Page 1© Crown copyright 2004 Data Assimilation at the Met Office Hadley Centre, Met Office, Exeter.CTCD Workshop. 8 th Nov, 2005 Chris Jones
Page 2© Crown copyright 2004 Outline Intro 2 interpretations of DA Current DA at the Met Office Plans/hopes for future
Page 3© Crown copyright 2004 Met Office Data Assimilation The Met Office has a long history of data assimilation in an operational (NWP) framework Currently running operational 4D-Var scheme. DA now appealing to a much wider audience Ocean forecasting Seasonal/decadal forecasting Carbon cycle/ecosystem research Keen to make most of existing expertise internal and external to Met Office Central part of CarboEurope Not my area of expertise, despite a chequered past…
Page 4© Crown copyright 2004 What does DA mean? 2 different interpretations of the phrase “data assimilation” i. Conventional, “NWP” style: Model formulation is fixed Uses “current” or “new” data Constrains model prognostic variables Product is Analysis – best estimate of snapshot of reality Initialsed state from which to produce best forecast ii. Parameter optimisation: A form of what used to be called “tuning” (which was highly subjective) Generally uses “historical” data or climatology Multiple model runs to constrain internal parameters Product is the model itself (optimised)
Page 5© Crown copyright 2004 Data Assimilation applications at the Met Office Climate/Carbon cycle related data assimilation does already exist at the Met Office. 1. “DePreSys” (Decadal Prediction System) Run climate model, nudging sea surface temperature and salinity to observations Assimilate atmospheric variables too (better first season, better surface fluxes into ocean) Run climate model for next 10 years as a forecast Hindcasts show some skill relative to persistence
Page 6© Crown copyright 2004 Data Assimilation applications at the Met Office Climate/Carbon cycle related data assimilation does already exist at the Met Office. 2. NCOF/CASIX ocean biogeochemistry modelling Run ocean carbon cycle model (HadOCC) in operational (“FOAM”) hi-res ocean model. Assimilate physical variables (in-situ + satellite) Drive with atmos fluxes from NWP model produces realistic ocean state Allows simulation of ocean pCO2, and air-sea CO2 flux Aim to also assimilate ocean colour obs from satellite (as a proxy for chlorophyll concentration) Better constraint on biological variables
Page 7© Crown copyright 2004 pCO2 in North Atlantic Climate model (no DA) FOAM (HadOCC + physical DA) obs Better ocean simulation (through DA) improves C-cycle simulation
Page 8© Crown copyright 2004 Attribution of mechanisms Given confidence in simulation we can learn from the mechanisms in the model Spring draw-down of pCO2 is biologically driven. Rest of year is physically driven (mainly response to SSTs)
Page 9© Crown copyright 2004 Data Assimilation applications at the Met Office Climate/Carbon cycle related data assimilation does already exist at the Met Office. 3. CAMELS (Carbon Assimilation and Modelling of the European Land Surface) Optimisation of terrestrial carbon cycle models using observed carbon cycle data Results promising to date Will feed into integration component of CarboEurope
Page 10© Crown copyright 2004 Prior and optimised diurnal cycle at Bray in Orchidee: means over 1997 Growing Season (days ) Black: data with uncertainties Green: prior Red: optimized
Page 11© Crown copyright 2004 BETHY simulated carbon flux
Page 12© Crown copyright 2004 A success… and a caveat TRIFFID modelled NEP saturates very quickly with light levels Can’t represent diurnal cycle of productivity Requires better treatment of light penetration into canopy
Page 13© Crown copyright 2004 “light mod” allows better treatment Parameters optimally determined using observed data from Loobos Original model could be optimised too Get decent looking performance Parameters outside “physical” range Danger of optimising deficient model… A success… and a caveat
Page 14© Crown copyright 2004 Future Plans: CarboEurope-IP CarboEurope aims to, “ understand and quantify the present terrestrial carbon balance of Europe” 4 components. “continental integration” Make use of all different data streams to constrain models Make use of models to give full coverage of European land surface Essentially a Data Assimilation problem