Data assimilation in land surface schemes

Mathew WilliamsUniversity of Edinburgh

Globalforecast

capability

Keyprocesses

Challenges for the JULES team

Initialconditions

Parameters

Errors

EO &sensor

networks

Continualtesting &

development

JULES

The philosophy of data assimilation

• Observations and models contain useful information about the target system

• But ALL observations and models are subject to error - and may be subject to bias

• Models and observations can be combined to optimise information

Improving estimates of land surface process

MODELS OBSERVATIONS

FUSION

ANALYSIS

MODELS+ Capable of interpolation

& forecasts─ Subjective & inaccurate?

OBSERVATIONS+ Clear confidence limits

─ Incomplete, patchy─ Net fluxes

ANALYSIS+ Complete

+ Clear confidence limits+ Capable of forecasts

+ Data consistency

Approaches to data assimilation

• Sequential (predictor-corrector)

The Kalman Filter

MODEL At Ft+1 F´t+1OPERATOR

At+1

Dt+1

Assimilation

Initial state Forecast ObservationsPredictions

Analysis

P

Ensemble Kalman Filter

Drivers

Ft+1Ft+1At

At

The Ensemble Kalman Filter

MODEL At Ft+1 OPERATOR F´t+1

At+1

Dt+1

Assimilation

Initial state Forecast ObservationsPredictions

Analysis

P

Drivers

Observations – Ponderosa Pine, OR (Bev Law)

Flux tower (2000-2)Sap flowSoil/stem/leaf respirationLAI, stem, root biomassLitter fall measurements

GPP Croot

Cwood

Cfoliage

Clitter

CSOM/CWD

Ra

Af

Ar

Aw

Lf

Lr

Lw

Rh

D

DALEC

Time (days since 1 Jan 2000) Williams et al (2005)

Time (days since 1 Jan 2000) Williams et al (2005)

= observation— = mean analysis| = SD of the analysis

Time (days since 1 Jan 2000) Williams et al (2005)

Time (days since 1 Jan 2000) Williams et al (2005)

= observation— = mean analysis| = SD of the analysis

0 365 730 1095-4

-3

-2

-1

0

1

2

0 365 730 1095-4

-2

0

2

Time (days, 1= 1 Jan 2000)

b) GPP data + model: -413±107 gC m-2

0 365 730 1095-4

-3

-2

-1

0

1

2

c) GPP & respiration data + model: -472 ±56 gC m-2NE

E (

g C

m-2 d

-1)

0 365 730 1095-4

-2

0

2

a) Model only: -251 ±197 g c m-2

d) All data: -419 ±29 g C m-2

Data brings confidence

Williams et al (2005)

= observation— = mean analysis| = SD of the analysis

Approaches to data assimilation

• Sequential (predictor-corrector)• Inversion techniques

– Monte Carlo – Adjoint

Monte Carlo Inversion

MODEL F F´OPERATOR

Forecasts

ObservationsPredictions

P

Drivers

Initial

J

D

Cost function

REgional Flux Estimation eXperiment (REFLEX)

• To compare the strengths and weaknesses of various MDF/DA techniques

• To quantify errors and biases introduced when extrapolating fluxes

• www.carbonfusion.org

REgional Flux Estimation eXperiment (REFLEX)

FluxNet dataMODIS

MDF

Full analysisModel parameters

DALECmodel

Training Runs

Deciduous forest sites

Coniferous forest sites

Assimilation

Output

REgional Flux Estimation eXperiment (REFLEX)

FluxNet dataMODIS

MDF

Full analysisModel parameters

DALECmodel

Testing site forecastswith limited EO data

MDF

MODIS

Analysis

FluxNet data

testing

Assimilation

Scaling

A multiscale approach

www.abacus-ipy.org

Process models-upscaling

EO data:-landcover-phenology

Flux towers-processes-parameters

Geostats-Spatial drivers-Uncertainty

Tall tower /aircraft-Check on upscaling-Inversions

Linking fluxes to atmospheric [CO2]

Figure by Paul Parrish, data from J Moncrieff & J Grace

The Orbiting Carbon Observatory (OCO)

Approach: • Collect spatially resolved, high resolution

spectroscopic observations of CO2 and O2 absorption in reflected sunlight

• Use these data to resolve spatial and temporal variations in the column averaged CO2 dry air mole fraction, XCO2 over the sunlit hemisphere

• Employ independent calibration and validation approaches to produce XCO2 estimates with random errors and biases no larger than 1 - 2 ppm (0.3 - 0.5%) on regional scales at monthly intervals

OCO will acquire the space-based data needed to identify CO2 sources and sinks and quantify their variability over the seasonal cycle

Source: David Crisp, JPL

Spanning measurement scales

OCO will make precise global measurements of XCO2 over the range of scales needed to monitor CO2 fluxes on regional to continental scales.

Spatial Scale (km)

1

2

3

4

5

6

CO

2 E

rror

(pp

m)

1 10 100 1000 10000

OCO

FlaskSite

AquaAIRS

Aircraft

0

FluxTower

Globalview Network

NOAATOVS

ENVISATSCIAMACHY

Tall tower

Source: David Crisp, JPL

A strategy for JULES?

• LOCAL: DA for local parameter PDFs, process testing, C-water interactions, full state descriptions. FluxNet, IPSL, NCAR, ACCESS.

• REGIONAL: upscaling, coupling to/inverting atmospheric data/models. CarboEurope, ABACUS.

• GLOBAL: Global assimilation with optical, CO2, water, temperature remote sensing, flasks. NCEO & CCDAS.

THANKS!

Tim Hill, PhD thesis

Monte

Carlo

Inversion