I’m interested in understanding how climate behaves on decadal to centennial timescales. My research focuses on paleoclimate, in particular the past 2,000 years (a.k.a the Common Era). I use paleoclimate reconstruction to understand the temporal evolution of temperature, and spatial patterns of the temperature field. The former puts the recent global warming in a broader context, and the latter lends insight into the understanding of dynamics of the climate system.
Instrumental temperature records are only available for the past 150 years or so. In order to get a longer history of temperature, we turn to proxies and look at how they express climate signals. We treat proxies as observations of past climate, and with the help of statistical tools we extract climate information of the past. This process is called reconstruction, and has been widely used in geosciences across all timescales.
The crediblity of paleoclimate reconstructions largely depends on two aspects: proxy data being used and statistical methods being employed. How reliable is a climate reconstruction? To what extent can we trust the reconstructed climate? The answers are not simple. In fact, the shortness of instrumental data and scarcity of proxy data make it extremely challenging to assess how well a reconstruction can be. In my research, I use pseudoproxy experiments to isolate the methodological impact, and conduct proxy quality control experiment to evaluate the data impact.
Data challenges: In a multi-proxy setting, different types of proxies are assembled in the same dataset. They differ in length, resolution, the degree of their sensitivity to temper- ature and they may respond to temperature of different seasons. Furthermore, the sparsity of proxy network adds an additional level of complexity to reconstructing past climate based on multi-proxies. All these reveal the challenge of conducting multi-proxy reconstructions, and the difficulty of isolating individual factors that contribute to reconstruction uncertainty.
Statistical challenge: From a statistical point of view, deriving CFRs is challenging because of the data’s high dimensionality: yearly instrumental temperature observations (T) are only available for the past ∼150 years, and the number of proxies (P) is on the order of several hundreds to thousands. In other words, there are very few degrees of freedom available to estimate the variables. Statistically speaking, it is a rank-deficiency problem, where covariance matrix (which quantifies relationships between proxies and temperature) is singular (non-invertible), and traditional regression methods are no longer applicable.
We show that the choice of staistical method for reconstruction can drastically change reconstruction results. In particular, a reconstructed La Niña-like pattern during the transition from the Medieval Climate Anomaly (MCA) to the Little Ice Age (LIA) has been widely tied to medieval droughts in southwest North America. This pattern is now used as a key benchmark for global climate models (GCMs), which have yet to reproduce it. I tested the pattern's robustness by using four different CFR methods and two proxy networks. With the older network, the reconstructed patterns is highly method-dependent, with the La Niña-like pattern not reproduced by two of the CFR methodologies. With an updated proxy network (PAGES2k), a globally uniform MCA emerges with all methods, in agreement with simulations from the Paleoclimate Model Intercomparison Project Phase 3 (PMIP3) ensemble.
Wang, J., Emile-Geay, J., Guillot, D., Smerdon, J.E.: Evaluating climate field reconstruction techniques using improved emulations of real-world conditions. Clim. Past, 10, 1-19, 2014
Wang, J., Emile-Geay, J., Guillot, D., Mckay, N. P., Rajaratnam, B.: Fragility of reconstructed temperature patterns over the Common Era. Implications for model evaluation., in revision, Geophys. Res. Lett.
Wang, J., Emile-Geay, Mckay, N. P., J., Guillot, D., Vaccaro, A.D.: The climate continuum in a global multi-proxy temperature reconstruction spanning the Common Era, in prep
4th Asia 2k Workshop, Kyoto, Japan, Talk, March 2015
AGU Fall Meeting, San Francisco, CA, Talk, December 2014
4th International Workshop on Climate Informatics, Boulder, CO, Poster, Sept 2014
COAA-SCC 2014 Annual Workshop, La Jolla, CA, Talk, Sept 2014
3rd Asia 2k Workshop, Beijing, China, Talk, May 2014
EGU General Assembly, Vienna, Austria, Talk, April 2014
Advances in CFR Workshop, Woods Hole, MA, Talk, April 2014
AGU Fall Meeting, San Francisco, CA, Talk, December 2013
7th Graduate Climate Conference, Woods Hole, MA, Poster, November 2013
10th Urbino Summer School in Paleoclimatology, Urbino, Italy, Poster, July 2013
12th International Meeting of Statistical Climatology, Jeju, Korea, Talk, June 2013
PAGES 4th OSM and 2nd YSM, Goa, India, Poster, February 2013
AGU Fall Meeting, San Francisco, CA, Poster, December 2012
AGU Fall Meeting, San Francisco, CA, Poster, December 2011