Terrestrial Sciences Section
The goal of the Terrestrial Sciences Section (TSS) is to increase scientific understanding of land-atmosphere interactions, in particular surface forcing of climate, through model development, application, and observational analyses and to represent that understanding in climate models. Research in TSS spans a broad knowledge of the relationships among the biosphere, hydrosphere, cryosphere, and atmosphere. Scientists in TSS develop and use appropriate multi-scale models, remote sensing, advanced analytical techniques, and observations to study the role of the terrestrial biosphere in the climate system. Topics of study include the regulation of planetary energetics, planetary ecology, and planetary metabolism through exchanges of energy, momentum, and materials (e.g., water, carbon, mineral aerosols) with the atmosphere and ocean and the response of the climate system to changes in land cover and land use.
Scientists in TSS are involved in developing the land model used in the Community Atmosphere Model (CAM) and the Community Climate System Model (CCSM). This model, the Community Land Model (CLM), includes biogeophysics and hydrology, the traditional physical core components of land models, and is being further developed to include river routing, biogeochemistry (carbon, nitrogen, mineral aerosols, biogenic volatile organic compounds, water isotopes), and vegetation dynamics. Gordon Bonan (TSS) co-chair’s the CCSM Land Model Working Group (LMWG), Natalie Mahowald (TSS) co-chairs the CCSM Biogeochemistry Working Group (BGCWG), and other TSS scientists actively participate in both working groups, providing strong input to model development and implementing and testing model parameterizations. Model development is based on process studies of the relevant physical, chemical, and biological mechanisms and the numerical modeling techniques required to represent these mechanisms. TSS scientists compare model output with observed atmospheric, ecological, and hydrological data to validate and improve the model on a wide range of spatial and temporal scales. TSS provides a focal point for CGD and university ecological and hydrological research and serves as a resource to these communities in their use of CCSM.
Members of TSS participate in many NCAR strategic initiatives: Biogeosciences, Weather and Climate Impact Assessment, Water Cycle, and Wildland Fire.
In collaboration with the CCSM LMWG, Bonan, Keith Oleson (TSS), Mariana Vertenstein (CCSM Software Engineering Group), Peter Thornton (TSS), and Sam Levis (TSS) oversaw the scientific and software engineering development of CLM. A new under-canopy turbulence scheme, developed by Xubin Zeng (University of Arizona) and colleagues, was adopted to reduce the excessively warm daytime ground temperatures in sparsely vegetated areas. A new parameterization of fractional snow cover on the ground, developed by Zong-Liang Yang (University of Texas), was tested to improve the low fractional snow cover in CLM with deep snowpack. The proposed parameterization uses different relationships between snow depth and fractional snow cover during the snow accumulation and snow melt phases. The transition between these two phases was problematic, and only the accumulation phase was accepted for implementation in CLM. This new parameterization increased the fractional snow cover on the ground and therefore increased surface albedo in the arctic during winter. This cooled surface temperature and helped eliminate a prominent high latitude winter warm bias in CCSM2. However, the parameterization was not formally adopted for the next version of CCSM because the surface cooling led to excessive sea ice in the arctic.
Levis and Bonan conducted climate model experiments that examined the impact of prognostic leaf phenology on simulated climate. With prognostic leaf phenology, which was developed as part of a global dynamic vegetation model for use with CLM, the emergence and senescence of foliage depends on air temperature and soil water. Observations show that emergence of foliage in springtime slows surface air temperature warming as a result of greater transpiration. Model simulations with the CAM coupled to CLM confirm that evapotranspiration contributes to this pattern and that this pattern occurs more reliably with prognostic leaf area as opposed to prescribed leaf area. With prescribed leaf area, leaves emerge independent of prevailing environmental conditions, which may preclude photosynthesis from occurring. In contrast, prognostic leaf area ensures that leaves only emerge when conditions are favorable for photosynthesis, and thus transpiration. These results reveal a dynamic coupling between the atmosphere and vegetation in which the observed reduction in springtime warming trend only occurs when photosynthesis, stomatal conductance, and leaf emergence are synchronized with the surface climate.
In collaboration with the LMWG, active research was undertaken to understand and improve biases in the model related to high evaporation of intercepted water. Downscaling of rainfall within a grid cell is thought to be key to improving the interception of rainfall. The implementation of sunlit and shaded leaves in CLM is also deficient. Changes to this are needed to improve the simulation of gross primary production and also alleviate some of the low transpiration bias in CLM. Runoff generation based on a topographic index was also advocated, but not yet adopted for CLM.
New capabilities being developed for CLM by TSS scientists and their collaborators include: carbon and nitrogen cycles, mineral aerosols, dynamic vegetation, prognostic canopy air space, water isotopes, and land cover and land use change including an urban land cover parameterization.
A considerable amount of time over the past year was devoted to software engineering. The CLM code was revised to use a nested hierarchy of sub-grid types (grid cell, land unit, column, plant functional type). This allows multiple plant functional types to co-occur on a single soil column and to compete for below-ground resources, such as water, with several separate soil columns in a grid cell. It was revised, yet again, to support vectorization. The vectorization project, headed by Vertenstein with support from Forrest Hoffman (Oak Ridge National Laboratory), Dave Parks (NEC Solutions America), and Hideyuki Kitabata (Central Research Institute of Electric Power Industry), provides a single code for scalar and vector platforms, maintains the scientific functionality of the model, is portable to various machines, and does not significantly degrade performance on existing supported platforms.
Oleson continued to develop a state-of-the-art diagnostics package for CLM. New features including time series analyses, regional analyses, significance testing, and comparison of simulated river flow with station observations were added to the diagnostics package.
TSS scientists conducted several projects to implement biogeochemistry in CLM and CCSM. This research broadly addresses how biogeochemical coupling of carbon, nitrogen, iron and sulfur cycles affect climate, air quality, radiative forcing, and ecosystem function on regional to global scales. It involves two specific research agendas related to mineral aerosols and the terrestrial carbon cycle.
Mahowald and collaborators identified the sources, transport, and sinks of mineral aerosols using NCAR models and available observations, especially with regard to how humans may modulate this ‘natural’ aerosol. Model simulations of dust for many years using National Centers for Environmental Prediction (NCEP) reanalysis datasets showed that dust climatologies using NCAR models are good (in collaboration with Chao Luo at University of California, Santa Barbara (UCSB)). This work, though, also showed that dust sensitivity to meteorology and source parameterization are larger than the differences between including and excluding a land use source in the model. Further work with Masaru Yoshioka (graduate student, UCSB) suggests that even the TOMS AI (Total Ozone Mapping Spectrometer Aerosol Index), which has been used to argue that mineral aerosols have a largely natural source, is unable to distinguish between land use and natural sources included in the model. Collaborations with Jean-Louis Dufrense (UCSB and Laboratoire de Meteorologie Dynamique) have suggested that TOMS AI sensitivities to atmospheric boundary layer height may be responsible for some of the previous results identifying natural dry lake beds as the major sources of dust. This is due to the sensitivity of TOMS AI to aerosols which are higher in the atmosphere, and the fact that natural dry lake beds in the middle of desert regions tend to have higher boundary layer heights than marginal areas on the edges of deserts, where land use is likely. Collaborations with Rob Bryant (University of Sheffield) showed that hydrology in dry lake basins in Africa plays a complicating role in modulating dust emissions, and imply that human extraction of water from dry lake bed systems may also have a role in causing mineral aerosol source regions. The source of the temporal variability in mineral aerosols was addressed by studies in collaboration with Charles Jones (UCSB) and Chao Luo (UCSB) that quantified the role of African easterly waves and diurnal variability in driving mineral aerosol variability. Aerosol assimilation efforts by Phil Rasch (CMS) and Bill Collins (CMS) use the mineral aerosol model developed by Charlie Zender (University of California, Irvine) and Mahowald.
Using six different scenarios, Mahowald in collaboration with Luo (UCSB) has made the first estimates of future and preindustrial dust sources using CCSM simulations. These simulations suggest that mineral aerosols may decrease by 20-60% in the future, a result which depends on the model simulation used for the calculation. An important point from these simulations is that ‘natural’ aerosols such as mineral aerosols are very sensitive to human impacts on carbon dioxide, land use, and climate change, and because of the important role of mineral aerosols in modulating climate and biogeochemistry, may cause important feedbacks.
This figure shows mineral aerosol loading under different climate regimes using six scenarios of human/dust interactions in the source areas: no source changes (TIMIND), source changes due to precipitation and temperature (BASE), source changes due to precipitation, temperature and carbon dioxide fertilization (BASECO2), and the last three include a 50% landuse source (from Mahowald and Luo, 2003).
Efforts have also been directed at understanding the role of mineral aerosols on impacting climate and biogeochemistry. Mahowald and Lisa Kiehl (graduate student, UCSB) showed that mineral aerosols may interact with clouds in North Africa and the North Atlantic in both the liquid and ice phase, suggesting for the first time large scale interactions between mineral aerosols and clouds. Mahowald with Greg Okin (University of Virginia) and others have shown that mineral aerosol inputs to ecosystems such as the Amazon play an important role in supplying nutrients. Mahowald and Jenny Hand (Advanced Study Program (ASP)) have investigated the changes to iron solubility which occur while mineral aerosols are transported in the atmosphere due to clouds and solar radiation in collaboration with observationalists Ron Siefert and Yin Chen (University of Maryland). Work with Claire Mahaffrey and Ric Williams (University of Liverpool) has shown the role of mineral aerosols in supplying iron for nitrogen fixation in the North Atlantic during field campaigns, while work with Dave Siegel and others at UCSB have looked at the role globally of mineral aerosols in fertilizing ocean biota using satellite data.
This figure shows mineral aerosol and cloud interactions. Correlation between monthly anomalies (monthly mean - climatology monthly mean) of surface desert dust at Barbados and total Cloud Amount from International Satellite Cloud Climatology Project (ISCCP) for all clouds (a), low thin clouds (b) and high thin (cirrus) ice clouds (c) and high cirrostratus clouds (d). Assuming independent variables and normal distributions, correlations are significant at the 95% level at 0.2. These significant correlations are consistent with mineral aerosols interacting with both liquid and ice phase clouds in the North Atlantic.
In collaboration with Levis, Rasch and Zender, interactive dust has been put into the latest version of the CAM model. This dust will enable us to examine in more detail the interactions between climate and biogeochemistry in the CCSM framework.
Thornton continued research to develop the carbon and nitrogen biogeochemical algorithms for CLM. Thornton added fully prognostic terrestrial carbon and nitrogen cycles to CLM, building on the development of hierarchical data structures and below-ground resource competition described in the 2002 Annual Scientific Report. This work included bringing existing biogeochemistry algorithms from the Biome-Biogeochemistry (BGC) model into CLM and the development of several new algorithms to address the demands of sub-daily time stepping and geographically generalizable phenology algorithms. Scientific highlights of the new biogeochemistry version of CLM include: (1) Improvements to the treatment of sunlit and shaded canopy fractions and canopy integration of leaf morphology (specific leaf area) which dramatically improve the simulation of global total gross primary productivity (GPP); (2) Fully prognostic carbon and nitrogen state variables; (3) A phenology algorithm for global application, allowing realistic representation of growing seasons for drought deciduous vegetation (warm-season grasslands and tropical deciduous forest) and seasonally deciduous vegetation (cold-season grasslands and temperate deciduous forest); (4) Competition between plants and soil heterotrophs for common soil mineral nitrogen resource, as well as competition among multiple plant functional types for both mineral nitrogen and water resources. These interactions have important consequences for future evolution of co-limitations between the carbon and nitrogen cycles under a changing climate and changing atmospheric composition and chemistry; and (5) Integration of carbon and nitrogen biogeochemistry with disturbance effects such as mortality and fire. These processes are represented at a simple level now, with this as a major avenue for further research and development.
This figure shows results from two experiments modifying the treatment of canopy integration of radiation (Experiment 1, Figure 1a) and specific leaf area (Experiment 2, Figure 1b). Experiment 2 also includes the effects of Experiment 1. In both figures the upper left panel shows the annual maximum leaf area index (LAI) based on remote sensing observations, and the upper right panel shows the diagnosed gross primary production (GPP) required to sustain the observed LAI. The middle left panel in 1a shows GPP from the CLM control run, and the middle right panel shows the effect on GPP of improving the treatment of sunlit/shaded canopy radiation integration. The middle panels of 1b show the results from Expt 1 and the additional effect on GPP of improved treatment of vertical distributions of specific leaf area (a measure of leaf thickness). The bottom panels in both 1a and 1b show the differences of control, Expt1, and Expt2 GPP against diagnostic GPP.
Thornton is participating in the Coupled Climate Carbon Cycle Model Intercomparison Project (C4MIP). Initial model experiments focus on the terrestrial carbon cycle components of partially and fully coupled systems and represent the international state of the art in exploring and understanding these coupled dynamics. The CLM is currently the only participating model that includes coupling between carbon and nitrogen cycles. During the September, 2003, Hamburg workshop, Thornton demonstrated that the behavior of the carbon-only models is mimicked closely by the CLM with nitrogen dynamics switched off, but that the climate coupling dynamics are significantly different with the nitrogen dynamics turned on, pointing to an important dimension of uncertainty in the current best estimates of coupled system behavior.
This figure illustrates the influence of nitrogen cycle dynamics on the carbon cycle response to forced climate variation. The same model has been run with (solid line) and without (dashed line) nitrogen cycle dynamics. Results shown for a single grid cell in the temperate deciduous broadleaf forest region of eastern U.S.
Thornton continued to collaborate actively with researchers in the AmeriFlux and FluxNet communities, to make use of eddy covariance observations of water and carbon fluxes and other detailed site-level observations to evaluate the performance of the stand-alone version of the CLM biogeochemistry model (Biome-BGC). Working closely with Bev Law and colleagues at Oregon State University, Thornton compiled a database of leaf-level gas exchange observations and soil respiration chamber observations across multiple sites spanning wide gradients in climate and plant functional type. They are using these observations to help separate the autotrophic versus heterotrophic components of net ecosystem carbon exchange, providing additional constraints for model evaluation. A new cross-site collaboration was initiated at the 2003 AmeriFlux workshop held in Boulder in early October to use observations from all of the available deciduous broadleaf forest sites to evaluate that aspect of the Biome-BGC model. Additional collaborations just in the planning phase now are: to perform detailed evaluations of nitrogen fertilization response using observations from a manipulation experiment in evergreen forest in Maine; and to perform cross-site evaluations of the model performance in tropical evergreen forest in collaboration with the NASA-sponsored Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA).
Thornton and Nan Rosenbloom (TSS) continued to explore new spin-up approaches for the Biome-BGC model, using numerical multivariate optimization methods. A combination of rate constant acceleration and the Downhill Simplex optimization method gives the best results, considering both computational cost and final deviation from true steady-state conditions. More sophisticated methods such as the conjugate gradient approach can give better answers, but the computational expense outweighs these benefits.
A major research focus for TSS is natural and human-mediated changes in land cover and ecosystem functions and their effects on climate, water resources, and biogeochemistry. TSS scientists worked on several projects to implement land cover and land use change in CLM and to use climate models (CAM, CCSM) to study the impact of these processes on climate.
Bonan, Oleson, Johan Feddema (University of Kansas), and Linda Mearns (Environmental and Societal Impacts Group (ESIG)) studied the effects of historical land cover change on global climate. Climate model simulations were performed with the Parallel Climate Model (PCM) and examined the sensitivity of simulated climate to different specifications of present-day land cover and natural potential vegetation. Uncertainty in the classification of present-day vegetation can produce large differences in the simulated climate.
This figure shows sensitivity of simulated springtime climate (March-May) to two different datasets of present-day land cover. Simulation B07.43 uses a dataset that recognizes only a single boreal forest vegetation type (needleleaf evergreen tree). This results in a lower surface abledo than simulation B06.62, which recognizes needleleaf evergreen and needleleaf deciduous trees. As a result, simulated surface air temperature is several degrees warmer in northern Russia and Siberia compared to B06.62. These simulations use PCM as the climate model.
Present-day vegetation has generally cooled surface climate, especially in the mid-latitudes due to the higher albedo of croplands compared to natural vegetation.
This figure shows simulated summer climate (June-August) with present-day vegetation (B07.43), which includes croplands, and natural vegetation (B07.09), without croplands. By increasing surface albedo, croplands in North America, Europe, and parts of Asia have cooled summertime climate. These simulations use PCM as the climate model.
Bonan, Oleson, Feddema, and Vertenstein also worked to develop and implement an urban land cover parameterization for CLM. The parameterization uses concepts from urban canyon models to simulate the radiative balance of a city, turbulent energy fluxes, and the hydrologic cycle.
Levis and Bonan applied a model of natural vegetation dynamics, developed for use with CLM, to understand how past land cover change in response to climate change has affected climate. Simulations with the fully coupled CCSM examined the impact of vegetation on the North African climate 6000 years before present. A stronger African monsoon at this time allowed vegetation to expand northwards into desert landscapes. The climate model simulations showed that the greener North Africa, with more vegetation and an associated change to richer soils, strengthened the monsoon by reducing surface albedo and increasing soil water holding capacity.
Through the activities of Dave Schimel (TSS), the section holds a leadership position in the development of data assimilation techniques for biogeochemistry and carbon cycle studies. The carbon data assimilation activity expanded to include models on three scales.
A local scale model, developed with R.H. Braswell (University of New Hampshire), assimilates CO2 flux observations and has been used to analyze seasonal and interannual controls at the Harvard Forest in Massachusetts and at flux sites in Brazil as part of the LBA. This system uses a Markov Chain-Monte Carlo approach and ensembling to estimate parameters as distributions and forms the basis for rigorously estimating local ecosystem model parameters and functional relationships for use in coupled models.
At the mesoscale, the Colorado State University assimilation system, (Regional Atmospheric Modeling System and Data Assimilation System - RAMDAS) has been ported to NCAR and is being used to design optimal approaches for sampling CO2 concentrations in order to retrieve surface fluxes in complex terrain. RAMDAS has a sophisticated land surface scheme, but over the next year, the CLM, which incorporates CO2 fluxes will be added as an option. The RAMDAS system will support design and execution of surface and airborne field studies for an NSF Biocomplexity-sponsored research program on carbon fluxes in mountain landscapes, in collaboration with the University of Colorado, Colorado State University and University of Miami, Florida.
The global Carbon Data Assimilation Model, developed by David Baker (TSS visitor) and collaborators, is now operational and couples a simple atmospheric transport model, the transport model adjoint and an optimization scheme to estimate surface fluxes globally from concentration observations. Current studies emphasize analyzing optimal sampling schemes for estimating global surface fluxes and high-resolution regional fluxes. Studies using the 4-dimensional variational assimilation approach are beginning to unravel the interactions between the atmospheric sampling scheme, assumptions about spatial and temporal coherence of concentrations in the atmosphere, and the resolution of the estimated flux field. Preliminary results strongly suggest that 1) relatively high sampling density is needed to resolve regional fluxes, independent of model skill, 2) that continuous observations are far more influential on the model solution that episodic (e.g. flask) measurements, and that 3) assumptions about time and space correlations have a large impact and must be based on careful analyses.