Land Surface Modeling: A Mini-Workshop
4. Workshop Presentations
The science discussions were introduced with several overviews, and organized into three broad areas, each with several invited speakers: 1) Traditional Land Surface Modeling Approaches, discussing the origins and current status of land-surface modeling in GCMs, especially the soil-vegetation atmosphere transfer schemes, 2) Present Focus, discussing some current developments, especially dynamic vegetation modeling, and 3) Future Directions, discussing some ideas for possible new thrusts in global modeling.
An overarching question for the workshop was: What are the land-surface modeling requirements for studying interannual to multi-decadal climate variations? This issue was reviewed by Robert E. Dickinson, who emphasized that land schemes must provide 1) land-surface temperatures; 2) land-surface moisture and hydrology; 3) feedbacks on the atmosphere; and 4) terrestrial biogeochemical cycling. For those purposes, land schemes must determine resistances to sensible and latent heat fluxes which in turn depend on soil moisture. They need to adequately move water across the landscape and into stream flow, provide temporally varying albedos (considering in particular the effects of snow), represent profiles of soil moisture and soil temperature, and, account for the spatial heterogeneity of these processes and characteristics. This can only be achieved if the atmospheric model provides correct inputs to the land scheme. Dickinson suggested defining the success of land schemes based on process validation and whole-model validation (e.g., impact on entrainment, boundary layer clouds, thermodynamic environment for moist convection).
Jon Foley emphasized the 20th century's rapid increase in world population, the impact of this growth on grain and water use, and consequences of forest and grassland conversion into cropland and pasture. He described the Integrated Biosphere Simulator (IBIS) consisting of the following interacting modules in a hierarchical framework: land-surface (forced by observed or modeled atmosphere), carbon balance, vegetation phenology, and vegetation dynamics, and below-ground carbon and nitrogen cycling. This scheme simulates physical processes (energy, water, and momentum balance) and physiological processes (photosynthesis, respiration). The dynamic vegetation component allows for transient changes in the vegetation cover, and the sensitivity of biological processes to CO2 concentrations. Foley demonstrated the behavior of this model in paleoclimate simulations, and emphasized the need for such models in long-term climate simulations.
4.2 Traditional Land Surface Modeling: Soil Vegetation Atmosphere Transfer Schemes (SVATS)
Randy Koster described the catchment-based approach for modeling land-surface processes, i.e., using the hydrological catchment as the fundamental land unit rather than the regular grid adopted in atmospheric models. He emphasized (using PILPS results) that both evaporation and runoff formulations control annual evaporation. The less accurate of these formulations determines the evaporation error. He emphasized the need to improve the treatment of subgrid-scale soil moisture heterogeneity and its effects on runoff generation. His preliminary results illustrate the key role of spatial heterogeneity in determining runoff.
Cynthia Rosenzweig described the land scheme currently implemented in the GISS GCM. The basic structure is defined by Abramopolous et al. (1988) and Rosenzweig and Abramopolous (1997). Each gridbox has specified fixed fractions of up to ten vegetation types, bare soil, ice and water. The scheme is conceptually similar to other SVATS, but includes six soil layers with parameterized underground runoff.
Gordon Bonan described general aspects of the land surface model (LSM) developed for the NCAR GCM (CCM). LSM, which encompasses atmosphere-biosphere exchanges of water, energy, and carbon, has been used to assess the role of vegetation in boreal climates, as well as the impacts of land cover change during the last century, on surface energy fluxes and air temperature over the US. Bonan reported a cooling, associated with land-use change, resulting from higher albedo of snow-covered farmland in winter and a decline in diurnal temperature range in summer. Net radiation declined in the eastern half of the US with the introduction of cultivation due to modest increases in sensible heat fluxes and large declines in latent heat fluxes over the region.
Marc Stieglitz provided an overview of the climatic importance of snow as a controller of albedo and energy fluxes, as a ground insulator that impacts hydrologic and biologic processes, and as a water source. He described his snow model which is currently being implemented in the GISS and GSFC GCMs. The 3 snow layers allow for sensible and latent heat exchange between the atmosphere and snow surface, and downward movement of water within the snow pack and to the ground surface. The initial version of the snow model adds substantially to the computation time of the global model, but it is anticipated that this problem can be improved with some attention to the snow model's time step, especially when the snow layer is thin. Evaluations of the snow model performance in the GCM and offline are being carried out.
David Robinson discussed a suite of snow data available for evaluation of integrated land-snow parameterizations. This set comprises data from 1972 to the present integrated from NOAA weekly snow charts derived from visible satellite observations, and daily charts using visible and microwave satellite observations and ground-station observations. The Robinson data set, at 1° latitude/longitude spatial resolution and pentad temporal resolution, includes surface albedo, snow cover extent, and station snow depth from 1972 onward. Robinson stressed the importance of the 18 month overlap period built into efforts currently underway at NOAA to switch from a weekly to a daily hemispheric snow product.
Michael Coe presented the global HYDrological Routing Algorithm (HYDRA) which builds on previous work representing annual states by simulating seasonal river discharge, and seasonal changes in lake and wetland area and volume, on a horizontal resolution of 5' latitude/longitude. HYDRA inputs are monthly mean runoff, precipitation, and evaporation from a GCM, reanalysis data, or observations. Initial studies forcing HYDRA with NCEP reanalysis data indicate that 1) discharge is overestimated for most rivers in the northern high-latitudes, mid-continental North America, eastern Europe, central and eastern Asia, India, and northern Africa, 2) including lakes reduces the amplitude of the seasonal cycle of river discharge as well as the magnitude of annual discharge from rivers, and 3) using Lake Chad as a test case, HYDRA captures realistically the seasonal fluctuation of lake levels and wetland areas.
Elaine Matthews provided an overview of the types of data required for initializing and validating land-surface models including data that are integrative with respect to processes, synoptic so as to minimize instrument merger for measurements for single periods, representative of seasonal, interannual, and decadal variability, and calibrated so as to provide long-term records from a succession of instruments. In summarizing the types of data required (hydrology, vegetation, disturbance, carbon stocks and exchanges, surface energy) she indicated that aside from several global data sets such as temperature, precipitation, and radiation, many of the data are in the form of individual or suites of field measurements, or are proxy and derived measurements. Eric Wood noted the availability of the ISLSCP CD which represents the latter types of data, and Alan Robock brought attention to the Global Soil Moisture Data Bank, a cooperative project to assemble field measurements of soil moisture and related variables. The soil moisture data set comprises field measurements of soil moisture at different depths for locations in the former Soviet Union, India, the United States, Brazil, China, and Mongolia.
Eric Wood summarized results of several phases of the Project for the Intercomparison of Land-surface Parameterization Schemes (PILPS). The first phase focused on a basic comparison of scheme behavior under artificial (GCM-generated) forcing. The second phase focused on validation by comparing model results with field observations. Based on PILPS intercomparisons, Wood concluded that most SVATS have similar latent heat parameterizations, although substantial differences in surface energy balance were obtained for the numerous schemes used in the intercomparison. He questioned whether these differences were produced by the parameterization of temperature and/or ground heat flux. He also noted that these schemes give very different results regarding runoff and evaporation, a difference he attributes to soil column hydrology, especially the drainage runoff component. From this analysis, it appears that 3-4 vertical soil layers and some representation of horizontal heterogeneity are needed in the SVATS approach.
4.3 Present Focus
Andrew Friend described the dynamic vegetation model based on his HYBRID model (Friend et al., 1997). The model focuses on interactions among the atmosphere, vegetation, and soils that impact seasonal-to-centennial climate predictions. Based on developmental and biophysical properties of plants, HYBRID operates on a diurnal cycle at any spatial resolution, and dynamically models combinations of general plant types, photosynthesis, respiration, plant competition, nitrogen uptake, energy balance and transpiration, soil organic matter decomposition, and soil mineral N availability. HYBRID is being merged with the soil physics of the current GISS scheme and further developed for mechanistic incorporation of land use effects. Improved predictive capability is expected due to simultaneous treatment of multiple feedbacks between carbon, nitrogen, and water cycles. The model also allows inclusion of transient anthropogenic and natural disturbance, response, and recovery, which is useful for climate, carbon and nitrogen studies.
4.4 Future Directions
Chris Field described CASA, a global biogeochemical model driven with climate and a suite of remotely-sensed data. He stressed the importance of incorporating full nutrient cycles (above- and below-ground), influences of water and temperature stress, dynamic allocation of photosynthate, and transient biospheric responses to environmental change.
Jim Hansen and Elaine Matthews presented material on the declining growth rate of atmospheric methane concentrations, especially during the early 1990s. Underlying causes for this trend could be decreasing emissions, increasing chemical destruction, or some combination of the two. About 75% of annual methane emissions are from anthropogenic sources; the largest natural source is wetlands whose responses to climate change or to the temporary Pinatubo cooling are unknown. Recently, several global models of methane emission from natural wetlands have been developed and tested; M. Coe has reported on a linked global model of surface water including lakes, rivers, and wetlands; and techniques to correct microwave data to improve their use in characterizing spatial and temporal variations in wetland area have been developed. This new suite of models and data forms the basis on which to begin incorporating wetlands into GCMs in order to model their short- and long-term impacts on biochemistry (methane) and hydrology in current, paleo, and future climates.
Roni Avissar indicated that among all land-surface characteristics considered in SVATs, only stomatal conductance, soil-surface wetness, leaf area index, surface roughness, and albedo play a major role in the redistribution of energy at the ground surface and that the relation between these characteristics and the surface heat and momentum fluxes is strongly nonlinear. Furthermore, he argued that landscape discontinuities resulting from spatial heterogeneity of surface characteristics could induce atmospheric mesoscale circulations which have a strong impact on the structure of the planetary boundary layer, clouds, and precipitation. He indicated that SVATS for GCMs need to provide higher statistical moments and characteristic length scales of the spatial distribution of the land-surface characteristics mentioned above, and proposed a higher-order SVAT scheme that provides the variance of land-surface heat fluxes. This scheme is designed to directly use remotely sensed parameters.