Parametrizing cloud and precipitation in today s NWP and climate models. Richard Forbes

Transcription

1 Parametrizing cloud and precipitation in today s NWP and climate models Richard Forbes (ECMWF) with thanks to Peter Bechtold and Martin Köhler RMetS National Meeting on Clouds and Precipitation, 16 Nov 2011

2 Luke Howard In December 1802, a pharmacist called Luke Howard presented his paper, "On the modification of clouds" ('modification' meaning 'classification ), published in 1803 and still very much used today. Howard optimistically predicted that by his classification, meteorology would be rescued from empirical mysteriousness, and the reproach of perpetual p uncertainty But all cloud formation and dissipation processes follow the same fundamental But all cloud formation and dissipation processes follow the same fundamental laws of physics and it is this that we strive to represent in models.

3 Talk Outline Cloud and precipitation in NWP and climate models: 1. Representing clouds in GCMs 2. Physical principles and cloud parametrization 3. Approximating complexity

4 1. Representing Clouds in GCMs Microphysics - ECMWF Seminar on Parametrization 1-4 Sep

5 Why do we need clouds in GCMs? Hydrological (precipitation) Radiative Diabatic (shortwave, longwave) (condensation, evaporation)

6 Microphysics Parametrization: The category view Water vapour Condensation Evaporation Deposition Sublimation Con ndensation Ev aporation Cloud water Freezing Melting - Bergeron Cloud ice Deposition Sublimat tion Autoconversion Collection Collection Autoconversion Collection Rain Freezing - Melting Snow Surface Precipitation

7 Microphysics Parametrization: The hydrological perspectivep Water vapour Condensation Evaporation Deposition Sublimation Con ndensation Ev aporation Cloud water Freezing Melting - Bergeron Cloud ice Deposition Sublimat tion Autoconversion Collection Collection Autoconversion Collection Rain Freezing - Melting Snow Surface Precipitation

8 Microphysics Parametrization: The radiative perspective p Water vapour Condensation Evaporation Deposition Sublimation Con ndensation Ev aporation Cloud water Freezing Melting - Bergeron Cloud ice Deposition Sublimat tion Autoconversion Collection Collection Autoconversion Collection Rain Freezing - Melting Snow Surface Precipitation

9 Microphysics Parametrization: The diabatic process perspective p Water vapour Condensation Deposition Con densatio on Eva aporation Cloud water Autoconversion Collection Evaporation Freezing Melting Collection Sublimation Cloud ice Autoconversion Collection S ublimati on Depositio on Rain Freezing Melting Snow Surface Precipitation

10 Microphysics Parametrization: The diabatic process perspective p Water vapour Condensation Evaporation Deposition Sublimation Water Phase Freezing Melting Ice Phase Surface Precipitation

11 Why do we need clouds? The cloud and precipitation parametrization is there to be able to predict the correct impacts on hydrology, radiation and dynamics for NWP and climate.

12 2. Physical Principles and Cloud Parametrization Microphysics - ECMWF Seminar on Parametrization 1-4 Sep

13 From global to micro-scales Hugely complex system. Need to simplify! emu.arsusda.gov

14 Building a cloud parametrization scheme Microphysics Discrete representation of hydrometeors (categories), warm- phase/cold phase, cloud/precipitation, split ice/snow, graupel, hail? Prognostic or diagnostic (dependent on timescales involved)? Number of moments for each category: mass only (single moment), +number concentration (double moment), or size resolved (binned)? Microphysical i processes, what complexity?

15 Cloud droplet nucleation Key concepts for water cloud nucleation: 1. Equilibrium saturation vapour pressure for bulk water as a function of temperature: Clausius-Clapeyron. 2. Drop nucleation cloud droplets form when air just reaches saturation for bulk water due to presence of abundant cloud condensation nuclei. Aitken, Wilson, Thomson (Kelvin), Raoult, Köhler. Aitken without dust particles in the atmosphere there will be no haze, no fog, no clouds and therefore probably no rain. Cloud scheme approximations: 1. The above means that GCM cloud schemes can generate clouds/precipitation by removing all supersaturation with respect to water. 2. In fact, early cloud schemes with one prognostic variable (water vapour) didn t explicitly represent clouds but generated precipitation by removing all supersaturation in the atmosphere.

16 Precipitation formation Key concepts for precipitation formation: 1. Collision-coalescence of a population of particles of differing sizes leads to rapid growth of larger particles. Complex system with uncertainties in particle size distributions and collision efficiencies. Cloud scheme approximations: 1. Precipitation generation in many GCMs is a simple function of cloud water content (single moment scheme, e.g. Kessler 1969). Termed autoconversion. 2. Next level of complexity is to include the number concentration as well as the cloud water content (double moment schemes). Con nversion rate Cloud water content 3. More complex schemes also represent the spectral dispersion of the cloud droplet size distribution (bin schemes).

17 Cloud/precipitation particle diffusional growth and evaporation Key concepts for diffusional growth/evaporation: 1. Diffusion of water vapour to/from a drop or ice particle: Maxwell (1890) (electrostatics analogy), e.g. for ice particle, rate of change of mass: m t = 4πsCF L L RT s RT 1 s k a T + χe si Cloud scheme approximations 1. In many GCMs, the above equation is integrated over the (uncertain) particle size spectrum (which may be simple function of ice content). Additional uncertainties in the particle shape (C) and ventilation (F). 2. In more complex schemes, the particle size spectrum may be predicted (double moment, bin).

18 Complexity versus simplicity 1. Hugely complex system with many uncertainties (aerosols, nucleation, ice particle shapes, particle size distributions, turbulent forcing, spatial and temporal heterogeneity.) needs to be greatly simplified. 2. Some basic physical principles described in classic papers allow a first approximation. 3. Can we get improved simulations with increasing complexity? e.g. single moment > double moment > binned 4. Perhaps surprising how far we can get with simple schemes! (but always room for improvement )

19 Example: Satellite Infra-red Image South-east Pacific Satellite infra-red and simulated IR image sequence from the ECMWF global model for the south-east Pacific off the coast of Chile during the VOCALS observational campaign 10/2008

20 Example: CloudSat Radar Reflectivity Cross-section through a mid-latitude front

21 Physical principles and cloud parametrization Cloud and precipitation parametrizations are physically-based y approximations of the effects of numerous complex microscale processes.

22 3. Approximating Complexity Microphysics - ECMWF Seminar on Parametrization 1-4 Sep

23 Approximating Complexity The real world is highly heterogeneous The graph represents variability in any meteorological quantity, but is particularly relevant for cloud and precipitation. For example: 1. Ice water content 2. Precipitation rate 3. Optical depth of liquid water cloud uantity Q 4. Vertical profile of radiative heating due to ice cloud Space or Time

24 Approximating Complexity Improving the model fit to observations 1. A simple parametrization may give a reasonable first approximation. 2. A parametrization of intermediate complexity may improve the fit to observations. uantity Q 3. A complex parametrization may ygive the best approximation to the data. However. Space or Time

25 Approximating Complexity Getting the right answer for the wrong reasons 1. Aspects of NWP/climate models are often tuned to agree with observations, e.g. top of atmosphere radiation. 2. But can get the right answer for the wrong reasons compensating errors. 3. For example, cloud liquid water content and effective radius could both be wrong but give the correct radiative impact of cloud on average. 4. But compensating errors will limit further improvements. 5. We need to constrain with observations as many aspects of cloud properties as possible. uantity Q Incorrect R eff, correct LWP Incorrect R eff,incorrect LWP Correct R eff, incorrect LWP Space or Time

26 Approximating Complexity A complex model is not necessarily the best model 1. Complexity usually brings more degrees of freedom needed to capture the true variability, but may also bring more opportunities to go wrong! 2. In some cases a simple parametrization can be as good as or a better approximation than a complex parametrization. 3. In other cases a more complex parametrization may seem better at first sight, but a simple parametrization i may be just as good if modified. uantity Q 4. Do we know the observations well enough to constrain the complex model parametrization? Space or Time

27 Approximating Complexity Improving the model fit to observations 1. As we increase model complexity to improve our fit to the real world, we need to ensure the more complex model is properly constrained and gives better results for the right reasons. 2. This is a particular challenge for global models that have to represent ese all the different e meteorological regimes. uantity Q 3. Also, for NWP and climate the choice of complexity is often a compromise with other constraints such as computational expense. Space or Time

28 Summary Microphysics - ECMWF Seminar on Parametrization 1-4 Sep

29 Summary 1. Cloud and precipitation parametrizations for NWP and climate are there to be able to predict the correct impacts on the hydrological cycle, radiation and dynamics and we need to understand these impacts. 2. Cloud and precipitation parametrizations are physically-based approximations of fthe effects of numerous complex microscale processes and we should continue to strive to improve them. 3. Cloud and precipitation parametrizations need to be of appropriate complexity which can be constrained with observations (a challenge for global models!) and we therefore continue to need even more comprehensive cloud and precipitation related observational data.

30 Some key challenges for cloud and precipitation parametrization Improving the microphysics, i from di drizzle to mixed-phase cloud to ice. Interaction of aerosols and microphysics (particularly ice nucleation), climate interest in aerosol feedbacks. Consistent representation across scales, sub-grid variability, wide range of model resolutions, small scale dynamics drives the microphysics. Understanding impacts of microphysical parametrization on hydrological cycle, radiation and dynamics. Data assimilation of remote sensing observations, dependent on microphysics for forward modelling of the observations.

31 Questions? Microphysics - ECMWF Seminar on Parametrization 1-4 Sep