1. Assessing the South American Land Data Assimilation System (SALDAS) Datasets for the LBA Model Intercomparison Project (LBA-MIP) Luis Gustavo G de Goncalves NASA-GSFC/UMD-ESSIC INPE-CPTEC Joao de Mattos, Dirceu Herdies, Juan Ceballos, Daniel Vila, Jim Shuttleworth, Dave Toll, Natalia Coupe, Scott Saleska

2. Background The LBA Model Intercomparison Project (LBA-MIP) The LBA Model Intercomparison Project (LBA-MIP) The South American Land Data Assimilation System (SALDAS) Forcing Datasets The South American Land Data Assimilation System (SALDAS) Forcing Datasets SALDAS x LBA sites validation SALDAS x LBA sites validation

3. LBA-MIP Responds directly to the LBA-ECO Phase 3 “Synthesis and Integration” Responds directly to the LBA-ECO Phase 3 “Synthesis and Integration” Leverage from previous PILPS and MIP experience Leverage from previous PILPS and MIP experience The goal is to gain comparative understanding of ecosystem models that simulate energy, water and CO2 fluxes over the LBA area. The task is to subject all the models to the same forcing and experimental protocol, and compare the output. Perform model intercomparison in 2 phases: 1) 1)Single point comparison (LBA sites) 2) 2)Regional comparison (model derived gridded datasets) 1)How does a comprehensive suite of models simulate vegetation function (especially carbon and water fluxes) across different sites, and in response to seasonal and interannual climatic variations? What are the different modeling approaches, mechanisms, and/or sensitivities that cause different models to behave differently? 2)How do model simulations compare to observations through time and across sites? 3)Can observations be used to improve comprehensive models, revise our predictions of the fate of Amazonian forests, and their impact on global carbon cycling, under a changing climate? Key questions

4. LBA-MIP  Acquire meteorological data from a network of eddy flux towers in the Brazilian Amazon, to provide a consistently filled continuous dataset for driving models.  Use data to drive a suit of models.  Compare simulation behavior across models.  Use the flux observations to test and improve vegetation models of Amazonian carbon and water cycling, for application to projections of climate effects on Amazonian forests.  LBA-DMIP, phase 2: Spatially continuous simulations. Focusing on ecosystem land-vegetation models that are embedded in global-scale coupled GCMs (i) drive a suite of models with spatially continuous meteorological fields; (ii) test outputs with observation data products; and (iii) use the results to improve models which may then be used make a new suite predictions for the future of Amazon forests under climate change. Obejctives

5. LBA-MIP Participant Models and Experiments

6. LBA-MIP II South American Gridded Datasets Preferably Reanalysis due to the larger number of observations relative to operational models Preferably Reanalysis due to the larger number of observations relative to operational models –NCEP/NCAR Reanalysis 2 and ECMWF ERA40: global, low temporal and spatial resolution –CPTEC SARR (South American Regional Reanalysis): 40 Km, 6 hourly: Data derived using the modified version of the Eta model (Chou and Herdies, 1996) and the Regional Physical-space Statistical Analysis System (RPSAS) data assimilation scheme applied at 40Km horizontal resolution and 38 vertical levels. This system integrates upper air and surface observations from several sources over South America, including vertical soundings from the RACCI/LBA and SALLJEX field campaigns over the Amazon and the low-level jet regions along the Andes, respectively.

7. LSM Atmospheric Forcing Data Quality of land surface model (LSM) output is closely tied to the quality of the meteorological forcing data used to drive the model Quality of land surface model (LSM) output is closely tied to the quality of the meteorological forcing data used to drive the model SALDAS project seeks to provide accurate, near-real-time and retrospective land surface states over South America SALDAS project seeks to provide accurate, near-real-time and retrospective land surface states over South America Model and observation-based data used to create high- quality atmospheric forcing datasets Model and observation-based data used to create high- quality atmospheric forcing datasets –Retrospective (2000-2004, CPTEC) –Real-time (2002-Present, CPTEC)

8. Forcing Data Specifics 3-Hourly files 3-Hourly files 1/8 th Degree (~12.5 km) over Equator 1/8 th Degree (~12.5 km) over Equator GRIB format GRIB format C-shell scripts, Fortran programs used to automatically generate and archive forcing C-shell scripts, Fortran programs used to automatically generate and archive forcing Quality controlled, adjusted for terrain height Quality controlled, adjusted for terrain height 15 Model and observation-based fields 15 Model and observation-based fields SALDAS Domain

9. Atmospheric Forcing Fields Model-Based Estimates of Standard Climate Station Data Model-Based Estimates of Standard Climate Station Data –Temperature( 2 m assuming grass) –Specific Humidity( 2 m assuming grass) –U East-West Wind Component(10 m assuming grass) –V North-South Wind Component (10 m assuming grass) –Surface Pressure( 0 m assuming grass) Observation-Based Data Observation-Based Data –Downward Shortwave Radiation –Precipitation

10. Atmospheric Forcing Fields Model-Based Estimates of Above Ground Data Model-Based Estimates of Above Ground Data –Temperature –Specific Humidity –U East-West Wind Component –V North-South Wind Component –Surface Pressure –Downward Longwave Radiation –Height, h, above ground at which data applies –Altitude assumed for location ETA coordinate represents topography as steps ETA coordinate represents topography as steps

11. Terrain Height Adjustment Terrain Height Adjustment ETA temperature, pressure, humidity and longwave radiation adjusted for differences in ETA versus LDAS terrain height ETA temperature, pressure, humidity and longwave radiation adjusted for differences in ETA versus LDAS terrain height Temperature and pressure corrected using standard lapse rate Temperature and pressure corrected using standard lapse rate Specific humidity and longwave radiation corrected by holding relative humidity constant Specific humidity and longwave radiation corrected by holding relative humidity constant

12. Observations Model-based data subject to model error, so observations used when possible Model-based data subject to model error, so observations used when possible Radiation Radiation –GOES-CPTEC downward shortwave –GOES-CPTEC PAR (not implemented yet) –GOES-CPTEC skin temperature (not implemented yet) Precipitation Precipitation –TRMM 3B42 –GPCC + CPTEC/INMET surface observations

13. Observed Radiation GOES data processed at CPTEC/DSA (Divisao de Satelites Ambientais: Environmental Satellites Division) to create 1/25 degree, hourly, instantaneous surface downward shortwave radiation, PAR and skin temperature fields GOES data processed at CPTEC/DSA (Divisao de Satelites Ambientais: Environmental Satellites Division) to create 1/25 degree, hourly, instantaneous surface downward shortwave radiation, PAR and skin temperature fields –Interpolated to 1/8 th degree GOES shortwave radiation is zenith angle corrected, used in place of ETA data when possible GOES shortwave radiation is zenith angle corrected, used in place of ETA data when possible GL 1.2 GOES downward shortwave radiation (W/m 2 ) SARR downward shortwave radiation (W/m 2 ) Merged SALDAS downward shortwave radiation (W/m 2 ) + =

14. Precipitation Make use of TRMM and raingauges analysis data to form best available product—a temporally disaggregated 3-hourly value Make use of TRMM and raingauges analysis data to form best available product—a temporally disaggregated 3-hourly value Temporal Disaggregation Process  Description: Use of rain gauge to correct satellite bias precipitation  Spatial Resolution: 0.25 degree  Temporal Resolution: total daily precipitation  Domain: South America  Methodology: Use TRMM sub-daily precipitation pulses to dissagregate the total daily amounts. Note: 3B42RT data used to derive temporal disaggregation weights Sum of hourly data values equals original total daily TRMM/gauge

15. SALDAS x LBA Validation 19992000200120022003200420052006 1. BAN 2. K34 3. K67 4. K77 5. K83 6. RJA 7. FNS 8. PDG

16. SALDAS x LBA Validation 19992000200120022003200420052006 1. BAN 2. K34 3. K67 4. K77 5. K83 6. RJA 7. FNS 8. PDG

17. Surf Pressure (hPa) Temperature (K) Wind (m/s)Spec Hum (g/Kg) BiasSTDBiasSTDBiasSTDBiasSTD BAN-0.081.50.0012.70.0231.4-0.340.088 PDG6.81.5-2.43.51.82.1-0.320.086 K67251.51.31.7-0.671.4-0.420.049 K344.26.90.372.30.211.1-0.410.065 FNS-1221.62.50.92.2-0.370.076 K83171.60.572.10.331.2-0.40.083 RJA-232.21.82.91.71.6-0.410.073 SALDAS x LBA Validation Pressure, Temperature, Humid and Wind

18. SALDAS x LBA Validation Precipitation

19. SALDAS x LBA Validation Precipitation

20. SALDAS x LBA Validation Precipitation

21. SALDAS x LBA Validation Precipitation

22. SALDAS x LBA Validation Precipitation

23. SWDown SALDAS x LBA Validation Radiation Bias increases with magnitude SWDown

24. Forcing Data Archive 2000 2001 2002 2003 2004 2005 2006 SALDAS Retrospective 3-Hourly SALDAS Real Time Hourly SARR 3 Hourly ODAS 6 Hourly DSA/GOES Hourly TRMM/Gauge Precip General Data Radiation Precipitation

25. Conclusions Model and observation based data merged to create robust, accurate 1/8 th degree 3-hourly forcing data set Model and observation based data merged to create robust, accurate 1/8 th degree 3-hourly forcing data set –CPTEC-SARR/ODAS/ETA data serves as base –CPTEC-DSA/GOES, TRMM/raingauge data used to augment data set Common set of forcing integral to SALDAS LSM intercomparisons (NOAH,SiB3,SSiB…LBA-MIP) Common set of forcing integral to SALDAS LSM intercomparisons (NOAH,SiB3,SSiB…LBA-MIP) Five years archived, with continuing production Five years archived, with continuing production Validation effort proceeding Validation effort proceeding

26. Conclusions Current SARR based forcing need improvement over LBA region based upon tower observations: Current SARR based forcing need improvement over LBA region based upon tower observations: –Surf Pressure average errors up to 25 hPa (K67) –Specific humidity errors up to -0.42 g/Kg –Shortwave downward radiation average errors up to 8.5 W/m2 over the region (Std=23 W/m2) –Temperature average errors up to 2.5 K (BAN) –SALDAS precipitation underestimates in most of the sites and seasons –SALDAS lower number of precipitation events higher than 20mm/3hrs relative to LBA

27. More Info… LBA-MIP LBA-MIP –http://climatemodeling.org/lba-mip http://climatemodeling.org/lba-mip “Modeleres” are welcome!! “Modeleres” are welcome!! South American LDAS Atmospheric Forcing Datasets South American LDAS Atmospheric Forcing Datasets –Atmospheric Forcing Datasets: LBA/DIS Beija-Flor Keyword = “SALDAS” LBA/DIS Beija-Flor Keyword = “SALDAS” –SALDAS Webpage: http://ldas.gsfc.nasa.gov/SALDAS http://ldas.gsfc.nasa.gov/SALDAS