U.S. Dept. of Commerce / NOAA / OAR / ERL / GFDL *Disclaimer  

 

1. CLIMATE DYNAMICS

GOALS

  1.1 BACKGROUND FOR COUPLED CLIMATE MODELING AT GFDL

          Two distinct coupled atmosphere-ocean climate models are actively being used in research on global warming and other aspects of climatic sensitivity and variability. This section summarizes the general characteristics of these models and the design of some key experiments using these models.

          The R30 coupled model, which has been under development for several years, has matured within the past year and is now being used for the generation of global warming scenarios and studies of decadal-to-centennial climatic variability. This model has an atmospheric horizontal resolution of 3.75° longitude and 2.25° latitude, with 14 levels in the vertical. It is coupled to an ocean model with 2° horizontal resolution, a simple free-drift ice model, and a "bucket" land hydrology. The atmospheric component of this model has been studied extensively, and a large ensemble of 40-year experiments with prescribed observed sea surface temperatures (SSTs) continue to be analyzed by the Climate Diagnostics group. The coupled model is run with adjustments to the air-sea fluxes of heat and fresh water to prevent climate drift. Several long control integrations have been performed with this coupled model, one of over 1,000 years, and including a suite of five simulation forecasts of the period 1865-2089.

          The R15 coupled model has only half of the horizontal resolution of R30 in both the atmosphere and ocean. The low resolution and simplicity of this model, and the flux adjustment strategy, are designed to allow integrations that would otherwise be computationally prohibitive. This model has been invaluable throughout the past decade as an initial probe into the variability and sensitivity of the coupled atmosphere-ocean system. In the past year it has been used to study an ensemble of global warming scenarios, the relative importance of temperature and salinity changes for the weakening of the Atlantic overturning, and, in a simpler configuration with a mixed layer ocean, the evolution of climate over the last glacial-interglacial cycle.

          Both the R15 and R30 models have been used to simulate the evolution of the Earth's climate due to past changes in greenhouse gases and sulfate aerosols, as well as to project climate change into the 21st century. In these "global warming scenario" integrations, the format follows that used in a previous R15 model study (1473) which, in turn, is based on Mitchell et al. (1995)1. An equivalent CO2 concentration is used to represent changes in all of the trace greenhouse gases, and changes in aerosol loading are modeled by changing the surface albedo. Most of the R15 scenario integrations begin in 1765; the R30 integrations begin in 1865. The runs proceed until the late 21st century with equivalent CO2 increasing at a rate of 1% per year after 1990. Integrations using new scenarios proposed by IPCC-2000 are now underway as well. All of these projections are clearly uncertain. The radiative forcing for the 20th century is also uncertain, due to uncertainties in aerosol loading (both anthropogenic and volcanic), indirect effects of aerosols on clouds, and possible changes in the solar irradiance. Any climate projections must be considered with these uncertainties in mind.

  1.2 LONG CONTROL INTEGRATIONS AND NATURAL
        CLIMATIC VARIABILITY

ACTIVITIES FY99

     1.2.1 Control Integrations with R30

          A version of the R30 model (R30V1) has been successfully integrated for over 1,000 years. The drift in global mean SSTs is less than 0.1K, and other indices of the model climate, such as the amount of sea ice and the strength of the North Atlantic overturning, are virtually free of drift as well after the first 100 years. Prior to this integration, simulations of this duration at GFDL have only been available from models of lower resolution with less vigorous atmospheric eddies, a less well-defined intertropical convergence zone, and a more diffusive ocean.

          In the past year, a new version of this model (R30V2) has been constructed, differing in the initialization procedure and in the choice of oceanic sub-grid scale diffusivities. In the control integration of R30V1, the Atlantic thermohaline circulation (THC) weakens initially when the model is coupled. An iterative step has now been included in the initialization procedure to bring the temperatures at the start of the coupled integration into closer agreement with those provided to the atmosphere-only integration that forms part of the flux adjustment computation. In addition, the simulated El Niño-Southern Oscillation (ENSO) in R30V1 has a frequency which is too low, and is too regular and of too large an amplitude. While the model may have too low a resolution to provide robust ENSO simulations, previous experience suggested that lowering the horizontal diffusivity in the oceanic component of the model would modify this simulated ENSO in the right direction, and R30V2 was constructed with a lower diffusivity for this reason.

          The R30V2 has now been integrated for over 400 years in "control" mode. Fig. 1.1 contains time series of the THC and global mean temperature from the control runs of R30V1 and R30V2. The initial weakening of the THC in R30V1 has been successfully corrected. Note, however, that this weakening in R30V1 has little effect on global mean temperature. The standard deviation of global mean surface temperatures is close to the observed value in both versions. The tropical Pacific variability (not shown) has improved over that in R30V1, but the simulated ENSO in R30V2 is still of too low a frequency and of too large an amplitude.

          Versions of the model without flux adjustments have been integrated as well. Although the model maintains a fairly realistic climate without substantial drift in the mean temperature, its deficiencies are significant enough that this unadjusted model is not used for global warming scenario generation. These deficiencies, particularly the lack of boundary layer stratus in the eastern subtropical oceans and the inability to maintain a THC of the observed magnitude, are being studied as input into the development of the next generation model.

     1.2.2 Multi-Decadal Variability in R30

          Analysis of the control integrations of both versions of the R30 coupled model has revealed a dominant pattern of variability linking the Arctic and North Atlantic on multidecadal to century time scales, substantially stronger than that found in the R15 model in previous work (1428), and with stronger links between the Arctic and North Atlantic. Variability with this time scale is of particular importance for the issue of climate change detection. The simulated variability involves large-scale exchanges of heat and fresh water between the Arctic and the North Atlantic.

          Sea surface salinity and temperature in the model's Denmark Strait provide useful indices of this pattern of variability. The spectrum of surface salinity and temperature in this region is shown in Fig. 1.2, and is characterized by a peak at 60-70 years which stands significantly above a red noise background. This time scale is also seen in both instrumental and proxy records of climate2. Shown in Fig 1.3 are the simultaneous regression coefficients of this Denmark Strait salinity index with the sea surface salinity field (color shading) and surface currents (vectors). These patterns depict typical anomalies associated with this mode of variability. The southward vectors in the western Greenland Sea indicate an enhancement of the East Greenland Current, transporting cold and fresh waters out of the Arctic, while the northward vectors in the Barents Sea indicate enhanced flow of warm and salty water from the Nordic Seas into the Arctic. Blue colors indicate anomalously fresh water in the Greenland Sea, Labrador Sea, and North Atlantic. The atmosphere appears to play an active role by modulating the wind stress over the Fram Strait and Nordic Seas. Preliminary analyses suggest that these stress anomalies are partly forced by ocean temperature and sea ice changes, so that this pattern of variability cannot be fully understood as simply the response of the ocean and ice fields to stochastic atmospheric driving.

     1.2.3 An Extreme Case of Simulated Natural Variability

          Analysis has continued of an exceptional episode of North Atlantic cooling produced by natural variability within a 12,000 year control simulation of the R15 coupled model. The last

9,000 years of this simulation are virtually free of climate drift in all components of the model. Near year 3100, the North Atlantic and surrounding regions experience a very anomalous event, lasting for 10-20 years, in which the surface air temperature near the sinking region in the models thermohaline circulation is colder than the mean by 2-4K. There are no instances throughout the rest of the 12,000 year integration with anomalies larger than 2K. The cold anomaly decreases in size as one moves away from the site of deep water formation, but extends from Northern Canada, across Greenland and the North Atlantic, to Western Europe. Associated changes in sea level pressure and surface winds appear to enhance the cooling of the North Atlantic. During the event, the strength of the thermohaline overturning in the North Atlantic decreases from 18 to 14 Sverdrups.

          The dynamics of this cooling event appears to be similar in some respects to that of the background multi-decadal variability in this model (1428). It is initiated by an influx of relatively fresh water from the Arctic. When this water reaches the site of deep water formation, convection stops, allowing surface waters to cool and freshen further. The search for the ingredient which makes this event such a remarkable outlier within the distribution of smaller events is underway.

          If this event is realistic, it has implications for the study of past and future climate. A sharp cooling event in the Holocene (~8,200 years before the present) bears some resemblance to the modeled event, but is generally thought of as associated with glacial meltwater discharge into the North Atlantic. The model result suggests that events of this magnitude can occur in the absence of meltwater pulses or other external forcing. The possibility that unforced variability may have a strongly skewed, non-normal distribution would also make the detection of forced climate change more difficult, at least in the North Atlantic region.

     1.2.4 The Arctic Oscillation and Northern Hemisphere Temperatures
              in R30 and Observations

          The Arctic Oscillation (AO) is a pattern of atmospheric variability extending from the lower stratosphere to the surface, characterized by a zonally symmetric redistribution of atmospheric mass between the Arctic and midlatitudes. The AO has been associated with changes in temperature on both the local and hemispheric scale. It has also been argued that a trend in the AO in the past several decades is related to the increase in Northern Hemisphere temperatures observed during those decades. To investigate the AO-temperature relationship, observational data and output from the control R30V1 integration have been compared. The regional temperature anomalies associated with a particular phase of the AO are diagnosed by regressing the local temperature anomaly on an AO index. The colors displayed in Fig. 1.4 depict these regressed temperatures, with the observational results (left panel) compared with the model output (right panel). The contour lines in both plots represent

the regression coefficients of local sea level pressure on the AO index. Both the near-surface circulation pattern and regional temperature anomalies associated with a particular phase of the AO are simulated well by the model.

          The thermal signature of the AO (both simulated and observed) is characterized by a quadrupole pattern. Negative centers are located over northeastern North America and from northern Africa through southwestern Asia, while positive centers occur over southeastern North America and northern Eurasia. As a consequence, the association between the mean temperature of the northern extratropics (20°-90°N) and the AO index can be thought of as the relatively weak residual that remains after the near cancellation of these regional temperature anomalies. A moving-window regression analysis is used to explore this relationship, in which the northern extratropical mean temperature is regressed on the AO index for all overlapping 50-year periods. This analysis, which is applied to both observational data and model output, yields a time series of regression coefficients. The temporal variations in these coefficients provide information about the robustness of the relationship between the AO and extratropical mean temperature. For the observed record, the regression coefficients are weakly positive, with magnitudes that are only about one-tenth as large as the regional temperature anomalies associated with the AO. A secular trend in the regression coefficients is evident, with values for the most recent 50-year windows being more than 50% larger than those from the beginning of the 20th century. For the coupled model, which provides much longer time series, the regression coefficients vary in sign, with considerable low frequency variability evident. Some (but not all) of the difference between the typical values of the simulated and observed regression coefficients is due to a spatial sampling bias in the temperature observations, in which the positive centers of the quadrupole pattern are better sampled than the negative centers, especially in the earlier part of the instrumental record.

PLANS FY00

          The control run of R30V2 will be continued over the next year. The natural variability of both control integrations will be analyzed in detail, with a particular emphasis on multidecadal variability. Experiments will be conducted with the R30 atmospheric model, forced by the patterns of SST and sea-ice anomalies that are generated by the coupled model on inter-decadal time scales, to help understand the atmospheric part of the pattern of variability on these time scales. The analysis of the extreme event in the R15 model will also continue, with the goal of finding the mechanism that creates such a dramatic outlier in the distribution of the model's low frequency variability.

  1.3 GLOBAL WARMING STUDIES

ACTIVITIES FY99

     1.3.1 Global Warming Scenarios with the R30 Coupled Model

          An ensemble of five global warming scenario integrations for the period 1865-2089 is nearing completion using R30V1. The five integrations differ only in their initial conditions, selected at roughly 50 year intervals from the control integration.

          Shown in Fig. 1.5 are the time series of global mean surface temperature (air temperature over land and surface temperature over ocean) from observations and from the model integrations. The spread in the simulations indicates the model's natural variability. When compared with R15 simulations, the ensemble mean warming is broadly similar, but slightly delayed (by roughly 10 years). The standard deviation of annual mean temperatures in the R30 model is roughly 40% percent greater than in R15.

          The geographical distribution of simulated surface temperature trends have been compared with the observed trends for the period 1949-1997 (ko). The simulated and observed trends are consistent in most regions, taking into account the internal variability of the trends, as estimated from the model. There are also several areas which are inconsistent, all of which are regions where substantial cooling has been observed (primarily the midlatitude North Pacific, and parts of the Southwest Pacific and the Northwest Atlantic). These regional inconsistencies are thought to result from deficiencies in one or more of the following: 1) the prescribed radiative forcing; 2) the simulated response to this forcing; 3) the simulation of internal climatic variability; and 4) the observed temperature record. Distinguishing between these alternatives is obviously a high priority for future research.

          The observed temperature trends in this same period have also been compared to trends generated by natural variability in the control integration. In nearly 50% of the areas analyzed (where data was deemed adequate for this purpose), the observed warming trends exceed the 95th percentile of the simulated distribution of trends for the same location. If the model's simulation of natural variability is accurate, these observed trends are very unlikely to have occurred due to internal dynamics of the climate system.

          Figure 1.6 shows the observed and simulated zonal mean surface temperature changes, as a function of latitude and time over the period 1880-1997. The observed record is characterized by high latitude Northern Hemisphere warming in the 1920s and 30s, followed by a period of nearly steady temperatures until the 1970s, when a more uniformly distributed warming commenced. The source of the warming in the 1920s and 1930s has been a source of controversy. One of the ensemble members captures this early century warming remarkably

well. This warming is clearly due to the model's natural variability, as comparison with the other realizations makes clear. In fact, the early 20th century warming in the simulation appears to be related to the dominant pattern of multi-decadal variability of this model, originating in the North Atlantic and Arctic (1.2.2). In contrast, the more uniform warming starting in the 1970s is broadly reproduced by every member of the ensemble.

          The working hypothesis consistent with these model results is that the rapid warming in the Northern Hemisphere in the early 20th century was primarily due to intrinsic natural climatic variability, but that the bulk of the more recent broadly distributed warming was directly forced by increasing greenhouse gas concentrations. Confidence in the validity of this hypothesis is dependent on the realism of the model's interdecadal variability. Furthermore, since low frequency variability is expected to increase with increasing climate sensitivity (fz), it may be important that the model used here is quite sensitive (~3.7°C for the equilibrium response to CO2 doubling), within the upper third of the canonical sensitivity range of 1.5°-4.5°C.

          A new series of ensemble integrations has been initiated using R30V2. Experiments using scenarios for future radiative forcing recently supplied by the IPCC have been initiated as well.

     1.3.2 Global Warming Scenarios with the R15 Coupled Model

          A series of global warming experiments with the R15 model has been designed to examine the importance in such integrations of the date at which the transition is made from constant to increasing level of greenhouse gases (1614). Experiments were performed using the dates 1766, 1866, and 1916 for this transition, with three realizations of each scenario. Fig. 1.7 illustrates how the strength of the model's North Atlantic thermohaline circulation decreases into the 21st century. Regardless of the time of transition to increasing greenhouse forcing, by 2065 the thermohaline circulation weakens by ~40% relative to the control case. The differences resulting from earlier starting times and a smoother transition from pre-industrial to 20th century forcing appear to be very subtle in this climate index and in several others examined. The implication is that one can perform global warming projection experiments for the 21st century with transient forcing starting in the early 20th century. Reduction in the length of the integrations is particularly important for experiments with higher resolution models.

          The cause of this reduction in strength on the Atlantic thermohaline circulation in climate model simulations of global warming has been a subject of controversy. This question has been examined with a set of five multi-century R15 coupled model experiments (kc). The experiments are designed so that the role of various surface fluxes in weakening the THC can be separated and quantitatively assessed. Changes in net surface fresh-water fluxes (precipitation, evaporation, and runoff from land) are found to be responsible for about two-thirds of the weakening of the model's thermohaline circulation in a global warming scenario. Increases in high latitude precipitation and runoff (Fig. 1.8) freshen high latitude ocean waters, lower water densities, and inhibit deep water formation. Surface heat flux changes associated with the warmer temperatures account for about one-third of the simulating weakening in the 21st century. Wind stress variations have negligible impact on the R15 model's overturning strength in these scenario integrations.

     1.3.3 Extratropical Circulation Trends in Observations and in
              R30 Scenario Integrations

          Greenhouse warming will impact global atmospheric circulation patterns, as well as temperatures. One of the key issues is the relationship between this forced change in circulation and natural modes of variability. One can speculate that the patterns of natural variability are the "directions" in which the atmospheric flow can move with relatively weak restoring forces, and that external forcing might preferentially push the system in these directions. In the R30 scenario integrations, changes in the Southern Hemisphere fit this description, at least superficially. However, in the Northern Hemisphere the model's response to global warming does not clearly resemble the dominant pattern of natural variability.

          In the Southern Hemisphere, the main feature of the extratropical circulation response is a poleward shift of the zonal winds, of the meridional overturning cells, and of the storm track. The shift in the zonal winds is seen as an increase of the westerly winds to the south of the jet maximum, and a decrease in the westerly winds to the north (Fig. 1.9, top panel). This pattern is very similar to one of the phases of the dominant pattern of month-to-month zonal wind variability in the Southern Hemisphere, found both in the model and in observations (middle panel). When this pattern is removed from the response to greenhouse warming, the remaining pattern is a much broader scale increase of the winds, primarily confined to the upper troposphere and lower stratosphere (bottom panel).

          Additional atmospheric integrations with fixed SSTs and sea ice suggest that the poleward shift of the extratropical circulation can be thought of as primarily the response to a global mean SST increase and secondarily a response to the presence of a local minimum in the SST warming over the Southern Oceans. The stratospheric wind increase, on the other hand, follows directly from the model's large-scale upper level temperature response, which is a warming in the tropical upper troposphere, associated with increased tropical SSTs, and a cooling in the extratropical stratosphere, associated with the increased longwave radiation emission by CO2.

          In the Northern Hemisphere, however, the forced circulation response has a more complex structure that does not closely resemble the dominant pattern of month-to-month wind variability, the Arctic oscillation (AO). This is the case despite the fact this coupled model does possess a realistic AO (1.2.3). Observations show a recent trend in the AO index, corresponding to a poleward shift of the Northern Hemisphere winds. Whether the observed trend in the AO is associated with climate change, global warming, ozone depletion, or whether it is natural variability, remains controversial. The R30 simulations bear a closer resemblance to observed trends at lower than at upper tropospheric levels, so it is possible that the upper level response is affected adversely by poorly modeled stratospheric forcing and dynamics.

     1.3.4 Comparison of Observed and Simulated Trends in Sea Ice Extent

          Surface and satellite-based observations3,4 show a large decrease in Northern Hemisphere sea ice extent during the past 25 years (Fig. 1.10). These observed trends have been compared to the trends found in the control and transient integrations from the GFDL R15 coupled model and the Hadley Centre model, HadCM2. The transient integrations were forced by identical estimates of the observed radiative forcing for the past century.

          Based on the estimates of natural variability obtained from the control integrations of the coupled models, the observed decrease in Northern Hemisphere sea ice extent is far larger than would be expected from natural (unforced) climate variations. The observed decrease agrees quite well with the retreat in sea ice extent simulated by the models during the second half of the century, when these models are forced by estimates of the past radiative (greenhouse gas plus sulfate) forcing. Consistent with the model results, most of the

observed decrease occurred during the last 25 years. The conclusion, which is dependent on the realism of the models' simulation of the natural variability of sea ice, is that the observed sea ice decrease is a fingerprint of anthropogenic global warming. Both models project that the decrease of Northern Hemisphere sea ice extent will accelerate into the next century.

     1.3.5 Equilibrium Response of a Coupled Model to CO2 Increase

          Because of the long time needed in reaching an accurate steady state in coupled atmosphere-ocean models, the equilibrium response to increases CO2 has generally been studied, at GFDL and elsewhere in the climate research community, with mixed-layer ocean models. In these models it is assumed that there is no change in the oceanic heat transport when the CO2 is perturbed, which is a severe limitation. Given that many models calculate a transient weakening of the Atlantic thermohaline circulation in global warming simulations, there is also interest in the final equilibrated strength of this thermohaline circulation with higher CO2 values.

          Accurate, steady-state simulations with a control, doubled CO2, and quadrupled CO2 have now been achieved with the R15 coupled model. The lengths of these integrations range from 12,000 years for the control to 4,000 years for the doubled CO2 run. The thermohaline circulation in the Atlantic Ocean, which is responsible for most of the oceanic poleward heat transport in the present climate, is remarkably similar in pattern and amplitude among the three integrations (Fig. 1.11). The maximum value is ~18 Sverdrups in all three cases. The depth of the outflow as seen by the placement of the zero contour is also very similar in each case. The contrast with the transient simulations with nearly identical models (1.3.2), in which the circulation weakens in the first few centuries following a buildup of CO2, is striking. The explanation for this constancy of the equilibrium overturning remains unclear.

          Oceanic heat transports are also nearly unchanged in the three simulations, consistent with an earlier study by Manabe and Bryan (1985)5 obtained with an idealized one-hemisphere coupled model with simplified land-ocean configuration.

PLANS FY00

          Additional global warming scenario integrations will be completed with R30 to extend the existing ensembles and to create new ensembles using radiative forcing consistent with those provided by IPCC-2000. A CFC tracer will be included in a subset of the new ensemble integrations for evaluation of the oceanic uptake of tracers and heat. Trend analyses and comparisons with 20th century observations will be extended. Analysis will continue of circulation changes in the scenario integrations, with particular attention on the dynamics of the poleward shift of the circulation in the Southern Hemisphere. Special focus will also be placed on the analysis of the existing R30 realization that most resembles the observed 20th century temperature record. Study of the R15 equilibrium climates will be directed towards understanding why the equilibrium thermohaline circulation is insensitive to increases in greenhouse forcing. As part of the development of the next-generation climate model, sensitivity experiments will begin with new sea ice, land, and atmospheric models coupled to a mixed-layer ocean to study the effects on climate sensitivity of model modifications.

  1.4 SIMULATION OF CLIMATE VARIATIONS DURING
        THE LAST GLACIAL CYCLE

ACTIVITIES FY99

          Changes in the Earth's orbit are the prime candidate for forcing glacial-interglacial fluctuations of climate. An R15 atmosphere-mixed layer ocean model was recently completed in which variations in the orbital configuration for the past 120,000 are prescribed. To reduce the computational requirements, the orbital variations are accelerated by a factor of 30, a strategy that is consistent with the relatively short response time of the atmosphere-mixed layer

ocean system. Variations in Earth's orbital configuration consist of changes in the relative timing of the seasons with respect to the date of perihelion (the "precession" of the perihelion), changes in the tilt of Earth's axis (or obliquity), and changes in the eccentricity of Earth's orbit. These orbital variations can exert a significant influence on the amplitude of the seasonal cycle, as well as the annual mean radiation received at a particular latitude.

          Analysis of the simulated climatic time series reveals a complex response to orbital forcing, involving interactions on many time and space scales. Of particular interest is the interaction between variations in precession and obliquity that occurs in the Sahel region of Africa. During northern summer, variations in the local radiative forcing in this region (Fig. 1.12, red curve) are dominated by the precession, as evidenced by the ~20,000-year periodicity. The radiative forcing is largest when the summer solstice coincides with perihelion. Yet the response of surface air temperature (Fig. 1.12, black curve) corresponds instead to variations in obliquity, which have a periodicity of approximately 40,000 years. The absence of the precessional signal in the temperature response results from the opposing influences of local radiative forcing and changes in monsoon strength. When the precession cycle favors warm northern summers, the enhanced thermal contrast between the African-Eurasian landmass and the tropical oceans yields a stronger summer monsoon. In the Sahel, this translates into more cloudiness and precipitation, with enhanced soil wetness. These changes compensate

for the increased solar radiation, resulting in little temperature response to precessional forcing. Such compensation does not occur in response to obliquity variations, since the enhanced latitudinal contrast in insolation associated with low obliquity favors both a weak monsoon (i.e., less cloudiness and soil moisture) and a positive local radiative forcing. Thus the simulated Sahel temperature time series is strongly correlated with variations in obliquity, with warm periods occurring when obliquity is low.

PLANS FY00

          This simulation of the response to orbital parameters over the past 120,000 years will be analyzed from a variety of perspectives to help shed light on the sensitivity of regional climates to changes in radiative forcing.

  1.5 COUPLED MODEL DEVELOPMENT

ACTIVITIES FY99

          Development of the next-generation coupled model for climate studies is taking place within the framework provided by the GFDL Flexible Modeling System (FMS). For information beyond that provided here, refer to Section 3.1 for a discussion of progress on the FMS framework itself, Section 4.1.4 for progress on the new ice model, and Section 1.6.3 for progress on a new land model.

     1.5.1 Development of the Ocean Component of the Coupled
              Climate Model

          An earlier version of the Modular Ocean Model has been superseded by the latest MOM3 version, resulting in important improvements in the simulation of the ocean circulation. This effort has focused on standardizing three different grid resolutions: 1) 3° horizontal resolution with 25 levels, primarily for paleoclimatic studies; 2) 2° horizontal resolution with 25 levels, for studies of natural decadal-to-centennial variability requiring integration of up to 1000 years; and 3) 1° horizontal resolution with 51 vertical levels, for the study of climatic variability on annual-to-decadal time scales and the next generation of greenhouse warming scenarios.

          All three models produce simulations which are improved over the 4° and 2° ocean model currently in use R30 and R15 coupled models. This improvement can easily be seen in a comparison of the water mass characteristics of control simulations (Fig 1.13).

          In addition to resolution changes, modifications to the model include: variable meridional resolution, with higher resolution in the tropics to improve ENSO simulations; a free upper surface that permits a more realistic treatment of surface fresh water fluxes; Gent-McWilliams mixing of tracers; an explicit mixed layer parameterization; and partial cells at the ocean bottom to better resolve bottom topography.

     1.5.2 Development of the Atmospheric Component of the
              Coupled Climate Model

          A new spectral dynamical core has been constructed within the FMS framework, and a message-passing version of this core with one-dimensional domain decomposition has been successfully tested. As a check of the new software environment, successful tests have been performed replacing the new spectral core with the B-grid dynamical core (3.1.1), with identical "column-physics".

          To insure continuity with previous work, a version of the R30 spectral model has been constructed within the FMS using physical packages very similar to those in use in the R15 and R30 coupled models. Initial tests of a mixed-layer model constructed with this R30 atmosphere coupled to new ice and land models, and of a fully coupled climate model using MOM3, have begun.

          A version of the spectral core has also been developed with grid point advection for water vapor and other tracers using the finite-volume Lin-Rood advection scheme. This model has far fewer Gibbs ripples than the standard version of the spectral model, and requires no polar filter or extra "hole-filling" algorithm to prevent negative mixing ratios. The model produces a different cloud field, but otherwise generates a circulation that is very similar to the standard model.

          A new web-based diagnostic package (5.4) has substantially streamlined model development by allowing rapid comparisons of modeled and observed atmospheric fields.

PLANS FY00

          The new ocean model, together with new land and ice models, will first be coupled to the FMS version of the spectral model with an atmospheric physics package similar to that in the existing R30, and with similar atmospheric resolution. Improvements in atmospheric physics and resolution will then build from this foundation.

          Improvements in the "column physics" of the atmospheric model will be coordinated with other groups throughout GFDL as part of the development of the FMS. Currently, the R15 and R30 climate models have no diurnal cycle, no land heat capacity, no stability dependence in drag coefficients or boundary layer mixing, and a fixed boundary layer depth. All of these deficiencies are being addressed in models, both grid point and spectral, that are under development. Within the next year, preliminary climatic sensitivity studies will be conducted with an atmospheric model in which all of these deficiencies are addressed. New convection and cloud prediction schemes will be evaluated for their use in coupled climate simulations as they are tested and become available (2.2).

  1.6 HYDROLOGY AND GLOBAL CLIMATE

ACTIVITIES FY99

     1.6.1 Observed Interannual Variability of River Discharge

          A dataset of monthly precipitation and discharge for almost 200 large (>10,000 km2) river basins has been constructed to support diagnostic studies of land water balance and evaluation of hydrologic and climate models. Discharge records were drawn from the U.S. Geological Survey and the Global Runoff Data Centre. Basin-mean precipitation amounts were estimated by spatial interpolation and averaging of normalized anomalies of station records from the Global Historical Climatology Network.

          The precipitation/discharge dataset has been employed in an analysis of interannual variability of river discharge. This analysis revealed (Fig. 1.14) that the sensitivity of annual runoff to annual precipitation in any basin, as computed by regressing annual mean anomalies in runoff against annual mean precipitation, is determined mainly by the basin's runoff ratio (mean runoff as fraction of mean precipitation). Furthermore, this functional relation is consistent with one that can be derived from M.I. Budyko's semi-empirical water-balance relation, under the assumption that interannual water storage is negligible.

     1.6.2 Water and Energy Balance Models

          An evaluation of the long-term water and energy balances of land in the control run of the R30V1 coupled climate model is underway. This evaluation will provide a basis for interpretation of climate-change experiments and for future model developments. A preliminary investigation has revealed a systematic, qualitative difference between modeled and observed water and energy balances. The evaporation from any river basin is limited both by water and energy availability. It is generally observed that the energy limitation on evaporation is well approximated by the net radiation at the land surface. In the model, however, evaporation exceeds that supportable by net radiation over large areas of non-glaciated land in the middle and high latitudes (Fig. 1.15). The additional source of energy is a substantial downward flux of sensible heat from the atmosphere to the ground. This heat flux evidently results from the climate model's lack of any stability dependence in the calculation

of vertical fluxes in the atmospheric boundary layer, a deficiency that is being addressed in the atmospheric model development effort.

     1.6.3 Development of New Land Model

          As part of the FMS development effort (3.1), a new land model is under development. The runoff-generating portion of the land model differs from previous GFDL models in that plant stomatal resistance and soil heat storage are recognized, as is geographic variability of physical characteristics of the land (soil type and vegetation type). The evaluation of this model includes investigations of its ability, in stand-alone mode (i.e., forced by near surface atmospheric observations and estimates of radiative fluxes and precipitation) to reproduce geographic and temporal variations in water and energy balances, as observed through long-term river discharge.

          One aspect of the model evaluation has been an assessment of the predictive value of estimates of the geographical distributions of plant rooting depth, non-water-stressed bulk stomatal resistance, soil texture, and surface roughness length. A series of numerical experiments in stand-alone mode reveals that the model reproduces the observed runoff distribution no better with the geographically varying parameter estimates than it does with globally constant values of all parameters. This unexpected result implies either that 1) errors in the observational data are much larger than those induced by neglect of geographically varying land characteristics, 2) parameter variations are inadequately estimated, or 3) the land model is inadequate for quantifying this sensitivity to land characteristics. These possibilities are now under investigation.

PLANS FY00

          Development and evaluation of the land model will continue. To facilitate computation of daily-to-monthly river discharge (for coupling to the ocean and for model evaluation), parameterizations of groundwater storage and the river network will be developed. These will be tested in stand-alone experiments and in preliminary coupled runs under FMS. Error estimates will be derived for precipitation in the observational precipitation/discharge dataset. These estimates will permit enhanced evaluations of land model performance.

  1.7 WAVES, TURBULENCE, AND THE GENERAL CIRCULATION

ACTIVITIES FY99

     1.7.1 Tropical Intraseasonal Oscillations

          The simulation of the TIO continues to be a severe test of a model of the tropical atmosphere. In previous work, the sensitivity of the R30 model's simulation of the TIO to various aspects of the model have been diagnosed, including most recently the dependence on cloud feedback (kl). The TIO in the R30 coupled model has now been analyzed and compared with the uncoupled atmospheric model in which the seasonal dependence of SSTs is prescribed. The atmosphere has a relative humidity-based cloud prediction scheme and a moist convective adjustment. An integration with the R30 coupled model in which CO2 is doubled has also been compared to the control R30 coupled model simulation.

          Space-time spectra are estimated from daily-sampled data taken from the two coupled model runs over 100 years and the uncoupled model over 35 years. All three model runs possess a space-time spectral peak for the wavenumber one component of the equatorial upper-tropospheric zonal velocity corresponding to eastward-moving waves with a period ranging from 40-60 days, in agreement with that observed (Fig. 1.16). While it has been argued that coupling might enhance power in the TIO band, this is not the case for these models. The coupled and uncoupled versions have nearly the same amount of power in the TIO band, with the coupled model having slightly less. The doubling of CO2 has almost no effect on the power in the coupled model.

          The power of the TIO in the upper tropospheric winds in all of these models is larger than actually observed by roughly a factor of two, calling into some question the significance of this insensitivity of the spectral peaks to air-sea interaction and doubled CO2. Reasons for this amplitude error are under investigation. The model's amplitude appears to increase with

increasing vertical resolution, as indicated by comparison with earlier 9-level models, and is also enhanced by cloud feedback in this model. One aspect of the model's convective parameterization, the absence of cumulus momentum transport, is a candidate for explaining this discrepancy in amplitude. The inherent noisiness of the convective adjustment algorithm may also be responsible.

     1.7.2 Moist Convective Turbulence

          Deep moist convection has several distinctive features that complicate attempts at constructing even the simplest scaling arguments for the relevant velocity scales in the convective cores and for the intermittency of the convection. Models of horizontally homogeneous radiative-convective equilibrium with cloud-resolving, non-hydrostatic models are useful laboratories for exploring some of these complexities. Work with this kind of model has recently highlighted a distinctive feature of moist convection that has implications for the kinetic energy budget of the atmosphere as a whole (je). Frictional dissipation in a moist atmosphere can occur as a result of the familiar cascade of energy to the small scales at which friction can act, but it can also occur in the microscopic shear zones around hydrometeors when they fall. The size of the latter can be estimated in a model by computing the net energy utilized in lifting water, since it can be shown that nearly all of this increase in the potential energy of water must be dissipated around falling drops. The size of this kinetic energy dissipation associated with precipitation is found to be nearly 4 W/m2 in a three-dimensional radiative-convective model, which is 3 times as large as the flux of kinetic energy into and out of the resolved scales in the model. This mechanism could account for as much as 1-2 W/m2 globally, making it competitive with the energy flux through the large-scale atmospheric circulation.

          The entropy budget of this radiative-convective model has also been found to have surprising features. Theories for convective velocity scales have been constructed6,7 based on the assumption that one can estimate the frictional dissipation of kinetic energy from an entropy budget. The radiative-convective simulations show that frictional dissipation (that is, diffusion of momentum) is not the dominant irreversible source of entropy in the system, rather it is diffusion of water vapor. This fact complicates attempts to use the entropy budget to estimate convective velocity scales. Experiments in which the latent heat of vaporization is varied from zero (the dry limit) to the value appropriate for water vapor show that this entropy theory does indeed work for dry convection, but that it begins to breaks down when the latent heat of vaporization is but 10% of the correct value.

     1.7.3 Geostrophic Turbulence

          Geostrophic turbulence generated through baroclinic instability is of central importance for the midlatitude troposphere, where it is resolved by the current generation of GCMs. It must be parameterized in non-eddy resolving ocean models of the sort used in all global change and paleoclimate studies. It is important to try to take advantage of dynamical similarities between the atmosphere and oceans, so that one can test ideas for closure schemes in the atmosphere, where data are abundant. A review of geostrophic turbulence in the atmosphere has recently been completed, focusing on theories for baroclinic eddy heat fluxes (1627). Theories of this sort provide a basis for the improvement of oceanic eddy closures. A study of the feasibility of using altimeter data to infer near-surface horizontal oceanic eddy diffusivities has also been completed (1583), in which atmospheric data is used to test the proposed algorithm, assuming that geostrophic eddy dynamics is similar in the two media. The algorithm works well for the atmosphere, suggesting that we can, in fact, infer these diffusivities from altimeter data.

          Previous work on a theory of eddy amplitudes and scales in baroclinically generated geostrophic turbulence (1362) utilized a two-layer model, which is justifiable in an atmospheric setting, but cannot describe oceanic vertical structures. Work has commenced on problems associated with generalizing this theory to the oceanic context using a multi-layer horizontally homogenous quasi-geostrophic model. Both spin-down and sustained turbulence simulations have been performed. The development of closure schemes revolve, in part, around the question of how the system stops the inverse energy cascade to larger horizontal scales. The issue arises as to whether it is possible, in the oceanic parameter regime, for surface heat exchange to absorb the energy cascading to large scales, or whether bottom friction is necessary, as in the atmosphere. The tentative conclusion is that bottom friction is, in fact, necessary for eddy equilibration, at least in a horizontally homogeneous flat-bottomed framework. In addition, unforced spin-down calculation with energy initially localized near the surface are proving to be a promising tool in the analysis of the vertical structure of oceanic eddy amplitudes.

          Work has also begun on simulations of baroclinic geostrophic turbulence with dissipative numerical schemes that require no explicit subgrid-scale diffusivity to determine if they have any advantages over standard hyperviscosity formulations.

     1.7.4 Mechanisms of Interdecadal Variability in the North Atlantic

          In addition to the research involving full GCMs, the mechanisms of interdecadal variability of the climate system have also been investigated using more simplified models. The physical mechanisms sustaining oscillations on these time scales remain unclear. While salinity is likely to play a role, similar oscillatory phenomena have been found in ocean models in which density is a function of temperature only.

          The robustness of the oscillations has been examined, without salinity effects, using simplified ocean models, both with the primitive equations in a rectangular basin and with the simpler planetary geostrophic equations. The factors examined include wind forcing, mesoscale eddies in the ocean, and atmospheric heat transport, the latter being simulated with a simple diffusive energy balance model. The interdecadal oscillations are found to be a fairly robust phenomenon. The presence of wind stress tends to damp these oscillations, as does the use of realistic bottom topography. Their amplitude also decreases if one decreases the amplitude of the oceanic vertical diffusivity. On the other hand, both the presence of mesoscale eddies in the ocean, and stochastic forcing from the atmosphere enhance the oscillations. These results are consistent with the picture of the thermohaline oscillations as a damped oscillator, with the forcing from deformation scale eddies or atmospheric variability serving as an energy source.

     1.7.5 Jovian Atmospheric Circulation

ACTIVITIES FY99

          The Jovian atmosphere involves geophysical fluid dynamics processes in novel arrangements and under different constraints than in the Earth's atmosphere and oceans. Understanding the Jovian circulation will help to generalize and clarify our theories for these processes.

          The main problem in defining Jupiter's meteorology, particularly the character of the jets and vortices, comes from the fact that the nature and extent of the motions are not generally known for the region below the clouds. To develop a theory for the circulation, 3-D primitive equation models are used to examine the formation and coexistence of the various turbulent and coherent phenomena for hypothetical vertical structures. The main hypotheses involved - that the atmospheric circulation occurs with a relatively thin upper layer and is driven by horizontal temperature gradients have been examined with a wide range of meteorological models over the years. Present studies extend these models by considering the role of the vertical structure more specifically.

          The main questions concerning the formation and maintenance of jets and vortices in a planetary context are: 1) why do vortices exist and what determines their lifetime; 2) what processes control the various single and multiple vortex states seen in the various Jovian anticyclonic zones; 3) how are the various jets generated in low- and mid-latitudes for an unbounded atmosphere in a way that is consistent with vortex genesis and maintenance; and 4) what causes the onset of the equatorial super-rotation in such a multi-jet, multi-vortex system. Possible answers to these questions have been found that are mutually consistent and suggest feasible vertical forms for the winds. The preferred vertical structures derived from such experimental calculations have recently been confirmed by the vertical wind and thermal distributions measured by the "Galileo" spacecraft probe of Jupiter's atmosphere (1454).

          The main vortex sets of Jupiter are made up of the singular Great Red Spot (GRS) at 22°S latitude, the three Large Ovals at 33°S latitude (after two of the original four merged in 1999), and the dozen or so Small Ovals that occur at 41°S latitude. To examine how such vortices are generated and behave, calculations have been made with the primitive equation model subject to various hypothetical and experimental forms of heating thought to be appropriate to the Jovian atmosphere. Realistic sets of equilibrated vortices have been produced in simulations of the three major groups as shown in the example in Fig. 1.17. They all exist under the same physical conditions and preferred vertical structures and parameter range derived earlier in basic vortex studies (1400). Solutions with interacting, merging, and collapsing vortices have also been derived. The main parameter values determining the characteristics of the three classes of vortices, derived from a wide range of calculations, are now well known.

          The circulation in Jupiter's atmosphere is made up of five westerly jets in each hemisphere in the region between 60° latitude and the equator, together with a super-rotating

flow at the equator itself. To examine how such multiple jets are generated, calculations have been made with the primitive equation model for various horizontal and vertical distributions of heating. Realistic jet sets, such as that shown in Fig. 1.18, have been produced from simple heating forms. In particular, they indicate that in low latitudes (10° ± 2°) a probe of the upper 100 km of the atmosphere, such as that made by the "Galileo" probe in 1996, would measure winds that vary little with depth, even though the jets do decay deeper in the interior. To examine the generality of multi-jet flows, a variety of calculations have been made to examine how the circulations depend upon the rotation rate and vertical extent of such flows.

PLANS FY00

          Work will continue on all of these theory-based projects. Analysis of the TIO will focus on factors that control the amplitude of this mode of variability, particularly cumulus momentum transport. Fundamental studies of moist convective turbulence will continue focusing on idealized inhomogeneous systems. Geostrophic turbulence research will concentrate on homogeneous simulations with oceanic stratification, the analysis of different mechanisms for halting the inverse energy cascade, and testing of dissipative numerical algorithms that do not require explicit subgrid-scale diffusion. Attention will also turn to high resolution studies of the simplest barotropic jet instabilities, so as to address connections between inhomogeneous and homogenous 2-D turbulence in idealized settings. Attempts will be made to close the gap between idealized models of interdecadal variability and the variability in comprehensive coupled models, on the one hand, and to better understand the oscillations that occur in the idealized models on the other.

          The solutions describing the genesis and maintenance of the multiple vortex sets and the multiple jets on Jupiter, as well as their parametric variations will be fully analyzed and documented. Additional calculations will be made to produce global circulations that achieve a more complete synthesis of the jets and vortices.


1. Mitchell, J. H., T. C. Johns, J. M. Gregory, and S. Tett, Climate response to increasing levels of greenhouse gases and sulfate aerosols, Nature, 376, 501-504, 1995.

2. Mann, M. E., and J. Park, Joint spatiotemporal modes of surface temperature and sea level pressure variability in the Northern Hemisphere during the last century, J. Climate, 9, 2137-2162, 1996.

3. Parkinson, C. L., D. J. Cavaleiri, P. Gloersen, H. J. Zwally, and J. C. Comiso, Arctic Sea Ice Extents, Areas, and Trends, 1978-1996, J. Geophys. Res. Oceans, in press, 1999.

4. Chapman, W. L., and J. E. Walsh, Recent variations of sea ice and air temperature in high latitudes, Bull. Amer. Meteor. Soc., 74, 33-47, 1993.

5. Manabe, S., and K. Bryan, CO2-induced change in a coupled ocean-atmosphere model and its paleoclimatic implications, J. Geophys. Res., 90(C6), 11,689-11,707, 1985.

6. Emanuel, K. A., and M. Bister, Moist convective velocity and buoyancy scales, J. Atmos. Sci., 53, 3276-3285, 1996.

7. Renno, N. O., and A. P. Ingersoll, Natural convection as a heat engine. A theory for CAPE, J. Atmos. Sci., 53, 572-585, 1996.

*Portions of this document contain material that has not yet been formally published and may not be quoted or referenced without explicit permission of the author(s).