LONG-TERM CLIMATE CHANGE/VARIABILITY
Anthropogenic Change GCM Experiments
My involvement in the area of long-term climate change began with my arrival at GFDL (since this is a topic of great interest here). I began with a project aimed at examining the response of the stationary waves and zonally averaged zonal component of the wind in GFDL GCM’s and comparing with atmospheric observations. Due to other demands, this work was never submitted for publication, although some of the results are presented in Lanzante (1992, 1993).
After this, a collaboration with my GFDL colleague Keith Dixon examined the output from an ensemble of low resolution CO2 + aerosols GCM experiments. Dixon and Lanzante (1999) examines the effects of the historical (radiative) starting time (i.e. how far back in time you go to begin) vs. the influence of the initial conditions for the behavior in the 21st century. We found that the two have comparable influence. The implications are that it may not be necessary to go back too far in time for starting coupled model simulations, however, the use of an ensemble of experiments would be wise in order to reduce the impact of the differing (random) initial conditions.
Another collaborative project (Shine et al. 2003) involved a comparison of stratospheric temperature trends from a wide variety of models. Our LKS radiosonde temperature dataset (Lanzante et al. 2003a; Lanzante et al. 2003b) was used as one of the observational references. The results indicate that while the models share similar patterns in the vertical structure of trends, the magnitudes of the trends vary considerably.
More recent work of mine (Lanzante 2007) was aimed at comparing the vertical temperature structure of trends in GFDL GCM climate change experiments with our homogenized radiosonde temperature dataset (Lanzante et al. 2003a; Lanzante et al. 2003b). In particular, this work sought to assess how much the adjustments made to our radiosonde data, intended to make them more temporally homogeneous, influence their agreement with GCM simulations. The adjustments were found to be important, almost universally increasing the agreement between models results and observations. This study also found that the impacts of climatically important volcanic eruptions can have an influence on the assessments.
A subsequent collaboration (Lanzante and Free 2008) with Melissa Free of NOAA’s Air Resources Laboratory (ARL), served as an extension of my earlier project and incorporates simulations from additional models from the IPCC AR4 archive.This project was aimed at comparing trends from several GCMs with those from both an updated versionof our homogenized radiosonde temperature dataset as well as those from a competing product produced by the British Met. Office. We found that systematic biases in the observed data have masked some of the greenhouse warming signal. Removal of some of these biases almost universally improves the agreement with climate models. As such this may have implications for detection and attribution studies. Using the same observed and model datasets we also examined volcanic temperature signals in the free atmosphere from both a short-term (several years after the eruption) and long-term (trend) perspective.(Free and Lanzante 2009).
Atmospheric Water Vapor
The relationship between temperature and water vapor has important implications for long-term climate change due to the positive feedback of water vapor in global warming. Some earlier studies at GFDL (Sun and Oort 1995; Sun and Held 1996) suggested that the relationship between temperature and water vapor was much stronger in the GFDL GCM than in the real atmosphere. I was involved in a later collaboration, with GISS researchers Mike Bauer and Tony Del Genio, which reexamined the conclusions of the earlier studies at GFDL (Bauer et al. 2002). We found that re-doing the calculations with a more consistent treatment of the GCM and observed data has a major impact, such that the difference between the model and real atmosphere is not nearly as large as was earlier implied.
An earlier collaboration with my former GFDL colleague Brian Soden examined variations in atmospheric water vapor using satellite and radiosonde measurements. This effort (Soden and Lanzante 1996) was aimed at comparing the climatologies of Upper Tropospheric Humidity (UTH) from these two sources.
Some years ago a collaborative project (Gaffen, Sargent, Habermann, and Lanzante, 2000) was completed, yielding new insights into the influence of instrumental (i.e. artificial) changes on the record of long-term climate variability derived from radiosonde temperature data. This effort was led by Dian Seidel (Gaffen) of NOAA’s Air Resources Laboratory (ARL), located in Silver Spring, MD. Dian’s earlier work on constructing a comprehensive data base of radiosonde station history information (metadata) proved quite valuable to us. Her radiosonde metadata is available from the ARL website. A more recent compilation of such metadata (station histories), which incorporates Dian’s along with metadata from other sources is now available from NOAA’s National Climatic Data Center.
One of the goals of this project was to use Dian’s metadata in conjunction with a statistical method developed by Ted Habermann as well as one which I developed [Lanzante (1996) Int. J. Climatolog. manuscript] in order to identify artificial changes (discontinuities) in long time series of upper-air temperature. Discontinuities can occur when a new temperature sensor is introduced or when new recording procedures are implemented. These discontinuities are potentially severe enough to corrupt estimates of long term climate variations, especially in the stratosphere and upper troposphere.
An example of the discontinuities that can occur in a long record of upper-air data is given by the time series of 700 hPa geopotential height anomaly at Veracruz, Mexico. The dots indicate points of discontinuity identified using the method of Lanzante (1996) and the horizontal lines are the means over each segment defined by these points. Another example is found in the time series of 100 hPa temperature at Valentia, Ireland.) In this case the metadata indicate that all except the second point of discontinuity occur near times corresponding to major changes in the observing system.
While the Gaffen et. al (2000) project narrowed the uncertainties, we were not able to provide any remedies. Furthermore, a subsequent comparison (Free et al. 2002) of the approaches taken by several different groups demonstrates that these competing methods yield very different results, with no indication of which, if any, are on the right track. Eventually we were able to make some progress both in diagnosing the biases in the radiosonde temperature data as well as creating an improved dataset. We utilized a limited network (87) of radiosonde temperature stations with relatively long periods of record. We employed a variety of tools to help us identify the major discontinuities, applied adjustments, and then made improved estimates of temperature trends in the free atmosphere. We have also compared our adjusted temperature data set with MSU satellite temperatures and demonstrated that our adjustments enhance data quality. Two manuscripts (Lanzante et al. 2003a; Lanzante et al. 2003b) report on our findings.
Because the dataset that we created terminates in 1997, and the method to produce it is quite laborious, we expanded our team and created a new dataset that incorporates our earlier one, but is much easier to update in real time. This new product, RATPAC (Free et al. 2005) is available online as an operational climate monitoring product distributed by NOAA’s National Climatic Data Center. However, in spite of our best efforts there is compelling evidence that we have not eliminated all of the systematic biases from our dataset (nor have others with competing datasets). Several subsequent studies have made this point, including one I was involved in, led by Steve Sherwood of the University of New South Wales, Australia (Sherwood et al. 2005)
From 2004-2006 I served as a Convening Lead Author in the preparation of a synthesis and assessment report (SAP1.1) for the U.S. Climate Change Science Program (CCSP). My chapter presents the observed temperature trends. The entire report details current knowledge of temperature trends in the lower atmosphere and their correspondence with similar trends computed from a large collection of climate models. Preparation of this report was a very demanding exercise that occupied the majority of my time for a period of more than two years. It included an organizing workshop at NCDC in October 2003, another workshop hosted by the British Meteorological Office in September 2004, and six meetings of the authors in Chicago. The result was a comprehensive report on the current state of knowledge as well as recommendations for future activities to serve as guidance for the international community. A major conclusion of this report was that a previously reported apparent discrepancy in the vertical structure of long-term temperature change between observations and climate models no longer exists at the global scale, although some discrepancies may exist in the tropics. I have created a brief summary of the current state of knowledge on this topic.
A related study that was used as input to the CCSP SAP1.1 report (Santer et al. 2005),led by Ben Santer of the Lawrence Livermore National Laboratory compared the vertical structure of warming in the tropical atmosphere in climate models with that from observations. It concluded that an inconsistency between results for short and long timescales suggests that there are residual biases in the observed data, since no such inconsistency exists in the model results. A follow-up study (Santer et al. 2008), again led by Ben revisited Santer et al. (2005) using additional newly created observed datasets and was able to push the conclusions of the CCSP SAP1.1 report further. It states that the previously reported apparent discrepancy between observations and climate models regarding the vertical structure of long-term temperature change no longer exists in the tropics, thus vanquishing the long-standing controversy.
Work with much of the same team has continued by comparing atmospheric temperature trends from satellite observations with climate model output. In one study, using model output from the CMIP3 archive, Santer et al. (20011), examined the time-scale dependence of the consistency in trends between models and observations, and found better agreement as the timescale (i.e. length of the period used for computing trends) increases. In another, Santer et al. (2012), a mulitmodel detection/attribution study using CMIP5 models, identified the signal of anthropogenic climate change in both the troposphere and the lower stratosphere.
In some unrelated work (Seidel and Lanzante 2004) Dian Seidel and I found that some of the long-term changes in atmospheric temperature could be characterized as having occurred nonlinearly in the form of a step change. However, there is sufficient ambiguity that a linear change is also plausible.
In the spirit of historical review, several of us (mostly CCSP SAP1.1 report authors) have collaborated on 2 manuscripts (Thorne et al. 2011; Seidel et al. 2011) for a new Wiley publication featuring review articles on climate change topics for an interdisciplinary audience. In the first of these manuscripts we provide a comprehensive history of temperature trends in the troposphere, in observations and climate models. A motivation and focal point is the controversy that erupted 20 years ago regarding the different rates of warming at the surface and in the troposphere, and it’s resolution during the last few years. The second manuscript is a companion paper on trends in the stratosphere.
Statistical downscaling will be the main focus of my research for the foreseeable future. Although these efforts relate primarily to “Climate Change”, I have a sepatate page that covers research in this area: Statistical Downscaling Page. Unfortunately one ongoing project with my GFDL colleague Keith Dixon has gotten “back-burnered”. It is aimed at demonstrating to a less technical audience how some of the manifestations of climate change vary by spatial scale. This involves examining temperature time series from climate models averaged over different domains. The basic idea is that for more localized regions climate change is not as readily obvious for some time into the future. By contrast, when temperature is averaged over large regions climate change is noticeable much sooner. We are hoping this project can be revived in the future.
Return to John Lanzante’s Home Page