The Weather and Climate Impact Assessment Science Initiative
    Home Current Projects Climate Variability in Past Centuries Stationarity in the Climate System
  About the Program
Overview
People
Contact Information
Program Documents
2004 Review Documents
  Research
Current Projects
Publications
Presentations
  Links of Interest
Supporting Institutions
Other NCAR Initiatives
  For Program Staff
Upcoming Events
Mailing List
Website Statistics Initiative Staff Only
  Search This Website

...more search features  
Stationarity in the Climate System
 

Real world and simulated proxy series: Teleconnection fidelity

Proxy climate reconstructions are generally based on a regression relationship between the proxy and particular climate variable(s) over a calibration period. Subsequently, it is a fundamental assumption in paleoclimate research that this relationship remains constant (and often linear) in time. This issue is not limited to individual proxy representation of local climates, but the correlation approach is often also used for interpreting dynamical links to large-scale climate circulation. This is particularly the case since the importance of large-scale climate modes has been recognized. Potentially, these so called teleconnections provide one of the strongest tools in paleoclimate research as climatic changes are thought to project onto the major modes of climate variability and an integrated signal with a particular spatial component could help improve the signal to noise ratio. So far, little research has dealt with verifying the fundamental assumption of stationarity, both for local climate as well as for teleconnetion fidelity. Here, it is tested if within a coupled model framework the stationarity assumption for the relationship between known proxy locations and the tropical El Nino/Southern Oscillation system uniformly holds or if the 150-year instrumental record is too short for a calibration. In an expansion, the strength of combined multi-proxy networks is evaluated if the common signal in different locations performs better than the individual local records alone.

 
Click on image to enlarge
(Click on image to enlarge)
Conceptual layout of stationarity test: simulated proxy series and their relationship to NINO3 over the full Millennium.
 

Output from fully forced transient simulations with the PaleoCSM of last 1000 years was used to create simulated climate proxies for key global climate variables and modes. These simulated proxies (see below for details) can be directly compared to the "real" (within model) values of the variables and modes over the entire millennium. The model framework offers an opportunity to test the stationarity of these relationships, something that cannot be reproduced in the real world where the instrumental record is very short.

This comparison extends the period over which proxy climate reconstructions can be tested for validity by a factor of approximately six times. This extension is crucial, because in the real world it is not possible to determine how reliable modern calibrations of the proxy/climate relationship are over centuries-to-millennial length periods. To my knowledge, his work is the first use of transient-forced long climate runs to examine proxy-reconstruction uncertainty issues of this kind. Previous work has used control runs, and has focused on the number of proxy sites needed to reconstruct N. Hemisphere and global average surface temperatures.

Work in this subcomponent has been directed to understanding proxy reconstructions of the ENSO climate mode, which are understood less than global and hemispheric climate reconstructions. The PaleoCSM has shown good capability to reasonably represent ENSO behavior, both in the ocean component and in surface teleconnections, which makes it an appropriate experimental vehicle for this work.

 
Click on image to enlarge
(Click on image to enlarge)
Long-Term Verification Stability of Proxy-based Reconstructions of ENSO: 1000-year Simulations of ENSO Teleconnections and Tree Ring-based Precipitation Reconstructions in Central Chile.
Click on image to enlarge
(Click on image to enlarge)
 
 

Results show that reconstruction reliability over long periods of time (e.g., long-term significance of statistics that compare real NINO3 SST's with proxy-based reconstructions of NINO3) is more deeply affected than previously recognized by the underlying tightness of the proxy/ENSO-teleconnected climate variable relationship (e.g. tree ring width vs. precipitation in a teleconnected region). This underappreciated need for robustness is necessary so that as teleconnection strength waxes and wanes over long periods, the noise added to the reconstruction process by the proxy/climate relationship is not enough to "pitch" the NINO3 reconstructions into insignificance. A related, and highly important, result is that it is possible to have statistically significant calibration and verification of NINO3 reconstructions based on relatively poor-quality proxy data (e.g., a relatively low explained variance in the tree ring/precipitation relationship) during the period 1850-2000, which then fail to pass significance tests over much of the full 1000 year period. This result indicates that modern-period validations of reconstructions based on relatively poor-quality proxies can give a strongly false sense of security about the likely long-term reliability of these reconstructions. This work was done using Monte Carlo replication to be able to examine mean behavior as well as that of individual realizations.

 
Click on image to enlarge
(Click on image to enlarge)
Central Chile tree-ring reconstruction simulated in climate model by applying noise to precipitation in corresponding grid boxes along Chilean coast. Comparison of 'gold-standard' with 'noise-added' series as well as running correlation over the millennium with NINO3 series. Highly significant variance is explained in 20th century, but prior to this 'calibration' period, often unstable correlation into insignificant relation. (Details: See Wahl and Ammann (in prep).
 

These outcomes are indicated robustly by the results, and have generated significant excitement among paleo-reconstruction experts. This kind of experimentation provides a much more scientifically rational procedure for determining proxy reliability and locations for likely successful reconstruction than current methods, which are frequently subjective and heuristic.

Team/Collaborators: E. Wahl, C. Ammann (NCAR), N. Graham (Scripps and HRC), D. Nychka (NCAR), M.E. Mann (University of Virginia)

 

Characterization and optimal reduction of uncertainty in climate and ecosystem reconstructions

Receiver Operating Characteristics analytical methodology is used to systematically explore the joint minimization of false positive and false negative identifications of modern analogs for fossil assemblages (pollen, forams, etc.). For the first time, this procedure offers a rational basis for accepting more of one kind of error to reduce the other in specific situations. These analyses are being recognized in the pollen-based paleoecological community as contributing one of the most important steps forward in application of the modern analog technique (MAT) in the last 20 years.

 
Click on image to enlarge
(Click on image to enlarge)
Modern Pollen Calibration Samples from the Southern California Mountains: Discrimination of like-Vegetation vs. unlike-Vegetation Samples using Receiver Operating Characteristic Analysis.
Click on image to enlarge
(Click on image to enlarge)
 
 

Development of Receiver Operating Characteristic (ROC) and Monte Carlo replication methods for characterization and optimal reduction of uncertainty in microfossil-based climate and ecosystem reconstructions (based on pollen, foraminifera, diatoms, etc.)

These analyses are being recognized in the pollen-based paleoecological community as contributing one of the most important steps forward in application of the modern analog technique (MAT) in the last 20 years. This step forward uses the ROC analytical methodology to systematically explore the joint minimization of false positive and false negative identifications of modern analogs for fossil assemblages of pollen, forams, etc.; and it gives a rational basis for accepting more of one kind of error to reduce the other in specific situations. This work is the first of its kind in this field, and utilizes existing theory and practice from signal testing, medical testing, and weather forecast evaluation as its foundations. We have also used Monte Carlo techniques to examine the interactions of optimum analog selection, analyst effort given to counting microfossils (sample size), and reconstruction precision. This work has identified unrecognized costs of lost information by hugely increasing false negative analog misidentifications in order to sharply decrease false positives. Use of optimizing analog selection thresholds (i.e., a proper balance between low false positives and low false negatives) allows analyst effort given to sample counting to be reduced by nearly an order of magnitude, with no loss of precision!

Team/Collaborators: E. Wahl (NCAR)

 

Publications

Mann, M.E., C.M. Ammann, R.S. Bradley, K. Briffa, T.J. Crowley, M. Huges, P.D. Jones, M. Oppenheimer, T. Osborn, J.T. Overpeck, S. Rutherford, K.E. Trenberth, and T.M.L. Wigley, 2003: On past temperatures and anomalous late 20th century warmth. EOS, 84(27), 256ff.

Naveau P., Jomelli V., Cooley D. and Rabatel A. Extreme value theory in lichenometry and paleoclimate. (Submitted to Quaternary Research.)

Wahl, Eugene, "Assigning Climate Values to Pollen Surface Sample Sites and Validating Modern Analog Climate Reconstructions in the Southern California Region", Madroño, 50:4, (in press).

Wahl, Eugene: "A General Framework for Determining Cutoff Values to Select Pollen Analogs with Dissimilarity Metrics in the Modern Analog Technique", Review of Palaeobotany and Palynology, (in press).

Wahl, E.R. and Ammann, C.: "Stationarity and Fidelity of Simulated El Niño-Southern Oscillation Climate Proxies over the Last Millenium in Forced Transient AOGCM Output". (In preparation, to be submitted to Climate Dynamics).

  Initiative Staff Only Denotes Initiative Staff Only ©2007 UCAR   |   Privacy Policy   |   Terms of Use   |   Top of Page  
NCAR Weather and Climate Impact Assessment Science Initiative