On the relative humidity of the Earth s atmosphere

Transcription

1 On the relative humidity of the Earth s atmosphere Raymond T. Pierrehumbert The University of Chicago In collaboration with Remy Roca, LMD 1

2 The Many Roles of Water in Climate IR radiation, heat transport, lapse rate, clouds, land ice, sea ice, chemical weathering... (see Pierrehumbert Nature 2003). This talk will focus on radiative feedback of water vapor What determines the degree of subsaturation of the free troposphere? How do we know if GCM s are representing this correctly? How can we represent these processes in idealized climate models? (like EBM but better...) 2

3 Water Vapor Basics Water vapor is a greenhouse gas Greenhouse gases warm the planet by reducing the IR cooling to space ( OLR ) corresponding to a given T (z) For a given T (z) OLR varies logarithmically with specific humidity For typical midlatitude conditions, doubling q reduces OLR by 6W/m 2 Saturation vapor pressure increases (roughly) exponentially with temperature. This matters, but because the atmosphere is unsaturated, its influence on water vapor content is subtle. 3

4 Consequences of Logarithmic Dependance Dryness of dry air is important (matters whether rh is 5% or 10%) Fluctuations matter. OLR(q) > OLR( q ). Dry/moist variations allow more cooling than the uniformly mixed state 4

5 How Much of Greenhouse Effect is Due to Water Vapor? OLR computed with January climatology No Atmosphere CO2 alone (330 ppmv) CO2 & NCEP relative humidity CO2 & saturated troposphere OLR, W/m Latitude 5

6 Sensitivity of Climate to WV Greenhouse Effect GCM simulations, FOAM, mixed layer ocean Water vapor in radiation code only over-ridden with a specified profile Three simulations: Dry Atmosphere GCM q reduced by factor of 2, relative to control parameterization. Saturated below 300mb 6

7 Dry Atmosphere, year 5 ORO ocean (0), land (1), sea ice (2) flag Mean Max 2 Min

8 Dry Atmosphere, year 10 ORO ocean (0), land (1), sea ice (2) flag Mean Max 2 Min

9 Dry Atmosphere, year 10 TS1 Surface temperature (level 1) K Mean Max Min

10 WV reduced by half ORO ocean (0), land (1), sea ice (2) flag Mean Max 2 Min

11 WV reduced by half TS1 Surface temperature (level 1) K Mean Max Min

12 Saturated Atmosphere, year 10 TS1 Surface temperature (level 1) K Mean Max Min

13 The real atmosphere is very unsaturated 7

14 Meteosat Free Tropospheric Humidity 16 Jan, 1992 (Brogniez retrieval)

15 What do GCM s do? GCM s all agree on magnitude of WV feedback, and behave as if the RH were held roughly constant as the climate changes. Why does complex WV physics produce such simple and consistent behavior? Is this behavior right, or are the models all (consistently) wrong? Caution: Part of model agreement is due to compensation between variations in lapse-rate effect and water vapor content (B. McAveney) 8

16 What determines the atmosphere s water vapor content? The water vapor increases until there is a balance between the rate of removal and the rate of addition of water vapor. 9

17 Cold Warm

18 Dry air is manufactured by bringing it to a cold place, then returning it to a warmer one. The longer you go (since last resaturation) the colder are the places you have visited, and the drier you get. Prob. of coldest place visited after time τ vs. Prob of waiting time between resaturation events 10

19 The Maximum Negative Excursion Statistic T (t) be a 1D random process. Let P min (T 1, τ) be the probability that the minimum T visited during the time interval [0, τ] is T 1. For Brownian motion, P min (T 1, τ) exp[ (T 1 T (0)) 2 2στ ] for T 1 < T (0), zero otherwise. Identical to probability of final position, provided T 1 < T (0). Reflection Principle Remark: Statistics of back-trajectories are the same as statistics of forward trajectories. 11

20 Minimum statistic for bounded random walk 0-2 Minimum temperature probability, T Bounded between -10 and Temperature Time since saturation Probability 12

21 The Diffusion-Condensation equation t q = κ 2 q S(q q s ( r)) S(θ) = 0 for θ < 0. S(θ) increases rapidly with θ 13

22 Failure of eddy diffusivity: The cold trap example q(0) = q s (0) q s (x) q(x) Diffusive solution 14

23 vs. Stochastic Simulation Random walk 100,000 particles in 1D Saturate (set q=1) on collision with left hand boundary Remove all water at right hand boundary Reset supersaturated particles to local saturation (q=.5) at cold trap 15

24 1 Stochastic Simulation, particles, cold trap at x= <q> x 16

25 5000 N(1.) N(.5) N(0.) Stochastic simulation: Humidity probabilities x 17

26 A real trajectory from NCEP data 18

27 300 NCEP Trajectory, January 2003; Midlatitude mid-pacific Launch 1100 Temperature (K) 290 Pressure (mb) Temperature (K) Pressure (mb) Time (days) 400

28 Back Trajectory Reconstructions of Tropospheric Water Vapor Use NCEP 3D 4xdaily winds to compute back trajectories. Re-saturate only when trajectory encounters boundary layer Compute q sat along trajectory, and condense when supersaturated. 19

29

30 Midlatitude Moistening with Fixed Dynamics Repeat calculation with trajectories held fixed, but temperature increased (or decreased) everywhere by 1K Addresses question of thermodynamic control of WV, but uses the correct (lagrangian) temperature statistics. In a real climate change scenario: The statistics of the trajectories might change The vertical structure of temperature might change 20

31 Back-trajectory reconstructon 500mb Mixing Ratio Histogram, Dec Count P(logq,-1K) P(logq,control) P(logq,+1K) log (q) 10 21

32 Back-trajectory reconstructon 500mb Relative Humidity Histogram, Dec Count P(rh,-1K) P(rh,control) P(rh,+1K) Relative Humidity 22

33 Comparison with observation mb Relative Humidity Histogram, Dec 1994 ERA-40 Trajectory model Count Relative Humidity 23

34 Sources of Water Vapor Encounter with a (nearly) saturated boundary layer Moist air detrainment from convective towers (Really a form of advection, but too small scale to be explicitly resolved). Evaporation of precipitation falling through dry air. A major means of moistening the vicinity of convective towers. 24

35 Dr y Air

36 The Big Questions How much subgrid mixing is there between moist and dry air? (esp. vertical) How moist is the air near tropical convective systems? In the ascending (condensing) branch of midlatitude synoptic eddies? How far does moistening extend from a region of active condensation? How sensitive are the preceding to microphysical assumptions and convective parameterizations, and how well do models do at reproducing the correct behavior? (cf. Emanuel and Z-R in 1D). How can we diagnose and understand the combined cloud/wv effect on OLR in regions of active condensation? The problem of cloud fraction behavior! How can we formulate simplified models of the humidity, for use in idealized climate models (i.e. next step beyond EBM)? What is the effect of climate change on stratospheric water vapor? How strong are feedbacks due to stratospheric clouds? 25

37 Longwave flux from ScaRaB-2 and METEOSAT-5 ScaRaB-2 LW channel (4-100 m) METEOSAT IR channel ( m) (from Duvel et al., 2000) Objective Convert narrow band radiances (Wm -2 sr -1 ) into long wave flux (Wm -2 ) -> 2 steps : angular and spectral integrations METEOSAT WV channel ( m) Statistical relationship between Meteosat radiances (IR and WV) and ScaRaB LW flux

38 Conclusions The zero order effect of warming is indeed to keep relative humidity constant. Deviations from fixed rh arise from changes in trajectories or changes in vertical structure of temperature. Are these really first order effects? Eddy diffusion is not a suitable way to represent water vapor in simplified models. 26

39 Prospects in situ observations of WV isotopes (e.g. HDO) will soon tell us a great deal about the pathways leading water to the atmosphere. How to make the best use of such data? Microwave satellite data (SSM/T2) will provide direct information on water vapor in very cloudy regions. Post-ERBE radiation budget satellites will reinvigorate the study of allsky OLR in predominantly cloudy regions. The Southern Hemisphere response of the Earth during the LGM can be used to provide a crucial check on our understanding of cloud/wv feedbacks and sensitivity of climate to changes in CO 2 27