GC43B-01 INVITED
Data assimilation over the last millennium using a large ensemble of simulations
A simulation is performed over the past millennium using a three-dimensional climate model of intermediate complexity that is forced to follow temperature histories obtained from a recent compilation of well-calibrated surface temperature proxies. This is achieved using a simple data assimilation technique that could be briefly described as follows. For each year, a large ensemble of simulation is performed (96 here). The member of the ensemble that is the closest to observations is then selected as representative for this particular year and used as the initial condition for the subsequent year. The distance between the model results and the proxy record is measured by a cost function using reconstructed and simulated temperatures at the locations where the proxies are available. The best simulation retained is the one that minimizes this cost function. The simulation obtained by this technique provides a continuous record over the past millennium that is compatible with model physics, with the forcing applied and with the available proxy-records. In a first step, the simulated temperatures over the last 150 years are compared to estimates obtained using only information from thermometers in order to demonstrate the validity of the techniques as well as to provide estimates of the errors bars on this new reconstruction. In a second step, the simulated temperatures over the whole millennium are compared at quasi- hemispheric-scales with the reconstructions based on statistical techniques. The simulated pattern of anomalies are then analyzed as well as the mechanisms responsible for the regional changes, with a particular focus over Europe and North America. The results are then compared to the ones of other data assimilation techniques used to study the climate of the last millennium. http://www.astr.ucl.ac.be/index.php?page=Wokshop_assim
GC43B-02
A new approach to global climate reconstructions of the last 1000 years
Efforts to reconstruct global gridded climate datasets from sparse paleoclimatic observations face special challenges. Extrapolations to data void regions based on observed present-day linear correlations (i.e. modern analogs) are too simple for this purpose, and are not strictly justified in a different and evolving climate. Modern data assimilation methods used for generating the "Reanalysis" datasets of the past half-century are also inappropriate for several reasons, and would require expensive re-tuning to assimilate the much sparser and temporally lower resolution paleoclimatic observations. Most importantly, such methods would make reconstructions of the key tropical SST fields problematic. In this talk, we will explore a different approach to the assimilation of paleoclimatic observations, in the standard framework of modern "3D-VAR" data assimilation methods but uniquely tailored to the assimilation of pre- industrial observations. The basic idea is to exploit important evidence from recent studies that global atmospheric climate changes (including those ultimately forced by radiative forcing changes) are strongly constrained by global SST changes, and the global response to global SST changes is well approximated by the linear response to tropical SST changes in a relatively low-dimensional dual space of tropical SST and global climate anomaly patterns. This relationship may be formally expressed as y = Gx, where x is the anomalous tropical SST field, y is the global climate response field, and G is a linear operator. We have found that G can be well approximated by retaining as few as 5 terms in its singular vector expansion series, representing contributions from optimal forcing/response singular vector pairs in decreasing order of importance. Our proposed 3D-Var approach to paleoclimatic data assimilation will amount to estimating a mean climate state for each 50-yr period in the last millennium that minimizes deviations from both a "first guess" mean state and the available proxy-based observations for that 50-yr period, taking into account the expected errors in each, while also maintaining the y=Gx relationship. The G operator appropriate to the pre-industrial climate could be estimated by forcing an atmospheric GCM coupled a mixed layer ocean with pre-industrial radiative forcings, and then determining its global response to prescribed localized SST perturbations on a regular tropical grid. The same GCM could also be used to provide a better "first guess" (as well as its expected error) for each 50-yr period than the modern climate as the first guess. Additionally, such a 3D-Var approach would account for errors associated with both the local and proxy nature of the paleoclimatic observations in a flexible and straightforward manner, and also incorporate "forward" multivariate models of those proxy observations that are not related solely to either local temperature or local precipitation. One could thus assimilate all available continental and marine paleoclimatic observations for the last 1000 years to produce a 1000-yr time series of global gridded 3- dimensional atmospheric and SST fields at 50-yr temporal resolution and a spatial resolution consistent with that of our reduced dimensional space. One could also provide error estimates of the reconstructions as a natural by- product of the analysis.
GC43B-03
Reconstructing Spatial Patterns of Climate Change During the Last Millennium: The Challenges of Method and Data
Understanding synoptic patterns of climate change during the last millennium is fundamentally tied to our ability to faithfully reconstruct such patterns from sparse data networks and regression-based climate field reconstruction (CFR) techniques. Here we demonstrate several challenges associated with CFR methods and the consequences for estimates of reconstructed spatial patterns. We focus on two widely applied CFR techniques that have been used to reconstruct temperature and hydrologic variables during the last millennium: regularized expectation maximization (RegEM) and canonical correlation analysis (CCA). These methods are tested using a pseudo-proxy framework that reflects real-world proxy observing sites and is derived from General Circulation Model simulations of the last millennium. We demonstrate that that the skill of reconstructions varies spatially and can be quite poor over important regions such as the equatorial Pacific Ocean. These regions of reduced skill appear most notably associated with the distribution of the proxy network. Widespread variance losses are also noted, resulting in significant underestimates of variability in many regions of the reconstructed field. Furthermore, the spectral fidelity of the reconstructed climatic fields is shown to have altered ratios of high (annual to decadal) and low (multi-decadal and lower) frequency variability. Collectively, these characteristics of CFRs should be a fundamental consideration of work that seeks to interpret reconstructed spatial patterns of climate change during the last millennium.
GC43B-04 INVITED
The Fingerprint of Persistence: The Remarkable Character of Medieval Climate Change over the Western US
The body of evidence for persistent aridity and recurrent severe drought across the western US during Medieval times (more or less 800-1300 AD) has grown from suggestive tree-ring chronologies developed during 1970s to a variety of convincing proxy records available today. These records come from tree-rings (moisture availability), lake level reconstructions (precipitation), sediment charcoal and fire scar records (wild fire), pollen and diatom assemblages (moisture availability, river discharge), soil mobilization (rainfall, wind) and archeological studies. Two recurring themes that arise from this collected work are that a) the instrumental climate record for the West is not representative of the past millennium, and b) Medieval climate in the West was marked more by persistent precipitation deficits than by remarkably dry years. Such change requires persistently altered circulation patterns which, in turn, imply important and long-lasting changes in boundary conditions. Recent proxy-model statistical studies and dynamical model experiments have attempted to reconstruct circulation changes and mechanisms that may have produced these precipitation changes. Not surprisingly, the results are consistent with the idea that Medieval climate change over the western US resulted from persistently cool sea surface temperatures in the eastern tropical Pacific. Notably, such cooling is supported by proxy records from the tropical and mid-latitude Pacific. While this congruence of proxy and model results is not conclusive, alternative candidates for producing such circulation changes have not yet been put forward and tested. The presentation will review the much of evidence discussed above and frame a general picture of current hypotheses.
GC43B-05
Spatial covariance of water isotope records in a global network of ice cores spanning 20th- Century climate change
Estimating the spatial extent of past climate changes has been an ongoing challenge for paleoclimatology. For such estimates to be made with confidence, it is important to establish an understanding of the spatial coherence of proxy records during an interval of known climate change. We use water stable isotopes from high- resolution ice cores and 20th-Century observations of sea-level pressures and sea surface temperatures to assess the covariance among isotopic records and its link to organized patterns of climate variability. Covarying signals in the cores are identified using empirical orthogonal function analysis. Results from regression analysis show that the leading signals are consistent with key climate patterns including the Northern Atlantic Oscillation and Southern Annular Mode, and variability in tropical Pacific sea surface temperatures associated with the El Nino-Southern Oscillation. Patterns that have recently been identified in instrumental data, such as positive tropical Pacific SST anomalies associated with the negative phase of the SAM, are evident in the ice cores. These explanations for the variance of stable isotopes are consistent with recent studies using isotope-enabled general circulation models, and provide a physical basis for interpreting the observed isotopic signals. While there is also a global change signal that is evident when analyzing the records collectively, there are some limitations in reconstructing global temperatures due to the geographic coverage of the available records and the current lack of modeling studies to explain the observed global-scale changes. Still, water stable isotope ratios preserved in ice cores provide a sufficiently rich sampling of large-scale climate variability that they can be more widely used in physically-based paleoclimate reconstructions covering the last millennium and other periods.
GC43B-06
The record and forcing of paleodrought in western North America: extrapolation from the western US to western Canada
Much of the classic paleodrought research is based on networks of tree-ring chronologies from the western US and in particular the southwest. This extensive work has established the timing, severity and causes of drought across the region over the past millennium. Even though some of this literature implies that the findings extend to western North America, they may or may not apply to western Canada which is beyond the influence of the North American monsoon. On the basis of this previous work, we hypothesize that in western Canada (i) severe drought is linked to negative phases of ENSO (La Nina) and the PDO, and (ii) the growing season of about three months limits the tree-ring proxy to a signal of summer climate. We test these hypotheses using data from our network of more than 60 tree-ring chronologies extending from northeastern Montana to the southern Northwest Territories across Canada's western Interior; a subhumid region where drought represents Canada's most damaging natural hazard. Our results do not support these hypotheses. At many sites the tree-rings capture a signal of spring and even winter precipitation. The spatial and temporal association between these seasonal moisture signals and large-scale climate forcings is quite complex and reveals that, in contrast to the US southwest, moisture deficits correlate in general with positive phases of the ENSO and PDO, and there is a strong non- stationary multidecadal hydroclimatic signal likely related to the AMO and/or PDO. These periodic low-frequency hydroclimate signals are also evident in the climate variability simulated by GCMs for this region. Our ongoing research is focused on the time-varying relationships between the climate forcings and drought, and in the changes in the pattern of teleconnections of drought in the western interior of Canada and ENSO between phases of PDO and AMO. http://www.parc.ca/urtreelab/
GC43B-07
Validating Historical and Future GCM Simulations of Climate Moisture Variability with Observed and Dendroclimatic Records
Validating the decadal, multi-decadal or lower frequency climate variability of GCM output is limited by the scarcity of long observational records. Tree-ring proxy data analyses carried out in western North America has proven valuable to quantify natural climate variation over centuries to millennia; therefore, providing a unique opportunity to validate GCMs. Many of the drought events during the 20th century have been linked to natural climate variability modes such as ENSO, PDO, PNA and AMO. Reconstructions of annual and seasonal climate moisture from tree-rings for the past 500 years for sites in Montana, Alberta, Saskatchewan and the NWT show drought events in previous centuries more extreme in magnitude, frequency and duration than recorded during the instrumental period. These drought events may be linked to the different natural climate forcings. The key objectives of this work are to: 1) determine the relationships between observed data and natural climate variability modes; 2) identify the climate responses (i.e. precipitation and temperature) and influences of the large-scale climate forcings in tree growth as derived from the tree-ring analysis; 3) compare and validate climate variability modes identified in GCM control runs to the observed and proxy climate variability; and, 4) derive a composite time series of future climate moisture variability for the study area.