ASR 2002

SIGNIFICANT ACCOMPLISHMENTS

Oceanography Section

Yeager and Large completed an analysis based on 1958-1997 hindcast simulations which shows how interannual anomalies in the Pacific extratropics are communicated to the equatorial Pacific via isopycnal pathways, thereby contributing to subthermocline decadal variability on the equator. The figure shows the propagation of salinity (top panel, in psu) and temperature (bottom panel, in deg C) anomalies computed as annual deviations from the mean isopycnal property at horizontal pathways north and south of the equator.


OS_highlightfigure.jpg (109254 bytes)

Climate Change Research Section

Recent simulations of PCM were done using the new volcanic forcing data put together by of Caspar Ammann (postdoc, Advanced Study Program). These simulations show the additional factor of volcanic forcing, using the same method described in the above paleoclimate simulations, improves agreement of simulated globally averaged temperature with observations.

This figure shows the globally averaged surface air temperature anomaly from 1880 to present. Also, the range of observed temperatur estimates from several sources is shown in the gray color. Note that the simulated variations of temperature change closely match the increase of the temperature from the 1880s to the present, and the cooling effects of particular volcanoes are striking. The addition of volcanic forcing helps explain the short-term cooling events that are well observed.

Climate Analysis Section

Aiguo Dai (Climate Change Research Section) and Kevin Trenberth (Climate Analysis Section, CAS) have estimated annual and monthly mean values of continental freshwater discharge into the oceans at 1 degree resolution using several methods.  The most accurate estimate is based on stream-flow data from the world’s largest 921 rivers, supplemented with estimates of discharge from unmonitored areas based on the ratios of runoff and drainage area between the unmonitored and monitored regions.  Simulations using a river transport model (RTM) forced by a runoff field were used to derive the river mouth outflow from the farthest downstream gauge records.  Separate estimates are also made using RTM simulations forced by three different runoff fields, including those from estimates of precipitation P minus evaporation E computed as residuals from reanalyses.  Compared with previous estimates, improvements are made in extending observed discharge downstream to the river mouth, in accounting for the unmonitored stream-flow, in discharging runoff at correct locations, and in providing an annual cycle of continental discharge.  Snow accumulation and melt are shown to have large effects.  The use of river mouth outflow increases the global continental discharge by 19% compared with unadjusted stream-flow from the farthest downstream stations.  The river-based estimate of global continental discharge is 37288 ± 662 km3/yr.  The river flow and discharge data are available from the CAS data catalog (http://www.cgd.ucar.edu/cas/catalog/).

This figure shows annual discharge rate (103 m3 s-1) from each 4° lat by 5° long. coastal box.  The numbers are the total discharge (in 103 m3 s-1) from the coasts behind the solid lines.  Blank coastal boxes have zero discharge.

 

This figure shows annual discharge into the global ocean smoothed using 5° lat running-mean from the new four different cases, compared with that of Baumgartner and Reichel (1975).  The one based on the 921 rivers is believed to be best.

 

Clara Deser (CAS), along with Michael Alexander (Climate Diagnostics Center, CDC) and Michael Timlin (CDC/NOAA), documented the recurrence mechanism over the North Pacific and North Atlantic in which winter SST anomalies created by atmospheric circulation changes persist beneath the shallow summer thermocline and become re-entrained into the mixed layer during the subsequent fall and early winter, thereby “re-emerging.”  This process extends the persistence of winter SST anomalies beyond the timescale associated with the thermal inertia of a fixed-depth mixed layer and has the potential to affect the winter-to-winter persistence of the overlying atmosphere as well.  In related work, Deser, Alexander, and Timlin have proposed an extension to the theoretical stochastic climate model of Hasselmann to incorporate the effects of the re-emergence process on the predicted winter-to-winter persistence of SST anomalies.  By comparing the extended Hasselmann theory to a more sophisticated coupled model (a bulk ocean mixed layer model and atmospheric GCM) and to observations, they were able to demonstrate that the extended theory is an excellent approximation to the more complex coupled model and agrees well with the observed persistence characteristics of SST anomalies in both the North Pacific and North Atlantic.  They propose that the extended Hasselmann paradigm should be adopted as the relevant “null hypothesis” for midlatitude SST variability.

 

Persistence characteristics of sea surface temperature (SST) anomalies in the North Pacific.  This figure shows monthly lag autocorrelation curves for SST anomalies in the western North Pacific (35-45N, 155E-170W) from observations (dashed black curve) and a bulk ocean mixed layer model coupled to an atmospheric GCM (solid black curve) starting from March.  Note the high winter-to-winter persistence and low winter-to-summer persistence, consistent with the reemergence mechanism.  The grey exponential curve is based upon a proposed extension of the Hasselmann theory for SST anomalies created by atmospheric forcing.  It is shown to provide an excellent fit to the observed winter-to-winter memory of SST anomalies (the original theory predicts no winter-to-winter persistence).

 

James Hurrell (CAS), Martin Hoerling (CDC/NOAA) and Adam Phillips (CAS) continued their work that links North Atlantic climate change since 1950 to a progressive warming of tropical SSTs, especially over the Indian and Pacific Oceans.  Additional analysis has been made of CCM3 simulations forced by monthly, observed SST and sea ice variations during 1950-1999, as well as similar experiments with other AGCMs.  Further work tests the atmospheric sensitivity to regional patterns of tropical SST variability.  The forcing patterns include the observed 1950-1999 trend in tropical SSTs, and an idealized pattern of warming covering only the Indian Ocean.  The model-based solutions reveal an annular response to Indian Ocean forcing, in the sense that increased convection (diabatic heating) over that region drives a strengthening of the stratospheric polar vortex and an annular response in the troposphere that includes the positive index phase of the NAO.

This figure shows the linear trend in January-March geopotential heights (gpm per 50 years) in response to the observed linear trend component of the SST change over 1950-1999. In the left panel, the observed monthly evolution of the SST trend over the tropics (20S-20N) provided the anomalous lower boundary forcing, while in the right panel the SST trend over the Indo-Pacific region (45-170E, 20S-20N) was specified.  The height anomalies are obtained from 25-member ensembles of CCM3.6 relative to a control (climatological SST) simulation. The simulated changes over the Arctic and North Atlantic closely resemble observed changes, arguing that observed climate change in these regions is linked to a progressive warming of tropical SSTs, especially over the Indian and western Pacific Oceans.

 

Hurrell is the volume editor for a forthcoming AGU monograph “The North Atlantic Oscillation.”  This is a principal outcome of a Chapman Conference held in December 2000 on the NAO.  The monograph is thematically organized and provides a comprehensive (multidisciplinary) overview of material (theory, observations, models) related to the NAO.  There are 12 chapters, each presenting a thorough overview of a topic as well as new research.  Each chapter was subjected to critical peer review and revised accordingly.  Hurrell, Martin Visbeck (Lamont Doherty Earth Observatory, LDEO), Yochanan Kushnir (LDEO), and Geir Ottersen (University of Oslo and Institute of Marine Research) authored the introductory chapter of the volume, and they also published a short article in CLIVAR Exchanges announcing the book and outlining its contents.

 

 

This figure shows the book cover of the AGU monograph “The North Atlantic Oscillation.”

 

Terrestrial Sciences Section (TSS)

Members of the Terrestrial Sciences Section, in collaboration with the CCSM Land Model and Biogeochemistry Working Groups, participated in and oversaw the development and implementation of the new Community Land Model (CLM2). CLM2 significantly reduces the Northern Hemisphere summer surface temperature cold bias compared to the NCAR LSM land surface model, reduces annual precipitation, improves simulated snow cover, and improves the seasonality of runoff (figure). The improvements to runoff are particularly evident in cold regions, where NCAR LSM had near constant runoff throughout the year while CLM2 has low runoff in winter when the soil is frozen and peak runoff in spring during snow melt. A grid cell-based river routing scheme routes the runoff generated by the column biogeophysics and hydrology downstream into oceans.

Members of TSS have expanded the capability of CLM2 to include biogeochemistry (carbon and nitrogen cycles, mineral aerosols, biogenic volatile organic compounds), vegetation dynamics, soil degradation, and urbanization. Applications of CLM2 coupled to climate models demonstrated the importance of land use (croplands) in cooling the climate of the United States and of vegetation as a determinant of the climate of North Africa during the mid-Holocene 6000 years ago. An active research program in carbon data assimilation has been formed within the section.

Community Climate System Model (CCSM)

Figure caption:.  Seasonal ice extent simulated by the sea ice component of CCSM averaged over the last 20 years of the control run.  The black line represents the10% contour from SSMI observations and indicates that the simulated sea ice extent agrees well with the observations.

Figure caption:  The forty year running trend in the simulated NAO index.  Also shown is the maximum observed forty year trend of the NAO index from  the NCEP reanalysis. b) The percent of time that a trend of a certain   length is present in the model simulations. The 0.1%, 0.5%, 1.0%, 2.0%,  3.0%, 5.0%, and 10.0% contour intervals are shown. The maximum observed  trends from the observations are shown as asterisks. These results indicate  that the observed trends in the NAO index are unusual compared to those in the CCSM control run. One hypothesis is that the difference is due to the changes in climate forcing that have taken place during the past 50 years.  This hypothesis will be tested using the CCSM 20th Century simulations.

Geophysical Statistics Project (GSP)

Claudia Tebaldi (GSP, ESIG and Research Applications Program) in collaboration with Linda Mearns (ESIG) and Richard Smith (University of North Carolina) has developed a Bayesian, random effects model to combine the results of different climate model simulations and quantify the uncertainty.  Different General Circulation Models (AOGCM) produce different climate change projections, these differences being especially striking when trying to evaluate climate change on a regional (subcontinental) scale. Studies of multi-model ensemble predictions try to reconcile the AOGCM’s responses by combining them into (weighted) means and use the ensemble spread as an envelope of uncertainty around them. In it natural to embed this analysis in a statistical context in order to formalize these otherwise heuristic procedures.  A statistical model also delineates the hypotheses and optimization criteria leading to a final estimate of climate change and uncertainty surrounding it.

 

This figure summarizes the posterior distributions of change in average temperature across the 22 regions for the 9 AOGCM experiments.  Top panel, winter season.  Bottom panel, summer season.  The distributions for winter climate change have lower variance within region, but higher variance among regions than the distributions for summer climate change, some regions show distinctly larger climate change signals than others, especially high latitude regions of the Northern experiments such as Alaska (ALA), Greenland (GRL) and Northern Asia (NAS).  Notice also how all the distributions’ supports are bounded away from zero, on the positive side of the real line, making the case for a significant global warming effect, for all regions and all seasons.

This project is lead by Bengtsson in collaboration with Nychka and Chris Snyder (MMM). Several aspects of numerical weather prediction (NWP) make forecasting and data assimilation particularly challenging: very high-dimensional systems, strongly non-linear (possibly chaotic) dynamics, and real-time requirements for assimilating data and physical models.  In practice one must:  address multi-modal forecast distributions, specify spatial covariance structures, use severely rank-deficient matrices, devise sampling schemes, and understand the properties of sample based filtering algorithms. In this project we implement Bayes theorem through an approximation based on a discrete sample from a mixture of Gaussian distributions. To handle non-linear systems and still have a stable filtering method, we have found that it is important to use nearest neighborhoods of states to derive updates.  Building on the success of this NonGaussian filter in Lorenz '63 3-dimensional system an extension was formulated for higher dimensional systems. The basic idea is to consider mixture approximations that are local both in the state space and also in the physical space. As a starting point it is demonstrated that this approach can be effective in handling distinctly NonGaussian distributions produced by the 40-dimensional Lorenz '93 system.

 

 

This figure gives a perspective rendition of 50 ensemble members for the first three state variables of the 40 variable Lorenz '93 system at particular time step in its evolution.  The surprising feature is the clear departures from a multivariate normal distribution in these point clouds. Usual methods of data assimilation would update the posterior of the state under the assumption that the ensemble has a joint Gaussian distribution.  When the distribution is approximated by a more flexible model (mixture of Gaussians) the improvement in RMSE for forecasting from this time step is approximately 15%.

 

Climate Dynamics and Predictability (CDP)

Ramalingam Saravanan (CDP) and Prof. Ping Chang (Texas A&M) have been exploring numerical seasonal prediction using a model where only the thermodynamic coupling between the atmosphere and the ocean is considered. In this study, the ocean is assumed to be a motionless slab, contributing only through its heat capacity. The dynamic role of the ocean, associated with the ocean circulation, was deliberately ignored. The coupled model consisted of the NCAR atmospheric model, CCM3, coupled to a slab ocean model, with spatially varying slab depth representing the oceanic mixed layer.  A series of numerical experiments analogous to the hindcast experiments of the Dynamical Seasonal Prediction (DSP) intercomparison project (Shukla et al., Bulletin of the Atmospheric Sciences, 2000) were carried out. In contrast to the hindcast DSP experiments, these were true forecast experiments. The slab ocean was initialized with the observed SST, and the coupled model was integrated forward in time for nine months starting from December 15th. A ten-member ensemble of experiments was carried out, for the 16 years from 1981 to 1996. Analysis of the forecasts shows that thermodynamic coupling alone is sufficient to produce skill comparable to the hindcast experiments, both in the Pacific-North American region and in the Tropical Atlantic.  This suggests that such a coupled configuration can provide a “low-cost alternative” for coupled prediction, while awaiting the resolution of problems such as the severe systematic errors in fully coupled predictive models of ENSO. 

saravanan_fig.gif.gif (19990 bytes)

This figure shows predictability in an atmospheric model coupled to a mixed layer ocean: El Nino and north tropical Atlantic sea surface temperatures.

Climate Modeling Section (CMS)

This year, the Climate Modeling Section in collaboration with the Atmospheric Modeling Working Group released a major new release of the NCAR Community Atmosphere Model called CAM2. This model incorporates improvements in the physics, dynamics, and numerical formulations developed in partnership with colleagues at universities and national laboratories. Some of the major enhancements include better treatments of cloud condensate, radiation processes, and sea ice. Through NCAR's ongoing collaboration with NASA, CAM2 incorporates a new formulation for atmospheric circulation known as the Finite Volume dynamical core. CAM2 will be the atmospheric component of the coupled climate models used by NCAR in upcoming national and international climate assessments.

This figure shows the outgoing longwave radiation (OLR) simulated by the new Community Atmosphere Model (CAM2, left panel) and measured by the NASA Earth Radiation Budget Satellite (ERBS, right panel). The region is the tropical Pacific between 100E and 80W. The OLR has been averaged between 5N and 5S. The mean value for the period 1985-1999 has been subtracted to give anomalies, i.e. differences relative to the long term mean. CAM2 was integrated using observed sea surface temperatures.

Positive anomalies east (to the right of) the date line show periods when convection is suppressed relative to the long-term mean. These roughly correspond to La Ninas during the observational record. Negative anomalies east of the date line show period of enhanced convection corresponding to El Ninos. Notice that the anomalies east and west of the dateline generally have opposite sign, corresponding to the dipole in convective cloud cover driven by ENSO.

The figure shows that the new CAM is able to reproduce the tropical fluctuations in OLR with considerable fidelity in their timing, strength, and geographical distribution. This is a marked improvement compared to the previous version of CAM.