Click for next page ( 22


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS SUMMARY Cloud feedback and its association with water vapor feedback and lapse rate feedback appear to be the largest contributors to uncertainty in climate sensitivity and is therefore one of the key uncertainties in projections of future climates. Improvements are particularly needed in the treatment of marine boundary layer clouds and tropical convective clouds. Progress on better understanding cloud feedback seems possible now because: (1) current climate models have predictive cloud schemes that produce important effects on climate sensitivity; (2) new data are becoming available that can be used to test these new climate models; and (3) cloud-resolving models have emerged as a new tool for understanding and testing cloud feedback processes in climate models. An accelerated, focused effort to test the simulation of cloud, water vapor, and lapse rate feedbacks in climate models, and their role in climate sensitivity should be initiated. Existing and planned observations should be used in this new emphasis to test the simulation of clouds, water vapor, and lapse rate in climate models and the response of these variables to known forcings. Effective collaboration among efforts to diagnose observations, to model cloud systems, and to model the global climate is essential. A set of observable metrics should be used to evaluate the success of these activities. WATER VAPOR Water vapor feedback is the most important positive feedback in climate models. It is important in itself, and also because it amplifies the effect of every other feedback and uncertainty in the climate system. Most modeling 21

OCR for page 21
22 UNDERSTANDING CLIAL1 TE CHANGE FEEDBACKS and observational studies suggest that the water vapor feedback in current climate models has the correct sign and magnitude (Held and Soden, 2000~. The magnitude of water vapor feedback is so large, however, that modest uncertainty in water vapor feedback can still have a significant effect on the magnitude of climate change. It is known from basic physical principles that the vapor pressure in equilibrium with a water surface increases exponentially with temperature at a rate such that a 1 percent change in absolute temperature, a change of about 3C, is associated with an approximately 20 percent increase in saturation vapor pressure. Because water vapor is the most important greenhouse gas in Earth's atmosphere, the dependence of vapor pressure on temperature forms the basis of one of the strongest positive feedbacks in the climate system. If the relative humidity distribution remains approximately constant as temperature and specific humidity increase, then water vapor greenhouse feedback nearly doubles the sensitivity of climate above what it would be in the absence of water vapor feedback. On the largest spatial scales, existing data and current climate models are basically consistent with the assumption that on interannual time scales, relative humidity is more or less constant (Soden et al., 2002; Wentz and Schabel, 2000~. However, local diurnal and seasonal relative humidity variations are significant, and analysis of climate model simulations of these features is needed. Furthermore, the relationship between temperature and humidity on interannual and longer time scales shows substantial vertical and regional structure, which models are only partly successful in simulating (Bates and Jackson, 1997; Bauer et al., 2002; Ross et al., 2002~. As shown by modeling and observational studies (Del Genio et al., 1991; Harries, 1997; Held and Soden, 2000; Shine and Sinha, 1991; Soden et al., 2002), water vapor variations in the tropical upper troposphere seem to have the strongest effect on outgoing long-wave radiation. However, the relative importance of water vapor in different regions of the atmosphere is sensitive to the assumptions made about clouds and about the variations (or lack thereof) of relative humidity with temperature. In fact, according to Harries (1997), "Uncertainties of only a few percent in knowledge of the humidity distribution in the atmosphere could produce changes to the outgoing spectrum of similar magnitude to that caused by doubling carbon dioxide in the atmosphere," underscoring the importance of reliable upper tropospheric water vapor observations. Uncertainty about water vapor feedback rests primarily on the question of whether the relative humidity distribution might change in an altered climate state. Several hypotheses have been put forward describing mechanisms that could alter the relative humidity distribution in a warmed

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS 23 world. The mechanisms that seem most likely to be meaningful are those that may govern the relationship between the area of moist and dry regions in the upper troposphere of the tropics (Lindzen et al., 2001, Pierrehumbert, 1995~. In the tropics the greenhouse effect is strong, and large contrasts in upper tropospheric relative humidity are sustained between regions of large- scale ascent and descent. So far, no mechanism has been demonstrated to operate that would provide a significantly more reliable projection than an assumption of constant relative humidity distribution. Nonetheless, the factors that influence water vapor distribution need further study. One useful metric for evaluating the question of whether relative humidity will change was put forward by Inamdar and Ramanathan (1998~. Using Earth Radiation Budget Experiment (ERBE) data they examined the relationship between outgoing long-wave radiation at the top of the clear atmosphere and surface temperature. They found the slope of the regression line between these two variables to be consistent with an assumption of fixed relative humidity (vs. absolute humidity). Using this approach it is possible to compute a gain factor of the clear-sky water vapor feedback. Gain factors determined using this and other observational approaches should be compared with the factors similarly derived from models. This approach is discussed here for to illustrate only one of the many approaches that can be used to assess the ability of models to faithfully represent water vapor feedbacks. Understanding of the water vapor distribution is being hindered by a lack of accurate measurements of water vapor concentration with sufficient spatial and temporal resolution and global coverage (Kley et al., 2000~. Accurate measurements of the water vapor distribution can be used to test understanding of the mechanisms that determine its distribution, and also test to see if the increase of water vapor with time is consistent with models of climate change. An integrated water vapor observing system should be developed, which has sufficient accuracy to measure decadal trends in the water vapor distribution and sufficient spatial resolution to test mechanisms by which that distribution is maintained. It should include a network of in situ sounding systems capable of measuring water vapor throughout the troposphere and lower stratosphere, complemented by ground-based remote sensors (such as have already been deployed at Atmospheric Radiation Measurement Project Cloud and Radiation Testbed (ARM CART) sites). These observations would allow quantification of temporal and vertical water vapor variations and would allow calibration and validation of satellite observations, which would extend the global coverage of the observing system.

OCR for page 21
24 UNDERSTANDING CLIMATE CHANGE FEEDBACKS The global radiosonde network cannot be relied upon for precise water vapor observations, unless substantial improvements are made to ensure higher quality observations in the upper troposphere and in other cold (and dry) regions, and to ensure the long-term continuity of the observations. Expansion of the program for special water vapor soundings of the troposphere and stratosphere (e.g., Oltmans and Hofmann, 1995) to more sites (currently only Boulder is routinely observed) would be very beneficial. These should include both oceanic and continental regions, at a variety of latitudes. Efforts to consolidate and quality-control water vapor observations from different sources (e.g., the NASA Water Vapor Project [NVaP], Randel et al., 1996) should also be encouraged, so that water vapor variability can be examined in conjunction with variations in other atmospheric variables, particularly temperature and radiation. The water vapor observing system should be closely linked to a global cloud, aerosol, and precipitation observing system. Many of the issues mentioned above are discussed in greater detail in a report by the National Research Council (NRC) on the Global Energy and Water Cycle Experiment (GEWEX) Global Water Vapor Project (NRC, l999c). LAPSE RATE FEEDBACK The strength of Earth's greenhouse effect depends on the fact that the temperature decreases with height in the troposphere, so that emission from water vapor and clouds in the colder upper troposphere is less than that from the surface. A stronger lapse rate (the rate of decrease of temperature with altitude) gives rise to a stronger greenhouse effect and a warmer surface, all else being equal. If the lapse rate changes systematically with the surface temperature, then a potentially strong lapse rate feedback may exist. Radiative processes, large-scale dynamical processes, and convection determine the lapse rate. Radiative processes generally cool the atmosphere and heat the surface, and convection and large-scale motions in the atmosphere generally move heat upward. In the tropics the lapse rate generally follows the moist adiabatic lapse rate, the rate at which saturated air parcels cool with altitude as they are raised adiabatically. The moist adiabatic lapse rate decreases with increasing surface temperature, so by itself lapse rate feedback is expected to be negative in the tropics (Hansen et al., 1984; Wetherald and Manabe, 1986~. If the assumption of fixed relative humidity is a good approximation, then the water vapor feedback is partially cancelled by the lapse rate feedback (Cess, 1975~. If the lapse rate is reduced, then the air at altitude is

OCR for page 21
CLOUD, WA TER VAPOR, AND LAPSE RA TE FEEDBACKS 25 warmer. The warmer air contains more water vapor. The decreased greenhouse effect caused by a weaker lapse rate is offset by the increased greenhouse effect from larger amounts of water vapor at higher altitudes. Patterns of vertical temperature structure change are one of the few parameters widely used to detect and attribute climate change to particular forcings, or to natural variability (e.g., Tett et al., 2002~. (Surface temperature changes are the other main detection parameter.) If climate models correctly simulate climate feedback mechanisms, they should correctly reproduce the change in vertical temperature structure associated with different climate forcings. Thus, changes in lapse rate are indicators both of the strength of lapse rate feedback and of the response to climate ~ . torclngs. Climate models generally reproduce the observed lapse rate in the tropics and elsewhere through the incorporation of large-scale dynamics and parameterized convection and radiation. Some observations suggest relationships between surface temperature trends and temperature trends in the free troposphere that seem inconsistent with the behavior of current climate models (NRC, 2000a; Santer et al., 2000~. It is still unclear whether these apparent inconsistencies are the result of a measurement problem or a failure of our understanding of the climate system. Unfortunately, current upper-air temperature observations are not well suited to determining lapse rate changes. The vertical resolution of satellite observations is too coarse for accurate lapse rate computations, although newer instruments (e.g., the Advanced Microwave Sounding Unit) provide better vertical resolution than older ones (e.g., the Microwave Sounding Unit). Both satellite and radiosonde observations are hampered by time- varying biases, which are very difficult to remove (NRC, 2000a). Lapse rate trends are particularly sensitive to attempts to remove these biases (Lanzante et al., in press). Similarly, trends in measures of atmospheric instability and convection that are related to lapse rate (e.g., Convective Available Potential Energy and Convective Inhibition) are affected by radiosonde data inhomogeneities (Gettelman et al., in press). Thus, to improve our ability to diagnose lapse rate feedback and to detect changes in the vertical temperature structure of the atmosphere, improved long-term upper-a~r temperature soundings are required. The observations must be of sufficient precision to measure decadal trends in temperature (and water vapor) distributions and sufficient spatial resolution to test mechanisms by which those distributions are maintained. More information concerning upper-air temperature monitoring requirements can be found in NRC (2000c). Using the improved observations that are recommended here, correlation statistics of temperature, water vapor, and clouds on various time

OCR for page 21
26 UNDERSTANDING CLIMATE CHANGE FEEDBACKS and space scales should be employed to rigorously diagnose the ability of models to simulate the feedbacks that underpin interannual variability of the lapse rate and water vapor distributions. Extending the work of, for example, Ross et al. (2002), Sun and Held (1996), and Sun and Oort (1995), these analyses should be focused not only on improving understanding of the feedback processes and their representation in models but also on deriving new, parsimonious model representations of these processes. Several existing national and international programs (e.g., ARM and GEWEX) could be very helpful in facilitating this work. CLOUD FEEDBACKS Because clouds are generally colder than the surface they overlie and because they absorb and emit terrestrial radiation, the presence of clouds generally reduces the energy emitted to space from Earth relative to the emission from Earth when clouds are absent. For terrestrial radiation, clouds thus act very much like greenhouse gases and warm the surface of Earth. Clouds also reflect solar radiation very effectively, which reduces the amount of solar energy reaching the surface of Earth. This tends to cool the surface. Different cloud types have different effects on the energy balance of Earth (Hartmann et al., 1992~. If the structure or area coverage of clouds change with the climate, they have the potential to provide a very large feedback and either greatly increase or decrease the response of the climate to human-caused forcing. At this time both the magnitude and sign of cloud feedback effects on the global mean response to human forcing are uncertain. It has been well documented that climate models are sensitive to the representation of clouds and their radiative properties (e.g., Cess et al., 1990; Paltridge, 1980; Schneider, 1972, Senior and Mitchell, 1993; Stocker et al., 2001; Webster and Stephens, 1984~. A relatively modest change in cloud properties can have a significant effect on Earth's energy balance. In addition to their influence on the radiative processes that define the energy balance of the planet, clouds processes are integral to the cycling of water between the surface and the atmosphere. A striking example of the contribution of cloud feedbacks to uncertainty in climate sensitivity is exhibited by comparison of the current climate models at the Geophysical Fluid Dynamics Laboratory (GFDL) and the National Center for Atmospheric Research (NCAR). The GFDL model has a rather high sensitivity (near 4C for doubled CO2) while the NCAR model has a rather low sensitivity (near 2C). The primary reason for this difference

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS 27 is in the response of low marine boundary layer clouds in the two models. As the climate warms, marine boundary layer clouds decrease in the GFDL model and increase in the NCAR model. Comparison of these two models with observations, theory, and cloud-resolving model computations should lead to much greater understanding of the response of marine boundary layer clouds to changing climate, and a consequent reduction in uncertainty of climate sensitivity. Another key uncertainty in cloud-climate interactions is the response of anvil clouds to surface temperature. It is unknown whether anvil clouds expand or contract when surface temperature warms. A combination of detailed observational studies and cloud-resolving modeling studies can shed light on this issue. Some models incorporate a cloud optical thickness feedback that assumes cloud water content will increase with temperature following the saturation vapor pressure, but satellite and in situ data do not show an obvious signal of this nature, and low clouds show an apparent signature in the opposite sense (Tselioudis et al., 1992~. Clouds couple many feedback processes in the climate system. Some of the interactions of clouds with other feedback processes are illustrated below. Clouds and Water Vapor Feedback The formation and evaporation of clouds are intimately tied to the amount of water vapor in the atmosphere. The amount of water vapor and its vertical distribution are also influenced by the amount and distribution of clouds. For example, a number of studies have shown very clearly how the water vapor in the middle to upper troposphere is sensitive to the presence of ice crystals, the nature of the microphysical properties of these ice crystals, and the way these crystals fall in the atmosphere. (e.g. Donner et al., 1997; Stephens et al., 1998~. Vertical transport of water in both vapor and ice form by convection in the tropics is an important source for upper tropospheric water vapor (Pierrehumbert and Roca, 1998; Salathe and Hartmann, 1997; Udelhofen and Hartmann, 1995~. The broad role of water vapor feedback in climate change and the specific importance of upper tropospheric water vapor cannot be divorced from the associated role of clouds and cloud feedbacks.

OCR for page 21
28 UNDERSTANDING CLIME TE CHANGE FEEDBACKS Clouds, Lapse Rate, and Precipitation The vertical distribution of clouds is an important factor in determining radiative heating. In turn, radiative heating is closely coupled to the temperature profile, convective heating, and precipitation. A number of modeling studies have illustrated how the radiative effect of cloudiness, the vertical profile of temperature, convection, and precipitation are tightly coupled (e.g. Fowler and Randall, 1996; Liang and Wang, 1997; Ma et al., 1994; and Slingo and Slingo, 1988~. Clouds and Sea-Ice Albedo Ice albedo feedbacks that may occur in polar regions are tightly coupled to the surface energy balance and to clouds. Clouds can change the heat balance of the surface and influence surface ice formation and melting, and overlying clouds can mask the effect of surface ice on the albedo of Earth. A complex coupling thus exists between cloud feedbacks and ice-albedo feedback processes (see Chapter 4~. Clouds and Soil Moisture The feedbacks involving soil moisture and evaporation are intimately tied to the hydrological cycle over land (see Chapter 6~. Clouds are central to soil moisture feedbacks both through their profound influence on the surface energy balance and through their association with precipitation. The relationships among soil moisture, boundary layer humidity, and cloudiness serves as a possible mechanism for a strong, coupled feedback between clouds and the underlying land surface. Clouds, Chemistry, and the Marine Biosphere The effect of changing concentrations of cloud and ice condensation nuclei (CCN and IN, respectively) on clouds and precipitation has received much attention recently (e.g., Durkee et al., 2000~. The association between aerosol forcing, cloud nuclei, and cloud processes provides a path that links clouds to oceanic emissions of dimethyl sulphide (DMS) and to gas phase chemistry (e.g., Charlson et al., 1987; Coakley et al., 1987~. The consequences of these links are twofold. In the case of DMS emissions they

OCR for page 21
CLOUD, WA TER VAPOR, AND LAPSE RA TE FEEDBACKS 29 provide additional feedback mechanisms if the production of nuclei depends on temperature or solar radiation reaching the surface. In the case of increasing aerosols the relation between aerosols and condensation nuclei connects cloud processes to the broader problem of estimating climate forcing through the so-called indirect aerosol forcing. Cloud Radiation Processes Clouds affect both the radiation balance at the top of the atmosphere and the distribution of radiative heating between the atmosphere and surface. The Effects of Clouds on the Top-of-the-Atmosphere Energy Budget The radiation budget of Earth is the difference between solar radiation absorbed by the planet and terrestrial infrared (JR) radiation emitted to space. Clouds affect this budget by reflecting sunlight back to space (the albedo effect of clouds), thereby decreasing the solar radiation absorbed by the planet, and by absorbing thermal radiation emitted by the surface and lower atmosphere (the greenhouse effect of clouds), thereby reducing the radiation emitted to space. The balance between these negative and positive effects on the radiation balance depends on the type and location of the cloud in question (Hartmann et al., 1992~. The albedo effect of low clouds over ocean, for example, tend to dominate over their greenhouse effect and produce a negative impact on Earth's energy balance, whereas the reverse is generally true for high, thin cirrus. Satellite experiments like ERBE and Clouds and the Earth's Radiant Energy System (CERES) provide a quantitative measure of the instantaneous effects of clouds on the top-of-the- atmosphere (TOA) radiation balance and confirm our understanding of the effect of different cloud types on this budget. Although data collected from these satellite experiments provide an important source of information for testing models, they do not sufficiently constrain critical assumptions about the treatment of cloud processes in climate models. Measurements of radiative properties and inferred column-integrated cloud optical properties as have been made over the past 20 years are insufficient to advance understanding and modeling of cloud feedbacks. What is needed are measurements of those key variables prognosed from models that describe the underlying cloud physical processes. These variables include the mass of liquid water and ice in clouds and precipitation and how these water masses mutate, passing from the cloud to the

OCR for page 21
30 UNDERSTANDING CLIMATE CHANGE FEEDBACKS precipitation state. New global observations that will be relevant to understanding cloud feedbacks and validating global climate models are becoming available from new satellite measurements obtained from NASA's Earth Observing System. New global data on cloud properties, water vapor, and aerosols instruments is expected from instruments such as the Moderate Resolution Imaging Spectroradiometer (MODIS), the Multiangle Imaging SpectroRadiometer (MISR), and the Advanced Infrared Sounder (AIRS) (Aumann et al., 2003; Diner et al., 2002; King et al., 2003) (See also the subsequent, broader description of the global-scale observations that are required.) The Effects of Clouds on the Partitioning of the Radiation in the Atmosphere and at the Surface The reflection of solar radiation by clouds causes a strong reduction in the energy balance of the surface because most of the solar radiation that is not reflected is absorbed at the surface. At high latitudes where insolation is weak and the atmosphere is relatively dry, the addition of clouds can heat the surface through increased downward IR emission by the atmosphere. Whether cloud layers heat or cool the atmosphere relative to clear skies, and the amount of this heating or cooling that takes place is largely determined by the vertical location and distribution of the clouds. High clouds tend to warm the atmosphere relative to surrounding clear skies, whereas low clouds tend to enhance the cooling of the atmosphere. While the total incoming and outgoing radiation at the TOA can be measured, the amount of radiative heating that occurs within the atmosphere versus how much heating occurs at the surface cannot be directly measured. Thus, model parameterizations of the internal heating of the climate system cannot be tightly constrained by observations. This shortcoming is a significant source of uncertainty in understanding cloud feedbacks. Clouds and the Large-Scale Circulation: The Cloud Parameter~zation Problem Developing and testing understanding of cloud fields and their interactions with the larger-scale environment has proven to be difficult. Many of the processes that control cloud feedbacks occur on scales smaller than those resolved by large-scale models in use today. These processes are thus parameterized, meaning that they are expressed in terms of large-scale

OCR for page 21
CLOUD, WA TER VAPOR, AND LAPSE ~ TE FEEDBACKS 31 quantities that are resolved by models. The influence of these large-scale properties on smaller-scale parameterized processes, and the subsequent feedback of the latter to the large scale, is referred to as the cloud parameterization problem. Three types of processes are critical: (1) cloud physical processes, including processes that govern the life-cycle of cloud- scale phenomena, (2) cloud radiative process, and (3) large-scale cloud thermodynamical processes that determine the heating in the climate systems and the associated atmospheric circulation. Quantifying Processes That Govern Life Cycles of Large-Scale Cloud Systems Although the cloud processes that influence the radiation budget in principle are numerous and occur over a vast range of scales, the dominant scale of variability of cloudiness is the synoptic scale (e.g., Rossow and Cairns, 1995~. Therefore, key first steps in understanding cloud feedbacks in the global climate system require understanding processes that organize clouds on this same large scale. These processes involve connections between the general circulation of the atmosphere and the weather systems that are a manifestation of this circulation, the formation and evolution of the large cloud systems associated with these weather systems, and the latent heating and radiative heating distributions organized on this larger scale. Because the processes that govern cloud evolution are modulated by the weather systems in which they are embedded, a fruitful strategy should embrace the study of weather. Numerical weather prediction models, related data assimilation activities, and synoptic data on weather and clouds should be used to understand and model cloud and precipitation evolution over a range of time scales from hours to weeks. Day-to-day weather variations are carefully observed, assimilated into models, and used to make predictions. Because day-to-day weather variations include variations in cloud amount and type, these variations should be used to test the ability of climate models to predict cloud variations on these time scales (e.g., the testing of cloud simulations at the ECMWF tHogan et al., 2001; Klein and Jacob, 19993~. This strategy has several advantages. . Cloud feedbacks are currently diagnosed primarily by using coarse resolution climate models and even simpler one-dimensional equilibrium models. The use of Numerical Weather Prediction (NWP) models and forecast validation will allow day-to-day weather variations and their association with cloud variations to be used to validate models.

OCR for page 21
32 . UNDERSTANDING CLIMATE CHANGE FEEDBACKS Connecting feedback diagnostic studies to NWP and data assimilation efforts introduces a certain rigor to the exercise of model-data comparison by tying the analysis methods more tightly to the observations and allowing many more realizations. NWP and the data assimilation process offers a consistent way of obtaining integrated datasets necessary for understanding processes deemed important to cloud feedback. Quantifying the Relationship Between Cloudiness and Radiative and Latent Heating A strategy for understanding the relationship between clouds and precipitation in more quantitative detail requires a change in current research practices. Research activities and observational practices for clouds and precipitation are typically designed in isolation from each other. The parameterization of the radiative effect of clouds is often treated separately from the parameterization of precipitation. Advances will occur with the adoption of a more integrated approach toward developing global cloud and precipitation observing and projection systems. Global precipitation, water vapor, and cloud-observing systems must be designed in concert with one another so that the interconnectivity of these processes can be better observed and understood. Similarly, parameterization and projection systems must address these variables as part of an interconnected system. WHY HAS PROGRESS ON CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS BEEN SO ELUSIVE? The 1979 NRC report on carbon dioxide and climate contains the following statement in reference to cloud effects on climate change: Trustworthy answers can be obtained only through comprehensive numerical modeling of the general circulations of the atmosphere and oceans together with validation by comparison of the observed with the model-produced cloud types and amounts. This strategy remains valid today, but it has not yet been executed. Three obstacles have heretofore limited the advancement of understanding

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS 33 of cloud, water vapor and lapse rate feedbacks: inadequate data, incomplete theories, and untested projections. Inadequate Data for Developing and Testing EIypotheses Measurements of Earth's energy budget with sufficient accuracy for climate studies began about 1985 with the Earth Radiation Budget Experiment. Measurements of water vapor and upper tropospheric temperature need to be improved in both accuracy and sampling. At the present time cloud data come from two sources: surface visual observations (Hahn and Warren, 1999) and meteorological imaging instruments on operational satellites (Rossow and Schiffer, 1999~. The International Satellite Cloud Climatology Project (ISCCP) data provide estimates of cloud-top temperature and visible optical depth on spatial scales of tens of kilometers. These data have seen only limited use in climate model validation (e.g., Klein and Jakob 1999, Webb et al., 2001~. More of this kind of analysis is needed. In addition, more detailed global observations of clouds, including such things as vertical structure and cloud particle size, are needed to test climate model parameterizations and their relationship with precipitation, water vapor, and air temperature. Current satellite data give only rather crude estimates of cloud particle size and cannot readily distinguish cloud water from cloud ice. Passive infrared sensing of cloud-top height is imprecise when the clouds are not optically thick, which is an important constraint when studying high, thin clouds in the tropics and elsewhere. New cloud data are becoming available from instruments on satellites that use polarization of reflected sunlight and active scanning with cloud radars and lidars to probe the vertical structure and particle size of clouds. These data will provide an important new source of data on the global distribution of clouds that should be used to further constrain and test the simulation of clouds in climate models. Incomplete Theories Climate feedback hypotheses are necessarily concerned with large, complex, and coupled systems that do not necessarily obey simple laws. Most simplified feedback "theories" involving clouds consider only a limited set of the critical processes, even though the neglected processes are known to be important. For example, the rheostat hypothesis (Ramanathan and Collins, 1991) argues that the sensitivity of tropical

OCR for page 21
34 UNDERSTANDING CLIMA TE CHANGE FEEDBACKS convective cloud albedo will constrain tropical sea surface temperature (SST) below 303K. But the analysis implicitly assumes that spatial variations of clouds and SST are a useful analogy for climate change, when in fact they are not (Hardnann and Michelsen, 1993, Lau et al., 1994; Wallace, 1992~. And, the Iris hypothesis (Lindzen et al., 2001) speculates that the area of tropical anvil clouds will decrease with increasing SST, but again the observational evidence uses a gradient with latitude as an analogy for climate change, which it is probably not (Hartmann and Michelsen, 2002~. Simple theories for how clouds will respond to global warming are difficult to test using observations, since only a small global warming has so far been observed. It is easier to test the response to large forcings, such as the annual and diurnal cycle, which are well observed and the response amplitude is large. The treatment of clouds in climate models is still highly simplified, although current climate models are including more of the relevant physics of cloud processes. New data to validate these models is becoming available from measurement programs such as DOE's Atmospheric Radiation Measurement (ARM) program and new satellite observations. In most models, however, the key linkages between processes are often broken. For example, the separation of parameterized convection from large-scale cloudiness effectively decouples clouds from the model hydrological cycle. This artificial separation creates problems when attempting to use models to advance understanding on cloud and water vapor feedbacks. Additional problems with cloud schemes in climate models include . the introduction of model resolution dependence to the parameterizations of clouds. For example, the cloud processes represented by the large-scale schemes are typically microphysical in nature. The parameters that represent these processes have to be heavily tuned to the scale resolved by the model. This introduces an unavoidable degree of arbitrariness to the cloud feedback problem since global-scale cloud observations of these processes needed for tuning are lacking. a growing confusion between those processes that are really represented by the sub-grid-scale schemes and those processes that are represented by the resolved scales. This in turn creates a further degree of arbitrariness as to how to use existing observations (of precipitation, for example) to assess the merits of different parameterizations. cloud feedbacks currently addressed in global scale models chiefly articulated in terms of the resolved cloudiness and thus chiefly in terms of cloud radiation interactions. When averaged over time, the global energy balance of the atmosphere is fundamentally between latent heating

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS 35 associated with precipitation and the radiative heating, and notably including the contribution by clouds on the radiative heating (e.g., Stephens et al., 1994), so that latent heat release and radiative heating must be closely linked. The observed diurnal variation of convective precipitation is a manifestation of the interplay between radiation and convection. It is difficult to treat this kind of interaction in existing global models that deal with precipitation and related processes (by sub-grid-scale convection) in isolation from clouds and their radiative heating. Untested Predictions In general, quantitative tests of the role of clouds in global climate change are difficult to devise since one cannot observe a climate change. The best that can be done is to use a long record of climate, including cloud information, and test the models ability to simulate the observed variability. Such model evaluations require comparison datasets of relevant information accumulated over extended periods of time. Developing long-term datasets of even rudimentary parameters, let alone cloud parameters, has proven to be difficult for a number of reasons, including the lack of dedicated global monitoring and observing systems for this purpose (NRC, 1999a). Despite these difficulties a number of valuable global datasets have been compiled over the past two decades. That these datasets are underutilized is in part a reflection of the attention that has been paid to model intercomparison, but too little attention has been paid to testing models against data. The utility of model-to-model intercomparison exercises for characterizing and reducing uncertainty in climate change feedbacks is limited. Without rigorous and multifaceted comparisons of models to as much data as possible, model intercomparison activities tend to make the feedback processes behave similarly to one another while generating no evidence that their consensus behavior is any nearer to that of nature. DEVELOPING A SCIENTIFIC STRATEGY Despite the challenges described above, the potential for making important strides in understanding is very high at the present time, for the following reasons.

OCR for page 21
36 UNDERSTANDING CLIMATE CHANGE FEEDBACKS Improved Global-Scale Experimental Data In the most sophisticated cloud-resolving models (CRMs), NWP models, and climate models, the clouds are currently predicted in terms of three-dimensional distributions of cloud water and ice, using conservation equations for these quantities, so that fully prognostic cloud simulations have become the norm. The ability of these models to simulate the three- dimensional fields of water and ice correctly cannot be adequately tested at the present time because of a lack of sufficiently detailed global data, thereby thwarting model assessment and subsequent improvement. Datasets from satellite- and ground-based observations are currently being developed that would enable the validation of these more sophisticated models, if a sufficient effort is made to do so. Examples of datasets include cloud and aerosol data from Earth Observing System (EOS) instruments and cloud vertical structure data from the Cloudsat satellite. Surface data from the ARM provide a new capability to measure critical cloud properties at selected locations (Masher et al., 1998; Stokes and Schwartz, 1994~. The availability of global-scale data on precipitation, albeit confined to the global tropics (Kummerow et al., 2000), as well as the near-future availability of global cloud water and ice information from other planned satellite measurements (Stephens et al., 2002), provides the much needed datasets for evaluating cloud predictions under a variety of weather regimes. Evaluating Model Predictions Running models in a forecast mode is one way the link between heating and circulation can be examined, at least in the context of testing the shorter- time-scale feedbacks. Comparisons of the European Centre for Medium- Range Weather Forecasts (ECMWF) NWP predictions of cloud cover and occurrence, albeit limited in scope, show an encouraging degree of agreement with existing data (Hogan et al., 2001; Klein and Jakob, 1999; Miller et al., 1999~. These comparisons go beyond superficial comparisons of areal cloud amount by examining the vitally important vertical structure. Still missing are diagnostic studies of cloud property information such as liquid and solid water contents with corresponding quantitative precipitation. Information from CloudSat could help fill some of these observational gaps (Stephens et al., 2002~. Studies such as these highlight the utility of being able to run climate models in an NWP mode to perform diagnostic analyses of processes that operate on short time scales but that are critical to producing realistic

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS 37 projections of long-term climate change. With increasing computation power expected in the coming years and the higher spatial resolution expected of these global models, continued improvements in the representation of smaller-scale cloud processes with the subsequent improvement in predictions of cloud properties is anticipated. Thus, with improved resolution and improved global observations noted above, more probing testing of model parameterizations is possible, which is expected to lead to better parameterization methods and better cloud predictions. If adequately supported, the GEWEX Cloud Systems Study (Randall et al., 2000), which is focused on developing improved parameterizations for a wide variety of cloud types, is expected to contribute substantially to this effort. The approach generally combines observations, cloud-resolving models and global climate models. Better cloud predictions in turn will lead to more capable assimilation methods eventually expanding the use of existing and archived data, such as the archived but unused cloudy-sky radiance data derived from operational analyses. This then should feed back on model development with subsequent improvements. Validated cloud predictions should also expand diagnostic uses of new re-analyzed data expected from future re-analysis efforts that could be an integral part of the cycle of model evaluation, improvement, and data analyses. Toward Improved Theories Cloud-resolving models (CRMs) have evolved as one of the main tools for studying the links between key processes pertinent to studying cloud- related feedbacks (e.g., Browning, 1993; Grabowski, 2000~. As such, these models may be viewed as an essential tool for articulating the underlying theories of cloud feedbacks. These models continue to improve and are now being adopted more widely in a variety of cloud and precipitation research activities. CRMs are also being coupled experimentally ways into global models to serve as an explicit form of cloud parameterization, thereby overcoming the problematic separation between resolved cloudiness and unresolved convection (Randall et al., in press). CRMs embedded within GCMs should not be viewed as a panacea because they do not actually simulate the complete cloud dynamics in a GCM grid cell; rather, they provide a physical representation of the cloud statistics in the cell. They are also quite computationally intensive to run in this way. Moreover, it remains to be seen how difficult it is to develop a

OCR for page 21
38 UNDERSTANDING CLIMA TE CHANGE FEEDBACKS cloud resolving GCM with realistic sensitivity of global mean surface temperature. Despite improvements and increasing use of CRMs, their evaluation is far from extensive, being limited to a few test cases from a limited number of field campaigns. Future testing must examine the sensitivity of CRM simulations to assumptions in their microphysics and turbulence parameterizations and the limitations this sensitivity may impose. The cloud evolution predicted by these models is also sensitive to initial conditions (including the large-scale forcing that drives them). This sensitivity is problematic given that the source of this forcing usually derives from the analyses of large-scale operational models. Therefore, progress in CRMs has to be intimately tied to progress in NWP global models. Mutual improvements in turn can be expected to lead not only to better cloud prediction schemes in global models but also can be expected to promote new assimilation methods applied to CRMs and eventually a more penetrating way of testing and improving models with observations. These caveats should not overshadow the potential that CRMs present as tools to explore the interaction between the cloud physics and the general circulation of the atmosphere The cloud feedback problem and the indirect effect of aerosols are linked together. The provision of aerosols is hypothesized to affect the water budget of clouds through the indirect effect. But this affect cannot be understood without understanding the effects of dynamics and thermodynamics in providing moisture for clouds. In most cases one would expect the circulation and thermodynamics to have a much larger effect on the cloud properties than the provision of additional aerosols. Therefore one can argue that a good understanding of the relationship of cloud properties to the dynamics and large-scale thermodynamic environment of the clouds is necessary before the effect of additional aerosols can be convincingly predicted. To resolve these issues will probably require testing when the aerosol abundance is known as well as the dynamic and thermodynamic conditions. The effect of the dynamic and thermodynamic environment can then be separated from the aerosol effect and solved first. Direct measurements of aerosols and associated cloud properties may also provide critical information (Breon et al., 2002, Lohmann and Lesins, 2002~. Progress in understanding cloud, water vapor, and lapse rate feedbacks requires that an integrated effort with additional resources be developed that cross-cuts the interests of individual agencies. We propose that this effort be developed with the following elements:

OCR for page 21
CLOUD, WATER VAPOR, AND LAPSE RATE FEEDBACKS 39 Improving Datasets and Data Analyses. The fundamental problem is that the scientific community's efforts to model the basic physics of cloud, water vapor, and lapse rate feedbacks are much more advanced than the ability to measure the nature and evaluate the accuracy of their simulation in climate models. Therefore, a vigorous strategy should be implemented to promote and fund research that -maintains the important global datasets already under development but in jeopardy due to lack of support (e.g., GEWEX-related datasets such as the ISCCP and the Global Precipitation Climatology Project tGPCPJ); -uses existing datasets specifically to evaluate cloud, water vapor, and precipitation predictions in global-scale weather and climate models, as well as regional-scale cloud-resolving models. A focus should be placed on developing rigorous diagnostics methods and evaluation procedures; and -extends these activities to embrace the new improved datasets expected in the coming years. Testing predictions. A rigorous effort to test climate models against observational metrics must be initiated and coordinated among groups performing climate modeling, climate observation, and climate analysis. Metrics should include comparison of observed and simulated response of clouds, water vapor, and lapse rate to every well-observed forcing mechanism and time scale, including the diurnal and seasonal response, the response to ENSO and the response to volcanic eruptions. This intercomparison should include the estimation of global feedback parameters from seasonal variations (e.g., Tsushima and Manabe, 2001) and regional feedbacks as understanding warrants. Additional metrics should include cloud and water vapor variations associated with day-to-day weather changes. Weather prediction models and connected assimilation systems should be applied to the diagnosis of critical links between cloudiness, water vapor, precipitation, and weather variations. Within this effort, new methods for the assimilation of cloud, water vapor, and precipitation data must be promoted. Therefore, ongoing attempts to coordinate national climate modeling efforts must include an NWP component with data assimilation as well as a data assimilation effort using climate models. The time scales of relevance include diurnal, weekly (characteristic of weather systems), seasonal (characteristic of natural modes of variability; see Chapter 9), and decadal (characteristic of long-term climate change).

OCR for page 21
40 UNDERSTANDING CLIME TE CHANGE FEEDBA CKS Improving theory and models. A significant effort should be undertaken that builds upon the proceeding two elements with the specific goal of improving the representation of clouds, water vapor, and precipitation in NWP and climate models. This activity should use an integrated, hierarchical approach to model development connecting NWP model and assimilation developments, climate model parameterization developments, and cloud-resolving models. This effort must go significantly beyond the current model intercomparison projects, which have played an important role in identifying model errors and in developing uniform model diagnostics, but frequently have lacked an observational underpinning. The potential of this approach will not be realized without a more coordinated program of research and support. Progress on atmospheric hydrology feedbacks has been hindered by fragmented resources, which discourages crosscutting research in modeling, observational techniques, and diagnostic analyses. For example, research in collection and analysis of the global datasets of cloud and water vapor information, especially those derived from space-borne observations, are supported in large part by NASA, the development of NWP models by the National Oceanic and Atmospheric Administration (NOAA), and high-end climate modeling efforts by yet other agencies, each of which have their own objectives. A viable strategy for progress requires a thoughtful, efficient integration of observations, diagnostic research, global model development, data assimilation, and cloud-scale modeling. These elements have to be connected in one program as progress on any specific element of this strategy depends on progress on connected elements.