Report of the POTSDAM '95 IGBP NPP
Model Intercomparison Workshop (GAIM-DIS-GCTE)
Potsdam, Germany, June 20-22, 1995

Berrien Moore, Convener & GAIM Chair
Wolfgang Cramer, co-convener
Dork Sahagian, GAIM Executive Director
Ichtiaque Rasool, DIS Executive Officer
Will Steffen, CGTE Executive Officer
and the Participants* in "POTSDAM '95"

Global primary production of ecosystems on land and in the oceans is a crucial component of biogeochemical model development within IGBP. The second in a series of IGBP GAIM-DIS-GCTE workshops, "Potsdam '95," on global modelling of net primary productivity was held at Potsdam Institute for Climate Impact Research (PIK) June 20-22 1995. The scientific sponsorship of this workshop was jointly by IGBP-DIS, GAIM and GCTE, and it was hosted by the Potsdam Institute of Climate Impact Research (PIK), with financial support from NASA, the European Commission, and the U.S. Environmental Protection Agency. The purpose of Potsdam '95, like Potsdam '94 was to support a series of model intercomparisons by the various modelling teams around the globe that are curently modelling the terrestrial biosphere at large scales. A focus in the intercomparison was net primary productivity. Global primary productivity of ecosystems is a crucial component of biogeochemical model development within IGBP. The subject is complex and central to models of the global carbon cycle. The importance of this subject rests upon the fact that most terrestrial carbon cycle models as well as terrestrial models that treat energy and water fluxes must address in a fundamental manner either gross primary productivity (GPP) and/or net primary productivity (NPP). There are significant differences in the calculation of NPP within current global terrestrial models, and Potsdam '95 was held in order to compare model parameters and outputs.

Geographically referenced net primary productivity (NPP) and gross primary productivity (GPP) and their corresponding seasonal variation are key components in the terrestrial carbon cycle. They are needed to understand both the function of living ecosystems and also their effects on the environment. Productivity is also a key variable for the sustainability of human use of the biosphere by, for example, agriculture, animal production and forestry. Recently, it has become possible to investigate the magnitude and geographical distribution of these processes on a global scale by a combination of ecosystem process modelling and monitoring by remote sensing. Since agricultural and forestry production provide the principal food and fuel resources for the world, monitoring and modelling of biospheric primary production are important to support global economic and political policy making.

The main presentations made at the workshop were as follows:

1. Overview of the standard driving data sets assembled in the PIK '95
database and used with about 10 of the models. The standard data sets
included climate, soil texture, solar radiation, a weather generator, ISLSCP
Initiative 1 CD-ROM AVHRR data.

2. Updates on model developments since PIK-'94.

3. Results of model imtercomparisons carried out by 3 teams. These included
Model types (Institute for Physical and Theoretical Chemistry, Frankfurt),
Model products (PIK), and results for selected biomes (Woods Hole Research

4. Working groups on NPP results along climatic transects (N-S, E-W USA,
Sahel), NPP in areas of specified vegetation types (OTTER transect, Boreal
forest, grasslands), phenology and growing season length predictions, and
comparison of modelled and field measurements of NPP.

5. Discussion of presentation of PIK '94 and PIK '95 results; publications
and GAIM Science conference poster.

6. Global Primary Production Data Initiative (GPPDI) description and progress.

Models were compared on the basis of two primary salient criteria:
* Types of Models
* Types of Biomes
Several issues arose on the basis of the model intercomparisons.
* CASA and SDBM had contrasting results which may have been caused by FPAR differences. It must be determined which is the correct FPAR to use, and if FPAR can account for the entire difference between the model reuslts.
* Why are taiga and tundra NPPs similar despite differneces in FPAR?
* Significant differences were revealed between forest-grassland transitions in different models, and between forest-forest transitions (N-S transects) as well.
* Is seasonal timing related to water residence and depletion rate? Is it more sensitive to monthly mean vs. start of month? The seasonal timing variations are large between models and may not be fully acounted for.

Input datasets-
In addition to comparing model results, the participants assembled and compared an extensive number of input datasets during and after Potsdam '94. The standard data sets included climate, soil texture, solar radiation, a weather generator, ISLSCP Initiative 1 CD-ROM AVHRR data. This intercomparison of input data set the stage for a set of "Standard Experiments" which were a new contribution of Potsdam '95. In this exercise, there was a standaad data base for temperature, precipitation, solar irradiance and soil texture, a weather generator, AVHRR, and NDVI. Differences still exist in vegetation cover, LAI, and the parameterization of the models for processes such as decomposition, nutrient cycling, and evapotranspiration.
The workshop was divided into three working groups:
1) NPP results along climatic transects (N-S, E-W USA, Sahel), and at individual sites;
(Berrien Moore, Alex Haxeltine, Bill Parton, Annette Schloss)
The developing set of IGBP terrestrial transects offer an excellent opportunity to validate simulations of NPP from the various global models. The transects are defined as a series of study sites distributed along an underlying gradient of a major environmental factor (e.g, temperature, precipitation) that influences ecosystem structure and function. The spatial context in which the study sites are located is also important. In most cases the underlying environmental gradients are sufficiently orderly to support an approximately linear gradient in physical space, and are physically continuous. The length of the transects has been specified as of the order of 1000 km, to ensure that they overlap sufficiently with the spatial realm of global climate, atmospheric, and biospheric models. The transects do not have a prescribed width, but it is often fixed to facilitate the application of remotely sensed data. An important part of transect studies is the extrapolation of the understanding gained from intensive patch-scale studies at key points along the transect to landscape and regional levels.

2) NPP in areas of specified vegetation types (OTTER transect, forests, grasslands);
Forests- (Rolf Otto, Steve Running, Blandine Lurin)
Oregon- the available 4 transect is too short and the climate smoothing is too coarse for NPP model intercomparison needs. Orographic functions will allow various parameters as well as NPP to be corrected for altitude.
Boreal- There is a problem in the transition from water-limited to energy-limited systems because LAI is formulated by some models using water and fails when the system is energy limited as in Boreal biomes. If LAI is set too high, it does affect NPP estimates, as do hydrologic imbalance, respiration, and other factors. These problems need to be addressed and models need to treat water and energy consistently.

Grasslands- (C. Field, Kicklighter, Plochl)
Grasslands are characterized by high water use efficiency (C4 vs. C3). Without the need to support woody tissue, there is hgihly efficint NPP. Grasslands are generally water-limited. Once there is excess water, the biota has the luxury to begin investing in wood, etc. There is low transpiration in places and times and high evaporative demand.
* Many areas with low NPP/precip. ratios are now used for agriculture.
* All maps used for average NPP map used potentiali vegetation EXCEPT SDBM.

Sites - (Steve Prince, Dick Olson, Dork Sahagian)
There are significant differences between NPP model results even when considering the same specific site locations. This was revealed after matching model results from sites in several contrasting regions and ignoring sites that were not common to all model results. Differences in lat/long grid cell definitions were corrected for in the analysis.

3) Phenology- (J. Collatz, E. Kohlmaier, A. Fischer)
Phenology and growing season length predictions were considered, and comparisons were made between modelled and field measurements of NPP for the various models. Both phase and amplitude of seasonality were compared. The strategy was to consider four selected gridcells (consistent vegetation type) in Europe and Central Africa in terms of NPP, FPAR, and LAI. Sites were chosen for temperate deicuous forest (50N 12E), temperate mixed forest (50N 7E), tropical grassland (11N 20E), and tropical forest (1N, 27 E). Models results were compared to each other as well as to FSIR and NDVI. Seasonality in NPP and NDVI does not match between models potentially because:
1. NDVI does not reflect acute stress in NPP.
2. Mismatches exist between vegetation classification and actual vegetation.
3. Characteristics of FASIR NDVI may render them non-representative. Examples are monthly compositing, averaging actual vegetation to 1x1 vs using homogeneous potential vegetation, non-linear tranformation of NDVI to FPAR, transformation of FPAR to LAI, etc.
4. Temperature-limitation is not adequately accounted for in some models. The match in seasonality between NDVI and NPP is generally better in water limited systems (grasslands) than temperure limited system (temp. deciduous forests).

Workshop Results-
During and after the workshop, NPP was investigated at the global, regional and biome level. These results form the basis for a poster presented at the GAIM Science Conference. At the global level, a map of global NPP mean of all models was generated. In addition, the standard deviation for all models was mapped, and a table of the global NPP value predicted by each model was created.
At the Regional level, a monthly series of maps for each model was generated. The regional analysis highlights the seasonal sensitivity of NPP and the related differences between NPP models.
At the Biome level, a series of plots was generated of average model NPP vs. climate drivers (e.g. temperature and precipitation). It was found that different models are based on different definitions of biomes and steps are being taken to construct a standard set of biomes, independent of model application. In the present analysis, 14 distinct biomes are identified.

Models used in the intercomparison included:

After breakout groups reconvened and presented the results of their comparisons, general discussion included

1. Asessment of reltaive strengths and weakness of various existing NPP models.
2. Ideas on how to improve the models
3. The importance of GPPDI.
4. Collaboration between NPP and IGBP terrestrial transects.
5. Potsdam '96- It was agreed that in a year's time, the problems delineated in Potsdam '95 would have become clear enough that a small working group should convene to run and compare models without the more general format presentations. Also, comparisons with field data sould be made.

Products emerging from the Potsdam '95 MPP Workshop include;
* Posters for GAIM Science Conference
* Full-length journal article(s) for special GAIM issue of Global Biogeochemical Cycles
* On-line NPP data bases (administered at PIK)

The Workshop was followed by an open meeting of the GPPDI Steering Committee. The goal of GPPDI is to provide the parameterization and validation data needed in support of modelling global primary production and other applications. Main topics discussed were as follows:

1. The usefulness or otherwise of individual site data for the purpose of parameterizing and validating global grid cell based models. Inclusion of further, existing compilations of NPP data for individual sites in GPPDI had strong support, but doing this quickly depends on funding. ORNL were originally earmarked to do this type of work but they have run out of funds for 1995 and none of the other constituent research groups have yet received funds. Various alternative methods of acquiring these data were discussed including engaging expert consultants for specific biome types, focussed workshops, and an individual to continue the approach used for grasslands by ORNL. Each of these will require some small funding support from DIS, since none of the constituent groups is able to use their research funds for this. The immediate need is to maintain the literature search and personal contacts to find individual site data sets, as has been started for grasslands by ORNL. Dr. Jonathan Scurlock is currently available for the next 6 months and could continue the work from his office at King's College London, UK. The anticipated approximate cost to retain Dr. Scurlock would be $20,000 for 4 months, by which time ORNL expect to receive funds from NASA for FY 1996. It was noted that the work to be done to take GPPDI to a suitable point might be greater than the 8.0 person-years estimated at the Paris 1995 meeting.

2. It was noted with pleasure that CESBIO, France have identified funds from CNRS and have appointed a post-doctoral assistant (Sofie Moulin) to start work on the NPP of agriculture in Europe and N. America.

3. PIK are waiting for the assistance of the Director of DIS to identify funding for their component of the project, but the Director had no progress to report so far.

4. ORNL and UMCP are awaiting the response of NASA to their two cross-referenced unsolicited proposals sent to the Ecosystem Processes and Modeling branch in May.

5. ORNL have developed a global grasslands data base and have set up a World Wide Web site which contains the information. They request comments on the www data base before it is expanded with additional data sets. Their funding for FY 1995 is exhausted and anticipate a gap in activity until new funding arrives, possibly by the end of 1995.

6. The need to indicate data quality or at least the degree to which this can be determined. It was decided to provide a quality index as a summary of the metadata to aid users who are disinclined to make their own judgements.

7. The issue of the vast quantity, but doubtful utility, of forestry inventory and rangeland production data was discussed again.

8. The issue of the minimum size of areal units for model validation was briefly discussed. As suggested by S. D. Prince in the GPPDI presentation, this might be significantly less than 0.5 x 0.5 degree, but would depend on the heterogeneity of the landscape, the climate and other controlling variables.

9. The offer by GCTE to contribute NPP and other data sets for their proposed and existing transects, and the suggestion to approach the USA LTERS to request similar information were noted.

10. Next meeting of the Steering committee was planned for PIK '96 (June 1996). The Project group (PIs of the 4 contributing projects) would meet in early 1996 to coordinate the program. Some resources may be needed to facilitate this meeting which would ideally be held at the DIS office.

Data Management- PIK and DIS have created a World Wide Web site at PIK which contains the GAIM-DIS-GCTE NPP model intercomparison data sets, results of model runs and descriptive materials on each of the models. The need for continued scientific management of the data base is clear; maintenance by communications engineers is provided by PIK, but there is no provision for the scientific management needed to make necessary changes, support preparation of the planned publications, incorporate remaining data sets and make changes as models develop, to correct errors in the data base and to harmonize the model descriptions. It was estimated that 50% of a person's time is needed to perform this function - duties to include correction of errors in the data base, addition of new data sets, preparation of summaries for presentations and papers, harmonization of the model descriptions and preparation for future NPP workshops. PIK may be able to offer temporary help but, in the medium term, new funds will be necessary to provide this function. The funds for this might be requested in a modified PIK proposal for involvement in GPPDI so 1.5 people would work on GPPDI and 0.5 on the data base, thus expanding the GPPDI effort as suggested by the Steering Committee. The source of funds for PIK were to be sought by the Director of DIS, but a direct approach to the German Government by W. Cramer may also be possible. Another possibility would be the removal of the data base to the GAIM office in University of New Hampshire and the provision of scientific support for the data base by GAIM funds, however it would be undesirable to move from PIK having once established it there. Another possibility would be to move the data base to the DIS office if and when it is established in Toulouse and to provide for the data base management through the existing source or some other source of funds.

* G. Churkina, G. Colinet, J. Collatz, W. Emanuel, G. Esser, C. Field, A. Fischer, A. Friend, A. Haxeltine, M. Heimann, J.Hoffstadt, C. Justice, J. Kaduk, L. Kergoat, D. Kicklighter, W. Knorr, G. Kohlmaier, B. Lurin, P. Maisongrande, P. Martin, R. McKeown, B. Meeson, R. Olson, R. Otto, W. Parton, M. Pleochl, S. Prince, J. Randerson, B. Rizzo, A. Ruimy, S. Running, B. Saugier, A. Schloss, J. Scurlock, P. Warnant, U. Wittenberg