Speaker: Marcus van Lier-Walqui
NASA/GISS & CCSR, Columbia University 

Weather and climate models have well-known biases in their representation of physical processes. A prime offender is cloud microphysics, owing to the complexity of hydrometeor interactions as well as the approximations that underpin bulk parameterizations. To some extent, models can be improved by finding optimal values for tunable model parameters, and estimating the uncertainty in these parameters — “parametric” uncertainty. Radar observations, including polarimetric radars and radar Doppler spectra, have shown much promise in providing information related to microphysical processes and can thus be leveraged via, e.g., Bayesian estimation, to probabilistically constrain model parameters. A deeper problem is that structural assumptions are typically hard-coded into parameterization schemes, and thus cannot be systematically improved in the same manner, nor can uncertainty associated with these choices be quantified. This fundamental shortcoming of traditional parameterizations motivates the use of multi-physics ensembles in probabilistic weather forecasts — these are, in essence, attempts at spanning both parametric and structural uncertainties in physical parameterizations, but they typically cannot span these uncertainties smoothly or probabilistically. I will present work on a new microphysics scheme, the Bayesian Observationally-constrained Statistical-physical Scheme, or BOSS,  and describe how it was developed specifically to facilitate characterization of parametric and structural uncertainties in a Bayesian framework. An additional benefit of BOSS is that it is “smooth” and therefore amenable to adjoint methods. I will also present work on applications of Bayesian parameter estimation to ice microphysics, cloud property retrievals, and climate model tuning.

Refreshments: 3:15 PM

Charles Knight
NCAR/MMM  

“The box” represents classical nucleation theory, CNT, a conceptually simple and at first appealing mechanism in which the interfacial energy between an initial, unstable phase (liquid water, here) and a stable one (ice) constitutes an energy barrier against the stable one’s first appearance.  Ice nucleation obviously involves crystal growth, but the theory of CNT in general has been based upon thermodynamics and chemical reaction theory, independent of crystal structure.  Outside the box here is treating crystal growth and nucleation in terms of growth of the known hydrogen-bond network of ice: the ice crystal structure and its tetrahedral bonding.  The initial context here was trying to explain the observed correlations between ice crystal growth in liquid water and the ice crystal structure, part of which appeared to involve two-dimensional nucleation of new molecular layers at an interface between ice and liquid water.  This explanation turned out not to work well whereas a simple model of growth of the bonding network does seem to provide conceptual understanding.  A bonding-network approach to homogeneous nucleation is unwieldy but interesting, and from that point of view, CNT (for nucleating ice) seems dubious.  The actual mechanism may be dominated by structural effects, not interfacial energy.

Refreshments: 3:15 PM

Andy Wood
Martyn Clark
NCAR/RAL/HAP 

Ensemble hydrologic (streamflow) prediction provides critical inputs for water, energy and hazard management, particularly in the face of extremes such as floods and droughts.  Following steady advances in operational ensemble numerical weather prediction since the 1990s, US and international operational prediction groups have invested heavily in developing datasets, methods, and models to enable a seamless suite of probabilistic hydrologic predictions spanning timescales from hours to seasons.  Ensemble hydrologic forecasting systems are now operational in a number of countries (including the US), and are enhanced by an increasingly crowded field of operational continental and global ensemble hydrologic prediction services.  In this presentation, we provide background describing the evolution of ensemble hydrologic prediction systems, and highlight the role of the HEPEX (Hydrologic Ensemble Prediction Experiment; www.hepex.org) initiative since 2004 in defining and promoting an integrative, scientific view of the elements of a hydrologic ensemble prediction approach.  These include methods for the probabilistic downscaling and calibration of meteorological forecast ensembles, hydrologic model parameter estimation and uncertainty quantification, hydrologic model data assimilation, model output post-processing, and ensemble forecast verification and communication for use in risk-based decision-making.  We summarize the current state of practice in applying these methods to achieve reliable ensemble streamflow forecasts (locally and globally), and discuss long-standing and new challenges identified by the ensemble hydrologic prediction community.

Refreshments:  3:15 PM  

Shane Keating
University of New South Wales
Australia  

The era of earth-observing satellites has revolutionised our understanding of our planet and the dynamical processes that shape it. In many real-world geophysical systems, however, estimates of turbulent mixing and transport are limited by the resolution of available observations. In this talk, I will describe a suite of stochastic filtering strategies for estimating mixing in turbulent
geophysical flows from “superresolved” satellite imagery obtained by combining coarse observations with an efficient stochastic parameterization for the unresolved scales.

The method enhances the effective resolution of satellite observations by exploiting the effect of spatial aliasing and generates an optimal estimate of small scales using standard Bayesian inference. The technique is tested in quasigeostrophic simulations driven by realistic climatological shear and stratification profiles. Two applications are considered: calculating poleward ocean eddy heat flux from satellite altimetry, and estimating the three-dimensional upper ocean velocity field from superresolved sea-surface temperature imagery. In each case, the superresolved satellite observations result in a considerable improvement in estimates of turbulent fluxes compared with the raw observations.

Refreshments: 3:15 PM