glass panel meeting 25-27 august 2003 tucson, arizona, usa sahra

GLASS Panel Meeting25-27 August 2003Tucson, Arizona, USA

GLASS Panel Meeting25-27 August 2003Tucson, Arizona, USA

SAHRA

GLASS Framework2 x 2 Matrix:

Spatial scale vs. Land-atmosphere coupling strategy

2 x 2 Matrix:

Spatial scale vs. Land-atmosphere coupling strategy hydro.iis.u-tokyo.ac.edu/GLASS/hydro.iis.u-tokyo.ac.edu/GLASS/

Current GLASS Actions• Local Offline

– PILPS C1 (finishing)– San Pedro (starting)– Isotope PILPS (proposed)

• Large-Scale Offline– GSWP2 (underway)

• Large-Scale Coupled– GLACE (underway)

• Local Coupled– LOCO (proposed)

• Other issues

Summary of the first phase of the Summary of the first phase of the PILPS C-1PILPS C-1 project projectComparison of both « biophysical » and « biogeochemical » flux from different types of models with observations at one EUROFLUX site: Loobos

The site:The site:-Temperate « mature(100 years) » coniferous forest-Climate: 700 mm precipitation , 9.8 °C mean temperature- Planted on a sand no soil carbon at the beginning of the plantation- Measured fluxes: NEE, LE,H, Rn- Meteorological parameters: incoming SW rad., precipitation, temperature, wind speed, relative humidity, pressure-Period covered: 1997-1998

Models:Models: Including SVAT with and without carbon cycle

Simulations:Free equilibrium simulations:-Models are run until equilibrium of state variables using years 1997-1998 « in loop »Free 100 years run:-simulation of « realistic scenario »: Beginning with a soil with no carbon, the models are run for 1906 (plantation of the forest) to 1998 using observed climate.

www.pilpsc1.cnrs-gif.fr/ www.pilpsc1.cnrs-gif.fr/

Main preliminary conclusion

Taking into account that models was not calibrated, the models reproduce relatively well the observations

Sensible heat flux is overestimated at night. High net CO2 and latent heat fluxes are underestimated

The 100 years simulation was very interesting since if all models give relatively similar NEE and are all able to reproduce the difference of sink between 1997 and 1998, trajectories of models carbon fluxes and pools are very different !

For more details on results go to:


CLASS-MCM CLASS-UAORCHIDEE-1 ORCHIDEE-2SWAPIBIS

VISA

Total Soil carbon (Kg C/m-2)

MCAVIM

Main preliminary conclusionTaking into account that models was not calibrated, the models reproduce relatively well the observations

Sensible heat flux is overestimated at night. High net CO2 and latent heat fluxes are underestimated

The 100 years simulation was very interesting since if all models give relatively similar NEE and are all able to reproduce the difference of sink between 1997 and 1998, trajectories of models carbon fluxes and pools are very different!

For more details on results go to:


PILPS Semi Arid Experiment(PILPS San Pedro)

USA

Arizona New Mexico

Phoenix

Albuquerque

Kendall

Lucky Hills

Tucson

Sevilleta

Sponsors

SAHRA

www.sahra.arizona.edu/pilpssanpedro/ www.sahra.arizona.edu/pilpssanpedro/

Locations

SiteLongitude

WestLatitude

NorthElevation[m.a.s.l.]

Precipitation[mm/year]

Temperature[C]

Lucky HillsShrubland

11003’05’’ 3144’37” 1372 340 18.6

KendallGrassland

10956’28” 3144’10” 1526 340 19.3

TucsonShrub/cacti

11149’48” 3213’01” 730 305 20.2

Sevilleta Grassland

10643’30” 3420’30” 1730 270 17.2

SevilletaShrubland

10644’39” 3420’05” 1776 270 16.9


1996 1997 1998 1999 2000

1996 1997 1998 1999 2000

Forcings

Outputs

Spin up PeriodData Supplied

CalibrationPeriod

Data Supplied

EvaluationPeriod

Data NOT supplied

2001 2002Forcings 2000

Outputs 2001 20022000

Forcings 19941993

Outputs 19941993

San Pedro Shrub & Grass

Sevilleta Shrub & Grass

Tucson Mixed Shrub & Cacti

Split Sample Test


Science Questions• What is the ability of the models to reproduce the water,

energy, and carbon exchanges in semi-arid environments?

• This is the first PILPS study in a semi-arid region, with study sites in Arizona and New Mexico.

• Does model calibration reduce the among-model range in the model simulations?

• Code for performing a multi-criteria optimization procedure developed by Gupta and Bastidas will be made available to all model participants for calibration of their schemes.

• Are the current (usually single) representations of semi-arid lands in the models enough to reproduce the different environments that exist in those areas?

• The experiment will be a direct test of spatial transferability of surface parameters, as is commonly practiced but not validated in weather, climate and hydrologic models.


Implementation• What is the ability of the models to reproduce the water,

energy, and carbon exchanges in semi-arid environments?

• Uncalibrated simulation of 3 Arizona sites (validation data withheld)

• Does model calibration reduce the among-model range in the model simulations?

• Validation data provided for Arizona sites – modelers can calibrate their models to improve simulations

• Are the current (usually single) representations of semi-arid lands in the models enough to reproduce the different environments that exist in those areas?

• Use calibrated parameters from grassland and shrub sites in Arizona for similar sites in New Mexico (NM validation data withheld)

• Validation data provided for Arizona sites.


Isotopes in PILPS : IPILPS

• Stable and radioactive isotopes are valuable for both tracing and dating

• Stable water isotope measurements – Demonstrated that the Amazon recycles its water (3 or 4 times)– Prompted early GCM deforestation experiments (e.g. Henderson-

Sellers & Gornitz, 1984)

• Carbon isotopes differentiate C3 and C4 photosynthesis• Off-line (PILPS) inter-comparison is timely e.g.

– “Heavy” water (HD16O & H218O) at the land surface

– River routing schemes tested against GNIR – the Global Network for Isotopes in Rivers

•QUESTION: Are the stores and fluxes of two different “heavy” water isotopes (HD16O & H2

18O) correctly characterised at the land surface?

Basin Isotope Modelling

• Isotopes provide an independent climate and water resource validation tool

• Weakened signal in central Amazon (Manaus) indicates increased water recycling

• Logarithmic enhancement in Darling reveals irrigation impacts

H-S, Stone, McGuffie etc GRL, 2003,

Manaus

Enhanced D/H ratio

Darling

ndnd Global Soil Wetness Global Soil Wetness ProjectProject

Multi-model investigations into variability and predictability of the global surface water and energy cycles

This phase of the project takes advantage of:

•The10-year ISLSCP Initiative 2 data set (1986-1995)

•The ALMA data standards developed in GLASS

•The infrastructure developed in the pilot phase of GSWP

www.iges.org/gswp/ www.iges.org/gswp/

ndnd Global Soil Wetness Global Soil Wetness ProjectProject

GSWP-2 represents an evolution in multi-model large-scale land-surface modeling with the following features:•The ten year length of the ISLSCP Initiative 2 allows for a better investigation of interannual land surface climate variability.•The ISLSCP Initiative 2 data set contains more than one rendition of many global fields, produced by different methods and scientists. This gives us a straightforward means to investigate LSS sensitivity to the choice of forcing data sets.•Application and further development of the methods of calibration and validation of LSSs with in situ and remote sensing data, including direct simulation of brightness temperatures as observed by satellite.


GSWP-2 data sets

GSWP-2 data sets for parameter specification, meteorological forcing, and validation have been produced.

The data sets are based on the ISLSCP-Initiative II data, but many of the fields represent additional processing, such as the production of “hybrid” data sets combining gridded observations (low temporal resolution) with model reanalysis (high time resolution). This hybridization removes systematic errors in the reanalysis data, providing a superior set of forcing data for the land surface models.


Hybrid minus reanalysis

Hybrid minus reanalysis

Data servers

The data sets have been posted online for community access. There are three DODS servers for accessing the data directly over the internet:

North America: http://www.monsoondata.org:9090/dods/Europe: http://dods.ipsl.jussieu.fr/ Asia: http://ftp.tkl.iis.u-tokyo.ac.jp:9090/dods/

The North American and Asian servers are GDS.

There is also FTP access to the individual files at:ftp://monsoondata.org/ (password required) and direct HTTP access to files at:http://dods.ipsl.jussieu.fr/


AMS 2004• There will be a GSWP-2 session at the AMS 18th

Conference on Hydrology, AMS Annual Meeting (Seattle, Washington, USA, 11-15 January 2004) to present preliminary results of the experiment.

• 12 Abstracts submitted to GSWP-2 session.

• 24 November deadline for baseline runs.• Start on QC, comparison, and sensitivity test cases.• Public presentation of preliminary results at AMS

(Seattle, Jan 2004).

Progress


GRP/GMPP exchange via GSWP-2

Plans are being made for a collaborative effort between GRP and GMPP.

•Bill Rossow has been communicating with Paul Dirmeyer and Jan Polcher about using GSWP estimates of surface (latent and sensible) heat fluxes over land for helping to close the global surface energy budget. •GSWP will also pursue sensitivity studies using ISCCP estimates of surface radiation for model forcing, compared to SRB and reanalysis estimates, to understand how uncertainty in our estimates of radiation propagate into the terrestrial hydrologic cycle.


K02 strategy, part 1: Establish a time series of surface conditions (Simulation W1)

Step forward thecoupled AGCM-LSM

Write the valuesof the land surface prognostic variablesinto file W1_STATES


Write the valuesof the land surface prognostic variablesinto file W1_STATES

time step n time step n+1

(Repeat without writing to obtain simulations W2 – W16)

GLACE: Global Land-Atmosphere Coupling Experiment

This experiment is a broad follow-on to the four-model intercomparison study described by Koster et al. (2002)*, hereafter referred to as K02.*J. Hydrometeorology, 3, 363-375, 2002

glace.gsfc.nasa.gov/ glace.gsfc.nasa.gov/

K02 strategy, part 2: Run a 16-member ensemble, with each member forced to maintain the same time series of surface prognostic variables (Simulations R1 – R16)


Throw out updated values of land surfaceprognostic variables; replace with values for

time step n fromfile W1_STATES


time step n time step n+1

Throw out updated values of land surfaceprognostic variables; replace with values for

time step n+1 fromfile W1_STATES


All simulations in ensemblerespond to the land surface boundary condition in thesame way is high

Simulations in ensemblehave no coherent responseto the land surface boundary condition is low

Define a diagnostic that describes the impact of the surface boundary on the generation of precipitation.


Plan for GLACE

K02

Step 1: 16-member ensemble,with prognostic states written out at each time step by oneof the members.

Step 2: 16-member ensemble,with all members forced to use the same time series ofsurface prognostic states.

All simulations are run over July.

NEW

GLACE

Step 1: 16-member ensemble,with prognostic states written out at each time step by oneof the members.

Step 2: 16-member ensemble,with all members forced to use the same time series ofsurface prognostic states.

Step 3: 16-member ensemble,with all members forced touse the same time series ofdeeper (root zone and below)soil moisture states.

All simulations are run fromJune through August


Bonan14. NCAR with CLM2

Kanae/Oki13. U. Tokyo w/ MATSIRO

Xue12. UCLA with SSiB

Koster11. NSIPP with Mosaic

Lu/Mitchell10. NCEP/EMC with NOAH

Taylor9. Hadley Centre w/ MOSES2

Sud8. GSFC(GLA) with SSiB

Gordon7. GFDL with LM2p5

Verseghy6. Env. Canada with CLASS

Viterbo5. ECMWF with TESSEL

Kowalczyk4. CSIRO w/ 2 land schemes

Dirmeyer3. COLA with SSiB

Hahmann2. CCM3 with BATS

McAvaney/Pitman1. BMRC with CHASM

ContactModel

Participating Groups

Status

Dropped out

15. LMD w/ ORCHIDEE Polcher Not yet submitted

Not yet submitted

Not yet submitted

Not yet submitted

Prelim.ResultsLarge variety

coupling strengths

evident among

models, with no systematic

patterns.


In principle, imposing land surface boundary states should decrease the intra-ensemble variance of the atmospheric fields.

pdf of precipitation ata given point, acrossensemble members

correspondingpdf when land boundaryis specified

We are examining this in GLACE by looking at the variance ratio: 2

P (S)

2P (W)


Strawman Proposal: LOcalCOupled

• Phase 1: Synthetic column – extract column boundary conditions from atmospheric model– Proof of concept

• Phase 2: Case Studies with Past Field Campaigns– PILPS sites: Cabauw, FIFE, Boreas, others?– GABLS sites: which?– Others: SGP/ARM/IHOP, LBA, Wangara, others?

• Phase 3: Future Field Campaigns– Upcoming Sites: HEAT (Houston, TX)-Urban;

others?– New Site(s): Design “local coupled testbed(s)”

Local-Coupled modeling experiments – GABLS and GLASS?Multi-LSS + Common-SCM to understand land-PBL feedbacks

Field Site for Local-Coupled Testbed

• GOAL: Collect a specifically targeted dataset to meet the objectives of the GLASS local-coupled action.

– A well-defined mesoscale hydrologic catchment (for surface water balance observations)

– Observations from groundwater table through PBL

– Include heterogeneity in elevation, vegetation, soils, climate (inc. snow), etc.

– Observation platforms:• In-situ: weather, fluxes, snow, soil moisture, groundwater, vegetation• Aircraft: fluxes, temperature, moisture, microwave, etc.• Soundings: tethered • Satellite: various available

Questions:

What should the spatial scale and heterogeneity be?

When, where, and how long?

Is this experiment of interest?

Parameters•Soil Properties•Vegetation Properties•Elevation & Topography•Subgrid Variation•Catchment Delineation•River Connectivity

Forcing•Precipitation•Wind profiles•Humidity profiles•Radiation•Air Temperature profiles

Fluxes•Evapotranspiration•Sensible Heat Flux•PBL fluxes•Radiation•Runoff•Drainage•Isotopes/carbon

States•Soil Moisture•Groundwater•Surface Water•Temperature (soil, veg, air)•Humidity•Wind•Pressure•Snow•Carbon•Nitrogen•Biomass

Other Issues

• Initialization of soil wetness in climate models– Lack of transferability of soil moisture data sets exists from one model to

another. – GLASS is preparing a summary paper to educate the broader modeling

community as to the pitfalls of treating soil moisture as a uniformly defined quantity across models

• Application of LSSs in regional models. – The regional modeling community is not well connected to the global

modeling community who seem to do much of the land model development work.

– LSSs in regional models are often not well initialized, and the panel perceives that the regional modeling community underestimates the severity of this problem.

– A workshop on the topic is being considered to bring the land surface and regional modeling communities together. A potential stage for such a workshop may be the African Monsoon Multidisciplinary Analysis (AMMA).

• Urban modeling• iLEAPS• ISLSCP

glass panel meeting 25-27 august 2003 tucson, arizona, usa sahra

Documents

different types of models

years simulation

latent heat fluxes

soil carbon

carbon cyclesimulations

observationssensible

spatial scale

edupilpssanpedro locationswww