specific curve Minimum model TTL dynamics (large scale vs mesoscale) Cloud microphysics Theory with predictive skill? New information about climate (dyn+chem) of past few decades (attribution to processes). Remarkably poor state of theory re. global fields of water vapour and clouds. What controls mean RH? Why nearly constant? What controls global cloud fields? Why albedo 0.3?

7 This talk some more theoretical aspects; concepts, ideas.

8 Relative humidity conceptual perspective Relative humidity: IF the partial pressure (or mixing ratio) can be understood as the saturation vapour pressure at some (unknown) location X, we have

9 And it follows (using approx. expression for Clausius- Clapeyron) (beta = L/R; latent heat/gas constant)

10 Dehydration concept for air parcel Total water in parcel Point of Last Contact with Condensate / Last saturation Lagrangian Dry Point

11 For stratospheric water vapor we are not interested in relative humidity in the stratosphere and hence don't care much about 'T'; but what we are interested is the relation between "Eulerian average temperature changes" in the "x/ LDP"-domain, and the average of dt x (i.e. Lagrangian perspective, only considering locations and times of last dehydration).

12 Seems simple need to worry only about this small region. But -

13 The importance of transport and zonally asymmetric temperature (changes) In tropical upwelling region, saturation mixing ratios vary by one order of magnitude. Circulation is important. [Data: ERA-40, 370K pot. Temperature, January 2000] [ppmv]

14 -> Need objective way to determine the relevant positions/times (i.e. the Lagrangian Dry Point, LDP). One way to do so is to use trajectories (whether this captures the relevant transport mechanism is an open question).

15 The LDP distribution (time/pressure) LDP s for back trajectories started at 82hPa(30S-30N) in February and October (each point 1 trajectory). -> LDP distribution is clustered around cold point, but highly variable in time.

16 The climatological mean LDP distribution (lon/lat) Color: LDP temperature; Contours: LDP density. -> Clear tendency for LDP to be preferentially at coldest locations.

17 Interesting/perplexing result: [Fueglistaler and Haynes, 2005] Color: "Eulerian" temperature anomalies Dark contour lines: LDP density. -> The LDP distribution is highly variable, and responds strongly to changes in temperature structure.

18 "Bias" towards coldest regions True diff: -1.2K w/o weighting: identical T (There's a catch here discussion?)

19 LDP temperatures and 100hPa average temperature [Fueglistaler and Haynes, 2005] (This figure looks very similar for calculations based on ERA- Interim, and/or comparison with cold point Temperature.)

20 -> Like in troposphere, first guess that the average temperature change in "x" (or LDP) is well captured by the average temperature change in the whole domain works very well, despite the large changes in the spatial distribution of the "x"/ldp-field.

21 -> Instead of an "exciting" story of how humidity change "decouples" from mean temperature changes, we are (also) confronted with the problem of analyzing a problem that works out in such a way that many (most?) don't even perceive it as a problem.

22 Approach GFD or probabilistic problem? Not quite clear at present problem has clearly components of a stochastic problem (probability to experience a certain temperature), but it's not quite clear whether a random-walk approach captures the problem well (recall that flow is well structured). -> Perturbation experiments (analytical perspective) as a semi-empirical approach to understand the key characteristics of the system, which then can help formulate the idealised problem.

23 Problem is nonlinear / nonlocal Temperature perturbations have different impact on <T LDP > depending on where, and sign/magnitude of perturation [ppmv] versus

24 (I) Dispersion Kinematic diabatic ERA-40 Temperature ERA-Interim Temperature [Liu et al. 2010]

25 (II) "Exposure time" Artificial modification of upwelling by multiplication of model diabatic heating with a factor k. -> Average LDP frostpoint temperature proportional to ln(k). -> clue for statistical description? [Liu et al., 2010]

26 (III) Variance Method: For given trajectories, perturb T-field and re-evaluate LDP. (Spatial scale 30deg/10deg; time scale variable.) Quasi-stationary T-perturbation (LDP dist. not uniform -> large std-dev between error patterns.) Fast fluctuations -> maximum bias. [Liu et al., 2010]

27 (IV) Variations around mean state

28 Summary Atmospheric humidity: - Problem is very non-linear. - Stratospheric water may be a case where the problem is somewhat simpler. - Up to now, focus of work on physical processes. - Perhaps time to think about theoretical problem even if not all physical processes are correct. To bear in mind: - Careful identification of properties that simplification should reproduce.