Regional Model-Data Comparison

1
Regional Model-Data Comparison

• An NACP Interim Synthesis Project
• Andy Jacobson, Mac Post, Debbie Huntzinger, Bob
  Cook
• Coordinators

2
Synthesis of Interim NACP Results

• Ecosystem Models
• Contribute in hand regional, continental results
  (including ones cut from global analyses)
• A range of temporal and spatial resolutions
• No standardization of model runs!
• Inversion Models
• Contribute North America results in hand from
  TRANSCOM or from other relevant activities
• Spatial scales
• TRANSCOM regions, and
• 1º grids centered on half-degrees
• Temporal scale - monthly

3
Regional MDC Objectives

• Development of benchmark data sets and approaches
  for model-data evaluation.
• Evaluation of strengths and weaknesses of various
  model formulations, both inverse models and
  ecosystem models resulting from the comparison to
  data.
• Formal comparison of inverse and forward
  ecosystem model results for enhancing
  identification and diagnosis of temporal and
  spatial patterns of carbon fluxes.

4
Regional MDC Questions

• Interannual Variation
• What is the spatial pattern and magnitude of
  interannual variation in carbon fluxes during
  2000-2005?
• What are the components of carbon fluxes and
  pools that contribute to this variation?
• 2002 Drought
• Do model results and observations show consistent
  spatial patterns in response to the 2002 drought?
  
• From measurements and ecosystem models, can we
  infer what processes were affected by the 2002
  drought?
• Identification of Sources/Sinks
• What are the magnitudes and spatial distribution
  of carbon sources and sinks, and their
  uncertainties during 2000-2005?

5
Conversion to Common Grid - both Data and Model

• Spatial Resolution 1x1, centered at (x.5, y.5
  for each grid cell)
• Domain 50 to 170 W longitude, 10 to 84 N
  latitude
• Temporal Resolution monthly (or annual)
• netCDF files, CF-1 convention

6
Observations and Measurements

• Satellite based
• In Hand
• MODIS GPP
• MODIS NPP (annual)
• In Development
• LAI (gap filled, smoothed MODIS for NACP)
• Survey
• In Hand
• NASS crop yield based annual NPP
• FIA based biomass
• In Development
• Soil C (CONUS-SOIL, http//www.soilinfo.psu.edu/),
  30cm, 100cm
• Site based
• Eddy flux NEE, estimated GPP, NPP
• All of these are not strictly direct measurements

7
FIA Based Forest Biomass
From Deborah N. Huntzinger (University of
Michigan)
From Blackard et al. (2008), G. Moisen,
contact (Rocky Mt. Res. Sta., USFS)
Example Regional Data Sets for Model Data
Comparison
8
Spatial Aggregation of Forest Biomass

• 250m resolution aggregated to 1 degree
• Average biomass of forest on forested land
  computed
• Area of forested land in 1 degree cell kept

9
Model Submissions
10
Additional Inversion Models

• Transcom3 IAV inversion (D Baker),13 models
• Rödenbeck Jena, 3 different networks
• CarbonTracker
• FRCGC Japan (Patra)
• U Michigan geostat, 2 models
• LSCE France (Peylin)
• LSCE France (Chevallier)
• LSCE France (Rayner)
• Penn State (Butler), 2 models

11
Forward Model Metadata Tool
12
Forward Model Metadata Report
http//daac.ornl.gov/SURVEY8/survey_results.shtml
13
Some Example Forward Model-Data Comparisons
14
Analysis Approaches

• Comparison Techniques
• Time series plots for entire domain, vegetation
  regions
• Statistical point by point comparisons (Taylor
  plots, cumulative frequency distributions,
  Index-of-agreement, etc.)
• Spatial pattern comparisons (difference plots,
  variograms, etc.)
• Combinations of Comparisons
• Forward/Ecosystem model - Data inter-comparison
• Inverse and Forward model inter-comparison

15
Spatial Aggregation - North America
Convention (-) sink () source
16
Spatial Aggregation TransCom Regions
17
Taylor Diagrams
RMSD2sd_obs2sd_mod2-2sd_obssd_modR
18
Taylor Diagrams- Monthly
19
Time Series of Taylor Plot Components
20
(No Transcript)
21
Spatial Analysis

• Used to describe spatial correlation

1
2
3
4
Optimized
Anna Michalak University of Michigan
22
Seasonal Fluctuation of Spatial Variability
Deborah Huntzinger University of Michigan
23
Seasonal Fluctuation of Spatial Variability
Semivariance, ?2 ?moles/m2/sec2
Correlation Range, 3L (km)
Deborah N. Huntzinger University of Michigan
Optimized using Ordinary Least Squares
24
Seasonal Fluctuation of Spatial Variability

• where
• Vmax chosen to be
• 0.01 (?mols/m2/sec)2
• L and ?2 from previous slide

Deborah Huntzinger University of Michigan
25
Preliminary Observations

• Data are not strictly observations but involve
  results of some model
• Different data and different methods of
  comparison reveal various components of model and
  data compatability
• Inverse forward model comparison allows
  evaluation of inverse models indirectly with data

26
Proceeding from Here

• Need ideas for
• Additional regional/continental datasets
• Methods for model-data comparison the application
  of these methods
• List of graphics and analysis for working with at
  ORNL Workshop, January 7-9
• Outline of presentation for NACP Investigator
  Meeting, February 17-20