Tuesday 27 April

7:00 pm - Meet Franciscan Inn lobby for supper (optional)

Wednesday 28 April

9.00 am - NCEAS

1. Objective(s) of EMDI (validation vs. model/data enhancement)
2. NCEAS Workshop tasks and products
3. Participating Models
4. Experimental Design
5. NPP Data
6. Model Driver Data
7. Sensitivity Studies
8. Analysis/Visualization
9. Logistics (date/place/agenda for EMDI, Web server)
10. Future: EMDI-II (flux data, temporal)
11. EMDI products

1. Objectives and Protocol

1. Define the questions to be asked and who is interested in the answers
2. Discuss validation vs. model/data enhancement
3. Define Criteria: e.g., 90% of model values are within 20% of the field data mean and variance values

1. General approach

1. Release of model driver data
2. Modelers do first-cut = "Blind" simulations
3. Receive model output/release of 1/2 NPP "Answers"
4. Interim analysis and visualization (prototype for PIK/EMDI)
5. Modelers review results and model algorithms
6. Receive revised model output
7. Perform analyses and visualization for workshop

1. Problems: disturbance, land use, managed systems
2. Hierarchy of the experiments

1. Global - run models for 3000 Class III
2. Point - run models for 150 Class I/II (with monthly climate and more detailed ancillary information)
3. Spatial - run models for ~1000 grid cells (data promised but not yet available)
4. Flux - run models for 15-20 flux towers, predicting NEP
5. Regional - run models at 0.5º grid cell resolution for broad regions (Eastern US or China or globally) and compare to NPP in environmental space
6. Transient - run models for series of years for Great Plains and compare distribution of NPP over years to Sala data or to range of NPP in several grassland series

1. GPPDI: an update

1. Class I/II
2. Class III
3. Grid cells (Great Plains, Sala, Queensland, FIA, St. Petersburg)

1. New Data: feasibility with flux towers?

1. Flux - 15-20 flux towers

1. Minimum criteria for model initialization:

1. Veg-type - biome (select biome classes), dominant species
2. Climate - annual/monthly temp-precip
3. Soils
4. NDVI

1. Required/Desired output variables

1. ANPP (gC/m2)
2. NPP (gC/m2)
3. Others?

1. Spinup protocols for models- lead times the models need (e.g. climate), assumptions we MUST make

1. Sensitivity analyses

1. Remote Sensing: Implications of remote sensing quality? (e.g. long-term vs. actual NDVI’s)
2. Veg models: Implications of climatologies: long-term vs. actual climatologiesà all variables, select variables
3. Models with an open N-cycle: implications of pre-industrial through to current Ndep?

pm - Continue discussion as necessary for model inputs and outputs:

• FPAR vs. NDVI needs for remotely-sensed driven models
• Additional output variables: e.g. allocation (above, belowground)
• C, H₂O ‘budgets’.

1. Kick (TEM); Chris (IBIS); Bev (Flux data)
2. Peter (BiomeBGC); Steve (GLOPEM); Annette (data)
3. Bill (CENTURY); Ben (LPJ); Dick (data)

• Define rules/criteria to filter NPP data for site selection (e.g. +/- 2 stdev from a climate or biome driven partitioning of the data)
• Discuss the issues associated with validating NPP simulations from remote sensing-based models - perhaps a core group of RS models (e.g. GLOPEM, CASA, etc.) that use VI’s
• Define a set of Class I-II temperate forest sites (OTTER, a few LTER sites, perhaps some Bartlett forest data).

Thursday 29 April

9:00 am - NCEAS

1. Visualization (suitable graphics software, e.g., Spyglass??)
2. Simple quantitative and spatial techniques (e.g. r² by biome or climatic regions environmental)
3. Environmental relationships

1. Assume empirical relationship between environment and NPP field measurements should be reproduced between driving variables and model simulations
2. Covariate analysis using climatic driving variables

1. Environmental controls

1. Assume within that mean and variance of model simulations should be similar to mean and variance of NPP field measurements within homogenous blocks such as biomes, species, soils, climatic
2. Nonparameteric permutation analysis of blocks

1. Testing the analysis strategies for PIK/EMDI workshop (e.g., use a simple model (e.g., the MIAMI model) with data from Day 1 to simulate a ‘model/data intercomparison’ exercise)

• Visualization techniques
• Simple quantitative and spatial techniques
• Environmental Space techniques

pm - Continuation of breakout groups

Friday 30 April

9:00am - NCEAS

1. EMDI: date/place/agenda
2. Schedule: pre EMDI, post EMDI
3. Web server (need NCEAS staff to participate)

• Mirroring NCEAS and PIK
• Information and communication

1. FTP server

• Volume of data
• "User support" for workshop participants uploading/downloading)
• Access control?

1. Organization of outputs

• Testing data filtering methods
• Testing simple quantitative and spatial techniques
• Testing environmental space techniques

pm - Field Trip — Salt Marsh, day hike, sea kayaking?

Saturday 1 May

9:00 am?NCEAS

1. Flux data
2. Other time-series data (crop and forest inventories)
3. Dendroecology data
4. MODIS data

• Discuss products we would like to see from the EMDI — papers, data sets, etc.
• Do we want to ‘deal’ with another GCP-type special issue?

3:00pm - adjourn

Participants:

Hibbard, Kathy k.hibbard@unh.edu University of New Hampshire

Kucharik, Chris kucharik@bob.soils.wisc.edu University of Wisconsin, Madison

Kicklighter, Dave dkick@lupine.mbl.edu Marine Biological Laboratory

Law, Bev lawb@ccmail.orst.edu Oregon State University

Olson, Dick rjo@ornl.gov Oak Ridge National Laboratory

Parton, Bill billp@nrel.colostate.edu Colorado State University

Prince, Steve sp43@umail.umd.edu University of Maryland

Schloss, Annette annette.schloss@unh.edu University of New Hampshire

Smith, Ben ben@planteco.lu.se University of Lund, Sweden

Thornton, Peter peter@ntsg.umt.edu University of Montana