impact of weather and climate variability on terrestrial fluxes: integration of eo data into models...

Impact of weather and climate variability on

terrestrial fluxes: integration of EO data

into models

Jean-Christophe Calvet, Gianpaolo Balsamo, Alina Barbu,

Alessandro Cescatti, Sébastien Lafont, Fabienne Maignan,

Dario Papale, Camille Szczypta, Balazs Szintai,

and Bart van den Hurk

Context

WWOSC, Montreal, 18 August 2014 2

Heritage: geoland2 FP7 project (2008-2012), macc2 (2011-2014)

Copernicus Atmosphere service (http://copernicus-atmosphere.eu/)Global atmospheric CO2 forecasting (ECMWF).

Copernicus Global Land service (http://land.copernicus.eu/global/)Near-real-time production of satellite-derived LAI, FAPAR, surface albedo (SA), land surface temperature (LST), and surface soil moisture (SSM) products at a global scale, together with other vegetation indices, burnt areas, water bodies.Started 1 January 2013.

ImagineS FP7 project (2013-2016)-From biophysical variables to agricultural indicators.-Performs the research needed in support to Copernicus GLS.-Use of European space-borne instruments (SPOT-VGT, PROBA-V, METOP-ASCAT, … future Copernicus Sentinel missions) and of geostationary satellites for LST

Data assimilation

WWOSC, Montreal, 18 August 2014 3

Numerical models contain errors that increase with time due to model imperfections and uncertainties in initial and boundary conditions. Data assimilation minimizes these errors by correcting the model stats using new observations.

Paul R. Houser, Figure from http://www.hzg.de/institute/coastal_research/cosyna

Data assimilation

WWOSC, Montreal, 18 August 2014 4

Numerical models contain errors that increase with time due to model imperfections and uncertainties in initial and boundary conditions. Data assimilation minimizes these errors by correcting the model stats using new observations.

Paul R. Houser, Figure from http://www.hzg.de/institute/coastal_research/cosyna

Data assimilation

WWOSC, Montreal, 18 August 2014 5

Numerical models contain errors that increase with time due to model imperfections and uncertainties in initial and boundary conditions. Data assimilation minimizes these errors by correcting the model stats using new observations.

Paul R. Houser, Figure from http://www.hzg.de/institute/coastal_research/cosyna

LDAS objectives

WWOSC, Montreal, 18 August 2014 6

From reanalyses to large scale land monitoring: integration of satellite products into a land surface model

Use of a Land Data Assimilation System (LDAS) able to perform the sequential assimilation of LAI and SSM satellite-derived observations.

Assess several products jointly and evaluate their consistencyThe LDAS is able to integrate/simulate most of the Copernicus GLS products and is used to:

- perform the cross-cutting quality control of LAI, SSM, FAPAR, surface albedo, LST- detect possible product anomalies (comparing recent months to past multi-annual analyses)- produce agricultural indicators (above-ground biomass, GPP, …) and drought indices

LDAS-France

WWOSC, Montreal, 18 August 2014 7

Within the SURFEX modeling platform of Meteo-France- Interoperable with operational real-time applications: weather forecast, hydrology, atmospheric CO2 inversions- Shared by many meteorological services in Europe and North Africa- Used in CNRM-ARPEGE climate model (IPCC simulations)- Version 8 will be open-source (end 2014)

ISBA-A-gs land surface model- Photosynthesis-driven phenology (no growing degree-days)- All the atmospheric variables impact phenology- Simulates the impact of long-term changes of atmospheric CO2

- Interannual variability of LAImax is modeled- LAI is flexible and can be analyzed at a given time- FAPAR and SA are modeled- SSM is modeled

(LAI, soil moisture, + FAPAR, surface albedo, surface temperature)

Extended Kalman filter

H observation operator

K Kalman gain

LDAS-France

WWOSC, Montreal, 18 August 2014 8

Joint assimilation of LAI and surface soil moisture (8km x 8km)

Barbu et al. 2014, HESS

LDAS-France

WWOSC, Montreal, 18 August 2014 9

Able to cope with uncertainties in the model and in the meteorological inputs

Example: no-precipitation test (model error 3)

Parrens et al. 2014, HESS

FORECAST

ANALYSIS

LDAS-France

WWOSC, Montreal, 18 August 2014 10

Accounts for sub-grid heterogeneity (8km x 8km)

Barbu et al. 2014, HESS

LDAS-France

WWOSC, Montreal, 18 August 2014 11

Accounts for sub-grid heterogeneity Grid cell near Toulouse (8km x 8km)

Barbu et al. 2014, HESS

LDAS-France

WWOSC, Montreal, 18 August 2014 12

LAI analysis (mean value for France)

Barbu et al. 2014, HESS

LDAS-France

WWOSC, Montreal, 18 August 2014 13

Soil moisture change rate in 2011 (extreme spring drought)

Barbu et al. 2014, HESS

LDAS-France

WWOSC, Montreal, 18 August 2014 14

10-daily GPP change rate in 2011 (extreme spring drought)

Barbu et al. 2014, HESS

Difficulties

WWOSC, Montreal, 18 August 2014 15

Assimilation of vegetation products: LAI vs. FAPAR

- FAPAR has very little sensitivity to LAI changes for LAI > 2- FAPAR is a radiative product (has a diurnal cycle, depends on atmospheric variables)- New LAI products are designed to limit the saturation effect (ex. GEOV1, Baret et al. 2013)- However, FAPAR is highly informative at wintertime

Assimilation of surface soil moisture

- ASCAT product has seasonal and interannual issues (vegetation impact on the C-band signal should be better described)- Decoupling in dry conditions- Couple with assimilation of surface temperature (more sensitive to soil water content in dry conditions) ?

LDAS-France

WWOSC, Montreal, 18 August 2014 16

Assimilation of FAPAR Jacobians (sensitivity of FAPAR to LAI perturbations) Grid cell near Toulouse (8km x 8km)

LDAS-France

LDAS-France

WWOSC, Montreal, 18 August 2014 17

Date of lowest ASCAT SSM value in 2009

DoY

MODEL ASCAT CORRECTED ASCAT

Barbu et al. 2014, HESS

LDAS-France

WWOSC, Montreal, 18 August 2014 18

Increments

LDAS-France

WWOSC, Montreal, 18 August 2014 19

Increments

Too high SSM observations ?

LDAS-France

WWOSC, Montreal, 18 August 2014 20

Assimilation of SSM in a multilayer soil hydrology model Jacobian profiles (sensitivity of SSM to perturbations of deep layers)

Parrens et al. 2014, HESS

Model calibration/validation

WWOSC, Montreal, 18 August 2014 21

Surface soil moisture (ESA-CCI microwave-derived product) Correlations (1991-2008 day-to-day variability)

Szczypta et al. 2014, GMD

Model calibration/validation

WWOSC, Montreal, 18 August 2014 22

Leaf Area Index (GEOV1 Copernicus Global Land product)Correlations (1991-2008 10-daily interannual variability)

Szczypta et al. 2014, GMD

Model calibration/validation

WWOSC, Montreal, 18 August 2014 23

Agricultural production (1994-2010 interannual variability)

winter/spring cereal grain yield grassland dry matter yield

Canal et al. 2014, HESSD

Red: p > 1%Yellow: p < 1%Dark : p < 0.1%

Model calibration/validation

WWOSC, Montreal, 18 August 2014 24

Agricultural production (1994-2010 interannual variability)

Canal et al. 2014, HESSD

Global atmospheric CO2 forecasting

WWOSC, Montreal, 18 August 2014 25

ECMWF IFS 16 km, free-running CO2

Good skill for real-time synoptic variability

Excellent background state for - assimilating GOSAT and OCO-2 data- assessing/improving surface flux models

SURFACE FLUXESAnthropogenic emissions (EDGARv4.2) 2008 + trend (Janssens-Maenhout et al. 2012)Ocean climatology (Takahashi 2009)Near-real-time GFAS biomass burning (MACC, Kaiser et al. 2011)Real-time Net Ecosystem Exchange: CTESSEL (Boussetta et al., 2013), based on ISBA-A-gs

Agusti-Panareda et al. 2014, ACPD

Daily mean CO2 at Park Falls (396 m)

CO2 anomalies above 392 ppm (24 Sep 2010 00 UTC)

Daily mean CO2 at Park Falls (396 m)

CO2 anomalies above 392 ppm (24 Sep 2010 00 UTC)

Global atmospheric CO2 forecasting

WWOSC, Montreal, 18 August 2014 26

Importance of having online CTESSEL NEE in the ECMWF IFS CO2 forecast

Synoptic variability is better represented with online CTESSEL

Agusti-Panareda et al. 2014, ACPD

OBS (NOAA/ESRL)

CTESSEL online NEE

CTESSEL monthly mean NEE

Conclusions

WWOSC, Montreal, 18 August 2014 27

Demonstrated synergies between land and atmosphere Copernicus services

Sequential assimilation vs. parameter optimization - Joint sequential assimilation of vegetation variables and soil moisture permits:

- synergies between satellite observations in the visible and microwave spectra

- controlling the model trajectory (needed during extreme events)

- MaxAWC governs the vegetation interannual variability and can be retrieved

LDAS-France is operational - Cross-cutting validation reports are generated every 6 months for the Copernicus Global Land service- Possible application for land reanalyses and drought monitoring- Very good representation of grassland dry matter yield provided MaxAWC is known

Prospects

WWOSC, Montreal, 18 August 2014 28

Ongoing activities

Comparisons with WOFOST MARS simulations (wheat)Test the assimilation of FAPAR and other variablesMulti-layer soil hydrologyUse decadal LAI time series to better constrain MaxAWCFrom EKF to EnKFLink to hydrology (in situ river discharge observations used for validation)

Medium term objectives

Go global (LDAS-Monde)Build a multi-decadal global land reanalysis integrating the existing vegetation and soil moisture satellite-derived CDRsIntercomparison of global land reanalyses (eartH2Observe project)