Download presentation
Presentation is loading. Please wait.
Published byKerrie Powers Modified over 9 years ago
1
Soil moisture generation at ECMWF Gisela Seuffert and Pedro Viterbo European Centre for Medium Range Weather Forecasts ELDAS Interim Data Co-ordination Meeting 17./18.09.2003
2
Plans ( ELDAS 1 st progress meeting) Assimilation aspects: Minimize the combined errors in prediction of soil moisture, latent heat flux and screen level observations Further m w-Tb assimilation experiments (viewing angle, times) Assimilation of heating rates Technical aspects: Paper(s) focusing on the - new features of assimilation method - assimilation of m w-Tb - (assimilation of heating rates) Summer 2003: Build production system for the annual data base End of 2003: Start production Action: no further development Action: in production Action: SCM test runs Action: 2 papers: -published in GRL (T,RH,Tb) -Cond. accepted at JHM (OI, EKF) Action: still pending
3
Soil moisture analysis systems Optimal Interpolation: Used in the operational ECMWF- forecast since 1999 (Douville et al., 2000) Fixed statistically derived forecast errors Criteria for the applicability of the method - atmospheric and soil exceptions - corrections when T and RH error are negatively correlated Extended Kalman Filter: Used in the operational DWD- forecast since 2000 (Hess, 2001) * Updated forecast errors Criteria for the applicability of the method - no ‘direct’ atmospheric exceptions - soil exceptions still to be tested * Changes: - Assimilation of 2m- T and RH, m w-Tb -Model forecast operator accounts for water transfer between soil layers -Test adaptive EKF
4
Experiment Design Atm. initial conditions + dynamics forcing from ECMWF reanalysis (ERA40) Single-column model of the ECMWF NWP model + microwave emissivity model First guess: T 2m,RH 2m,HR(?) Soil moisture analysis scheme OI or Extended Kalman Filter Soil moisture Background error Increments (daily) Observations: T 2m,RH 2m,HR Observation of precipitation + radiation
5
Production system for soil moisture Starting point: Experiments based on Single Column version of the ECMWF’s NWP model (SCM) Requirements: 1.0.2 x 0.2 regular lat/lon grid for Europe (15W-38E, 35N-72N) 2.Computer time (cost efficiency) 3.Annual database for 1.10.1999 – 31.12.2000 control system Solutions: Add 1:run n x n SCMs over Europe (each SCM runs independently) Add 2:- run SCMs only for land points (about 25 000 SCMs) - I/O consideration - High degree of parallelisation in an easy way balance saving of computer time and time for programming Add 3:Supervisor Monitor Scheduler (SMS)
6
Production system for soil moisture(2) Progress of work: Changes to the SCM source code –SCM structure has been changed to run n x n SCMs in one run (single point area) –I/O netcdf I/O grib –OpenMP parallelization (up to 8 processes on one thread) Changes to the soil moisture analysis (SMA) –SMA has been changed to run n x n points in one run –I/O netcdf I/O grib Forcing data –Composition of forcing data changed from one point to n x n points –O netcdf O grib Control Structure –First SMS layout
7
1) Soil moisture analysis 1)Get forcing data from Mars archive 2)Prepare data for SCM INPUT 1) Background run 1)Get forcing data from Mars archive 2)Prepare data for SCM INPUT 1)Soil moisture perturbation 1)Final (soil moisture) trajectory 2)Check success of SMA (Costfunctions) 1)Forecast run 1)Final (soil temperature) trajectory 2)Check success of STA (costfunctions ) 1)Soil temperature analysis 1)Soil temperature perturbation
9
Production system for soil moisture (3) What is still missing? –Interpolation from gaussian grid to reg. 0.2 x 0.2 lat/lon grid –Incorporation of ELDAS maps (e.g. land cover) –Incorporation of ELDAS forcing data (precipitation, radiation) –Archiving of output in MARS –Observation (Re-analysis) data of 2mT and 2mRH for SMA +STA –Post-processing routines for parameters especially asked for by ELDAS validation –ECMWF orography problems (LW) Final tests
10
Time schedule(1) Estimated Production Time: Analysis for one day: - one SCM run for 1000 pixels needs 5 min on 8 nodes ~ 2 hours for 25000 pixels - 5 x SCMs are needed 10 hours for 25000 pixels approx. 5-6 months for annual database further parallelization needed (splitting Europe into boxes) (MPI, distributed memory)
11
Time schedule(2) Under normal circumstances: 6 weeks required to include missing bits and pieces 2 weeks final tests Start production by November/December Expected Start of production: ?
12
Assimilating SHR, T+RH, T+RH+SHR Soil moisture Days when SHR is available (50% data missing, 25% cloudy) Variable SHR observation error depends on cloud fraction flag (how many hours are cloud free): cloud fraction flag of neighbouring pixels cloud fraction flag of pixel
13
Extended Kalman Filter Time t+24h t0t0 t+9h t+12h t+15h Minimization 3 perturbed forecasts for each state variable Forecast (first guess) Analysed forecast for new soil moisture at t+24h Comparison with observations T 2m,RH 2m,Tb Simulated T 2m,RH 2m,Tb Opt. Soil moisture
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.