Climate Change Research Section

Bette Otto-Bliesner (CCR) and Esther Brady (CCR) show that the significant improvements in the component models of CCSM2 now result in improved simulation of the "Greening of the Sahara" during the early to mid-Holocene.   Mid-Holocene proxies for northern Africa suggest that lake levels were significantly higher and steppe and xerophytic vegetation grew in present-day desert regions.  PMIP atmosphere-only GCMs and previous atmosphere-ocean GCMs (including CSM1) underestimated both the northward shift and the magnitude of the precipitation increase required to maintain steppe vegetation.  A CCSM2 simulation for 8.5 ky BP gives a more realistic northward expansion of precipitation into the Saharan-Sahel region with both a longer African summer monsoon season and greater rainfall during the monsoon season.  

This figure shows the annual precipitation change (cm) over Africa for 8.5 ky BP compared to present simulated by CCSM2.  The insets give details of the monthly precipitation changes simulated for 8.5 ky BP.

Aixue Hu (CCR) and Gerald Meehl (CCR) conducted a set of CMIP coordinated experiments to investigate the sensitivity of the Thermohaline Circulation (THC) to the variability of the surface freshwater flux using NCAR’s CCSM2.0. The resulting THC variations show that when the northern North Atlantic receives a 0.1 Sv additional freshwater (hosing), the THC slows down by 4.4 Sv, and then recovers to its full strength 50 years after the end of the hosing. When atmospheric CO₂ is doubled (quadrupled) the present day value THC weakens by 1.5 (2.9) Sv. This weakening of THC is related to the CO₂ induced warming and the transport of melt ice water from Arctic into the Labrador Sea. When the surface freshwater forcing from the control run is used in the 1% CO₂ run, the THC is further weakened due to a higher freshwater input in the Labrador Sea. On the other hand, the use of the 1% CO₂ run’s freshwater flux in the control run leads to a stronger THC. This is due to an overall less freshwater flux input in the northern North Atlantic.

This figure shows 13-year low pass filtered maximum Atlantic meridional overturning. Black line is for the control run. Blue dashed line is for hosing run. Green dashed line is for 1% CO₂ run. Red dashed line is for the 1% CO₂ run with control run’s freshwater flux. Yellow line is for the control run with 1% CO₂ run’s freshwater flux.

Terrestrial Sciences Section

Members of the Terrestrial Sciences Section continued their work to implement biogeochemistry and its feedback on climate in the Community Land Model and Community Climate System Model. Natalie Mahowald continued her research to identify the sources, transport, and sinks of mineral aerosols and to understand the feedback of mineral aerosols on climate. Much of her work focused on distinguishing natural sources of mineral aerosols from anthropogenic sources due to land use. In addition, Mahowald made the first estimates of future and preindustrial dust sources using CCSM simulations.  These simulations suggest that mineral aerosols may decrease by 20-60% in the future. The simulations showed that mineral aerosols are very sensitive to human impacts on carbon dioxide, land use and climate change, and because of the important role of mineral aerosols in modulating climate and biogeochemistry, may cause important feedbacks.

This figure shows mineral aerosol loading under different climate regimes using six scenarios of human/dust interactions in the source areas: no source changes (TIMIND), source changes due to precipitation and temperature (BASE), source changes due to precipitation, temperature and carbon dioxide fertilization (BASECO2), and the last three include a 50% landuse source (from Mahowald and Luo, 2003).

Peter Thornton added fully prognostic terrestrial carbon and nitrogen cycles to CLM. Model experiments demonstrated that CLM replicates the behavior of other carbon cycle models, which do not include the nitrogen cycle, when nitrogen dynamics are switched off, but that the climate coupling dynamics are significantly different with the nitrogen dynamics turned on, pointing to an important dimension of uncertainty in the current best estimates of coupled system behavior.

Community Climate System Model

African Drying

A key unresolved question is how regional patterns of climate will respond to anthropogenic and natural forcings during the 21^st Century. A diagnosis and attribution for the observed regional changes during the 20^th Century is one prerequisite for gaining insights on this problem.

Among the most striking examples of a regional climate change has been the decreasing summer monsoon rainfall over tropical North Africa. This region receives most (80-90%) of its annual mean rainfall during July, August and September (JAS), and after relatively wet decades in the middle of the 20^th Century, it has suffered through severe and prolonged drought especially during the 1970s and 1980s. Furthermore, the trend toward drought has followed the north-south migration of monsoon rains, and a substantial drying trend has also occurred over southern Africa during February, March and April (FMA; Figure 1).

A 15-member ensemble of CAM2.0 simulations, forced with observed, global sea surface temperature (SST) and sea ice concentrations since 1950, largely reproduces the observed patterns of African drying, although regional differences from observations are notable (Figure 1). The results argue for a strong oceanic control on the emergence of African drought, including its spatial and seasonal dependency, since 1950. While it has long been recognized that variations in African rainfall are accompanied by coherent global patterns of SST anomalies, the CAM2.0 results represent a significant improvement over the ability of previous versions of the NCAR atmospheric model to capture the observed changes (not shown).

Figure 1. Linear trend (1950-2001) of February-April (left) and July-September (right) seasonal mean rainfall from gridded observed data (top) and the 15-member CAM2.0 ensemble mean (bottom). The units are mm day^-1 50yr^-1.  

Oceanography Section

Air-sea-ice Interactions

The influence of the North Atlantic Oscillation-Arctic Oscillation (NAO-AO) on northern hemisphere sea-ice variability has been investigated in the 1000-year control integration of the CCSM2 by Marika Holland (OS).  The model reasonably simulates the spatial structure and variance of the sea level pressure (SLP), surface air temperature, and precipitation associated with the NAO-AO. The sea-ice response to the NAO-AO compares well to observations.  Variations in the spatial structure of the SLP anomalies associated with the NAO-AO over the course of the integration are present and part of the natural variability of the system.   However, the magnitude of the observed trends over the last 40 years in the NAO-AO are never realized in the model simulations, suggesting that these trends may be associated with changes in anthropogenic forcing. 

This figure shows the (a) simulated and (b) observed winter (JFM) precipitation regressed on the NAO–AO index. Observed precipitation from 1979 to 1995 from Xie and Arkin (1996) is used for the analysis shown in (b).

Climate Modeling Section

New version of CAM

Members of the Climate Modeling Section (CMS) (Byron Boville, WilliamCollins, James Hack, Philip Rasch, and David Williamson), in collaboration with colleagues in the university and national laboratory community, contributed broadly to the development of the next generation Community Atmosphere Model (CAM). The new model is based on the CAM2, which was publicly released in June of 2002, and includes a large number of significant improvements and enhancements to the treatment of physical processes, new modeling and diagnostic extensions such as a Slab Ocean Model (SOM) configuration, and additional improvements to the software engineering implementation of the model. The simulation exhibits a number of important improvements with respect to the CAM2 simulation, including a reduction of the warm bias present in the CAM2 arctic simulations, a reduction of the cold bias at the tropical tropopause, and a significantly more realistic cloud response to tropical sea surface temperature (SST) variations. These improvements address several of the more important systematic biases identified in the coupled and uncoupled simulations with CAM2. It is expected that this new atmosphere will be incorporated in the version of CCSM used for NCAR's participation in the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report.

This figure shows changes in precipitation rate in response to ENSO as simulated by a prototype version of the new CAM3 in comparison with satellite-based observational estimates.

WACCM Developments

Byron Boville and Fabrizio Sassi, in collaboration with colleagues ACD and HAO have developed a preliminary version of WACCM2 which exhibits a very realistic simulation of the composition of the  middle atmosphere. The model includes fully-interactive chemistry and accurately reproduces the time evolution of many observed  features of the middle-atmosphere, including the "tape recorder"  of water vapor as shown in the figure below. 

This figure shows some aspects of the chemistry simulation obtained with WACCM2. The “tape recorder” of water vapor (in ppmv) at the Equator is compared to an observed climatology of HALOE.   It should be noted that not only the vertical extent of the tape recorder in the two panels is quite similar (a clear improvement compared to earlier simulations), but also the timing of the extreme dry/wet events is very good. The ozone column (Dobson Units, DU; contour interval is 20 DU from 130 DU to 570 DU) shows pronounced ozone depletion at southern latitudes during fall with a minimum ozone column of 130 DU in mid-October.

Climate Dynamics and Predictability (CDP)

Ocean sea surface temperature variations are shown to be responsible for decadal and secular variations in Sahel precipitation. Saravanan and Giannini show that the recent Sahelian dry period is the result of equatorial Atlantic warm temperature anomaly.

This figure shows:  Observed and modeled precipitation in Sahel. Model forced with observed SST.

Geophysical Statistics Project (GSP)

Sparse Approximations to Covariance Matrices: Approximate statistical methods based on compactly tapering covariance matrices were developed to handle large spatial prediction problems where observations occur at irregular locations. The potential speedup is striking, suggesting that a spatial prediction (Kriging) can be computed for several thousand points interactively in the R statistical computing environment. Also, as suggested by the theory, the tapering approximation is quite accurate sacrificing little statistical efficiency compared to an exact calculation.



This figure is an example of the computational time for the key step in a spatial prediction problem. Different lines indicate the time for solving a linear system for a 1-d Kriging problem based on two different high-level languages (Matlab and R) and different levels of tapering. For the R package the time difference between the exact solution verses using a tapered, sparse approximation is more that a factor of 100 for 1000 spatial locations.

Statistical Supercomputing with the Gibbs Sampler: The a software infrastructure was developed to complete a nontrivial statistical analysis of a large geophysical data stream. The current work has been able to efficiently blend QuikSCAT surface winds with analysis winds over a 48° by  128° (at 0.5 degree resolution) domain in six-hour intervals for a three-year period. Posterior means from this analysis along with 50 posterior samples are being made available on DVD in network Common Data Format (netCDF). To our knowledge, this spatio-temporal data product based on Markov Chain Monte Carlo is substantially larger than any other analysis attempted by a statistical research group.



This figure gives an example of the U component (A) and the divergence (B) of a wind realization based on blending NCEP and satellite wind measurements. It one of the 50 ensemble members in the GibbsWinds data product for this time point. The bottom plots indicate the energy spectrum in the realized fields as a function of wave number (C) and the spectrum for just the NCEP wind data product (D). The goal is to reproduce an approximate polynomial decay (red line) for the wind energy, matching the scaling law expected for turbulent flow. Due to issues of data the assimilation and resolution, the NCEP winds alone (D) lack the right amount of power for small scales and depart from this relationship.