Feedback processes over land are critically important to understanding the climate response over land and its effect on humans. The responses of the hydrologic and energy cycles over land play a critical role in determining the impacts of climate change on water resources, carbon stocks, and agriculture, yet these responses vary widely among different climate models. Unfortunately, basic climate processes such as the response of the land-atmosphere system to diurnal variations of insolation are poorly simulated in current climate models. Snow and ice melting and their associated hydrologic and radiative consequences tend to be poorly simulated, and dynamic vegetation modeling is in its very early stages.
We recommend an integrated analysis of the diurnal and annual cycles of the energy, water (in all its phases), and carbon budgets at the land-surface and through the atmospheric boundary layer for different ecosystems and climatic regimes, including managed ecosystems like irrigated cropland. This analysis—aimed at improving theoretical understanding and model parameterizations—needs to fully integrate land and atmosphere processes and use carefully designed observational metrics to test modeled processes, which must be robust in the face of time-varying land surface properties. Sustained multiyear observations of terrestrial ecosystems, their functioning, and their role in the climate system should be encouraged, to contribute to the development and improvement of process-oriented vegetation models for use with climate models.
The global, annual-mean surface temperature is the most widely used first-order measure of climate change. However, in assessing the impact of climate change the water balance over land is at least as important. Terrestrial surface hydrologic changes are important for human requirements such as drinking water, sanitation, agriculture, transportation, and energy supply. These changes are also important for the response of natural ecosystems on land to human-induced climate change. The variables used to measure changes in the surface water balance are precipitation, evaporation, and runoff rates, as well as soil and surface water storage. These quantities are related to temperature, wind, cloudiness, vegetation characteristics, and other climate system variables.
For this report the primary interest in terrestrial hydrology is its role in climate change feedbacks. In the tropics, interactions among land hydrology, vegetation, and surface energy balance can foster feedback mechanisms that may cause expansion or contraction of deserts, for example. In middle latitudes interactions between winter snowfall, spring snowmelt, and summertime convection can lead to potential changes in water availability during the growing season that may pose a substantial threat to agriculture. Earlier snowmelt can lead to more rapid drying, reduced summertime precipitation, and increased surface temperature over land. These feedbacks and the response of mid-latitude land hydrology to climate change and global warming are highly uncertain. Better characterized or reduced uncertainty in projections of the response of land hydrology to global warming would have important implications for the development of mitigation and adaptation strategies.
The feedbacks between soil water, evaporation, precipitation, and runoff are an integral part of the hydrological cycle over land, as are the interaction of vegetation and the frozen hydrology (ground and snow) at high latitudes, with their impacts on albedo and the availability of water for evaporation. Climate change, driven globally by the global rise of greenhouse gases, will have regional impacts on this hydrological cycle, differing across latitude and across continents. Currently our confidence in regional projections is limited by our lack of understanding of the processes and feedbacks controlling the precipitation-evaporation difference, both over land (where it is fundamental to the long-term drift of the hydrological cycle) and over the ocean (where it is one key component impacting the surface energy budget and changes in the thermohaline circulation). Improving confidence in regional temperature and freshwater resource projections is also intimately linked to a better understanding of the coupling between surface processes and the atmospheric boundary layer.
In addition to the above concerns, recent estimates are that over 30 percent of the discharge of the world's rivers is actively managed. This has occurred through the construction of some 40,000 major dams and diversion structures, which have been capable of changing the hydrologic regimes of the world's major rivers and potentially the global water cycle. The feedbacks between the managed portion of the terrestrial water cycle and other components have received virtually no attention, but are potentially important because the effects of water management on natural hydrographs are far larger than those projected to be caused by climate change. These changes in the discharge regimes of large rivers are known to have changed ocean circulation in the vicinity of river mouth estuaries, and perhaps at larger scales. Furthermore, changes in vegetation, many of which are related to water management, are known to have caused changes in the local cycling of moisture in the land-atmosphere system (e.g., Stohlgren et al., 1998), and anthropogenic changes in land cover due to management have been shown to have affects at global scales (Chase et al., 1996, 2000).
Reducing model uncertainty can be achieved in part through improved understanding and projections of the regional long-term drift of the hydrologic cycle over land. These improvements are fundamental to projecting ecosystem dynamics on decadal timescales. At present, coupled global models differ widely in their regional forecasts for future trends in the hydrologic cycle. The U.S. Water Cycle Initiative (USGCRP, 2001) has outlined several important science goals that need to be addressed to improve our ability to model the global and regional water cycle. A focused research effort is required to improve these models.
From a scientific perspective this area of research is ripe for progress. Indeed, in the past five years considerable progress has been made in understanding soil water feedback, in making soil water measurements, and in the development of land surface data assimilation systems that indirectly provide soil water fields on continental to global scales. The challenge ahead is synthesis, because the water cycle plays such a central role and it interacts directly with much of the climate system, and in particular over land with the energy and carbon cycles.
Overview of Terrestrial Hydrology Feedbacks Soil Water Feedbacks
The land-surface reservoirs of available soil water are small compared to the ocean reservoir, typically of order 0.1-0.6 m of water. However, their role is crucial to the surface climate over land because evaporation from the near surface soil layer and transpiration of water extracted by vegetation from their root zone is a major component of the surface energy balance. Over wet soils the daily mean Bowen ratio (the ratio of the sensible to latent heat flux) may be of order 0.5, while over dry soils when the vegetation experiences water stress, the Bowen ratio may exceed 1. In turn the increased evaporation over wet soils can lower maximum surface temperature by several degrees.
A feedback arises because increased surface evaporation over large land areas gives rise to increased precipitation (e.g., Beljaars et al., 1996), which maintains soil water levels. This is primarily a feature of the warm season, and wet regions of the tropics. Over the continents as a whole, precipitation minus evaporation (P-E) is positive, which contributes the runoff of fresh water to the oceans. Correspondingly E-P is positive over the oceans. However, the balance between P and E varies widely, both spatially and temporally. The monsoon circulations concentrate the flux of moisture from ocean to continents. Over large regions of land in summer (remote from the summer monsoon) P is more closely in balance with E. There is also a large seasonal cycle in which winter precipitation adds to soil water reservoirs in mid-latitudes and to snow accumulation at high latitudes, and this is drawn down in spring and summer by both runoff and evaporation. The diurnal cycle of precipitation and cloud is also involved, because transpiration depends on daytime solar radiation (and thus involves critically the shortwave cloud feedback), while the equilibrium temperature of the land-surface (which determines outgoing longwave radiation) is sensitive as well to the impact of clouds on the long wave balance, especially at night.
Although it seems clear that warming will accelerate the hydrological cycle, the net change of P-E and runoff over land, particularly in the warm season for specific regions, remains uncertain. At high latitudes it is possible that the likely increase in winter precipitation will lead to increased snowpack and spring runoff, but the impact on summer soil water, evaporation, and precipitation remains uncertain.
The complex couplings among precipitation, evaporation, soil water, runoff, the cloud fields, net radiation balance, and the vegetation on the diurnal timescale have not yet been modeled satisfactorily for the present climate. Typically this leads to fundamental errors in the diurnal cycle of precipitation (Betts and Jakob, 2002). The diurnal time scale, together with the continental scale circulation dynamics and the seasonal cycle, must be accurately modeled so that confidence can be assigned to projections of land hydrology changes associated with global warming.
One of the most important climatic characteristics of snow is its albedo. Fresh snow on a fully covered surface has an albedo of approximately 0.8. Aging of the snow will reduce this to about 0.4. Snow in tree-covered landscapes has an albedo of about 0.2 to 0.4 depending on the vegetation cover type.
In the fundamental snow-albedo feedback a decrease in snow extent decreases the surface albedo, which tends to increase surface temperature. These changes can affect large-scale circulation and planetary albedo which in turn, can affect subsequent snow precipitation and melt rates. There are many confounding factors to this picture, including the effects of vegetation and snow age.
Trees and other vegetation can protrude over snow and mask its high albedo. As a result treeless areas have a higher albedo when snow is on the ground than do forests. Numerous climate model studies have found that the presence of the boreal forest warms climate compared to tundra (Bonan et al., 1992; Douville and Royer, 1996). Forest and tundra ecosystems also differ in how they partition net radiation into sensible and latent heat fluxes. For example, albedo differences in snow-covered and adjacent snow-free forests can result in local energy circulations with advection of energy to the forests (e.g., Taylor et al., 1998)
At the regional scale the removal of snow cover may affect the thermal and dynamical structure of the atmosphere, but the temporal persistence of these effects is uncertain (Yeh et al., 1983). To date, one of the strongest pieces of evidence of a snow cover and weather feedback is the connection between springtime air temperature biases (5-10oC low in ECMWF weather predictions due to specified high-latitude snow albedo, which was biased high (by approximately 0.4) in the model (Viterbo and Betts, 1999). This result demonstrates the effect of large-scale snow cover on near surface air temperature, with the correction of the albedo bias correcting the air temperature bias.
In general the existence of a feedback mechanism between snow cover extent and continental- to global-scale weather and climate requires that snow processes affect atmospheric circulations at these scales. There are model results that suggest such an effect. For example, the interannual variations of the Asian summer monsoon rainfall have been significantly correlated with the tropical sea surface temperature and the Eurasian snow cover anomalies (Bamzai and Shukla, 1999; Corti et al., 2000).
In general, current models represent mean global snow cover fairly well but are less accurate in representing interannual snow cover variability. In general even off line terrestrial hydrologic models, forced with observed meteorology and radiation, tend to underestimate the observed variability in the record. The biggest difference in the predictions among models occurs in snow transition regions. Models do fairly well in the snow accumulation season but differ greatly from observations in the melt season, resulting in different predicted time of end-of-melt that varies by two to three weeks. This can affect the subsequent prediction of the onset of vegetation activity.
The largest modeling challenge related to snow melt is representing sub-grid snow cover at GCM grid scales. Accurately estimating the albedo of retreating snow cover involves accounting for factors such as snow patchiness and snow age. The modeling of these effects is well understood, but has mostly been carried out over idealized domains where the contrasts may not represent the variability observed in natural landscapes. The modeling of natural domains requires high-resolution modeling and the accompanying forcings.
In validating model predictions against observations a significant problem is the observational bias that results from the placement of instrumentation in clearings and the rather different snow dynamics of forested and cleared areas.
At the other end of the scale spectrum global modeling assessments of the snow-climate feedback have been rather limited and the results show discontinuous areas having correlations between snow extent and the Indian monsoon rainfall. The scope of these studies should be expanded to rigorously diagnose large-scale effects of snow cover on circulation and the planetary albedo. An important part of this work is the boundary layer coupling between the snow-covered surface and overlying atmosphere.
Modeling the melting of snow in springtime is important to correctly simulating the role of snow in climate feedbacks over land. Key issues in snow melt modeling include
• improving space-time distribution of snow. During the snow accumulation period most model prediction problems are largely attributable to precipitation and temperature surface forcing, while during snow ablation periods poor snow model predictions are more closely related to the model parameterizations related to surface energy transfer.
• evaluating errors in space-time extent. There is a need to evaluate and improve the quality of data; use this data for error diagnostic studies with a focus on transient zones at regional and continental scales; and better utilize offline evaluation methods.
• developing better global databases for model parameters (e.g., surface roughness, vegetation solar radiation extinction, canopy closure, snow patchiness functions).
• developing point or small area datasets for offline model evaluation across a range of snow climatologies and vegetation types, and for the evaluation of new model parameterizations.
The traditional view of terrestrial vegetation is that community composition and ecosystem structure are determined by climate. However, this is only part of the interaction of ecosystems with climate. Terrestrial ecosystems affect climate through exchanges of energy, water, momentum, CO2, and other radiatively important atmospheric gases. Changes in community composition and ecosystem structure alter albedo, surface roughness, stomatal physiology, leaf area, rooting depth, and nutrient availability and in doing so alter surface energy fluxes, the hydrologic cycle, and biogeochemical cycles. As a result, changes in ecosystem structure and function and the replacement of one ecosystem with another in response to climate change feed back to influence climate. The IPCC TAR has identified changes in land cover as a potentially important climate feedback.
Most studies of vegetation feedbacks have focused on biogeophysical processes related to energy, moisture, and momentum exchange with the atmosphere. Biogeochemical feedbacks are only now being included in climate models (Cox et al., 2000; Friedlingstein et al., 2001). This review focuses on biogeophysical feedbacks, considering a continuum of processes and time scales from physiological (minutes) to phenological (seasons) to vegetation dynamics (decades to hundreds of years).
Overview of Vegetation Feedbacks
The partitioning of net radiation into sensible and latent heat fluxes by vegetation is regulated in part by canopy conductance. Studies of the physiological response of plants to short-term exposure to enhanced CO2 concentrations routinely find reduced stomatal conductance and greater photosynthesis. Climate model simulations in which stomatal conductance decreases with a doubling of atmospheric CO2 routinely show decreased latent heat flux, increased sensible heat, and surface warming over large vegetated regions in summer (e.g., Sellers et al., 1996). In general, the physiological effects of doubled CO2 amplify the warming associated with the radiative effects of doubled CO2.
Previous climate model studies highlight the potential for physiological feedbacks from vegetation (e.g., Sellers et al., 1996). It is quite likely that changing atmospheric CO2 concentration will alter the physiology of plants and through this affect climate. However, we cannot yet quantify this feedback with certainty and rank it relative to other climate feedbacks. Uncertainty in its magnitude and importance arise for several reasons. First, physiological processes operating at the scale of an individual leaf need to be scaled to a canopy of leaves and then to a landscape of thousands of plants. There are few observations to guide this scaling, as most studies of stomatal conductance and its response to CO2 are obtained from leaf measurements. Second, most studies examine the short-term response of plants to CO2. Long-term acclimation to high CO2 may alter the short-term reduction in stomatal conductance. Third, the reduction in stomatal conductance observed in the laboratory may not be realized in the field, where many other environmental factors (e.g., dry soil, low nutrient availability) also limit photosynthesis. Finally, atmospheric CO2 is also known to alter the allocation of carbon to the growth of foliage, stem, and root biomass and the chemical quality of plant material. This is likely to affect climate by changing, for example, the amount of leaf area from which heat and moisture can be exchanged with the atmosphere or by changing the amount of carbon stored in the soil.
The seasonal emergence and senescence of leaves on deciduous trees alters albedo and sensible and latent heat fluxes and in doing so alters surface climate, including temperature and transpiration (Fitzjarrald et al., 2001; Schwartz, 1999). In the eastern United States, springtime air temperatures are distinctly different after leaves emerge (Schwartz, 1992, 1996; Schwartz and Karl, 1990). This temperature discontinuity over a period of less than a few weeks is related to increased transpiration upon leaf emergence that cools and moistens air. A similar distinct seasonal pattern to air temperature coinciding with the absence or presence of leaves on deciduous trees is seen in west central Canada (Hogg et al., 2000).
Because of the importance of foliage in regulating surface climate, improved representation of leaf area and its phenology are being implemented in climate models. In general, higher leaf area increases evaporation over vegetated regions in summer provided there is sufficient soil water (e.g., Buermann et al., 2001). As a result surface temperature cools and precipitation increases. Prognostic models of leaf area in which the amount of foliage depends on temperature, precipitation, and plant productivity are being included in the land models used with climate models. One study with interactive leaves found increased air temperature and reduced evaporation and precipitation over extratropical regions of the Northern Hemisphere in summer as result of lower leaf area (Dickinson et al., 1998).
As with stomata, leaf area must be considered a "known unknown" in its magnitude and importance as a climate feedback. Observations of temperature and leaf phenology demonstrate a change in temperature with leaf emergence, but prognostic leaf phenology is a new process for land-surface models. There is not a long history of climate model experiments to demonstrate the robustness of this feedback among climate models or to determine the key ecological processes regulating leaf area in a coupled climate-vegetation model.
Vegetation changes naturally over time in response to recurring disturbances and also in response to climate change. Fires, insect outbreaks, and windstorms that kill large tracts of trees initiate a process of revegetation and ecosystem recovery known as plant succession. A forest, for example, may undergo successive transformation from bare ground to herbaceous species to shrubs to young forest to mature forest following fire. Climate change that may, for example, convert a forest to grassland is superimposed on this successional development. This vegetation dynamics and change from one vegetation type to another alters numerous surface properties such as albedo, roughness, stomatal physiology, leaf area, and rooting depth and in doing so can alter climate.
The impact of vegetation dynamics on climate is seen regionally in the Sahel of North Africa and along the boreal forest-tundra ecotone. Precipitation limits the northward advancement of grasses and shrubs into the Sahara Desert. Temperature limits the northern extent of trees into tundra. In both these regions climate model simulations show amplification by vegetation of the climate response to changes in precipitation or temperature. Expansion of grasses and shrubs into desert in response to enhanced summer precipitation results in more precipitation (Claussen et al., 1999; de Noblet-Ducoudre et al., 2000; Kutzbach et al., 1996). The boreal forest warms climate compared to tundra as a result of the lower winter albedo of forest (Bonan et al., 1992; Foley et al., 1994).
Global vegetation models have been developed to allow interactive coupling of climate and vegetation. One approach, known as asynchronous equilibrium coupling, takes advantage of the relationships between climate and biogeography to interactively change vegetation cover (Claussen, 1994). Climate is simulated with an initial vegetation cover. This climate is used in a biogeography model to simulate the geographic distribution of vegetation. This map is then input to the climate model to obtain a new climate. Climate is iterated in this manner several times until a stable solution is obtained. Another type of model, known as a dynamic global vegetation model, explicitly simulates transient vegetation dynamics (Foley et al., 1998, 2000).
Coupled climate-vegetation models show that vegetation feedback amplifies the climate response to solar radiation or atmospheric CO2. For example, the colder climate as a result of reduced solar radiation and lower atmospheric CO2 some 115,000 years ago is not in itself enough to initiate an ice age. However, the associated reduction in the geographic extent of the boreal forest and the expansion of tundra due to the cold climate produces additional cooling that is sufficient to initiate an ice age (de Noblet et al., 1996). Coupled climate-vegetation models highlight the importance of the treeline in reinforcing the cold high-latitude climate of the last glacial maximum 21,000 years ago and the high-latitude warming 6,000 years ago (Kubatzki and Claussen, 1998; Levis et al., 1999; Texier et al., 1997). Other studies show that changes in the geographic extent of vegetation enhance the orbitally induced summer monsoon 6000 years ago in North Africa (de Noblet-Ducoudre et al., 2000; Doherty et al., 2000; Texier et al., 1997). The doubling of atmospheric CO2 from pre-industrial levels is likely to result in changes in ecosystem structure and function in response to altered temperature, precipitation, and CO2 fertilization. Climate simulations with coupled climate-vegetation models show large changes in climate as a result of vegetation changes (Betts et al., 1997, 2000; Levis et al., 2000).
As with stomata and leaf area, the inclusion of interactive vegetation in climate models is relatively new. Initial work with these models has demonstrated the potential for large feedbacks with climate. Future work must demonstrate the robustness of these feedbacks and reduce the uncertainty in these simulations.
Key aspects of the required research strategy are discussed below.
As described in the previous sections, several potential feedbacks exist between vegetation and climate, including radiative (albedo), physiological (stomata), micrometeorological (sensible and latent heat), hydrological (snow, soil water), biogeochemical (carbon and other greenhouse gases), and ecological (leaf area, biogeography). These are often viewed as separate areas of research. In particular, our understanding of fundamental vegetation processes and their inclusion in climate models suffers from the broad multidisciplinary scope of the potential interactions. There is not a coordinated research agenda to understand and model their potential feedbacks.
We still lack the simplified theoretical models needed to generalize our understanding of the feedbacks between the coupled energy and water cycles across different climatic regimes. Theoretical work on the diurnal cycle of the coupled land-surface-convective boundary layer system for different ecosystems and seasons would lead to improved understanding of this basic climatic control on cloud, radiation, and water cycle feedbacks over land. In addition, more work on the coupling of soil water, resistance to evaporation, lifting condensation level, and cloud base (observable from the ground by lidar ceilometers) would deepen our understanding of the land-surface and soil water controls on atmospheric subsaturation, cloudiness, and precipitation. Theoretical work on feedbacks and other interactions between ecosystems, biogeochemistry, and hydroclimatic processes at a very wide range of time and space scales also requires further development.
Progress in understanding terrestrial feedbacks depends critically on both systematic analysis of data generated by advanced, integrated observational datasets, and on careful testing and improvement of coupled modeling systems. Advancements in understanding of the processes responsible for terrestrial hydrology and vegetation feedbacks could be greatly facilitated by a program of integrated observations and analysis of the diurnal and seasonal cycles of the energy, water, and carbon budgets at the land-surface and through the atmospheric boundary layer. In addition, longer-term measurements and analyses of interannual ecosystem and hydrologic variability are important.
An important focus of research on terrestrial feedbacks should be on improving the parameterization of dynamic vegetation in climate models. This work must treat energy, water, carbon, and nutrients as a single system rather than as disciplinary components. Observations must be made to better understand the natural processes, improve the parameterizations, and test those parameterizations in coupled models. By focusing systematically on this joint observational and modeling problem, the various scientific communities that monitor, study, and model terrestrial vegetation may be spurred toward better integration in much the same way that coupled atmosphere-ocean models led to integration of atmospheric and oceanic sciences.
As discussed below, a global network of surface flux tower sites exists (Baldocchi et al., 2001), but many analyses have a narrow focus on, for example, the carbon balance at the site rather than the full energy, water, and carbon balance and their coupling to the boundary layer, and its cloud field. It is rare for example that sites measure boundary layer height, structure, cloud base and cloud cover (even though this can be done remotely) or the soil water profile. Yet the photosynthetic processes are tightly linked both to the soil hydrology, the surface energy balance (which depends on the clouds and the radiation field, whether direct or diffuse), as well as the coupling to the boundary layer over the diurnal cycle. Our ability both to measure (the fluxes) and to model the nighttime stable boundary layer is still unsatisfactory, and progress probably requires a careful study of the coupled water, energy, and CO2 budgets.
Therefore, an integrated analysis should evaluate and improve model representation of physical processes known to affect the diurnal, seasonal, and interannual cycles, using detailed field site data, as well as routine observations and simplified models.
Perhaps the most fundamental problem regarding the understanding of vegetation feedbacks is a lack of global datasets with which to evaluate existing land-surface processes in climate models. In addition to being critical for understanding terrestrial climate feedbacks, observations of essential ecosystem variables, such as biome type, net primary production, and carbon stores, are integrators of climate and therefore valuable diagnostic measures of models' overall ability to simulate surface climate. Unfortunately the existing observational efforts fall short of what is needed.
Field programs such as FIFE (First International Satellite Land Surface Climatology Project Field Experiment), BOREAS (Boreal Ecosystem-Atmosphere Study), and LBA (Large-Scale Biosphere-Atmosphere Experiment in Amazonia) provide tower flux data (e.g., sensible heat, latent heat, CO2) but only for particular locales. The inclusion of interactive vegetation provides additional ecological data, such as net primary production, carbon storage, leaf area, and biogeography, with which to test climate models.
The AmeriFlux network of permanent towers allows for sustained multiyear observation of particular ecosystems (Wofsy and Hollinger, 1998). It is part of a global network known as FLUXNET (Baldocchi et al., 2001). However, without the broad multidisciplinary focus of FIFE, BOREAS, or LBA many of these tower sites lack the suite of ancillary hydrological and ecological data needed to understand and model the observed fluxes. Most tower sites do not include measurements and analysis of the full energy, water, and carbon balance and their coupling to the boundary layer, and its cloud field. We recommend that these sites expand their focus to include such interactions, which are important for climate models on the boundary layer scale.
The National Science Foundation's Long Term Ecological Research (LTER) program allows the longest (in some cases multi decadal) sustained observation of particular ecosystems. These sites have been chosen to span the range of global biomes (e.g., tundra, boreal forest, grassland, desert). The focus of research is decidedly ecological, emphasizing community composition, ecosystem structure, and their response to environmental change. Some sites (e.g., Harvard Forest) have towers.
Observations of terrestrial ecosystems, their functioning, and their role in the climate system must be sustained over multiyear periods if they are to be of greatest use in the development and improvement of process-oriented vegetation models for use with climate models. A network of such observation sites spanning the range of global biomes is desirable. This suggests that coordination between the AmeriFlux and LTER programs is important to help ensure that a diverse set of biomes is observed.
Measurements using aircraft, such as the recent CRYSTAL and COBRA studies, should also be used to help diagnose the ability of global models to simulate the budgets of water, energy, and carbon for river basins and major ecosystems up to the continental scale. This work should involve a tight integration of land and atmospheric measurements and data assimilation with climate modeling.
To develop the observational basis to improve and test models, both the modeling and remote-sensing communities must work together to better define the vegetation parameters that are observable by satellite and that are critical to modeling vegetation feedback in the climate system. Some of the parameters that are emerging from a dialogue between these communities are leaf area and its phenology. Multiyear leaf area index datasets have been and are being developed for use with climate models (e.g., Buermann et al., 2001). These data products can be used as prescribed leaf area or as a validation of prognostic leaf area (Dickinson et al., 1998, Buermann et al., 2001). Another key emerging data product is fractional tree cover, which can also be used as an input to and validation of models (Bonan et al., 2002). Sustained monitoring of these parameters and extension of these records in the past should be encouraged to allow the modeling community to quantify the vegetation forcing of climate. At the global scale key satellite-derived data products, such as leaf area index, must have at least monthly temporal resolution to be of greatest use in improving and testing climate simulations. In addition, the data products should continue to expand the record to help better account for interannual variability in leaf area.
Much of the global evaluation of the surface climate is still based on 2-m air temperature, humidity, pressure, wind, and precipitation interpolated from station observations. However, many important components of the surface water and energy budget are not routinely measured, and some that are measured at selected experimental sites may not be freely available in the public domain. The shortwave and longwave radiation balances are only recorded at relatively few baseline radiation measurement sites, although satellite-based estimates of the surface short-wave balance have achieved a fair degree of accuracy. The surface fluxes of sensible and latent heat (together with some components of the radiation balance) are measured on flux towers at some 50 or more sites globally, although not all these data are freely accessible. Up-scaling these measurements over carefully selected stands of vegetation to give regional averages (of, say, evaporation) is not straightforward. Estimates of regional evaporation can be made using river basin hydrologic models from observations of precipitation and river runoff. However, regional estimates of precipitation can only be derived from point rain gages and calibrated radars, where available, or from satellite retrievals. Consequently precipitation estimates also have considerable uncertainty and may be biased low when precipitation is frozen.
Important subsurface variables, such as soil temperature and soil water, are also not routinely measured, although they are now being measured in a few important networks, such as the Oklahoma Mesonet and some AmeriFlux sites. The measurement of the freeze-thaw of the surface soil layer is now possible from satellite microwave sensors, but this product is not yet routinely available. Satellites can measure snow cover, but the important measurement of snow water equivalent still presents problems. All satellite measurements of the soil and surface layers have difficulties under forest canopies
A promising approach for developing global soil water fields is the extension of the land data assimilation system (LDAS) pioneered at the National Centers for Environmental Prediction (NCEP) for the regional Eta model. At present, two groups (one in the United States and one in Europe) are developing a global LDAS, using a mix of satellite and surface data to provide the surface radiation budget and precipitation needed to force offline a land-surface-vegetation-hydrology model, which will give subsurface fields of temperature and moisture. Global cooperation is here essential since not all the necessary data has been freely shared in real time in the past. The future availability from satellite of global maps of near-surface soil wetness will provide further useful input.
Because many of the key variables, especially below the surface, are not measured globally, the surface temperature, soil water, and surface energy balance in data assimilation systems is largely a product of a fully coupled land-surface-vegetation-atmosphere model, often constrained by the observed atmospheric diurnal cycle of temperature and humidity near the surface. In forecast and climate models the computed land-surface boundary condition depends on a large number of parameterized submodels, all of which are highly coupled and tend to exhibit considerable differences between each other and with observations. The relationships among the variables in these models is complex, and thus lack of knowledge in one area can have cascading effects. For example, the surface radiation budget depends on the model parameterizations for the cloud fields (which are not explicitly resolved in a global model), while the cloud fields depend on the dynamics, and the moisture field which in turn depends on moisture transports and the surface evapotranspiration. Soil water depends not only on model precipitation and evaporation (coupled to photosynthesis) but also on the subsurface hydrology, which is strongly dependent on the lateral heterogeneity.
The fundamental links between land-surface hydrologic processes, clouds, and precipitation depend in global models on the parametric representation of sub-grid scale boundary layer and cumulus convection. No completely satisfactory parameterization exists for convective clouds, which typically have organization on unresolved scales of 50 km and below. Boundary layer parameterizations are typically quite separate (with different formal closures) and poorly coupled to convective parameterizations in large-scale numerical models, when in nature there is a smooth continuum over the diurnal cycle. Over land in the tropics, for example, as the boundary layer grows after sunrise shallow clouds quickly form, deepen into cumulus congestus, and then organize into precipitating cloud bands, producing a wide range of mid- and high-level clouds, all of which impact the diurnal cycle of the surface radiation budget at the same time as they impact the surface hydrologic budget. From a climate perspective this diurnal cycle of convection plays an important role in the shortwave and longwave cloud feedbacks discussed in Chapter 3. Cloud resolving models are proposed in Chapter 3 as one tool with which to address some of these fundamental unresolved issues of the interaction between different time and space scales, although it is not yet possible to resolve simultaneously both boundary layer clouds and deep convection. To comprehensively characterize and possibly reduce uncertainty in the hydrological cycle of our climate models requires a major ongoing effort both in synthesis and in rigorous diagnostics that cuts across all modeling, theoretical, and observational communities.
Many areas of dynamic vegetation modeling, which are vitally important for simulating long-term feedbacks between the biota and climate, are still in stages of rapid development. It will be important to test the newly emerging prognostic ecosystem and leaf phenology algorithms in coupled climate-vegetation-land-surface models using both existing data and the new data sources outlined in the previous sections.
Many of the feedbacks associated with vegetation occur at longer time scales (centuries) than can be observed with existing observing systems. Paleoclimate research is an important activity to understand vegetation feedbacks on climate. The last glacial maximum and 6,000 years before present have emerged as key periods of focused research demonstrating that inclusion of interactive vegetation improves the simulated climate. Paleoclimate research must be integrated with and indeed is critical to the implementation, testing, and improvement of dynamic vegetation in climate models. The community should work to define standard paleoclimate experiments (e.g., last glacial maximum, 6000 B.P.) that are used to evaluate the coupled climate-vegetation model and highlight the importance of particular vegetation feedbacks (e.g., forest-tundra ecotone, green Sahara).
A clear metric for progress in the coming decade would be the accuracy with which our earth system models can reproduce, for example, the observed diurnal and seasonal variations of the hydrological cycle over land. The short-term modes of variability, like the diurnal, are well represented in even a few annual cycles, while interannual variability requires a longer statistical period. Reanalysis of the past 40 to 50 years of atmospheric data is now available (with new reanalyses in progress). The hydrological records of the past few decades are also being synthesized for global use. Flux site data records are approaching a decade in length. An accuracy of perhaps 5 percent in the key terms in the surface hydrology budget (precipitation and evaporation) would be a realistic target for the coming decade.
The ability of climate models to successful reproduce the terrestrial carbon cycle provides a clear metric to evaluate progress in vegetation models. The carbon cycle integrates across temperature, precipitation, energy fluxes, and the hydrologic cycle and the influence of these on various ecological processes. Tower flux data, ancillary ecological data, satellite-derived data products, and measurements of atmospheric carbon dioxide provide a wealth of critical data with which to constrain and evaluate the simulated carbon cycle.
Was this article helpful?