Simulateur mission end-to-end pour laltimétrie

Document Sample
Simulateur mission end-to-end pour laltimétrie Powered By Docstoc
					Development of an “end-to-end”
  altimeter mission simulator

              Alix Lombard - Juliette Lambin (CNES)
     Laurent Roblou – Julien Lamouroux (NOVELTIS)

Ø Debates        on future altimetry constellation design
  § need for continuity and complementarity between missions
  § variety of applications (climate, meso-scale, operational,…) but all need multi-mission
  § orbit : sun-synchronous or not, cycle/repetitivity, existing tracks or not, …
  § payload : bi-frequency or not, radiometer or not, platform stability (roll for wide-swath altimeter), …
  § data : sampling, latency/availability, …

Ø Need for a decision-making tool :”End-to-end”
mission simulator (R&D CNES funding)
  § objective : examine the merits of various observing configurations /
  discriminate among them
  § need for a simple, flexible, evolutive tool
                        Status : end-to-end altimeter mission simulator
                                          for storm surge observations
                          obs. systems        • Possibility of studying multi-missions
                                              altimetry configurations, easy tuning of
                                              orbit configurations parameters

                             Sampler                                            Altimetry
Storm surges                                         Analyzer
                              (pseudo-                                        configuration
   (model)                                          (assimilation)
                            observations)                                     performances

 Ø Framework of Observing-Systems Simulation Experiments (OSSEs, Arnold and
 Dey, 1986)   :   designed to evaluate the impact of observing system data in numerical analysis.

 Ø “Ensemble Twin Experiments” method (Mourre et al., 2004) :
     Ø pseudo-observations generated from a “control” simulation (oceanic model)
     Ø then assimilated in a “free” simulation
 The performance of the system is estimated in terms of model error (=ensemble
 variance) reduction performed via a data assimilation system.
Ø Model configuration : MOG-2D / T-UGO 2D (F. Lyard)
     § barotropic, non linear, finite element
     § zone : well known / studied and representative / varied
     (open ocean, shelf and coastal seas)
     § time period : 15 days, typical / varied winter storm surges
      conditions (16/11 to 01/12/1999)
     § atmospheric forcing: surf. pressure / 10m-wind (ARPEGE)
     § tidal forcing
                                        Ø Generation of pseudo-observations
                                             § Altimetry configuration set up by user (specify orbit
                                             § pseudo-obs. (Sea Level Anomaly) extracted from the
                                             model reference simulation (non-perturbed run), at the
                                             space-time altimetry positions
        Nadir               Wide swath       § then noise-added (gaussian noise of 0-mean and
                                             standard deviation specified by instrument noise level)

Ø Model errors computation (prior requirement for data assimilation)                   20/11
                                                                                                 11 cm²
    § estimated from a 100 Ensemble simulations of the model in response
    to atmospheric forcing errors (surf. pressure and 10m-wind perturbed)
    [Lamouroux, 2006]
    § error statistics thus estimated by the ensemble variance of the model
    at each analysis time step (daily) – errors variable in time and space                        0 cm²
Ø Data assimilation / Performance analysis method
   § s-EnROOI (simplified Ensemble Reduced Order Optimal Interpolation) configuration
   § “simplified” : no sequential control of the model (ensemble error reduction only estimated at
   analysis time, not propagated in time via the model) → quick execution / results obtained
   § Possibility to implement EnROOI, ROEnKF, EnKF (higher performance but longer
   computational time)… but idea to keep a simple / quick decision-making tool to
   discriminate between various observing scenario
   § SEQUOIA + MANTA codes used (De Mey, 2005)

                                                               Perturbed simulations

                                                           Model reference simulation

                                                Ensemble variance reduction estimation
                                                      at each analysis time step

Ø “Ideal” observing system
     § regularly spaced grid
     § pseudo-obs / analysis daily

q Results for Ta = 20/11/1999
(analysis time representative of model errors over the whole period)   q Time-averaged result

                          11 cm²                              11 cm²                               100 %



                           0 cm²                               0 cm²

     Ensemble variance                   Ensemble variance                  % of ensemble
        of the model                      after pseudo-obs.               variance reduction
      (before correction)                    assimilation                   over the period
                                                                         è Over the whole period
       è Strong and uniform reduction of variance,                       (synthetic gain ~ 78%),
       especially in the English Channel (gain Ta ~ 94%)                 methodology validated
                      Performance of various altimetry configuration
Ø Various performance diagnostics
   § at each analysis time step, mapped
   § synthetic over the period, space averaged …
                                    SWOT on a
                                    JASON orbit
                                                      è Efficient tool to estimate the
                                                      performances of various
                                                      altimetry configuration and to
                                                      discriminate among them.

                                                      è Allow to design orbit and
                                                      assess performances of multi-
                                                      satellite altimetry systems

                                                    Reduction of ensemble variance
                                                     time-averaged over the period
                                                   NB: the higher the percentage of variance
                                                   reduction, the more the altimeter mission will
                                                   provide helpful information to storm surges
                                                     Lamouroux et al, OSTST meeting, Hobart, 2007
                         Evolution : end-to-end altimeter mission
                  simulator for the study of tide aliasing question

Ø Context of possible sun-synchronous orbit configurations
(SWOT, Jason-3, Sentinel-3, …) → tide aliasing problem

                      obs. systems                                 Tides aliasing
      Tides                                                         diagnostics

  Storm surges          Sampler           Analyzer                 configuration

                                                 Existing module
                        Ø Extension work in progress
                             Evolution : end-to-end altimeter mission
                      simulator for the study of tide aliasing question
Ø Same methodology but some evolutions needed
→ some work done, some in progress

Ø Ocean tide model configuration : T-UGO 2D
    § 28 ocean tide components, model validated through
    comparisons with FES2004 / GOT00b
    § larger zone (long wave dynamics of ocean tides)
    § 1-year simulation
    § model dissipation parameters : topography, bottom friction coefficient, transfer coefficient
    towards barocline modes
Ø Generation of pseudo-observations :
    § ocean tide model reference simulation (non-perturbed run) → high frequency (HF)
    § lower frequency (LF) ocean circulation simulation (daily reanalysis from PSY2V2 global
    ocean model computed by MERCATOR-Ocean)
    § pseudo-obs. extracted from the sum of both simulations (ocean tide HF + ocean
    circulation LF), at the space-time altimetry positions
    → take into account the coupling HF aliased by altimetry sampling at LF / LF circulation
Ø Model errors computation :
    § estimated from ensemble simulations of the model in response to perturbed model
    dissipation parameters
    § work in progress
                                                     Conclusions and perspectives

Ø Work in progress for tidal analysis (end of R&D funding + SWOT PASO study)
    § ensemble model error statistics (ensemble variance) computation
    § implementation of specific tide aliasing diagnostics
    § more realistic observation errors to be defined (especially for wide-swath altimeter)
    § case studies (inferred from PASO SWOT instrument study)

Ø First prototype of the simulator (storm-surge model)
    § efficient tool to estimate the performances of various altimetry configuration and
    discriminate among them
    § simple, highly flexible and evolutive, first version of a powerful tool for designing orbit for multi
    -satellite altimetry systems (Jason-3, SWOT, Sentinel-3 …)

Ø Work plan / Perspectives
    § further tests of altimetry configurations : case studies, ≠ realistic mission scenario tests
    § implement other oceanic processes : ocean circulation, waves, …
    § implement more complex data assimilation scheme : e.g. for refined studies

Shared By: