Annual report ids

Doc File 3,544.00 KByte,

1. Introduction

The long-term objectives of this EOS-IDS project are multifold: (1) to improve our fundamental understanding of the biophysical and ecological processes governing the linked exchanges of energy, water, carbon and trace constituents between the terrestrial biosphere and atmosphere; (2) to provide the scientific basis for interpreting EOS and other satellite observations of the biosphere, the atmosphere and their interactions; and (3) to carry out credible projections of biosphere-atmosphere interactions for continuing greenhouse gas emissions and land use modification. To accomplish the goals, we bring together all approaches, from theoretical remote sensing, local field programs, tower measurements, to global forward modeling of the atmosphere and biosphere and inverse modeling of the carbon exchanges.

The tables below summarize project personnel and the institutions involved.

Table 1: Tasks and people at work on this project

|Task |Lead |Contributing |

|1. GCM development |Randall |Berry, Denning, Fung, Sellers |

|2. Biosphere model development |Berry/Field |Colla Denning Field, Randall, Sellers |

|3. Tracer model and carbon cycle |Fung |Denning, Field, Randall |

|4. Satellite data analysis |Tucker |Collatz, DeFries, Field, Fung, Mooney, Sellers |

|5. Canopy biophysics |Sellers/Collatz |Berry, Field, Ustin |

|6. Algorithm development |Ustin |Collatz, Field, Tucker |

|7. Biogeochemical models |Mooney |Field, Fung, Matson |

|8. Local model evaluation |Matson |Berry, Field, Vitousek |

|9. Mesoscale model development |Denning |Berry, Randall |

Table 2: Institutions involved in this project

|Lead Investigator |Institution |

|Randall, Denning |Colorado State University |

|Sellers |NASA Johnson Space Flight Center |

|Tucker |NASA GSFC |

|Fung |UC Berkeley |

|Berry, Field |Carnegie Institution of Washington |

|Mooney, Vitousek, Matson |Stanford |

|Ustin |UC Davis |

In FY98, our accomplishments include:

• Innovative development of the CSU GCM, including improved soil and snow models

• Inference of interannual variations in growing season length, photoysnthesis and respiration from atmospheric CO2 variations

• A new estimate of the contemporary deforestation source of CO2 using high-resolution Landsat data and NDVI-driven biomass estimates

• A historical analysis of deforestation source of CO2 using present-day NDVI-based land cover maps and potential vegetation maps

• Analysis of the climatic implications of interannual NDVI variations

• New self-consistent biophysical and biogeochemical model of climate, carbon and water cycling by the biosphere

• New mechanistic model of carbon/nutrient cycling and trace gas fluxes

• Innovative measurements and modelling of stable isotopes relevant to the carbon cycle

These and other accomplishments are detailed below. The group home page is at . We have had a very busy, exciting and rewarding year.


2.1 Progress and accomplishments

2.1.1 Flux Coupler

A flux coupler has been developed for the GCM that enables each of the climate system components to be run with spatial resolutions and time steps independent of the others. Previously, all surface processes were run with the same spatial resolution as the atmosphere, typically 4 degrees latitude by 5 degrees longitude. This had forced us to assign one biome type to a 4x5 grid cell, although in reality there is considerable heterogeneity in the biomes at that resolution. The flux coupler enables SiB 2 to be run at a resolution consistent with the scale of the available data. The flux coupler also permits the atmosphere model to be coupled to dynamical ocean models. Coupling with the POP 4/3 degree global ocean and a tropical Pacific basin model have been tested.

Briefly, in the flux coupler there are two independent grids, the atmospheric grid and the surface grid. The grids may be identical, or they may differ, in which case, one of the grids must be logically rectangular (representable as a two-dimensional array). Atmospheric forcing variables and surface fluxes are interpolated between these two grids. The interpolations are conservative and second order accurate, using a package provided from the LANL ocean modeling group. The grid cells of the surface grid can be aggregated so the calculations of the surface processes are performed once for the aggregated cell. The surface processes and the PBL physical parameterization (entrainment, TKE, and PBL top stratus clouds) are all performed on the surface grid. The temporal coordination of the atmosphere and surface is synchronous, and the most recent values of forcing and fluxes are communicated. The flux coupler also allows portions of the climate system to be replaced with data for forcing other parts of the system, e.g. SibDRV, where SiB2 is run with prescribed meteorological forcing. Thus, the flux coupler has merged previously separate methods of running SiB2 into one code.

A simulation has been done with 1x1 resolution prescribed sea surface temperatures and sea ice cover coupled to a 4x5 atmosphere. SiB2 was also aggregated to 4x5. It shows that the heterogeneities of the ocean surface can contribute to significant differences in surface heat fluxes compared to a 4x5 resolution control. This is especially apparent in regions of strong temperature gradient, such as the western boundary currents, and sea ice edges.

We expect land surface heterogeneities to similarly modify the surface fluxes of heat and CO2. A simulation with 1x1 resolution SiB2, with data from the 1x1 ISLSCP datasets is being run.

Figure 2.1: January prescribed FPAR for a) 4x5 degree resolution surface, and b) 1x1 degree resolution surface. In both cases, the atmosphere is run at 4x5 degree resolution.

Figure 2.2: January evaporation (mm/day) for a) 4x5 degree resolution surface, and b) 1x1 degree resolution surface. In both cases, the atmosphere is run at 4x5 degree resolution.

2. New soil and snow models

The soil and snow models of SiB have been updated. The soil/snow surface temperature is now an energy balance skin temperature rather than a bulk temperature for a surface layer. The simple one-layer patchy snow model has been replaced with a 3 layer snow model with prognostic temperature and density, and includes the retention of snow melt water and accounts for the thermal effects of snow melt drainage. Though a uniform snow layer, the effects of patchiness are parameterized in the snow albedo. At a certain snow depth limit, the 3 layer treatment reverts to one layer.

The 6 layer soil thermodynamic model has been extended to 10 layers. The thermal effects of draining water are also accounted for. The 3 layer soil hydrology has been replaced with a 10 layer model that includes melt/freeze effects. Instead of one root zone, roots now have a multiple-layer distribution, and the soil moisture stress calculation is modified accordingly.

We are beginning testing of the new snow and soil. SiB will be driven with prescribed meteorological data at specific sites where there is validation data such as FIFE and BOREAS. This will help us choose certain parameter values in the code. We will also test the new code with global SiBDRV, looking at snow cover and river runoff data.

2.1.3 A geodesic grid for the atmosphere model

An effort is underway to replace the formulation of the hydrodynamics in the atmosphere model. Currently, the hydrodynamics are computed using prognostic equations for the momentum. These are discretized on a regular latitude-longitude rectangular grid, with energy and enstrophy conservation, and with the mass and the momentum variables staggered on a C grid. Some of the shortcomings of this method include: the need to filter the equations near the poles where the meridians converge due to numerical stability criteria; two delta-x noise in the horizontal; and the inability of this discretization to satisfactorily represent gravity wave dispersion.

The new formulation of the hydrodynamics of the atmosphere is based on the discretization of the vorticity and divergence equations on an unstaggered geodesic grid based on the icosohedron. The grid cells are all hexagons except for a small number of pentagons. This grid requires no filtering since all grid cells are approximately the same area and it better represents the gravity wave dispersion relation. This formulation has been tested in a three-dimensional dynamical core and has performed better than the current atmospheric dynamics. Testing of the new hydrodynamics in the GCM with full physical parameterizations should be underway shortly.

With the new atmospheric grid, we have the option to run SiB 2 on an identical geodesic grid, or to run it on any rectangular grid with the flux coupler governing the interpolation between the two grids. One test of the new grid will be to run SiB 2 on the geodesic grid with prescribed meteorological forcing, verifying the flux coupler interpolations.

2.2 Publications:

Bounoua, L., G. J. Collatz, P. J. Sellers, D. A. Randall, D. A. Dazlich, S. O. Los, J. A. Berry, I. Fung, C. J. Tucker, C. B. Field, and T. G. Jensen, 1999: Interactions between vegetation and climate: Radiative and physiological effects of doubled atmospheric CO2. Journal of Climate (in press).

Bounoua, L., G. J. Collatz, S. O. Los, P. J. Sellers, D. A. Dazlich, C. J. Tucker, and D. A. Randall, 1999: Sensitivity of climate to changes in NDVI. Submitted to J. Climate.

Eitzen, Z., and D. A. Randall, 1999: Sensitivity of the simulated Asian summer monsoon to parameterized physical processes. Submitted to the Journal of Geophysical Research.

Zhang, C., D. A. Dazlich, and D. A. Randall, 1999: Simulation of soil moisture and surface water blanace using the Simple Biosphere Model 2. Journal of the Meteorological Society of Japan (in press).

2.3 Plans for 1999:

• Complete implementation of the new model based on the geodesic grid.

• Perform extended integrations with the new model including coupled ocean-atmosphere integrations.

• Implement isotope tracers in the model.

2.4 Personnel changes:

Changan Zhang left the project in May 1998.

3. UCSB/CSU – Scott Denning

3.1 Progress and Accomplishments

3.1.1 TransCom

The Atmospheric Tracer Transport Model Intercomparison Project (TransCom) has continued to develop. We have completed the analysis of the mixing ratio output from the Phase 2 (SF6) experiments, and have submitted a manuscript describing the results. The models were reasonably successful in simulating the meridional gradient of SF6 in the remote marine boundary layer, but had more difficulty in matching the observations at continental locations near emission sources. Overall, the differences among models was dominated by differences in parameterized vertical transport rather than by differences in resolved advection. More “convective” models (including the CSU GCM, GISS, and GISS-UVic models) underestimate SF6 over the north Atlantic, but are more successful at simulating continental mixing ratios. Less convective models perform better at remote marine locations, but systematically overestimate SF6 near source regions.

A third phase of TransCom is now being planned, which will involve full inversion calculation of the global CO2 budget by about a dozen tracer transport models. A workshop was held in Santa Barbara to draft an experiment plan, and a kick-off workshop will be held this fall in Berkeley. This activity will be coordinated at CSU. Recent inversion The experiment has been structured to allow maximum participation with low entry barriers, with a flexible design that allows groups with significant expertise and resources to perform more detailed calculations.

3.1.2 Stable Isotopes

Work continues on the implementation of stable isotopes of CO2 in SiB2 and the CSU GCM. We sponsored a visit by Philippe Peylin at UCSB, during which preliminary implementation of the online fractionation of 18O in CO2 was performed. This code uses prescribed δ18O of water in precipitation and water vapor, and calculates the isotopic ratio of water in soil and plant reservoirs, using these to calculate δ18O of CO2. Preliminary tests with this code show a decent match to observed ratios for the BOREAS site in midsummer, but winter and springtime values are poorly simulated due to continuing problems with soil temperature and snowpack in SiB. We plan to collaborate further with Peylin and Ciais on this work, and hope to have a working global code by the middle of 1999. We are recruiting a research associate to implement online calculation of δ18O of water in the CSU GCM and SiB2. This person will join the group in the fall of 1998, and will be funded separately from this project.

In collaboration with Sarmiento’s group at Princeton, we have obtained a map of simulated air-sea CO2 and 13C flux, which we are now using to drive a global atmospheric tracer simulation. This will be combined with a tracer calculation using the terrestrial 13C fluxes derived by Fung et al (1997) using SiB2 and CASA, and compared to similar simulations performed by the Princeton group using the SKYHI model.

Figure 3.1 The distributions of SF6 mixing ratios at the surface, as simulated by the models participating in TransCom2. Departures from observations are indicated by the triangles.

Figure 3.1 (continued)

3.1.3 SiB2-RAMS

We have successfully coupled SiB2 to the CSU Regional Atmospheric Modeling System (RAMS), and have tested the coupled model in both mesoscale and large-eddy simulations. This work (sponsored by DOE-NIGEC) will allow the IDS team to address questions of land-atmosphere interactions on a new range of spatial scales. We have focused on the analysis of forest-atmosphere coupling at the WLEF-TV tower in northern Wisconsin, where we have three years of continuous flux data and vertical profiles of scalar quantities to analyze. Research priorities include the local rectifier forcing, the relationship of ecosystem stress to turbulence and to boundary-layer clouds, and the development of methods to estimate regional-scale fluxes from boundary-layer concentrations.

SiB2-RAMS can be run on nested grids, with the outermost grid forced from gridded analyses, with inner grids allowing more detail at a study site. We have begun using this technique to analyze case studies at the WLEF tower. A priority for IDS involvement in this work will be to develop mesoscale data products from remote sensing inputs, which we might call “meso-mapper.” This work will require close collaboration with the Goddard and Maryland teams.

3.2 Publications

Peylin, P., P. Ciais, P. P Tans, K. Six, J. A. Berry, and A. S. Denning, 1997. 18O in atmospheric CO2 simulated by a 3-D transport model: A sensitivity study to vegetation and soil fractionation factors. Physics and Chemistry of the Earth, 21, 463-469.

Berry, J. A., G. J. Collatz, A. S. Denning, G. D. Colello, W. Fu, C. Grivet, P. J. Sellers, and D. A. Randall, 1997: SiB2, a model for simulation of biological processes within a climate model. In: P. van Gardingen, G. Moody, and P. Curran. (Eds.), Scaling Up, Society for Experimental Biology, Cambridge University Press, 347-370.

Pielke, R. A., R. Avissar, M. Raupach, H. Dolman, X. Zeng, and S. Denning, 1998. Interactions between the atmosphere and terrestrial ecosystems: influence on weather and climate. Global Change Biology, 4, 101-115.

Denning, A. S., M. Holzer, K. Gurney, Y. Balkanski, S.-M. Fan, P. Friedlingstein, I. Y. Fung, M. Heimann, R. Law, M. Maiss, P. Rayner, S. Taguchi, J. Taylor, 1997. Global tracer transport simulations of SF6 : Model intercomparison. Presented at 1997 Fall Meeting of the American Geophysical Union A11C-04.

Denning, A. S., M. Holzer, K. R. Gurney, M. Heimann, R. M. Law, P. J. Rayner, I. Y. Fung, S.-M. Fan, S. Taguchi, P. Friedlingstein, Y. Balkanski, M. Maiss, and I. Levin. Three-dimensional transport and concentration of SF6: A model intercomparison study (TransCom 2). Submitted to Tellus.

Denning, A. S., T. Takahashi and P. Friedlingstein. Can a strong atmospheric CO2 rectifier effect be reconciled with a "reasonable" carbon budget? Submitted to Tellus.

3.3 Plans for 1999

We will continue the work we've been doing in the past on stable isotopy in SiB and the GCM, on inversion intercomparison, and on local to regional scaling at the tall tower site. These projects are now or soon will be supported largely through funding outside the EOS-IDS project. We would like to use IDS support for two primary activities in 1999: (1) to develop a "meso-mapper" system that produces SiB2 boundary conditions at high spatial resolution from AVHRR, SeaWiFS, and MODIS input, which can be used in the SiB2-RAMS model; and (2) to work with the Carnegie teams on the development of the next-generation SiB3, in particular on the belowground modules. This will require a higher level of funding than we now receive.

4. Personnel Changes

Lara Prihodko is a Ph.D. student with expertise in land-surface modeling and remote sensing, and GIS. She has joined the team, and is supported by her own EOS Fellowship. Dr. Pier-Luigi Vidale, a mesoscale modeler with expertise in RAMS has joined the team. John Kleist is a scientific programmer and systems administrator. Connie Uliasz is a Research Coordinator responsible for technical graphics and editing and Web development. Vidale, Kleist, and Uliasz are not currently receiving any support from the IDS project, but we request a small increment in next year's budget (see below) to support a portion of their IDS-related work.

The entire team has relocated from UC Santa Barbara to Colorado State University.

4. NASA/GSFC - Jim Collatz, Piers Sellers, Jim Tucker

4.1 Progress and Accomplishments

The Goddard Modeling Group continues to focused on vegetation - climate interactions at continental to global spatial scales and seasonal to decadal time scales using i)satellite derived vegetation parameters from the IDS Team's Goddard Remote Sensing Group (Tucker, Los) and ii)the CSU SiB2-GCM climate model supported by the CSU Climate Modeling Group (Randall, Dazlich, Denning). Our work emphasizes the role of variability in the activity of vegetation (seasonal,interannual,spatial) as a factor in determining climate and carbon cycle variability. Two specific topics of interest this last year have been:

• The role of the diurnal and seasonal cycles in vegetation activity in determining the diurnal temperature range.

• The impacts of interannual variability in global vegetation cover on climate and productivity.

These studies are summarized below.

4.1.1 Diurnal Temperature Range (DTR)

Surface measurements of the diurnal temperature range during this century reveal that there has been a general increase in the mean and minimum temperatures(Tmin) while maximums (Tmax) have remained relatively constant resulting in a decreasing DTR trend over the last century. Previous modeling efforts - usually involving an increase in radiative forcing via elevated CO2

- have attributed cause to cloudiness, water vapor, aerosols, Clausius-Clapyeron dependence of evapotranspiration and soil moisture.

These past studies have used land surface models of various levels of complexity ranging from the simple bucket type to the more complex biophysical model BATS. The potential impact of land surface parameterization on the simulated DTR is substantial: i)Bucket models will tend to evaporate excessively at night because diurnal physiology is not included causing lower

minimum temperatures and larger DTRs (Figure 4.1b), ii)Biophysical models can exhibit unrealistic stresses because of miss-match between climate and prescribed parameters producing higher maximums and larger DTRs in response to increased radiative forcing (Figure 4.1c).

In addition to modeling studies we are also comparing multi-year time series observations of Tmax, Tmean, Tmin, surface short wave and NDVI. We have confirmed the existence of a late spring discontinuity (plateau) in the winter to summer upward march of the DTR in temperate regions that is associated with the amount and timing of vegetation cover. Comparisons with climate model results imply that increased dissipation of net radiation via latent heat flux

at the onset of the growing season contributes to the development of the DTR plateau.

Figure 4.1. Impact of canopy physiology on the diurnal temperature range. Simulations and measurements are for a rainless three-day period in Manaus, Brazil. SiB2 was parameterized for tropical rain forest and forced with observed and modified meteorological data. A) Time course of latent heat flux as measured by eddy correlation techniques and as simulated by SiB2 driven with the observed meteorological data. Note that in this forest transpiration dominates the total latent heat flux and that the simulated fluxes are close to observed fluxes. B) The change in latent heat flux and temperature caused by altering the observed meteorological driver data to produce plausible conditions for a doubled atmospheric CO2 scenario. Increasing down-welling longwave radiation (14 W/m2) and air temperature (1.7K) caused latent heat flux to increase but only during the day leading to lower maximum temperatures and DTRs. C) Simulations in which stomatal resistance remains fixed as in so call “bucket models” produced lower minimum temperatures and larger DTRs in response to the imposed hypothetical forcing. On the other hand, under severe physiological stress, the forcing scenario caused even greater stress during the day leading to higher temperature maxima and DTRs.

4.1.2 Impact of Observed Interannual variability of Vegetation on Climate

Satellite data from the 1980's indicate that globally, 1989-1990 had higher vegetation cover than the years 1982-1983. This difference is apparent in the FASIR products we use as boundary conditions for the vegetation in CSU SiB2-GCM. Figure 4.2 shows the difference in FASIR NDVI during June through August for those two time periods. Large areas of the Northern Hemisphere boreal forest show increases in NDVI of around 10%. Two 10 year simulations

were conducted using these two data sets as boundary conditions.

Results show that on average, gross carbon uptake by vegetation went up by about 1 PgC/year (Figure 4.3) compared to a total flux of over 100 PgC/year. Year to year variations of about 0.3 PgC/year are caused by the vegetation's response to the climate model's own internal variability. In general during the growing season, areas with more vegetation were cooler. Global mean

surface temperature is highly variable from year to year but there is a tendency towards a cooler climate with the higher vegetation coverage of the 1989-90 period relative to 1982-83 period.

4.1.3 Effects of Land Use Change on Climate

In collaboration with Ruth DeFries of UMDCP we are developing vegetation parameters for hypothetical global vegetation undisturbed by human activities. A first effort merges a literature survey based map and a satellite based map together with a set of rational rules for classifying each grid cell according to changes due to human use. Once classified, the time course and amplitude of seasonal changes in cover are prescribed by sampling neighboring unchanged

grid cells of the same type in the satellite based map. Multi-year equilibrium climate simulations for current and undisturbed vegetation scenarios are being run and results will be forthcoming by the next report.

4.2 Publications

Dang, Q-L, H.A. Margolis, S. Mikailou, M.R. Coyea,G.J. Collatz and C.L. Walthall. Profiles of photosynthetically active radiation, nitrogen and photosynthetic capacity in the boreal forest: Implications for scaling from leaf to canopy.1997 Journal of Geophysical Research 102: (D24) 28845-28859 DEC 26 1997.

Dang, Q-L, H.A. Margolis and G. J. Collatz, Parameterization and testing of a coupled photosynthesis-stomatal conductance model for boreal trees. Tree Physiology, 18, 141-153, 1998

Bounoua, L., G.J. Collatz, P.J. Sellers,D.A. Randall, D.A. Dazlich, S. O. Los, J.A. Berry, I. Fung, C.J. Tucker,C.B. Field and T. G. Jensen. Interactions between vegetation and climate: Radiative and physiological effects of double atmospheric CO2. Journal of Climate (in press)

Collatz, G. James, J.A. Berry and J.S. Clark. Effects of climate and atmospheric CO2 partial pressure on the global distribution of C4 grasses: Present, past and future. Oecologia 114, 441-545, 1998.

Doran JC, Hubbe JM, Liljegren JC, Shaw WJ, Collatz GJ, Cook DR, Hart RL, A technique for determining the spatial and temporal distributions of surface fluxes of heat and moisture over the Southern Great Plains Cloud and Radiation Testbed. J Geophys Res 103, 6109-6121, 1998

Kaufman YJ, Herring DD, Ranson KJ, Collatz GJ. Earth Observing System AM1 mission to Earth. IEEE Transactions of Geoscience and Remote Sensing 36,1045-1055.

Los SO, Sellers, PJ, Collatz GJ, Malmstrom CM, DeFries RS, Tucker CJ, Pollack NH, Dazlich DA, Bounoua L, Randall DA. A multi-year, global, monthly, 1 x 1 degree biophysical landsurface data set for climate studies derived from NOAA AVHRR data. Journal of Geophysical Research (conditionally accepted)

4.3 Plans for 1999

• Finish analyses of simulations and observations of the DTR and prepare publications.

• Continue work on impacts of interannual variability in global vegetation on climate using the FASIR NDVI times series and ensembles of GCM realizations.

• Run climate simulations using anthropogenic vegetation change scenarios as boundary conditions to test the hypothesis that such changes have influenced climate and primary roduction.

• Participate in the development of SiB3 along with other members of the IDS

• Work with Yongkang Xue of University of Maryland, College Park and Randy

• Koster of GSFC in two separate efforts to implement our canopy photosynthesis/conductance models within their land surface models. Both groups want our more mechanistic stomatal conductance model that uses the satellite products our IDS Team develops.

4.4 Personnel Changes

Sietse Los completed his PhD, and is a post-doctoral fellow associated with the project.

5. University of California, Berkeley – Inez Fung

5.1 Progress and Accomplishments

1. Deforestation (with C. Field and J. Tucker)

We have produced a new estimate for the deforestation source of CO2. The new estimate is based on high-resolution Landsat data, and CASA estimates of carbon inventories and fluxes. Our estimate is the first independent investigation of this source other than R. Houghton’s efforts. Our estimate is much lower than Houghton’s estimates – by at least a factor of 2. This would give an upper bound of 1.0 Pg C/yr to the deforestation source for the decade of the 1980’s. This compares to Houghton’s estimate of 1.6 PgC/yr.

Our lower estimate comes principally from the lower deforestation rate estimated from the Landsat data. The high-resolution Landsat data shows clearly that intense deforestation is concentrated in very small areas. Unless there is wall-to-wall coverage, any statistical means to estimate country-wide deforestation rate from a few sampled areas are likely to be biased.

Our analysis of the carbon dynamics using CASA shows also that the parameters in the Houghton model are very simplistic. In NPP, the NPP distribution reflects climate and NDVI variations. We converted NPP into biomass using turnover times for the live leaf, root and wood pools. The biomass distribution for the Amazonian rainforest is shown in Figure 5.1. It turns out that the deforested areas (according to Landsat data) are the regions whose potential biomass is 10-20% higher than the Amazonian average.


Our new estimate of the deforestation rate and biomass suggests that a lower (x60%) cumulative carbon emission than that estimated by Houghton.

For tropical rainforests, woody biomass comprises >90% of the total above ground biomass (see also Chris Field’s section on Biomas Allocation in this report). The fate of coarse woody debris dominates the timing of CO2 release to the atmosphere after deforestation. Since coarse woody debris has lifetimes of ~5 years in the tropics, the Houghton carbon model (not the areal estimates) is probably useful for decadally averaged carbon emissions. Amazonian deforestation rates in the 1990’s have shown large variations from year-to-year, because of fluctuations in the Brazilian economy. We are using CASA to produce a year-by-year estimate of carbon release from Amazonian deforestation for the 1990’s . This would be very useful for interpreting the atmospheric CO2 balance for any year or any 5 year period.

2. Atmospheric Transport

In the TransCom2 simulations of SF6 (see Scott Denning’s report), we are pleased that our new tracer transport model – termed the GISS-UVIC model – shows dramatic improvements over the original GISS tracer model. This is not to be unexpected, since the two models are based on the 1983 and 1997 versions of the GISS GCM, respectively. The new model will be the workhorse of all new tracer modelling work, while the 1983 version (used in the first 3D carbon simulation) goes into golden retirement. The new model is available to all team members for their investigations.

We have also run the GISS GCM for the period 1979-1994. For a seasonally invariant tracer, e.g. SF6, the year-to-year variations of the interhemispheric transport time (τ) varies by <15%, and the variations are related to the southern oscillation index. For a seasonally tracer, such as CASA monthly fluxes, the rectifier effect produced by the new tracer model is ~2 ppm, comparable to that produced by the CSU GCM. Interannually, the rectifier effect may vary by a factor of 2. This is non-trivial in the inferring sources/sinks from a few years of atmospheric CO2 data. We are in the middle of analyzing the model results to identify the processes for this large variation.

3. Carbon Studies (with C. Field, J. Randerson)

While we were the first to propose a northern hemisphere terrestrial carbon sink nearly a decade ago (Tans et al., 1990), we do not believe that the atmospheric CO2 data currently available are adequate to partition the sink uniquely between North America and Eurasia. There are a multitude of possible solutions that would match all the atmospheric CO2 observations. We have therefore carried out research to use C-13 as a constraint (Fung et al., 1997) and have sought the terrestrial sink signature in other aspects of the atmospheric CO2 data.

Jim Randerson has focused on using the trend in the amplitude of CO2 oscillations to infer changes in carbon processing by the biosphere. This work is described in Chris Field’s section of this report.

5.1.4 Iron

We initiated an investigation of the efficiency of iron utilization in ocean productivity. The study follows on our previous work (with I. Tegen) on the mineral aerosol cycle in the atmosphere. Dust deposition at the ocean surface is translated into iron availability via assumptions of the fractions of soluble iron in dust. The aeolian source is combined with the upwelling and riverine sources to estimate new production. Primary productivity in the ocean based on satellite data (see Chris Field’s report) is converted into iron utilization. The recycled source of iron is then estimated as the residual between utilization and new production. Analysis of the degrees of recyling of iron and nitrates lends support to the simple approach to modelling the iron cycle. The model produces some familiar features: in the known high-nitrate-low chlorophyll (HNLC) areas of the equatorial Pacific and southern oceans, iron recycling is high while nitrate recycling is low. In regions with intermediate aeolian deposition, new production and recycling are comparable sources of iron. In high dust deposition areas, there is little need for recycling of iron.

We have found in an earlier study (Tegen and Fung 1995) that approximately half the current loading of dust in the atmosphere is from recent disturbances. This new iron model paves the way for investigating related changes in the ocean carbon cycle and the link to atmospheric CO2 abundances.

1. Publications

Dai, A., A. D. DelGenio, and I. Fung (1997): Clouds, precipitation and temperature change. Nature, 386, 665--666.

Dai, A., I. Fung and A. D. DelGenio (1997): Surface observed global land preciptation variability during 1900--1988. J. Climate, 10, 2943--2962.

Fung, I.(1997): A greener north. Nature, 386, 659--660.

Fung, I., C.B. Field, J.A. Berry, M.V. Thompson, J. T. Randerson, C.M. Malmstrom, P.M. Vitousek, G.J. Collatz, P. J. Sellers, D. A. Randall, A. S. Denning, F. Badeck, and J. John (1997): Carbon-13 exchanges between the atmosphere and biosphere. Global Biogeochemical Cycles, 11. 507--533.

Randerson, J. T., M. V. Thompson, T. J. Conway, I. Fung and C. B. Field (1997): The contribution of terrestrial sources and sinks to trends in the seasonal cycle of atmospheric carbon dioxide. Global Biogeochemical Cycles, 11, 535--560.

Tegen, I., P. Hollrigl, M. Chin, I. Fung, D. Jacob and J. Penner (1997): Contribution of different aerosol species to the global aerosol extinction optical thickness: Estimates from model results. J. Geophys. Res. 102, 23895--23915.

Denning, A.S., M. Holzer, K. Gurney, M. Heimann, R. Law, P. Rayner, I. Fung, S.-M. Fan, S. Taguchi, P. Friedlingstein, Y. Balkanski, M. Maiss, and I. Levin (1998): Three-dimensional transport and concentration of SF6: A model intercomparison study (TransCom2). Submitted to Tellus.

Meyn, S., I. Fung and I. Tegen (1998): Iron supply and utilization in the world ocean: Estimates of annual Fe from atmospheric deposition, upwelling, rivers, and regeneration. Global Biogeochemical Cycles, submitted.

Reader, M. C., I. Fung, and N. McFarlane (1998): The mineral dust cycle during the Last Glacial Maximum. J. Geophys. Res. (submitted).

DeFries, R. S., C. B. Field, I. Fung, J. Collatz, and L. Bounoua. 1999. Combining satellite data and biogeochemical models to estimate global effects of of human-induced land cover change on carbon emissions and primary productivity. Global Biogeochemical Cycles (submitted).

Friedlingstein, P., G. Joel, C. B. Field, and I. Y. Fung. 1999. Towards an allocation scheme for global terrestrial carbon models. Global Change Biology (in press).

Randerson, J. T., C. B. Field, I. Y. Fung, and P. P. Tans. 1999a. Increases in early season ecosystem uptake explain changes in the seasonal cycle of atmospheric CO2 at high northern latitudes. Nature (submitted).

Randerson, J. T., M. V. Thompson, T. J. Conway, I. Y. Fung, and C. B. Field. 1997a. The contribution of terrestrial sources and sinks to trends in the seasonal cycle of atmospheric carbon dioxide. Global Biogeochemical Cycles 11: 535-560.

Randerson, J. T., M. V. Thompson, I. Y. Fung, T. Conway, and C. B. Field. 1997b. The contribution of terrestrial sources and sinks to trends in the seasonal cycle of atmospheric carbon dioxide. Global Biogeochemical Cycles 11: 535-560.

2. Plans for 1999

The development work revolves around modelling C-13 in the ocean. This will be done in collaboration with Scott Doney of NCAR, using the NCAR ocean model (NCOM). The goal is to determine the isotopic disequilibrium associated with air-sea exchange. Once this puzzle is in place, we will carry out a full inversion of atmospheric CO2, using both CO2 and C-13 observations as constraints.

We will also carry out inversion of the year-to-year change in the atmospheric CO2 budget. This will make use of our interannual transport model, interannual deforestation source, interannual biospheric fluxes (from CASA), among other things.

3. Personnel Changes

The group moved from the Universityof Victoria to the University of California at Berkeley in July 1998. The role of PI for the IDS project rotated from Dave Randall to Inez Fung on April 15 1998. Jasmin John also relocated to Berkeley and is the principal guardian and distributor of the codes and datasets with all IDS collaborators. She is also the keeper of the team webpage . Pierre Friedlingstein completed his post-doctoral fellowship, and assumed a permanent position with CNRS in Saclay, France. Jim Randerson, who graduated from Chris Field’s group, joined the Berkeley group as a DOE Hollaender Post-doctoral Fellow. Jim will also be working with Terry Chapin at the University of Alaska, Fairbanks.

6. University of California, Davis – Susan Ustin

6.1. Progress and Accomplishments

The EOS project effort at UC Davis has continued along three lines of work: 1) developing new methods for vegetation assessment using hyperspectral and hyperspatial information from EOS sensors, 2) regional applications of EOS data for assessment of vegetation condition, and 3) improved fundamental understanding of light interactions with leaves and plant canopies.

6.1.1 Spectral detection and quantification of biogeochemical composition

1) Improved accuracy in remote sensing of stress detection, vegetation mapping and other issues requires a better understanding of how scattering and absorption processes control leaf and canopy reflectance. In addition, fundamental questions remain on how the three-dimensional structure of leaves and plant canopies interacts with and affects their physiological functioning. The primary objective is to improve understanding of the interactions between cell structure and chemistry of leaves and canopy and the light environments. In collaboration with Dr. Stephane Jacquemoud, University of Paris 7, we developed the algorithm to produce virtual models of the three dimensional structure of dicot leaves. The model was constructed to allow modification of the anatomy, structure and biochemistry of the leaf cells so the photon absorption, scattering and transmission characteristics of leaf tissues of different morphologies can be studied. We expect to run these models in 1998-1999, using the linked PROSPECT and RAYTRAN models, on the supercomputer at the Joint Research Center, Ispra, Italy in collaboration Yves Govaerts and Michael Verstraete. Specifically, we developed virtual leaf models with the anatomical structures of two poplar clones having different anatomical characteristics, e.g., bilayer and monolayer palisade parenchyma) which are being tested against the field data acquired in 1997 in collaboration with Dr. Thomas Hinckley and Ms. Kim Brown, University of Washington. We have started to compare model predictions using PROSPECT in a canopy reflectance model against calibrated AVIRIS reflectance for these poplar stands and plan to continue this work next year.

2) Work with Dr. Scott Martens, CSTARS, in collaboration with Dr. Jiquan Chen, Michigan Polytechnic University, to compare modeled and measured PAR in the understory of an old-growth Douglas fir forest in the Gifford Pinchot National Forest (Martens et al., in prep.). Washington. All 3610 trees within an 8 ha area (400 m x 200 m) have been location mapped, height and crown dimensions measured, Estimate direct beam photosynthetically active radiation (PAR) at the ground surface throughout the year. The first step to study the impact of the three-dimensional light structure on ecophysiological responses was to develop a spatially distributed map of PAR for vertical and horizontal locations in the canopy. PAR was simulated for specific locations in the canopy using the three-dimensional ray-tracing radiative transfer model, SolTran, modified from PJSol, a model developed by Los Alamos National Laboratory (). In the model, canopy structure was characterized in three-dimensions using ellipsoids to represent each tree crown, using the dataset measured by Dr. Chen for the 8 ha study area who also provided the high resolution (1 m) terrain digital elevation model. Seasonal variation in PAR was simulated at monthly intervals over a year using the ray tracing model and assuming that canopy structure remained constant. The web site shows dynamic simulation results for PAR at different locations over a 24 hr period at the summer solstice and potential daily PAR for a year.

3) We have examined the use of hyperspectral data to detect surface expressions of soil organic matter and iron content in exposed soils from the Santa Monica Mountains. These soil properties were chosen for examination because they are sensitive to land use management and climate conditions, and are critical for determinations of “soil quality.” Analysis of laboratory measured spectra of soil samples showed that the HFBA technique (Pinzon et al., 1998a) could be used to classify soils and characterize the concentrations of these organic matter Palacios Orueta and Ustin, 1998). The vectors obtained from the laboratory study were applied to AVIRIS images of the region and we were able to characterize the areas of exposed soils in the images and concentrations of organic matter and iron (Palacios Orueta et al., 1998a,b).

4) In collaboration with Vern Vanderbilt and Vince Ambrosia, NASA Ames Research Center, we reexamined archived AVIRIS data over a ponderosa pine forest east of Mt. Shasta, California (Vanderbilt et al., 1998b). The over a three hour period, a flightline was repeatedly flown at about 15 minute intervals. The data were acquired at intervals that allowed us to match sun angles before and after solar noon. The data was classified for land cover types, images co-registered, and each type was evaluated for diurnal changes in spectral properties. Despite significant noise in the data, only forested land cover types showed diurnal changes that could not be attributed to factors other then canopy changes, possibly physiological or architecture or both. Changes suggested decreasing canopy water content over the period.

5) Spatial variation in canopy liquid water content was examined in coastal savanna woodlands (Ustin et al., 1998a), chaparral communities (Ustin et al., 1998b; Gamon et al., 1998), and in a poplar plantation of tree clones (Roberts et al., 1998c).

6.1.2 Algorithm Development

We have continued investigations into various types of spectral mixture modeling to address vegetation (and in some cases species mapping). We have continued to develop multiscaling of the hierarchical wavelets/foreground and background analysis and have presented papers and proceedings on this work (Pinzon 1998b-d). These studies include scaling between laboratory, field and image data sets (using Thematic Mapper, POLDER, AVIRIS and other images). The results of this work were presented at several meetings and in publications. We applied the unmixing technique to a POLDER data of the BOREAS study site using multiple view angle bands (Diaz Barrios et al., 1998; Vanderbilt et al., 1998a) to classify wetland land cover types. This approach could improve estimates of boreal wetlands and other land cover types, which could improve model estimates of methane and trace gas fluxes.

6.1.3 Analysis of prototype hyperspatial sensor data.

We have started to work with several one-meter four-band ADAR and three-band Kodak airborne digital data sets to begin to understand spectral mixing and scaling issues. We have examined some methodologies that have potential for application to natural hazards, contamination mapping, agriculture (). We have used AVIRIS to detect spatial variation in crop development in agricultural sites (Green et al., 1998). This work will be continued next year testing the high spatial resolution (4-5m) AVIRIS data over a several crop and soil areas.

6.1.4 Landscape Studies

1) We have continued to apply spectral mixture analysis at a variety of sites. We used linear unmixing analysis on AVIRIS data from Jasper Ridge to show seasonal changes in canopy water content, foliage and litter fractions (Ustin et al., 1998a). We used similar techniques to map boreal forest ecosystems in central Alaska using AVIRIS data and contrasted it with SPOT results (Ustin and Xiao, 1998). Dr. Roberts used a new mixing technique that allows each pixel to find the best fit endmembers and then maps the distribution of endmembers. This approach was applied to mapping chaparral communities in the Santa Monica Mountains (Roberts et al., 1998a), We presented preliminary results for the characterization of biomass from AVIRIS (Roberts et al., 1998b).

2) We have continued the watershed study in the Wind River Valley using the SPLASH model and anticipate producing a model results that will be compared to field data (Ameriflux site) and to SiBII and CASA results, from a related study by Dr. Chris Field and Dr. Joe Berry. This detailed analysis of forest structure derived from unmixing AVIRIS data and from the water content images, is used to provide input parameters into the model.

6.2.1 Reviewed Publications 1998

Diaz Barrios, M. C. S.L. Ustin, G.L. Perry, V.C. Vanderbilt, L.A. Morrissey and G.P. Livingston, F.-M. Bréon, S. Bouffies, M.M. Leroy, M. Herman and J.-Y. Balois. Discrimination of wetland and non-wetland community types with multispectral multiangle polarized POLDER data. Submitted to IEEE Transactions on GeoScience and Remote Sensing (in review).

Pinzon, J.E., S.L. Ustin, C.M. Castaneda, and M.O. Smith. 1998a. Investigation of leaf biochemistry by hierarchical foreground/background analysis. IEEE Trans. Geosci. and Remote Sensing (in press).

Palacios-Orueta, A. and S.L. Ustin. 1998. Remote sensing of soil properties in the Santa Monica Mountains. I. Spectral Analysis. Remote Sensing of Environment 65:170-183.

Palacios-Orueta, A., J.E.Pinzon, S.L. Ustin, and D.A. Roberts, Remote sensing of soil properties in the Santa Monica Mountains. II. Hierarchical Foreground and Background Analysis. Remote Sensing of Environment (in press).

Ustin, S.L., D.A. Roberts, and Q.J. Hart. 1998a. Seasonal Vegetation Patterns in a California Coastal Savanna Derived from Advanced Visible/Infrared Imaging Spectrometer (AVIRIS) Data. in Remote Sensing Change Detection: Environmental Monitoring Applications and Methods, Elvidge, C.D, and Lunetta, R. (Eds.), Ann Arbor Press, MI (in press).

Ustin, S.L., M.O. Smith, S. Jacquemoud, M.M. Verstraete and Y. Govaerts, Geobotany: 1999. Vegetation Mapping for Earth Sciences, in Manual of Remote Sensing: EARTH SCIENCES VOLUME Andrew Rencz, editor Chapter 4 (in press).

Ustin, S.L. and Q.-F. Xiao, 1998b, Mapping of forested ecosystems in interior central Alaska. Submitted to International Journal of Remote Sensing (submitted).

Vanderbilt, Vern C. Guillaume L. Perry, Joel A. Stearn, Susan L. Ustin, Martha C. Diaz Barrios, S. Zedler, Jon L. Syder, Leslie A. Morrissey and Gerald P. Livingston, François-Marie Bréon, Sophie Bouffies, Marc M. Leroy, Maurice Herman and Jean-Yves Balois. 1998a. Sunglint Allows Wetlands Discrimination. IEEE Transactions on GeoScience and Remote Sensing (in press).

6.2.2 Publications in Press last review (1997)

Roberts, D.A., M. Gardner, R. Church, S.L. Ustin, G. Scheer, and R.O. Green. 1998a. Mapping chaparral in the Santa Monica Mountains using multiple endmember spectral mixture models. Remote Sensing of Environment 65:267-279..

Sanderson, E.W., M. Zhang, S.L. Ustin, and E. Rejmankova. 1998. Geostatistical scaling of canopy water content in a California salt marsh. Landscape Ecology 13:79-92.

Ustin, S.L., D.A. Roberts, J.E. Pinzon, S. Jacquemoud, G. Scheer, C.M. Castaneda, and A. Palacios. 1998c. Estimating canopy water content of chaparral shrubs using optical methods. Remote Sensing of Environment 65:280-291.

Xiao, Q.F., G.E. McPherson, J.R. Simpson, S.L. Ustin. 1998. Rainfall interception of Sacramento’s Urban Forest. Horticulture and Urban Planning 24: 235-244.

6.2.3 Proceedings and Symposium Publications 1998

Gamon, J. L.-F. Lee, H.-L. Qiu, S. Davis, D. Roberts, and S. Ustin, 1998. A multi-scale sampling strategy for detecting physiologically significant signals in AVIRIS imagery. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

Green, Robert O., Dar Roberts, and Susan Ustin, 1998. Mapping agricultural crops with AVIRIS spectra in Washington state. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

Pinzon, J.E., J.F. Pierce, S.L. Ustin and Claudia M. Castaneda, 1998b. Image Registration by Nonlinear Wavelet Compression and Singular Value Decomposition. NASA Publication 2 CP-1998-206853, Proceedings of the Image Registration Workshop, J. LeMoigne (editor). NASA Goddard Space Flight Center, Greenbelt, MD.

Pinzon, J.E., S.L. Ustin, C.M. Castaneda, J. F. Pierce, and L. A. Costick, 1998c. Robust spatial and spectral feature extraction. Robust Spatial and Spectral Feature Extraction for Multispectral and Hyperspectral Imagery, Algorithms for Multispectral and Hyperspectral Imagery, IV, Shen, S. and Descour, M., (editors). AeroSense98, SPIE Proceedings 3372 (in press).

Palacios-Orueta, A., J.E. Pinzon, D.A. Roberts, and S.L. Ustin, 1998d. Remote sensing of soil properties in the Santa Monica Mountains observed with AVIRIS. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

Roberts, D. M. Gardner, J. Regelbrugge, D. Pedreiros, and S. Ustin , 1998b. Mapping the distribution of wild fire fuels using AVIRIS in the Santa Monica Mountains. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

Roberts, D. K. Brown; R. Green, S. Ustin, T. Hinckley, and K. Keightley, 1998c. Investigating the relationship between liquid water and leaf area in Clonal Populus. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

Sanderson, E. W. and S.L. Ustin, 1998. Evaluation of Landscape Structure Using AVIRIS Quicklooks and Ancillary Data. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

Vanderbilt, V.C., V. Ambrosia, and S.L. Ustin, 1998b. Diurnal reflectance changes in vegetation observed with AVIRIS. Proceedings of the Seventh Earth Science Airborne Workshop, Jet Propulsion Laboratory, Pasadena, CA, January 12-14, 1998. (in Press).

6.3 Plans for 1999

• Extend the HFBA approach to the analysis of POLDER airborne data to additional data sets and to the POLDER I satellite sensor archive data (November 1996-June 1997) in preparation for launch of the next ADEOS.

• Extend the analysis of AVIRIS data using the new hyperspatial (4-5m) data to examine spatial scaling issues and to evaluate canopy stress detection methods.

• Continue to evaluate high spatial and spectral resolution data sets to address emerging issues about regional climate impacts and “local” effects having broad (global) distributions. Examples are spectral detection of contaminated soils and detection of vegetation condition, landscape fragmentation, losses of biodiversity or invasions of weedy species. One example we will pursue is the use of AVIRIS to detect leafy spurge infestations, one of the more serious invasive species in western rangelands.

6.4. Personnel Changes

PhD students Graduated:

Eric Sanderson (Wildlife Conservation Society)

Jorge Pinzon (NASA Goddard Space Flight Center)

Alicia Palacios (Univ. Valencia, Spain)

Qingfu Xiao (Univ. California Davis)

Larry Costick (Univ. California Davis)

Rene Lobato (Institute of Water Technology, Mexico)

Ms student Graduated:

Martha Diaz Barrios (Colombia)

Dr. John F. Pierce joined the lab this year as a Research Scientist

7. NASA/GSFC - University of Maryland: Jim Tucker, Ruth DeFries

7.1. Progress/Accomplishments

In previous years, the University of Maryland has contributed to the EOS/IDS project by deriving global land cover data sets from AVHRR data to be used as boundary conditions in the models developed by the team. Currently, the land cover data set derived from AVHRR data at a one by one degree spatial resolution is being used as a boundary condition in SiB-2 and CASA. We have subsequently produced more accurate thematic land cover data from AVHRR data at higher spatial resolutions. In addition, we have derived alternative approaches to use satellite data to more accurately describe the spatial heterogeneity of the land surface as “continuous fields” of vegetation characteristics.

An additional effort over the past year has been to simulate a data set of land cover distributions prior to human disturbance in order to conduct experiments on the changes in terrestrial carbon storage from land cover disturbance.

These efforts are described in more detail as follows:

1. Thematic land cover distributions

As part of the effort in 1997, we derived a global network of training and validation sites through interpretation of over 150 Landsat scenes coregistered with AVHRR data. In this year, we have used these training data to derive a global land cover classification at 8km resolution for the year 1984. The classification approach uses reflectances in red, near-infrared and surface temperature in addition to NDVI with an automated decision tree algorithm (DeFries et al. in press a). The algorithm has also been applied to AVHRR data at 1km resolution to produce a higher resolution global land cover classification (Hansen et al. submitted). Though it is not currently feasible to runs the models developed in this team at these higher spatial resolutions, the higher resolution land cover data sets provide the possibility of improved representation of the land surface through land cover mosaics and other aggregations of the higher resolution land cover data sets.

7.1.2 Continuous fields of vegetation characteristics

As an alternative to the conventional approach for describing land cover as a discrete number of land cover types, we are developing data sets that describe the land surface as continuous variables of important vegetation characteristics (e.g., growth form, leaf type, and leaf duration). Continuous variables allow more accurate description of mixtures and gradients in the vegetation. Over the past several years, we have developed a method using linear discriminants and the training data discussed above to derive continuous fields from AVHRR data. In the past year, we have applied the approach to each year in the 12 year time series of the 8km AVHRR Pathfinder record. Analysis of the interannual variability of the continuous fields indicates that the algorithm gives fairly stable estimates of the percent woody, herbaceous, and bare proportions in locations where these proportions are not expected to change (DeFries et al in press b).

7.1.3 Use of continuous fields in CASA

Based on the continuous fields results, we are testing approaches to use them in CASA. The purpose of this work is to provide more realistic representations of the land surface to improve model results. Currently, we are carrying out an experiment to modify the scheme in CASA for allocating NPP to root, stem, and leaf based on the continuous fields. Comparison with field measurements are underway to determine if the revised allocation scheme improves the model estimates of biomass..

7.1.4 Effects of land cover change on carbon storage and climate

Spatially explicit data sets of land cover prior to human disturbance are required to estimate 1) carbon fluxes resulting form land use change using CASA and 2) biophysical effects on climate through changes in surface roughness, albedo, and water exchanges using SiB2. In the absence of such data sets, we have constructed a map estimating land cover distributions prior to human disturbance based on the satellite-derived one by one degree land cover classification and maps of natural (potential) vegetation. Experiments using these maps in CASA suggest that human-induced land use change has contributed approximately 180 Gt carbon to the atmosphere, with about one-third contributed from land use change prior to 1850 (DeFries et al., submitted). The experiments also indicate that effects of land use change on NPP are substantial in some regions. NPP is currently up to 100 percent higher than that of the natural vegetation in the areas with temperate agriculture and up to 90 percent lower than the natural vegetation in areas where land degradation has occurred in the tropics.

7.2. Relevant Publications from This Reporting Period

DeFries, R., Hansen, M., Townshend, J. R. G. and Sohlberg, R., in press a, Global land cover classifications at 8 km spatial resolution: The use of training data derived from Landsat Imagery in decision tree classifiers, International Journal of Remote Sensing,

DeFries, R., Hansen, M., Townshend, J., in press b, Global continuous fields of vegetation characteristics: A linear mixture model applied to multiyear 8km AVHRR data. International Journal of Remote Sensing.

DeFries, R., Field, C., Fung, I., Collatz, G.J., Bounoua, L. submitted. Combining satellite data and biogeochemical models to estimate global effects of human-induced land cover change on carbon emissions and primary productivity. Global Biogeochemical Cycles.

Hansen, M., DeFries, R., Townshend, J., and Sohlberg, R. submitted. Global land cover classifications at 1km spatial resolution using a classification tree approach. International Journal of Remote Sensing.

7.3. Plans for 1999

• carry out experiments on the effects of land cover changes on climate - Using the maps of current land cover and land cover prior to human disturbance as described above, we will assess the effects of land cover change on climate through the use of SiB2 coupled with the GCM.

• develop spatially explicit estimates of dynamic land cover changes for the previous several hundred years and apply these data sets in CASA to derive estimates of carbon fluxes from land use change - A pressing need to advance understanding of the terrestrial sources and sinks of carbon is data on historical land cover over the past several hundred years, with emphasis on the past few decades. In the absence of a detailed historical reconstruction of land cover, we will generate approximations of land cover distributions and apply these maps in CASA to obtain initial model estimates for carbon fluxes attributable to land use change.

• incorporate land cover heterogeneity in models - This effort will involve a continuation of the work to incorporate continuous fields in CASA and assessment of the results. We will also explore methods for incorporating the continuous fields in SiB2. In addition, we will test the effectiveness of using land cover mosaics and other means to represent heterogeneity in the models.

7.4 Personnel Changes

There are no personnel changes.

8. Stanford University A. --Harold Mooney

8.1 Progress/Accomplishments

The EOS project effort within the Department of Biological Sciences has been focused on different aspects of ecosystem phenology, the study of recurring vegetation cycles that is essential for terrestrial ecosystem production, and assessment and prediction of ecosystem dynamics. The study, a focus of Xia Li’s PhD thesis, aims at improving estimates of current and future atmosphere-biosphere carbon exchange, specifically the algorithms of the inter- and intra-seasonal canopy dynamics. In the past, we have reported our study of ecosystem phenology on budburst and litterfall, and the development of a new version of the BGC model with dynamic algorithms describing the processes of carbon and nutrient allocation and turnover to predict ecosystem response to changes in climate and atmospheric composition. This year we have been following our study on ecosystem phenology further emphasizing primarily on canopy longevity and root/shoot growth synchronization. The study attempts to establish phenological routines for each phenological processes, including budburst, longevity, litterfall, and root-shoot synchronization, then integrate these phenological constrains into physiological-based ecosystem models.

Leaf longevity is an important life history feature that affects ecosystem carbon allocation and nutrient-use efficiency. It is closely coupled with many plant traits, such as specific leaf area, mass-based leaf N concentration, daily photosynthetic photo-flux density, or ratio of construction cost to carbon gain (Howard 1969, Bazzaz and Harper 1977, Williams et al. 1989). The co-dependent leaf longevity and plant traits interact to both influence and reflect ecosystem productivity through complex whole-plant processes (Chapin 1980, Coley et al. 1985, and Reich et al. 1991). Due to the complex interactions between leaf longevity and ecosystem physiological processes, many efforts, trying to estimate growth cession using environmental signals such as temperature and photoperiod, have failed (Hanninen 1990, Kramer 1995). Further (1995, 1991) developed a conceptual longevity model based on ecosystem carbon balance.

We adopted Kikuzawa’s carbon cost-benefit hypothesis on leaf longevity in our analysis for ecosystem canopy longevity. The time of growth cessation was estimated when maximum marginal carbon gain was achieved ([pic], where t*is leaf longevity, p(t) is photosynthetic rate, m(t) is maintenance respiration rate, and C is construction cost). Ecosystem marginal carbon gain was evaluated via model simulation using a revised version of the BGC model based on a detailed ecosystem model (BGC2, Pierce et al., 1996). The allocation and decomposition routines in BGC model were modified to accommodate dynamic seasonal changes with added phenological processes of on/off set of greenness of ecosystem.

The approach to apply the maximum marginal carbon gain hypothesis via model simulation (BGC-Phenology) was tested against observational data at Harvard Forest. Daily CO2 flux, obtained from model simulation driven by observational climate data and fed with biological and environmental parameters rooted on field study, accorded with the eddy flux observations at the site. Annual NPP fell within the range of the NPP observations as well. Days of off-set of greenness at Harvard Forest from 1989 to 1994 predicted by the BGC-phenology corresponded well with the NDVI and the eddy flux observations (Figures 8.1a and 8.1b).

Figure 8.1. Daily CO2 fluxes at Harvard Forest, 1991-1993, as calculated by GBC phenology (a); and observed by eddy flux method (b).

Root-shoot synchronization and root dynamic seasonal fluctuations were studied through field observation and model simulation. This study focused on the synchronous growth of above- vs. belowground and root longevity in correlation with soil nutrients and physical conditions (temperature and moisture). Field observation was conducted at The Wind River Canopy Crane Research Facility (WRCCRF). This summer, 80 minirhizotron tubes were put into the soil at 45 degree (with reference to horizon) along a pathway under the crane station to observe the dynamic growth of roots in the Wind River Experimental Forest (dominated by 400 to 500 year old Douglas-fir and western hemlock). Following on before this winter, resin bags will be buried aside to the tubes to determine the net mineralization and nitrification of the soil synchronously with the root growth. Along with these field work, cooperative efforts by other researchers at the site including photographing of shoot initiation, sap flow and other below-ground soil study, are conducted to allow an integrated inputs and comprehensive study at the site.

Model simulation of root longevity will be based on the optimal root lifespan hypothesis proposed by (Eissenstat and Yanai 1997). The optimal root lifespan hypothesis assumes that root lifespan will optimize the efficiency of nutrient acquisition which is calculated by the rate of nutrient acquisition divided by the rate of carbon expenditure for root growth and maintenance. This hypothesis will be test via model simulation (presumably using a revised version of BGC-Phenology) with fed field observations at WRCCRF.

Synchronous with these studies, daily NDVI data from 1995 to 1998 are being retrieved from the Satellite Active Archive of NOAA () and manipulated to estimate more precisely the on/off set of greenness of ecosystems for the study sites. Based on these new data, budburst and leaf longevity studies will be revised to yield better results.

8.2 Publications

X Li and HA Mooney, A Phenology-Based Ecosystem Physiological Model, in preparation.

8.3 Plans for 1999

Field root observations will be continued with data analysis and model development for the belowground. A comprehensive physiological and phenological ecosystem model based on BGC2 will be constructed which simulates dynamic canopy seasonal variations and root growth. The model will be tested against field observations and contemporary satellite records at selected ecosystems.

4. Personnel Changes

There are no personnel changes

9. Stanford University/Carnegie Institution of Washington: Pamela Matson

9.1 Progress and Accomplishments

We are developing and testing mechanistic models of carbon and nutrient cycling, trace-gas production and efflux, and N leaching losses in natural and agricultural systems. With these tools we can: 1) address the roles of climate, and soil characteristics (e.g., texture, diffusivity, and carbon content) in controlling N cycling processes in the soil, N availability to plants, and losses of the gases NO, N2O, and N2; and 2) draw on the models to identify key variables or predictors of nutrient cycling and N loss potential of soil systems, to be applied in CASAII, SiB, and other global models.

The mechanistic model we have developed, called NLOSS, simulates the transport of water and heat in soil; the microbiological processes of denitrification, nitrification, and decomposition; plant growth and net production; soluble compound leaching; gaseous compound diffusion; and transport of gases across the soil-atmosphere interface. The following section outlines the methods used to simulate these processes in NLOSS. We are currently comparing mathematical formulations used in other models (e.g., in CASA, for example, gas fluxes are a function of water filled pore space) with NLOSS’s more mechanistic formulation of gas. NLOSS’s mechanistic treatment of processes allows us to carry out sensitivity analyses and to develop simplified trace-gas emission models and predictors based on available climatic and edaphic information. With NLOSS we can also examine the role of alternate fertilizer and water management agricultural practices in reducing N losses via both trace-gas emissions and soluble compound leaching. Given that nitrous oxide from agriculture will be subject to climate convention accounting, this model and hybrids of it may prove useful in both estimating the role of agricultural practices in emission of N2O and in evaluating alternatives to reduce emissions.

NLOSS has thus far been tested using data collected in the Sonora, Mexico. We are preparing to carry out a regional-scale comparison of NLOSS (aggregated to the regional level using GIS data bases and TM data obtained in Sonora under a separate NASA grant) with regional scale CASA estimates of N cycling and loss. We also will utilize a newly collected data base on 15N signatures of different components of the Sonoran soil-water-plant system to evaluate the extent to which NLOSS can simulate appropriate processing of nitrogen in the soil-water system.

We are also currently modifying CASAII to incorporate N and P biogeochemistry and productivity limitation, and the impact of aerosol/cirrus clouds on NPP, and will carry out analyses at the global scale on spatial variability of N and P limitation and on tropical aerosol impacts on NPP. Most of our effort in the next year will focus on applying elements of NLOSS, especially those related to water and gas flux through the soil layers, and the elements of the N and P biogeochemistry represented in CASAII to SiB.

Manuscripts on NLOSS are in preparation; manuscripts on the CASAII-NLOSS comparisons and on CASAII biogeochemistry changes will be prepared in the next several months. Eight presentations (seminars and contributed papers) that include model development and output have been given since Sept 1997.

Appendix: NLOSS Model Description

Soil Mass and Heat Transport. NLOSS simulates the transport of liquid water, vapor, and heat with a finite-difference discretization of a one-dimensional soil column. The liquid and gaseous components of the water flux are coupled either by 1) calculation of a latent energy loss with the FAO Penman-Monteith technique (Allen et al., 1994); or 2) vaporization within the soil column and subsequent transport (following Kondo and Saiguso (1994)). In this study we apply the FAO Penman-Monteith method to compute the surface energy balance.

We have described our implementation of the FAO Penman-Monteith method elsewhere (Matson and Riley, 1997; Riley et al., Submitted). Briefly, this submodel applies a Richard’s equation solution for transport of liquid water in the unsaturated soil, and uses hourly measurements of net radiation, ground heat flux, and air temperature and humidity to estimate an evapotranspirative flux. The bypass model of Eckersten and Jansson (1991) is included to account for water flow through macropores. We apply the method of Rosenberg (1974) to reduce the evapotranspiration rate for very dry soils. The resulting “actual” evapotranspiration is partitioned into soil evaporation and transpiration by the methods described in Ritchie (1971) and Hanks (1991).

Denitrification Submodel. We have described our submodel of denitrification in our previous report (Matson and Riley, 1997). Our goal in developing this portion of NLOSS was to improve on previous process models, such as DNDC (Li et al., 1992), by including details of trace-gas fluxes within the soil profile, estimating gas flux to the atmosphere using the surface concentration gradient, determining the soil anaerobic fraction, and estimating the model’s predictive uncertainty based on input parameter and initial condition uncertainty.

Nitrification Submodel. The nitrification submodel predicts the rate and byproducts of the microbial conversion of NH4 to NO3 by applying the “hole in the pipe” conceptualization of Firestone and Davidson (1989). We model the nitrification rate as

|[pic] |( 1) |

where [pic] is an empirical constant, [pic] is a temperature correction factor, [pic] is a moisture correction factor, and [pic] is the concentration of NH4 in the soil. This method is similar to techniques used in Parton et al. (1996) and Li et al. (1992). A moisture and temperature-dependent factor is used to compute the fraction of the nitrification rate emitted as a gas, and a second, moisture-dependent factor is applied to fractionate the gas emission between NO and N2O.

Trace-Gas Transport and Efflux. Details of the methods used to simulate trace-gas transport within the soil column and emission to the atmosphere were described in our previous report (Matson and Riley, 1997). Briefly, NLOSS simulates gas transport within the soil column assuming Fickian diffusion modified by an empirical correction to the diffusion coefficient based on tortuosity and water occlusion of soil pores. We showed, by applying the conceptualization of Skopp (1985), that N2O transfer between the aqueous and soil-gas phases was not rate limiting.

Carbon Cycling Submodel. NLOSS’s prediction of soil carbon cycling is based on the decomposition model used in DNDC (Li et al., 1992), which is derived from the model of Molina et al. (1983). These models are structurally similar to the CENTURY model of Parton et al. (1987). Briefly, the soil carbon is divided into three organic matter pools: residues, microbial biomass, and humads. Each of these pools is further divided into labile and resistant fractions. Decomposition is modeled as pseudo-first order decay from each of these pools, where the rate constant is a function of the pool’s potential decomposition rate, soil temperature, and soil moisture. Soil N cycling is tied to carbon transformations, since each transfer of C requires a concurrent transfer of N. Because of the varying C:N ratios of the soil pools, C flow limitations will occur when there is insufficient N to complete the transfer. Conversely, surplus N from a C flow is returned to the soil mineral N pool.

Nitrogen Leaching. We have combined the hydrologic submodel of NLOSS with lysimeter measurements of NO2 and NO3 to predict the portion of applied N lost as leachate over the growing season for two experimental treatments. The leaching loss is computed as the product of the predicted downward water flux and the measured lysimeter N concentrations.

Productivity Model. Vegetation plays important roles in the soil moisture balance and in soil C and N cycling. We have designed NLOSS to accommodate productivity and growth submodels such as the field-scale CERES model (Hanks and Ritchie, 1991) and the coarse scale CASA model.

Model Uncertainty and System Sensitivity. We apply a Monte Carlo technique to estimate the impact of parameter and initial condition uncertainty on the model’s predictions of soil moisture, N trace-gas fluxes, and N leaching. This method involves performing many simulations, each with parameters and initial conditions chosen randomly based on the known means and deviations, and then computing a mean and uncertainty range from the ensemble results. The impact of parameter covariation (e.g., the relationship between the saturated matric potential and hydraulic conductivity) is ignored in this analysis. We also performed a sensitivity analysis of integrated N trace-gas fluxes and N leaching to individual parameters and initial conditions. This analysis allows us to better understand the dominant drivers of the trace-gas fluxes and to identify specific subsections of the model that require enhancement because of their relative importance to the simulation’s predictions.

10. The Carnegie Institution of Washington: Chris Field

1. Progress and Accomplishments

10.1.1 Carbon Dynamics Following Amazon Deforestation (with I. Fung and J. Tucker)

A summary of this effort is described in Inez Fung’s section of this report. A manuscript based on this project is nearing readiness for submission to GBC.

10.1.2 Human Caused Changes in Global Carbon Stocks (with R. DeFries and J. Collatz)

A summary of this effort is described in Ruth DeFries’ section of this report. A paper based on this project is submitted to GBC (DeFries et al. 1999).

10.1.3 Human Caused Changes in Terrestrial Primary Production (with R. DeFries)

A summary of this effort is described in Ruth DeFries’ section of this report. A paper based on this project is submitted to Science (DeFries and Field 1999).

10.1.4 Application of Continuous Fields in a Terrestrial Carbon Model (with R. DeFries)

Ruth DeFries and colleagues have recently developed techniques for describing the landscape using continuous distributions of vegetation attributes, in contrast to the traditional dominant vegetation types. With each pixel classified as a combination of woody vegetation, green vegetation, shade, and bare soil, it is potentially possible to make a terrestrial carbon model responsive to much finer gradations of vegetation structure, without the computational overhead associated with a traditional classification on a very fine scale. Since much of the world involves a number of vegetation types in a grid element of reasonable size (e.g. 1 x 1 degree), a representation with continuous field may be closer to reality. Of course, if the heterogeneity in vegetation is paralleled by heterogeneity in climate or soils, then the aggregation required for the continuous fields may deflate the potential value.

To assess these issues, we are currently evaluating possible applications of the map based on continuous fields, using the CASA model. The first exercise involves running CASA for a number of transects, with three kinds of simulations for each 1 x 1 degree area. One set of runs asigns the dominant vegetation type. One allows continuous fields. And the third is based on a separate calculation for each 1 x 1 km area. The first result from this analysis is that the continuous fields do a good job of detecting gradual transitions, better even than the separate analyses for each 1 x 1 km. The implication of this is that much of the important heterogeneity captured in the continuous fields exists at a spatial scale even finer than 1 x 1 km.

10.1.5 Large-Scale Application of 1 km Land Surface Classification (with R. DeFries)

Ruth DeFries and colleagues recently completed a new classification of global land cover at a spatial scale of 1 km. Application of land cover data at this spatial scale has the potential to dramatically improve our ability to model the terrestrial biosphere, including the development of estimates relevant to the Kyoto Protocols of the FCCC. At present, we do not have the computational resources, the complete suite of data sets, nor the necessary experience to run a model like CASA for several centuries at a 1 km resolution for the global scale. We are, however, in the process of building in all three areas, but especially the area of gaining experience with fine resolution data. Specifically, we have undertaken two kinds of activities. First, we are running CASA with 1 km data for selected 1 x 1 degree areas, using 1x1 degree data for everything except the cover class and NDVI. The output of these runs will be compared with surface observations and model runs for the aggregated 1 x 1 degree areas to assess the issues associated with moving to this spatial resolution. Second, we are conducting global CASA runs on mosaic data, with each 1 x 1 degree area classified with the relevant proportion of each contributing land cover class. This experiment should provide insight into the kinds and magnitudes of change we should expect from including the high spatial resolution land cover data.

10.1.6 Seasonal Trends in the Growth of the CO2 Amplitude (with J. Randerson, I. Fung)

After completing a comprehensive review of the changes in the CO2 amplitude at high latitude (Randerson et al. 1997a), Jim Randerson, in collaboration with Inez Fung, Pieter Tans, and Chris Field, undertook expanded studies on changes in the seasonality of northern atmospheric CO2. He has focused on asking what kinds of changes in the activity of the biosphere are consistent with the changes in the seasonality and the amplitude of atmospheric CO2. The conclusion, that the overall pattern strongly suggests increased terrestrial NPP in response to early-season warming, adds an important dimension to the debate on the nature and persistence of the northern hemisphere carbon sink (Randerson et al. 1999a). The results of this study were recently submitted to Nature.

We analyzed CO2 data from high northern latitude observation stations in the NOAA/CMDL network that had a mostly continuous record from 1980 through 1997. Monthly mean CO2 concentrations from Mould Bay (76(N), Point Barrow (71(N), Ocean Station M (66(N), and Cold Bay (55(N) were filtered (to remove the long-term secular trend) using a locally weighted regression procedure. We then calculated linear rates of change (in units of ppm per year) for each month in the combined time series using techniques described by Randerson et al.(1997). Negative change rates for the combined time series during June, July, August, and September indicate that CO2 concentrations decreased during summer, while positive rates from November to April indicate CO2 increased during winter and early spring (Figure 1a).

The monthly rates of change and the mean CO2 values (Figure 1a) were also used to calculate the peak-to-trough amplitude changes and the advance in the downward zero crossing time. The peak to trough amplitude increase for the combined time series from January 1980 to December 1997 was 0.60%/yr (p < 0.01) relative to the mean seasonal cycle for this period. The advance in the downward zero crossing time was 4.0 days.

Figure 10.1.

a) The filtered seasonal cycle of CO2 from 1980 to 1997 averaged for Mould Bay (76(N), Point Barrow (71(N), Ocean Station M (66(N), and Cold Bay (55(N) stations in the NOAA/CMDL flask network (solid line, left axis). Negative rates of change during the summer and positive rates during the winter and early spring indicate that the peak-to-trough amplitude increased and that the shape of the seasonal cycle changed from 1980 to 1997 (dashed line with circles and standard error bars, right axis).

b) The steady state seasonal cycle of NPP (solid line), heterotrophic respiration (dash -dotted line), and NEP (dashed line) as simulated by the CASA biogeochemical model for regions north of 50°N. Positive fluxes are out of the surface. The final year of the control simulation is shown (1997). The CASA model uses satellite-derived NDVI to estimate radiation absorbed by plants at a 1( x 1( resolution 35.

c) Monthly differences in NEP from the steady state seasonal cycle are presented for the five model runs in which NPP increased before July (dark blue), before August (light blue), during all months (green), after June (brown), and after July (red). Only differences during the final year of simulation are shown (1997). The differences between the steady state and the evolving NEP time series shown here represent possible seasonal distributions of a terrestrial sink. Each simulation started with carbon pools and carbon fluxes in steady state in 1980 using input data sets described by Randerson et al..

d) Modeled and observed seasonal cycles of atmospheric CO2 averaged for Mould Bay, Point Barrow, Ocean Station M, and Cold Bay. The modeled concentrations (dashed line) are from the GISS tracer transport model driven by steady state NEP fluxes from CASA (Figure 1b) and ocean exchange, biomass burning, and fossil fuel emissions as described by Randerson et al. The observed seasonal cycle is the same as in Figure 1a (solid line).

e) Observed and simulated monthly CO2 change rates from January 1980 to December 1997 for the combined time series of Mould Bay, Point Barrow, Ocean Station M, and Cold Bay. Observations are given by the dashed line with circles and standard error bars (same as Figure 1a). Simulated CO2 change rates from the GISS model are shown for the five CASA model runs in which NPP increased before July (dark blue), before August (light blue), during all months (green), after June (brown), and after July (red). Correlation coefficients between the observed and simulated change rates are as follows: before July (r = 0.92), before August (r = 0.91), during all months (r = 0.77), after June (r = 0.43), and after July (r = 0.12). To allow the seasonal cycle of simulated CO2 concentrations to reach steady state at the beginning of each GISS tracer model run, the first 4 years of input consisted of steady state NEP fluxes (each tracer model run lasted for 22 years). We removed the long-term secular trend from the modeled CO2 concentrations (and calculated the monthly CO2 change rates) using the same techniques that were applied to the station data.

In a series of forward simulations using CASA, we imposed increases in NPP during distinct seasonal periods. The seasonal NPP increases generated dynamic time series of net ecosystem production that were then used to drive the GISS atmospheric tracer model. We compared simulated monthly CO2 change rates from the GISS model with the observations shown in Figure 1a. Our goal was to identify the time of the year that changing surface fluxes from terrestrial ecosystems most closely reproduced the atmospheric observations. The seasonal timing of NPP increases had little effect on the magnitude of the annual terrestrial sink yet had important consequences for its seasonal distribution (Figure 1c). When NPP increases were confined to one edge of the growing season or the other (before July or after July), they significantly increased ecosystem uptake during the time of the imposed change. These model scenarios suggest that 1) the seasonal distribution of a terrestrial sink is likely to strongly depend on the driving mechanism, and 2) that long-term increases in NPP and subsequent increases in ecosystem uptake during one part of the growing season may be largely offset by increases in respiration at other times of the year.

To reproduce the rapid decrease in monthly CO2 change rates between April and July observed at the high northern latitude stations, increases in ecosystem uptake are required early in the growing season (Figures 10.1c). NPP increases before July most closely matched the observations, with an advance of 4.3 days as compared with the observed advance of 4.0 days.

We also conducted inversion analyses, using a 2-D model of atmospheric transport driven by climatological mean winds. Following a similar procedure to the CO2 concentration analysis, for each seasonal interval of modeled surface flux we estimated the linear rate of change from 1980 to 1997. For this analysis, the most dominant feature of change was an increase in uptake (or a decrease in release) during the first half of the growing season (before August). Surface fluxes changed most rapidly during June and July for the veteran station case and during May and June for the all stations case.

Surface fluxes from the all stations inversion analysis were used to investigate the effect of interannual variability in temperature. Fluxes at the start of the growing season show a significant negative correlation with spring temperatures. In contrast, at the end of the growing season, surface fluxes show a significant positive correlation with temperatures, consistent with the presence of a large pool of soil organic matter that significantly contributes to total ecosystem respiration at the end of the growing season.

No significant relationship was found between annual carbon balance and mean annual temperature, suggesting that the year-to-year carbon balance in northern ecosystems is controlled to some degree by an unstable balance between temperature driven uptake anomalies during the spring and release anomalies during the fall. If NPP increases from elevated spring temperatures are responsible for the increased early season uptake, then the NDVI increases (which show that a large fraction of the NDVI increases occur before July), mid and high northern latitude terrestrial sink estimates, amplitude increases, and the changes in the shape of the CO2 seasonal cycle reported here are all qualitatively consistent. Taken alone, none of the other possible mechanisms can account for all of these changes. More generally, our analyses suggest that any significant and persistent sink in mid and high latitudes of the Northern Hemisphere should leave a detectable signature in the seasonal cycle of CO2 if NPP increases are the driving mechanism.

10.1.7 A new Terrestrial Model Combining Surface Energy Balance, Photosynthesis, and Carbon Cycling (with J. Berry)

We (Jörg Kaduk, Joe Berry, and Chris Field) finalized the structure of the carbon cycle model based on CASA and SILVAN. The structure of the model carbon reservoirs is depicted in figure 10.2. SiB2, as implemented in the Colorado State University atmospheric general circulation model (CSU AGCM, Sellers at al. 1996a,b), was modified to facilitate point mode runs for single sites and the inclusion of the carbon cycle model within SiB2. The same model code allows for runs of multiple sites simultaneously, such as all Ameriflux sites. Since the same version of SiB2 is running in the mesoscale model RAMS, SiB2 and the nested carbon cycle model can now be run in global and regional climate models. This allows the simulation of local to global biosphere atmosphere interactions with simultaneous estimation of carbon fluxes in a coherent modeling framework.

We devised software for using observations from single sites to force the model and store them in a common format (netCDF), which allows easy access for visualization and investigation. Global gridded input data sets to force SiB2 with the carbon cycle model were prepared from the GEOS I climate reanalysis (Schuber S., C.-K. Park, C.-Y. Wu, W. Higgis, Y. Kondratyeva, A. Molod, L. Takacs, M. Seablom and R. Rood, 1995. A MultiYear Assimilation woth the GEOS-I System: Overwiev and Results. NASA Technical memorandum 104606, Vol.6.) and FASIR 2.1 NDVI data.

Obtaining a steady state for the reservoirs in a model with relaxation times in the order of hundreds of years running on a short time step in a land surface model presents a problem. It has to be addressed with a method that avoids simply running the coupled model until the reservoirs reach an equilibrium since computing costs of this approach are prohibitive. First tests with a simple estimation of the equilibrium indicate that the simulation time required to achieve a steady state of the carbon cycle can be shortened substantially (100-200 instead of more than 1000 years). The current implementation also allows us to run only the carbon cycle from stored climate data, reducing the required computing time. The still simple structure of the model allows direct analytical determination of the equilibrium, assuming linearity. Even though the model is not completely linear, we expect that this approach will provide a further major reduction of computing time required for the estimation of a steady state. We are currently working on an implementation, following Bolker et al. (1998) (B. M. Bolker, S. W. Pacala and W. J. Parton Jr., 1998. Linear analysis of soil decomposition: insights from the CENTURY model, Ecological Applications, 1998, 8, 2, 425-439.)

We presented posters on this new model at the International CO2 conference in Cairns Australia Sept 1997 and at the GCTE Science Conference in Barcelona, Spain, in March 1998.

10.1.8 Nutrient Effects in the CASA Model (with P. Matson and P. Vitousek)

We (Greg Asner, Jason Neff, Peter Vitousek, Pam Matson, and I) are modifying CASA to incorporate N and P biogeochemistry and productivity limitation. This involves a fundamental rethinking of the role of nutrient limitation in decomposition, a new approach to the nutrient competition between plants and microbes, and an increased role for interactions with the mineral soil. The model version with nutrient limitation will allow us to test ideas about the global impacts of N and P limitation as well as hypotheses about the consequences of anthropogenic N deposition. This approach may also facilitate development of an approach for using satellite data to assess nutrient and other limitations.

10.1.9 Biomass Allocation for Global Terrestrial Models (with I. Fung)

Plant biomass allocation is one of the most important parts of the terrestrial carbon cycle. It ism however, poorly understood at the mechanistic level and poorly represented in global models, with most models assuming either fixed allocation. As a contribution to what should be a continuing discussion of ways to model biomass allocation at the global scale, Pierre Friedlingstein, Geeske Joel, Inez Fung, and I developed a simple model that predicts the responses of biomass allocation to variation in the availability of light, water, and nutrients experienced by plants. The model also simulates changes in allocation in response to changes in atmospheric CO2. The basic idea of the model, which applies the same algorithms over all biome types, is that plants adjust allocation to leaves, roots, and stems to increase investments in the capture of limiting resources. When water or nutrients are limiting, root investments increase. When light is limiting, stem investments increase. Changes in allocation in response to altered CO2 follow the idea that the changes are driven through variation in the relative availability of light, nutrients, and water. A paper describing this work is in press in Global Change Biology (Friedlingstein et al. 1999).

When integrated with CASA and run globally, the results of this model look reasonable, though validation data are scarce (Fig. 10.3). Leaf and stem allocation are highest in the wet tropics, where water and nutrients are abundant. Root allocation is highest in the deserts (where water limitation is the driver) and at high latitudes (where nutrient limitation is the driver). Stem allocation is low across the world’s grasslands and tundra, in addition to the deserts. Stem allocation tends to rise wherever forests occur.

Though this model is only a first step at solving the allocation problem, it provides an attractive framework for testing a wide range of ideas.

10.1.10 Linking CASA with the VGPM Ocean Model

Working with Paul Falkowski and Mike Behrenfeld (Rutgers University), we (Chris Field and Jim Randerson) merged time-averaged results from CASA for the land and VGPM for the oceans, leading to the first estimate for global NPP based on parallel algorithms for the land and the oceans. The results of this integration (Field et al. 1998) provide a new level of spatial detail for global NPP, as well as a new perspective on the similarities and contrasts in the processes controlling land and ocean NPP.

Figure 10.4: Global, annual NPP (g C m-2 y-1) for the biosphere, calculated from the integrated CASA/VGPM model. The spatial resolution of the calculations is 1 x 1( for the land and 1/6 x 1/6( for the oceans. Input data for ocean color from the CZCS sensor are averages from 1978-1983. The land vegetation index from the AVHRR sensors is the average from 1982-1990. Global NPP is 104.9 Pg C y-1 (104.9 x 1015 g C y-1), with 46.2% contributed by the oceans and 53.8% contributed by the land.

NPP on land and in the oceans has been modeled using a variety of approaches with a range of fundamental mechanisms, specific details, and levels of integration. CASA uses an approach that is similar to that used in many ocean models, calculating NPP as a function of the driving energy for photosynthesis, APAR, and an average light utilization efficiency (ε). For both terrestrial and oceanic models, ε cannot be directly measured from space and must be parameterized using field measurements. We combined results from conceptually similar land and ocean NPP models, CASA for the land and the Vertically Generalized Production Model (VGPM) for the oceans. Both of these models are simple formulations designed with an emphasis on integrating spatially extensive satellite observations rather than describing the mechanistic details of NPP. Biospheric NPP was calculated based on observations averaged over several years; 1978-1983 for the oceans (from CZCS) and 1982-1990 for the land (from AVHRR). SeaWiFS data will, however, allow us to make comprehensive land/ocean runs for the current time period.

Using the integrated CASA/VGPM biosphere model, we obtain an annual global NPP of 104.9 Pg C, with similar contributions from the terrestrial [56.4 Pg C (53.8 %)] and oceanic [48.5 Pg C (46.2 %)] components. Average NPP on land without permanent ice cover is 426 g C m-2 y-1, while that for oceans is 140 g C m-2 y-1. On land and in the oceans, spatial heterogeneity in NPP is comparable, with both systems exhibiting large regions of low production and smaller areas of high production. In general, the extreme deserts are even less productive than the vast mid-ocean gyres (Fig. 1). Maximal NPP is similar in both systems (1000−1500 g C m-2 y-1), but regions of high NPP are spatially more restricted in the oceans (essentially limited to estuarine and upwelling regions) than in terrestrial systems (e.g., humid tropics) (fig. 2).

Next steps with this integrated model will involve working with SeaWiFS and EOS data to explore land-ocean issues in the coming years. These two models are not designed to be provide definitive accuracy or process representation, but their merger makes them well configured to address issues where land-ocean comparisons play a critical role.

10.1.11 Interacting Constraints from 13C: Terrestrial Turnover and the Land Fraction

Jim Randerson took the lead in a project to examine the interacting consequences of 13C data for the intensity of CO2 fertilization and the magnitude of the land fraction of the carbon sink. This research, in press in Tellus (Randerson et al. 1999b), worked from the fact that the residence time of carbon in a terrestrial ecosystem plays a critical role in determining the storage resulting from a given level of NPP stimulation. Residence time also plays a critical role, however, in the inferred land fraction of the carbon sink. This connection arises from the link between residence time and isodisequilibrium. A longer residence time implies that the carbon leaving in respiration is older and, therefore, heavier. Terrestrial exchange that consists of fixation and new, isotopically lighter carbon and the release of older heavier carbon, looks to the 13C budget like land uptake, because the atmosphere is becoming heavier.

Because these two aspects of residence time and 13C have contrasting implications, it is possible to use them in concert to tighten the constraints on key unknowns in the global carbon budget. Specifically, explaining the IPCC estimate of the terrestrial sink using only CO2 fertilization requires a long average residence time of carbon in the terrestrial biosphere. But if the residence time is long, then the fraction of the sink on land is small. If the residence time is short, CO2 fertilization is insufficient to explain more than a fraction of the sink, but the sink must be mostly terrestrial. CO2 fertilization can account for all of the missing sink only if carbon residence times are much larger than current estimates.


DeFries, R. S., and C. B. Field. 1999. Human Modification of the Landscape and Primary Production. Science (submitted).

DeFries, R. S., C. B. Field, I. Fung, J. Collatz, and L. Bounoua. 1999. Combining satellite data and biogeochemical models to estimate global effects of of human-induced land cover change on carbon emissions and primary productivity. Global Biogeochemical Cycles (submitted).

Field, C. B., M. J. Behrenfeld, J. T. Randerson, and P. Falkowski. 1998. Primary production of the biosphere: Integrating terrestrial and oceanic components. Science 281: 237-240.

Friedlingstein, P., G. Joel, C. B. Field, and I. Y. Fung. 1999. Towards an allocation scheme for global terrestrial carbon models. Global Change Biology (in press).

Malmström, C. M., M. V. Thompson, G. P. Juday, S. O. Los, J. T. Randerson, and C. B. Field. 1997. Interannual variation in global-scale net primary production: Testing model estimates. Global Biogeochemical Cycles 11: 367-392.

Randerson, J. T., C. B. Field, I. Y. Fung, and P. P. Tans. 1999a. Increases in early season ecosystem uptake explain changes in the seasonal cycle of atmospheric CO2 at high northern latitudes. Nature (submitted).

Randerson, J. T., M. V. Thompson, T. J. Conway, I. Y. Fung, and C. B. Field. 1997a. The contribution of terrestrial sources and sinks to trends in the seasonal cycle of atmospheric carbon dioxide. Global Biogeochemical Cycles 11: 535-560.

Randerson, J. T., M. V. Thompson, and C. B. Field. 1999b. Linking 13C-based estimates of land and ocean sinks with predictions of carbon storage from CO2 fertilization of plant growth. Tellus (in press).

Randerson, J. T., M. V. Thompson, I. Y. Fung, T. Conway, and C. B. Field. 1997b. The contribution of terrestrial sources and sinks to trends in the seasonal cycle of atmospheric carbon dioxide. Global Biogeochemical Cycles 11: 535-560.

Thompson, M. V., and J. T. Randerson. 1998. Impulse response functions of terrestrial carbon cycle models: Method and application. Global Change Biology (in press).

10.4 Personnel Changes

Jim Randerson received his PhD from the Department of Biology at Stanford University, and is now a DOE Hollaender Post-doctoral Fellow associated with Inez Fung at the University of California, Berkeley, and Terry Chapin at the University of Alaska, Fairbanks. Dr. Rebecca Shaw (PhD, UC Berkeley) has joined the group as a post-doctoral fellow.

11 Carnegie Institution of Washington – Joe Berry

11.1 Progress and Accomplishments

11.1.1 Modeling:

Over the past decade we have developed largely mechanistic models to predict how the rates of photosynthesis and transpiration of water by leaves of higher plants are controlled by environmental conditions such as light intensity, temperature, water availability and CO2 concentration. These models were first formulated at the scale of the fundamental metabolic unit, a leaf cell. Work together with Sellers, Collatz and others have developed methods for integrating these models to simulate these processes for canopies and landscapes. The difficult part of this work is not in writing the models but in testing these models against reality.

Ideally, we would like to have sets of measurements at two scales of observation, one of which is used to calibrate adjustable parameters of the model and another at a larger scale of integration that can be used to test the predictions of the model. We have developed robust calibration protocols based on measurements of leaves using portable gas exchange equipment that can easily be transported to the top of canopies. These measurements can be made with great accuracy at the leaf-scale. However, it has proven difficult to obtain corresponding canopy-scale measurements that are suitable for testing the calibrations and the theoretical arguments used to integrate from the leaf to the canopy scale. Canopy-scale measurements of net CO2 and water vapor exchange are now being conducted in many ecosystems; these measurements provide invaluable insight into the carbon balance seasonal and interannual variation in ecosystem processes. However, these measurements are noisy and the fluxes of CO2 and water vapor from the canopy are partially obscured by fluxes from other components of the ecosystem, particularly the soil. These problems are particularly severe in forest ecosystem, making it difficult to test the underlying assumptions and calibration measurements of canopy-scale models.

To address this problem, we constructed an apparatus to conduct measurements of photosynthesis and stomatal conductance on significant portions of forest trees. This apparatus uses an open flow gas exchange principle similar to that used by equipment for conducting leaf –scale measurements. The portion of the tree contained within the cuvette (the top meter) is exposed to environmental conditions that closely approximate the unenclosed portion of the tree, and measurements can be conducted continuously for several days or more. Therefore the measurements represent a larger and more complex structure and span a much longer time than the calibration studies. Yet, the accuracy of the measurements is similar to that of the leaf gas exchange studies

This year we have made use of a data set collected on a Pinus banksiana tree in the Boreal Forest of Canada to test the calibration of the physiological parameterizations used in SiB2. This data set includes a set of calibration measurements which were analyzed to fit the model to leaf-scale observations, and a set of observations of tree photosynthesis and stimulate conductance taken at 5 min intervals over five days. Biophysical parameters (leaf area and leaf display) and a record of environmental logged over these intervals were used to simulate the expected values of stomatal conductance and net CO2 uptake of enclosed tissue. A scatter plot of observations vs. the corresponding observations is shown in Fig. 11.1. The scatter around the diagonal line gives an indication of the simulation errors. This plot shows all data collected over the 5-day interval and spans a wide range of environmental conditions. The average deviation of the observed value from the simulated value is about 20% for stomatal conductance and about 30% for net photosynthesis for the data and simulations.

Figure 11.1. Scatter plots of simulated vs. measured values of stomatal conductance (gsw) and net photosynthesis (Pn) of the upper crown of a Pinus banksiana tree over a 5-day interval. A statistical index the normalized standard error of the estimate (NSEE) gives and index of the scatter and bias of the simulations.

We next asked how this fit might be effected by systematic variation in the value of key parameters of the model. A plot of such an analysis for 6 model parameters is shown in Fig. 11.2. Note that the error term (NSEE) generally shows a minimum. This indicates that there is an optimum value for each of these parameters, which tends to minimize the total errors observed over this 5-day interval. If our hypotheses concerning how to scale up from the leaf scale to that of the tree crown and our calibration experiments were correct, then we should expect to find that these minima should be fairly close to the parameter values that we obtain from analysis of physiological measurements. In each case, the empirical optimum at the tree-scale corresponds well with the value obtained from needle physiology studies. Studies like this are key steps in the testing and development of models used in global simulations.


Figure 11.2. Plots of the normalized standard error (NSEE) for simulations of stomatal conductance (filled circles), net photosynthesis (open circles) and a weighted average of stomatal conductance and photosysnthesis as influenced by changing the value of the indicated parameter. Minima of the curves are local optimum values for these simulations.

2. A new method for determining δ13C and δ18O of CO2 in air:

The flask sampling networks operated by NOAA, Scripps and CSIRO provide one of the most important sets of observations for analysis and modeling of the global carbon cycle. Flask samples of air are collected at frequent intervals at stations located around the globe and sent to central laboratories for analysis. Analysis of these samples for the δ13C and δ18O of CO2 provide important constraints on inversion calculations that attempt to determine the net uptake or release of CO2 from surface reservoirs and to locate these sources and sinks. Great care is taken in the measurement protocols used to analyze the flasks so that the measurement are of the highest possible accuracy and to provide standards so that measurements conducted at different times or in different labs are comparable. However, these methods require large samples (typically 2 liters) of air and the measurements are labor intensive

It is widely recognized that these global measurements need to be supplements by process-level studies at regional, ecosystem and plant scales. Such data is needed to calibrate and test the models of isotopic fractionation used to interpret the global observations. Work at these scales needs to be conducted with similar precision to the global-scale networks and even a modest level of experimentation could easily generate as many samples as are now processed by the global sampling programs. While the managers of these global programs are sympathetic to the need for local scale measurements, they just don't have the capacity to fill this need. Ecosystem scientists have been reluctant to make the large commitments of laboratory space, equipment and personnel to duplicate these facilities.

This report describes a new technology for measurements of CO2 concentration and isotopic composition of CO2 in air samples. This technology is specifically intended for ecosystem and plant or leaf-scale measurements. Significantly, the technology uses very small samples of air (less than 5 ml for a complete analysis) while achieving levels of accuracy and cross calibration that approach those of the flask networks. Most important, the entire procedure can be automated.

In this new procedure, the batch inlet system of a conventional isotope ratio mass spectrometer is replaced with a continuous flowing helium inlet system. Samples are introduced as plugs in the flow of helium and the isotope ratio is calculated from measurement of the "peaks" at m/e 44, 45 and 46 that are recorded when a sample of CO2 passes through the mass spectrometer source. Typically, the quantity of sample required for a measurement with the flowing helium inlet is about 10-5 that required for a measurement with a batch inlet system.

The system described here functions to isolate the CO2 from a given quantity of air; to release that CO2 into a stream of helium which carries the sample into the mass spectrometer. Additionally, by measuring the amount of air used and the size of the CO2 peak, one can calculate the CO2 concentration in a sample.

A schematic diagram of the system is shown in Fig. 11.3. Air containing CO2 is withdrawn from a flask and flows into an evacuated volume through a capillary tube immersed in liquid nitrogen. CO2 is quantitatively trapped from the air, and it is rapidly released when the trap is withdrawn from the liquid nitrogen.


Figure 11.3. Schematic of the new system for measuring CO2 and isotopic concentrations of CO2 in air samples.

A two position-six port Valco valve (1) is used to switch the connections between a "trapping state," with the capillary connecting the sample or standard flask to the vacuum chamber and a " CO2 release state" with the helium flow directed through the trapping capillary and on the mass spectrometer. An air actuated calendar (5) is used to submerge or withdraw the trapping capillary in a liquid nitrogen dewar.

With the 6-port valve in the trapping state, air is permitted to bleed through the trap and into the vacuum chamber. During this time the valve diverts the flow of helium directly to the mass spectrometer. Pneumatically activated micro-valves (2) provide on/off control on the flow of air from either the sample or standard flasks. The quantity of air sampled is measured manometrically with a high precision capacitance manometer (0-100 mbar). Once a preselected quantity of air is admitted the inlet valve is closed; the pressure is measured, and the residual air in the system is pumped away. The 6-port valve is then switched to the release state. The trap is withdrawn from the deware, and a micro switch is tripped permitting a current to flow through the stainless steel capillary. This current heats the capillary to about 80C, quickly volatilizing all of the CO2 and water that had been trapped. A nafion drier (7) downstream of the trap prevents the water from reaching the mass spectrometer. By alternately sampling air from the sample flask and one containing a standard both the CO2 concentration and the isotope ratio of the sample can be referenced to other measurements. CO2 concentration is determined with a precision of 0.5 ppm and the isotope ratios are determined to a precision of 0.05 o/oo.

All of the valves can be computer controlled, permitting the process to be automated. Each cycle of trapping and release typically takes about 4 min. A typical measurement sequence is comprised of ten cycles, five each of the sample and a flask containing a standard air. A multi position valve (not shown) can be used to permit analysis of several sample flasks over night.

11.2 Publications:

Colello, G. D., C. Grivet, P. J. Sellers, and J. A. Berry. 1998. Modeling of energy, water, and CO2 flux in a temperate grassland ecosystem with SiB2. 1998. J. Atmospheric Sciences 55:1141-1169.

Collatz, G.J., J.A. Berry and J.S. Clark. 1998. Effects of climate and atmospheric CO2 pressure on the global distribution of C4 grasses: present, past, and future. Oecologia,114: 441-454

11.3 Plans for 1999:

Modeling: We will continue work to test and improve models using data obtained from field experiments conducted at BOREAS and in agricultural and native prairie ecosystems in the Southern Great Plains of the US. We will also participate, together with Scott Denning and Jim Collatz in a major revision and enhancement of SiB, producing SiB3. The major goal of this revision are to provide improved soil and snow biophysics, improved radiation transport, add isotope fractionation, and develop the capacity simulate sub-gridscale heterogeneity of land surface properties.

Isotopic exchanges: Chris Still, as part of his thesis work will use flask sampling to assess the exchange carbon isotopes between the reservoirs of organic carbon contained in the soils and biomass of this ecosystem and the atmosphere. The ecosystem selected for this study supports a mixture of C3 and C4 species. It is expected that seasonal and interannual differences in the productivity of these components of the ecosystem will result in a significant and time varying isotopic disequilibrium between the carbon stored in the ecosystem and the atmosphere.

Radon exchange: We will initiate a study to evaluate the potential of using covariation of the concentrations radon gas and CO2 as a means for obtaining regional scale estimates of net CO2 exchange over the continents.

4. Personnel Change

The co-Investigators are Wei Fu, Miquel Ribas-Carbo and Chris Still.

12. Management

1. Communication within the team

The project effort has been divided into nine tasks, each of which has a lead person/organization to direct the work supported by contributing people/organizations. The tasks integrate across approaches, and unite the team across the geographic separation. Communication among team members occurs on a day-to-day basis via email, fax and telephone. Data and model exchanges occur via dedicated ftp accounts.

The full team meets twice a year to bring the whole team up to date, and to strategize about future directions. The team meetings are hosted by team members, on a rotating basis. Students and postdocs routinely present their work for comment by the entire team. Team collaborators and visitors are frequently invited to make feature presentations. In addition, task level meetings are convened to discuss joint publications and to plan new projects. The last three meetings were held: November 10-12, 1997 in Carlsbad, NM (hosted by Tucker); May 4-7 at Biosphere2 in Oracle, Arizona (organized by Berry); October 27-29 at University of California, Berkeley (hosted by Fung). The Biosphere2 meeting was held jointly with R. Dickinson’s and J. Foley’s EOS-IDS teams, with presentations by members of those teams as well. The next team meeting will be in April 1999, to be hosted by Ustin at UC Davis.

12.2 Collaborations with EOS, non-EOS, and non-NASA investigations

This project has formal collaborations with the EOS-IDS teams of R. Dickinson, J. Foley, and R. Koster. Inez Fung is also a member of J. Hansen’s IDS team, and Jim Tucker is also a member of C. Rosenzweig’s team.

Table 12.1 lists other significant science collaborations between this project and other scientists/institutions. As stated above, Dickinson and Foley are routinely invited to, and routinely attend, our semi-annual team meetings.

Table 12.1 Collaborating IDS or NASA Teams

|Team (Designated by Leader) |Nature of interaction |

|Foley |Biosphere model development |

|Dickinson |Model development and application |

|Reineke/Suzrez/Koster |Land-surface modelling, esp. hydrology |

|Schimel |Century soil module |

|Wielicki |Clouds and radiation |

|Hansen |Interannual variability; climate and carbon |

|Tegen |Dust modelling |

|Rosenzweig |El Nino |

|Moore |Carbon models; NDVI |

The datasets and models developed by this project are provided to collaborators and widely used by the community:

• SiB2 (or variations of SiB2) have been implemented in a large number of GCMs: COLA/JMA, INPE/CPTEC, PNL.ARM, DAO, NMC.

• CASA fluxes are used as starting points of a number of carbon inversion calculations (GFDL/Princeton).

• The carbon source/sink maps (e.g. fossil fuel emission and deforestation) are freely available to all CO2 inverse modelling work.

• The NDVI time series are distributed via the DACCs at EDC, LaRC, NSIDC

In addition to prodigious publications, team members are active in all aspects of scientific planning and outreach, from international bodies (e.g. ICSU, IGBP, WCRP) to national (e.g. Carbon Science Plan Working Group) to NASA ESAC and EOS. Furthermore, Randall, Fung (and Dickinson) are external co-chairs of the atmosphere, land-surface and biogeochemistry modelling groups of the NCAR Climate System Model. This collaboration with the NCAR CSM ensures that the lessons learned and codes developed by this team are available to the entire modelling community.






Figure 2.3: The geodesic grid for 2562 gridcells. This corresponds to approximately 4.5 degree resolution, and is comparable to a 4x5 degree latitude-longitude grid.


PI: David Randall (until April 15 1998)

Inez Fung (after April 15 1998)

Progress Report

December 1998

Figure 10.3: Annual allocation to leaf, stem, and root biomass, estimated with the new routine based on resource limitation.

Figure 4.2. Map of the difference in FASIR NDVI averaged over June-August between 1989-1990 and 1982-83., two time periods in the 1982-1990 FASIR data set that were most different.

Figure 4.3. The change in global monthly primary productivity (GPP) and mean surface temperature resulting from using NDVI from 1989-1990, a relatively green period versus 1982-93, which was less so, as vegetation boundary conditions. The dfiference in NDVI is the same each year so variations from year to yeare are an indication of the internal variability in the simulated climate while systematic deviations (positive in the case of GPP and negative for surface temperature) represent changes caused by NDVI forcing.

Figure 10.2. Reservoir and flux structure of the carbon cycle model. Displayed are the reservoirs for one tree and one grass type, the above- and belowground litter reservoirs in 5 soil layers and 3 SOM types in each soil layer. GPP is indicated in green, allocation in blue, plant respiration in red, and litter and SOM production and decay in black. The fluxes for the grass type and SOM transformation and translocation have been omitted.

Figure 5.1 . Distribution of live biomass in the Amazonian rainforest. The map is display at 1x1 resolution. The gridboxes with ongoing deforestation are outlined in black. Note the biomass of the deforestated areas are higher than the mean biomass for the entire forest.




Download Doc File