Jan Mandel/Blog/2011 Apr May

From CCM
Jump to: navigation, search

To the current blog page and the archive

<previous next>


April 4: Data assimilation seminar: Probability measures 1: Convergence in distribution

April 6: WRF-Fire released with WRF 3.3

April 9: GACR proposal

April 11: Data assimilation seminar: Probability measures 2: Proof of portmanteau theorem

April 18: Data assimilation seminar: Probability measures 3: Proof of Slutsky's theorem

  • This is the third part of the series of convergence theorems in probability to use in ongoing work on data assimilation in infinite dimensional spaces. This theory is generally presented in finite dimension in the literature. We are going over the proofs in detail and seek suitable variants to make sure that they still hold.
  • photos

April 20: Fireflux paper

April 25: Data assimilation seminar: Probability measures 4: Uniform integrability

Vitalli convergence theorem assumes only uniform integrability, which is weaker than the inequality assumed in the dominated convergence theorem. We show some interesting relations between uniform integrability, convergence in probability, and in Lp. This is the fourth part of the series of convergence theorems in probability to use in ongoing work on data assimilation in infinite dimensional spaces. This theory is generally presented in finite dimension in the literature. We are going over the proofs in detail and seek suitable variants to make sure that they still hold for random elements with values in Banach spaces.

April 25: Fuqing Zhang: Inter-Comparison and Coupling of EnKF with 4DVAR

April 26: WRF For Hurricanes Tutorial

Rob Rogers: Observations of hurricanes to improve numerical models

  • model evaluation, data assimilation, hypothesis testing (not in statistical sense?)
  • observations need to be in compatible format with models
  • airborne: expendables, remote sensors
  • transmitted from aircraft - doppler real time SOs trenamitted during P-3 mission and assimilated into HWRFx using HEDAS
  • impact of inner core observations: doppler data improves intensity error up to 72 hours but not later, frequency of superior forecast: better after 84 hours
  • contact

Fuqing Zhang: Assimilation of fine-scale hurricane observations

  • EnKF Meng and Zhang 2008a,b; data assimilated WSR88D at KCRP, KHGX,, KLCH radar radial velocity; superobservations, WRF-EnKF 3 domain, 40.5-4.5 km, 60 member ensemble
  • Katrina: WRF ARW 3.1 pertubations by WRFDA; EnKF_DF closely followed the best track except 6h delay; after the 2nd cycle EnKF analyses reproduce the observed with structure quite well; the differences are small with more than 60 members (tried 200, 300); most of the errors come from wave numbers 0, 1, maybe 2, 60 members capture that well; for the wavenumber 1-2 example the magnitudes of covariance are similar but in different locations; skill decay fig (how defines skill?)
  • Penn state WRF-EnKF realtime system
  • pseudo-ensemble hybrid data assimilation system (PEDA) for TC initialization with airborne doppler radar data
  • figure: max windspeed error remains about constant, WVAR,PEDA best
  • figure: abs error in position statrts at cca 20 -> 300km at 72 hours, keeps increasing NoDA best, WVAR PEDA worst
  • diagram: GFS analysis, very good; 60 ensemble pertiubations generated from B or WRF-Var; replace TC vortex; pseudo ensemble members; flow-dependent inner core; 3DVAR
  • TC vortex libary: TC vortices at different output times are binned according to vma
  • fig: errors after bias correction (NHC variable interpolator)
  • need 16000 cores for 1.5km ensemble in 6 hours (TACC Ranger)

Chris Snyder: Mesoscale Data Assimilation for Hurricanes

  • observations - relevant for environment (jm: additive), relevant for vortex (jm: positional) - sat images and recon flights, cloud-track winds, special dropsondes, doppler radar
  • observations are limited an intermittent, do not resolve all aspects of vortex structure or evolution
  • bogussing - assumed vortex structure; vortex removal and reloacation - (jm: much more than just deformation): extract from model forecasest thhourh spatial filter GFDL technique, spin up symmetric vortex, estimate asymmetry, insert the vortex
  • data ssimilation (DA): simplest: strong assumptions about prior covariance - independent in time, depends on distance only, no dependence on state. Sophisticated: relax assumptions, incorporate info
  • differences between EnKF and 4DVAR apparent when observations are incomplete
  • simple example vortex=maximum, single observation of vortex. 3d var puts in a bump (corrupts vortex structure), EnKF shifts the vortex coherently (jm: only if there is a suitable member, that's why they need so many), prerseves structure
  • analysis from WRF/DART: 96 members GFS 6hour forecast+spatially correlated noise for lateral boundary conditions; run continuously for 4 months, just using data, no artificial intervention (for position); crucial to include flow dependent covariance via model dynamics (fig: moved vortex position, two color blotches with wind difference caused by the move)

April 26: Revisiting the EnKF theory paper

April 27: WRF-Fire

  • photos: log profile, data assimilation, branding

April 30: Reply to referee 1 for the GMD paper

May 1: Whitepaper on wildland fire modeling

May 2: CCM Colloquium: Volodymyr Kondratenko

Ignition from a Fire Perimeter in WRF Wildland Fire Model

  • Abstract: The current WRF-Fire model starts the fire from a given ignition point at a given time. We want to start the model from a given fire perimeter at a given time instead. However the fuel balance and the state of the atmosphere depend on the history of the fire. The purpose of this work is to create an approximate artificial history of the fire based on the given fire perimeter and time and an approximate ignition point and time. Replaying the fire history then establishes a reasonable fuel balance and outputs heat fluxes into the atmospheric model that allow the atmospheric circulation to develop. Then the coupled atmosphere-fire model takes over. In this preliminary investigation, the ignition times in the fire area is calculated based on the distance from the ignition point to the perimeter, assuming that the perimeter is convex or star-shaped. Simulation results for an ideal example show that the fire can continue in a natural way from the perimeter. Possible extensions include algorithms for more general perimeters and running the fire model backwards in time from the perimeter to create a more realistic history.
  • Uses the fire replay feature in WRF-Fire, implemented earlier

May 2: Data assimilation seminar

Convergence of the EnKF revisited

EnKF Matlab classes for hurricane test

May 3: Revised EnKF theory paper submitted

May 4: Revising Harmanli fire paper

May 9:Data assimilation seminar: Bryan Smith: Statistics on manifolds 2

May 9: Wildfires project coordination meeting at the SCI Institute

May 10: Wildfires meeting at University of Utah Meteorology

May 11: Convergence study of the level set method

May 12: NASA ROSES 2011

May 13: WRF-Fire

May 14: Abstracts for 9th Symposium on Fire and Forest Meteorology, AMS

May 17: Data assimilation seminar: Osimorph Matlab classes

May 30: at CAS

WRF workshop perimeter ignition paper

Kalman filter in Hilbert space


May 31: at CAS

See also

Personal tools