Scrutinizing Forced and Unforced Variability in CMIP5 · 2016-02-17 · Scrutinizing Forced and...
1
www.PosterPresentations.com Accuracy of Model Projections During the Hiatus Period Results Scrutinizing Forced and Unforced Variability in CMIP5 Timothy DelSole, Xiaoqin Yan, Emerson LaJoie, and Laurie Trenary George Mason University and Center for Ocean-Land-Atmosphere Studies, Fairfax, VA 1960 1970 1980 1990 2000 2010 −5 0 5 10 20 CCCma.CanESM2 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 CCCma.CanESM2 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 20 IPSL.IPSL−CM5A−LR −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 IPSL.IPSL−CM5A−LR 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 20 MPI−M.MPI−ESM−LR −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 MPI−M.MPI−ESM−LR 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 20 25 MPI−M.MPI−ESM−MR −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 MPI−M.MPI−ESM−MR 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 NOAA−GFDL.GFDL−ESM2G −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 NOAA−GFDL.GFDL−ESM2G 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 20 25 NCAR.CCSM4 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 NCAR.CCSM4 1960 1970 1980 1990 2000 2010 −5 0 5 10 MRI.MRI−CGCM3 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 MRI.MRI−CGCM3 1960 1970 1980 1990 2000 2010 −6 −4 −2 0 2 4 MIROC.MIROC−ESM −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 MIROC.MIROC−ESM 1960 1970 1980 1990 2000 2010 −5 0 5 10 CSIRO−QCCCE.CSIRO−Mk3−6−0 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 CSIRO−QCCCE.CSIRO−Mk3−6−0 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 CNRM−CERFACS.CNRM−CM5 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 CNRM−CERFACS.CNRM−CM5 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 20 MIROC.MIROC5 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 MIROC.MIROC5 1960 1970 1980 1990 2000 2010 −6 −4 −2 0 2 4 MOHC.HadGEM2−ES −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 MOHC.HadGEM2−ES 1960 1970 1980 1990 2000 2010 −5 0 5 10 15 20 NCC.NorESM1−M −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 NCC.NorESM1−M 1960 1970 1980 1990 2000 2010 −8 −6 −4 −2 0 2 4 NOAA−GFDL.GFDL−CM3 −0.1 −0.07 −0.03 0.01 0.04 0.07 0.1 NOAA−GFDL.GFDL−CM3 Data: Surface air temperature (‘tas’) of the HadCRU4 dataset and pre-industrial control/ historical/rcp45 simulations of 14 models from CMIP5 project are interpolated onto common grid of 5°x5°. Mask is constructed based on data availability of HadCRU4. Hypothesis: ΔO = F Δa + Δu (10-yr mean obs) (Response pattern) (amplitude) (internal variability) Detection: Attribution: Δ denotes differences relative to climatology of 1961-1990. E and α are ensemble size and 5% significance level. denotes covariance matrix of internal variability. Comments: 1) The forced response pattern of each model was obtained by applying discriminate analysis (Jia and DelSole, 2012) based on 12-eof truncations to each model’s 10-year running mean ’tas’ of the 500-year pre-industrial control run and the historical runs. 2) Detection/attribution analysis was applied to tas anomalies of both HadCRU4 and each model relative to the 1961-1990 climatology. RCP4.5 used to extend historical runs. 3) The confidence interval accounts for the fact that the climatology of 1961-1990 has 1/3 the variance of a 10-year mean. 4) Years shown on the figure of curves are the center year of corresponding 10-year mean (e.g.,1995 for mean of 1991-2000). 5) Consistency is achieved if observations are within the 95% confidence interval (red shading band). Conclusions: 8 model simulations (blue) are inconsistent with HadCRU4 observations (black), 6 model simulations (blue) are consistent with HadCRU4 observations (black). The dominant forced response patterns differ significantly among some models. References: Jia, L. and T. DelSole, 2012: Optimal determination of time-varying climate change signals. J. Climate, 25, 7122–7137. 1650 1700 1750 1800 1850 1900 1950 2000 −4 −2 0 2 amplitude NOAA−GFDL.GFDL−CM3 Max NtC Discriminant for 30 EOFs: Ratio= 2.71 1650 1700 1750 1800 1850 1900 1950 2000 −6 −2 0 2 4 amplitude MIROC.MIROC5 Max NtC Discriminant for 30 EOFs: Ratio= 8.26 1650 1700 1750 1800 1850 1900 1950 −2 0 2 4 amplitude MRI.MRI−CGCM3 Min NtC Discriminant for 30 EOFs: Ratio= 0.351 1700 1750 1800 1850 1900 1950 2000 −4 −2 0 2 4 amplitude IPSL.IPSL−CM5A−LR Min NtC Discriminant for 30 EOFs: Ratio= 0.283 1650 1700 1750 1800 1850 1900 1950 2000 −4 −2 0 2 4 amplitude MIROC.MIROC−ESM Min NtC Discriminant for 30 EOFs: Ratio= 0.345 1650 1700 1750 1800 1850 1900 1950 2000 −4 −2 0 2 4 amplitude −CERFACS.CNRM−CM5 Min NtC Discriminant for 30 EOFs: Ratio= 0.323 Maximized Minimized GFDL-CM3 MIROC-5 MRI-CGCM3 MIROC-ESM IPSL-CM5A-LR CNRM-CM5 Changes in Internal Variability Due to 20 th Century Climate Changes This study investigates the possibility that internal variability responds to anthropogenic forcing on annual time scales. Current detection and attribution methods assume that internal variability does NOT change; our study tests if this assumption is valid. Previous studies assessed this question using univariate metrics such as the “residual consistency check” (Allen and Tett,1999). Unfortunately, this metric tests consistency in an aggregate sense, based on a suitably weighted total sum variability, and as such it allows errors in some components to be compensated by errors in other components, giving a potentially misleading impression of overall consistency. We examine consistency using more powerful discriminant analysis (DA) techniques, which effectively tests equality of the entire covariance matrix of variability. Log of Minimum and Maximum Noise−to−Control Ratio, Detrended CTR, and 30S to 30N 0 0.5 1 1.5 2 −−2 −1.5 −1 −0.5 Red = Raw Data Green = Detrended Blue = 30S to 50N GFDL-ESM GFDL-CM3 NCC-NorESM1 NCAR.CCSM4 MRI-CGCM3 MPI-ESM-MR MPI-ESM-LR HadGEM2-ES MIROC-ESM MIROC-5 IPSL-CM5A-LR CSIRO-Mk3.6.0 CNRM-CM5 Can-ESM2 We find that CNRM, IPSL-LR, MIROC-ESM, and MRI experienced a decrease in internal variability, while MIROC5 and GFDL-CM3 experienced increases in internal variability. The time series for the significant components are shown in fig. 2, while the corresponding spatial patterns are shown in fig. 3. These results suggest the following conclusions: • GFDL-CM3 experiences an increase in internal variability that is widespread throughout the globe. • The decrease in 20 th century internal variability tends to be concentrated in regions of sea ice. We conjecture that 20 th century losses in sea ice reduce annual mean variability because, without sea ice, the overlying atmosphere is exposed to the ocean’s surface for longer durations of the year and the ocean’s surface temperature would have a moderating effect on the near-surface temperature of the atmosphere. • MRI and IPSL-LR possess a low frequency oscillation in their respective control runs that does not persist in the historical period. • After masking out the polar regions outside 30S and 50N (see blue bars in fig. 1), we find that only MIROC5 yields a significant change in variability. However, the change in MIROC5 appears to be an artifact of a change in mean in the historical runs. • Some models breach the significance level only marginally, so we hesitate to identify them as exhibiting changes in internal variability because of the subjective nature of truncating EOFs. Our results show that some models exhibit a change in internal variability of annual mean near surface temperature in response to human influences. Much of this response is in polar regions and a plausible consequence of the loss of sea ice. In addition, changes in internal variability impact the assumptions underlying detection and attribution studies. We plan to continue our analysis to better address the sensitivity of our results to the choice of EOF truncation and the problem of overfitting when applying discriminant analysis. We are also working on applying our analysis to the 21 st century to determine if changes in internal variability continue in the future. Figure 1. Log of noise-to-control ratios for three case studies; before detrending the control run (red), after detrending the control run (green), and masking out the poles to analyze only the domain from 30S to 50N (blue). Black lines indicate significance at the 1% level. We find that most of the differences in internal availability happen in polar regions. Figure 2. Variates for the optimized noise-to-control ratios. Control period is in blue. One ensemble member from the historical period is in red. For the maximized cases notice the variance appears to increase in the 20th century. Conversely, in the minimized cases the variance decreases. 0 50 100 150 200 250 300 350 −50 0 50 −1.0 −0.5 1.0 NOAA−GFDL.GFDL−CM3 Max NtC Pattern for 30 EOFs 0 50 100 150 200 250 300 350 −0 50 −3 −2 −1 0 1 MIROC.MIROC5 Max NtC Pattern for 30 EOFs 0 50 100 150 200 250 300 350 −50 0 50 −−−−0.1 0.0 0.1 MRI.MRI−CGCM3 Min NtC Pattern for 30 EOFs 0 50 100 150 200 250 300 350 −0 50 −0.6 −0.4 −0.2 0.0 IPSL.IPSL−CM5A−LR Min NtC Pattern for 30 EOFs 0 50 100 150 200 250 300 350 −50 0 50 −0.2 −0.1 0.0 0.1 MIROC.MIROC−ESM Min NtC Pattern for 30 EOFs 0 50 100 150 200 250 300 350 −50 0 50 −0.2 0.0 0.2 0.4 0.6 CNRM−CERFACS.CNRM−CM5 Min NtC Pattern for 30 EOFs Minimized Maximized GFDL-CM3 MIROC-5 MRI-CGCM3 MIROC-ESM IPSL-CM5A-LR CNRM-CM5 Figure 3. The spatial pattern that corresponds to the variates in Fig. 2. We apply DA to annual mean, near- surface (2m) temperature from 14 CMIP5 climate models that simulated at least 500 pre-industrial control years and had a 3-member historical ensemble. The data was interpolated onto a common 5x5 grid and climatology was removed. DA finds the linear combination of variables that maximize (and minimize) the ratio of variances. The resulting ratio is called the noise-to-control ratio and given by Internal variability in the 20 th century is estimated by subtracting out the ensemble mean. Optimization is performed on the leading 30 EOFs. The upper panel in Fig. 1 shows the log of the maximized ratios for three separate applications of DA and the horizontal black line provides the 99% confidence level. In this figure, a significant maximized ratio means that an INCREASE in 20 th century internal variability was detected. Similarly, the lower panel shows the results for the log of the minimized ratios and the horizontal black line there indicates the 1% confidence level. Bars that stay between these two black lines indicate no changes in 20 th century internal variability are detected. We found that 2 models had significant maximized ratios and 4 models had significant minimized ratios. However, we noted trends in the control runs, so, we detrended all the control runs and repeated our analysis. The green bars indicate detrended data; detrending substantially reduced the change in variability for two models. Relation Between AMO and AMOC MOTIVATION: Research suggests that the Atlantic Multi-decadal Oscillation (AMO) is the most predictable mode of temperature variability. This predictability is often attributed to variations in the Atlantic Meridional Overturning Circulation (AMOC). Here we ask the following questions: 1. Is there a relation between the AMOC and AMO? 2. Is the relation consistent across climate models? 3. Is the maximum streamfunction the best variable for assessing the AMOC-AMO relation? Methodology Results Future Work DATA: Analysis is performed on 450 years of annual anomalies of sea surface temperature (SST) and Atlantic mass overturning circulation (AMOC) data from the following pre-industrial control runs: An index for the Atlantic Multi-decadal Oscillation (AMO) is defined as the area- weighted annual SSTA over the north Atlantic from 0-60 o N. The maximum strength of the AMOC is defined as the maximum in the AMOC between 30 o N-60 o N, and denoted as Ψ max . Control runs are split into two parts: a training part (225 years) for building the model and a verification part (225 years) for testing the model. OPTIMIZATION: The relation between the AMO and AMOC is diagnosed by finding the linear combination of AMOC principal components (PCs) that are most highly correlated with AMO in an integral sense (Jia and Del- Sole, 2011). This integral is called Average Predictability Time (APT): 1. AMOC MOST RELATED TO AMO 2. MODEL COMPARISION 3. AMOC and Ψ max • Only 3 APT are found to be significant across all models • No robust relation was found on decadal-to- multidecadal time scales Figure 1: The leading three components of the AMOC that are most related to the AMO, in a multi-model sense. Structures are for the training period. Differences in cross model AMOC-AMO relation are quantified by fitting an empirical model: Αn F-test is used to test equality of all models Figure 2: Diagram showing the clustering of the empirical models for 5 CMIP5 climate models. Clustering is based on F-Statistic for pairwise comparison of empirical models. Lines connecting models denote statistically equivalent models at α=0.05. The distance between unconnected models is proportional to F. • AMOC-AMO relations are statistically indistinguishable for MRI/MPI/NCC • CCC and NCAR are significantly different from all other models except MRI and NCC • APT patterns have larger correlation with AMO compared to Ψ max • Regressing out APT from Ψ max removes significant correlations with AMO Figure 3: Squared correlation for 3 leading APT-AMO (black), ψmax-AMO (red), and residual ψmax -AMO (blue). Residual ψmax is recovered after the 3 leading APT has been regressed out of ψmax. Negative lags indicate APT and ψmax lead AMO. • Extend optimization to include sea level pressure over the North Atlantic • Construct simple dynamical model to describe AMO-AMOC mechanism • Attribute differences in AMO variability to differences in regression model and AMOC variability Summary
Transcript of Scrutinizing Forced and Unforced Variability in CMIP5 · 2016-02-17 · Scrutinizing Forced and...
TEMPLATE DESIGN © 2008
www.PosterPresentations.com
Accuracy of Model Projections During the Hiatus Period
Results
Scrutinizing Forced and Unforced Variability in CMIP5 Timothy DelSole, Xiaoqin Yan, Emerson LaJoie, and Laurie Trenary
George Mason University and Center for Ocean-Land-Atmosphere Studies, Fairfax, VA
1960 1970 1980 1990 2000 2010
−50
510
20
●●
●● ●
●● ● ●
●
●●
●
●●
●
●●
● ● ● ●● ●
●
●● ● ●
● ●●
●●
●●
●●
●
●
●
●
●
●●
●●
●
●● ●
●●
●● ●
●
●●
●
●
●● ●
●●
●●
● ● ●●
●● ●
● ●●
●
●
●
●● ●
● ● ●
●●
●●
●
● ●
● ● ● ●
●● ●
CCCma.CanESM2
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
CCCma.CanESM2
1960 1970 1980 1990 2000 2010
−50
510
1520
●●
● ● ●●
● ● ●●
●●
●●
●
●
●
●●
●● ● ●
●●
●● ● ●
●●
●
●●
●●
●
●
●
●
●
●
●●
●●
●●
●●
● ●●
● ● ●●
●●
●
●●
●● ●
●● ●
●
● ●
●
●
● ●
●
●
●
●●
●●
●
● ● ●
●●
●●
● ●
● ●
●
●
●●
●
● ●
IPSL.IPSL−CM5A−LR
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
IPSL.IPSL−CM5A−LR
1960 1970 1980 1990 2000 2010
−50
510
1520
●● ● ● ● ● ● ●
●●
●
●●
●●
●
●●
● ● ●●
●●
●●
●● ● ●
●●
●●
● ●●
●
●
●
●
●●
●●
●●
●
●
●●
●
●
●●
●● ●
●●
● ●
● ●
● ● ● ● ●
●●
●● ●
●●
●
●
●
●●
●
● ● ●●
●●
●
●
●●
●●
●●
● ●
● ●●
MPI−M.MPI−ESM−LR
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
MPI−M.MPI−ESM−LR
1960 1970 1980 1990 2000 2010
−50
510
1520
25
●● ● ● ●
●● ● ● ●
●
●● ●
●
●
●
●● ● ●
●●
●●
●●
●● ● ●
●●
●● ●
●
●
●
●
●
●
●●
●
●●
●●
●●
●●
●●
●● ●
●
●● ●
●● ●
●●
● ●● ●
● ●●
●●
●
● ●
●
●
●
●
● ●● ●
●
●●
●●
●
●
● ●● ●
● ●●
MPI−M.MPI−ESM−MR
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
MPI−M.MPI−ESM−MR
1960 1970 1980 1990 2000 2010
−50
510
15
●● ●
●●
● ●● ● ●
●●
● ● ●
●●
●● ●
●
●●
●
●●
●
● ● ● ● ● ● ●●
●●
●
●
●
●
●
●
●●
●
●● ● ● ● ● ●
● ●● ● ●
●●
● ● ●
● ● ●●
●●
●●
●●
●
●
●
●●
●
●
●
● ● ● ●
●
●●
●● ●
●●
●
● ●
●●
● ●●
NOAA−GFDL.GFDL−ESM2G
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
NOAA−GFDL.GFDL−ESM2G
1960 1970 1980 1990 2000 2010
−50
510
1520
25
●● ● ● ● ● ● ● ●
●
●
●●
●●
●●
●● ● ●
●●
●●
●● ●
●●
●●
●●
●●
●
●
●
●
●
●
●
●
●●
● ● ●●
●●
●
●
● ●●
● ●● ●
● ●●
●
●●
●
●●
●
●
● ●●
● ●
● ● ● ● ● ●
●● ●
●● ●
●
● ●
●
● ●
●
● ●
●● ●
NCAR.CCSM4
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
NCAR.CCSM4
1960 1970 1980 1990 2000 2010
−50
510
●
●●
● ●● ● ● ● ●
●
●●
●
● ●
●●
●●
●
●● ● ●
●●
●● ● ● ● ●
●● ● ●
●
●
●
●
●
●●
● ●
●
●● ●
●
●●
●
●●
●
●
● ●
● ●
●
●
●●
●●
●●
●
● ●
●
● ●●
●
●
●
●
●●
●●
●
●
●
●
●
●
●
● ●
●
● ● ●
●
●
●
MRI.MRI−CGCM3
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
MRI.MRI−CGCM3
1960 1970 1980 1990 2000 2010
−6−4
−20
24
●●
●
●
●●
●●
●●
●
●●
●●
● ●● ● ●
● ● ● ● ●● ● ●
●●
● ●
● ●●
●● ●
● ●● ●
●● ●
●
● ●●
●
●
● ●
● ●
● ●
●
●●
●
●
●
●
●
●
●
●●
●
●
●
●
●
●
●
●
●
●●
●
●
●
●
● ●● ●
●
●
● ●
● ●
●
●
●
●
●
●
●
MIROC.MIROC−ESM
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
MIROC.MIROC−ESM
1960 1970 1980 1990 2000 2010
−50
510
●
●
●
●●
●● ●
● ●●
●
●●
●●
●
●● ● ● ●
●● ●
● ● ● ● ● ●● ● ● ● ● ● ●
●●
●
●
●
●●
●
●●
●
●●
●●
●
●●
● ●
●
●
●
●●
●●
● ● ● ●●
●● ●
●
●
● ● ●
●
●
●●
● ●
● ● ● ●
●
●
● ●●
●
●
●
●
●●
●
●
CSIRO−QCCCE.CSIRO−Mk3−6−0
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
CSIRO−QCCCE.CSIRO−Mk3−6−0
1960 1970 1980 1990 2000 2010
−50
510
15
●● ● ●
●●
● ● ●●
●● ●
● ●
●●
● ●● ●
●
●●
●●
● ●●
●●
●●
●●
●●
●
●
●
●
●●
●
●
● ●● ● ●
● ●●
● ● ●
●●
●
●
●
●●
●●
● ●
●
●
● ●●
●
●
●
●
●
●
●
●
●
●
●
● ●
●●
●●
●
●●
● ●
●●
●●
●
●●
CNRM−CERFACS.CNRM−CM5
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
CNRM−CERFACS.CNRM−CM5
1960 1970 1980 1990 2000 2010
−50
510
1520
●●
●●
●●
●● ● ●
●
●●
●
●
●●
●●
● ● ● ●●
●●
●●
● ●●
●●
●●
●●
●
●
●
●
●
●
●●
●
● ● ●
●
●
●
●
● ● ●● ●
● ●
●●
●
●
● ● ●● ●
●
●
● ●
●
● ●
●● ● ●
●
● ● ●
● ● ●●
●●
●
●
●
●
●● ●
●
●●
●
MIROC.MIROC5
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
MIROC.MIROC5
1960 1970 1980 1990 2000 2010
−6−4
−20
24
●●
●
●
●
●
●
●●
●
●
● ●●
●●
●●
●
● ● ● ●● ●
● ● ●●
●●
● ●
●
●
● ●
●
●
●●
● ●●
●●
● ●●
●●
●
●
●
●●
●
●
●
●
●
●
●
●
●●
●
●●
●●
●
●●
●
●●
●
●
●
● ●● ● ●
●
●
●
●
●
●
●
●
● ● ●
●
●
●
●
●
MOHC.HadGEM2−ES
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
MOHC.HadGEM2−ES
1960 1970 1980 1990 2000 2010
−50
510
1520
●● ●
● ●● ● ● ●
●●
●●
●●
●●
●●
●
●●
●●
●
●
●●
● ● ● ●●
● ● ● ●
●
●
●
●
●
●
●
●●
● ●●
●● ● ●
●
● ●●
●
●
● ●●
●
●●
●
● ● ●
● ●● ●
●
●●
●
●●
●
●
●
●
●
● ●
●
●
●●
●
● ●
●
●
●
●
●
●
●
●
NCC.NorESM1−M
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
NCC.NorESM1−M
1960 1970 1980 1990 2000 2010
−8−6
−4−2
02
4
●
● ●
●
●
●●
●
●●
●
● ●● ●
●●
●●
● ●●
●●
●● ●
● ● ●●
●
● ●●
●
●
●
●
●
●
●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●●
●
●●
●
● ●
●
●
● ●
●
●
●
●
●
●
●
●
●
●
●
●
●
●
● ●
●
●●
●●
● ●
●
●
●
●
●
● ●
NOAA−GFDL.GFDL−CM3
−0.1 −0.07 −0.03 0.01 0.04 0.07 0.1
NOAA−GFDL.GFDL−CM3
Data: Surface air temperature (‘tas’) of the HadCRU4 dataset and pre-industrial control/historical/rcp45 simulations of 14 models from CMIP5 project are interpolated onto common grid of 5°x5°. Mask is constructed based on data availability of HadCRU4.
Hypothesis: ΔO = F Δa + Δu
(10-yr mean obs) (Response pattern) (amplitude) (internal variability) Detection: Attribution: Δ denotes differences relative to climatology of 1961-1990. E and α are ensemble size and 5% significance level. denotes covariance matrix of internal variability.
Comments: 1) The forced response pattern of each model was obtained by applying discriminate analysis (Jia and DelSole, 2012) based on 12-eof truncations to each model’s 10-year running mean ’tas’ of the 500-year pre-industrial control run and the historical runs. 2) Detection/attribution analysis was applied to tas anomalies of both HadCRU4 and each model relative to the 1961-1990 climatology. RCP4.5 used to extend historical runs. 3) The confidence interval accounts for the fact that the climatology of 1961-1990 has 1/3 the variance of a 10-year mean. 4) Years shown on the figure of curves are the center year of corresponding 10-year mean (e.g.,1995 for mean of 1991-2000). 5) Consistency is achieved if observations are within the 95% confidence interval (red shading band).
Conclusions: 8 model simulations (blue) are inconsistent with HadCRU4 observations (black), 6 model simulations (blue) are consistent with HadCRU4 observations (black). The dominant forced response patterns differ significantly among some models. References: Jia, L. and T. DelSole, 2012: Optimal determination of time-varying climate change signals. J. Climate, 25, 7122–7137.
0 5 10 15 20 25 30
−0.2
0.2
0.6
1.0
log
ratio
NOAA−GFDL.GFDL−CM3 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
NOAA−GFDL.GFDL−CM3 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
ampl
itude
NOAA−GFDL.GFDL−CM3 Max NtC Discriminant for 30 EOFs: Ratio= 2.71
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
ampl
itude
NOAA−GFDL.GFDL−CM3 Min NtC Discriminant for 30 EOFs: Ratio= 0.412
0 50 100 150 200 250 300 350
−50
050
−1.0
−0.5
0.0
0.5
1.0
NOAA−GFDL.GFDL−CM3 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.4
−0.2
0.0
0.2
0.4
NOAA−GFDL.GFDL−CM3 Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
0.0
0.5
1.0
1.5
2.0
log
ratio
MIROC.MIROC5 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
MIROC.MIROC5 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−6−2
02
4
ampl
itude
MIROC.MIROC5 Max NtC Discriminant for 30 EOFs: Ratio= 8.26
1650 1700 1750 1800 1850 1900 1950 2000
−6−2
02
4
ampl
itude
MIROC.MIROC5 Min NtC Discriminant for 30 EOFs: Ratio= 0.443
0 50 100 150 200 250 300 350
−50
050
−3
−2
−1
0
1
MIROC.MIROC5 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.2
−0.1
0.0
0.1
0.2
0.3
MIROC.MIROC5 Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
MRI.MRI−CGCM3 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
MRI.MRI−CGCM3 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−20
24
ampl
itude
MRI.MRI−CGCM3 Max NtC Discriminant for 30 EOFs: Ratio= 1.91
1650 1700 1750 1800 1850 1900 1950 2000
−20
24
ampl
itude
MRI.MRI−CGCM3 Min NtC Discriminant for 30 EOFs: Ratio= 0.351
0 50 100 150 200 250 300 350
−50
050
−1.0
−0.5
0.0
0.5
MRI.MRI−CGCM3 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.4
−0.3
−0.2
−0.1
0.0
0.1
MRI.MRI−CGCM3 Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
IPSL.IPSL−CM5A−LR tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.5
0.0
log
ratio
IPSL.IPSL−CM5A−LR tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
IPSL.IPSL−CM5A−LR Max NtC Discriminant for 30 EOFs: Ratio= 2.33
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
IPSL.IPSL−CM5A−LR Min NtC Discriminant for 30 EOFs: Ratio= 0.283
0 50 100 150 200 250 300 350
−50
050
−1.0
−0.5
0.0
0.5
IPSL.IPSL−CM5A−LR Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.6
−0.4
−0.2
0.0
IPSL.IPSL−CM5A−LR Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
MIROC.MIROC−ESM tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
MIROC.MIROC−ESM tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
MIROC.MIROC−ESM Max NtC Discriminant for 30 EOFs: Ratio= 2.09
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
MIROC.MIROC−ESM Min NtC Discriminant for 30 EOFs: Ratio= 0.345
0 50 100 150 200 250 300 350
−50
050
−0.5
0.0
0.5
1.0
MIROC.MIROC−ESM Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.2
−0.1
0.0
0.1
MIROC.MIROC−ESM Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
CNRM−CERFACS.CNRM−CM5 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
CNRM−CERFACS.CNRM−CM5 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
CNRM−CERFACS.CNRM−CM5 Max NtC Discriminant for 30 EOFs: Ratio= 2.22
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
CNRM−CERFACS.CNRM−CM5 Min NtC Discriminant for 30 EOFs: Ratio= 0.323
0 50 100 150 200 250 300 350
−50
050
−0.20.00.20.40.60.81.0
CNRM−CERFACS.CNRM−CM5 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.2
0.0
0.2
0.4
0.6
CNRM−CERFACS.CNRM−CM5 Min NtC Pattern for 30 EOFs
Maximized
Minimized
GFDL-CM3 MIROC-5
MRI-CGCM3
MIROC-ESM
IPSL-CM5A-LR
CNRM-CM5
Changes in Internal Variability Due to 20th Century Climate Changes This study investigates the possibility that internal variability responds to anthropogenic forcing on annual time scales. Current detection and attribution methods assume that internal variability does NOT change; our study tests if this assumption is valid. Previous studies assessed this question using univariate metrics such as the “residual consistency check” (Allen and Tett,1999). Unfortunately, this metric tests consistency in an aggregate sense, based on a suitably weighted total sum variability, and as such it allows errors in some components to be compensated by errors in other components, giving a potentially misleading impression of overall consistency. We examine consistency using more powerful discriminant analysis (DA) techniques, which effectively tests equality of the entire covariance matrix of variability.
Log of Minimum and Maximum Noise−to−Control Ratio, Detrended CTR, and 30S to 30N
CanESM2 CNRM CSIRO IPSL MIROC5 MIROCESM HadGEM2 MPILR MPIMR MRI CCSM4 NCC GFDLCM3 GFDLESM
00.5
11.5
2−2.5
−2−1.5
−1−0.5
Red = Raw Data Green = Detrended Blue = 30S to 50N
GFDL
-ESM
GFDL
-CM3
NCC-
NorE
SM1
NCAR
.CCS
M4
MRI-C
GCM3
MPI-E
SM-M
R
MPI-E
SM-L
R
HadG
EM2-E
S
MIRO
C-ES
M
MIRO
C-5
IPSL
-CM5
A-LR
CSIR
O-Mk
3.6.0
CNRM
-CM5
Can-
ESM2
We find that CNRM, IPSL-LR, MIROC-ESM, and MRI experienced a decrease in internal variability, while MIROC5 and GFDL-CM3 experienced increases in internal variability. The time series for the significant components are shown in fig. 2, while the corresponding spatial patterns are shown in fig. 3. These results suggest the following conclusions: • GFDL-CM3 experiences an increase in internal variability that is widespread
throughout the globe. • The decrease in 20th century internal variability tends to be concentrated in regions of
sea ice. We conjecture that 20th century losses in sea ice reduce annual mean variability because, without sea ice, the overlying atmosphere is exposed to the ocean’s surface for longer durations of the year and the ocean’s surface temperature would have a moderating effect on the near-surface temperature of the atmosphere.
• MRI and IPSL-LR possess a low frequency oscillation in their respective control runs that does not persist in the historical period.
• After masking out the polar regions outside 30S and 50N (see blue bars in fig. 1), we find that only MIROC5 yields a significant change in variability. However, the change in MIROC5 appears to be an artifact of a change in mean in the historical runs.
• Some models breach the significance level only marginally, so we hesitate to identify them as exhibiting changes in internal variability because of the subjective nature of truncating EOFs.
Our results show that some models exhibit a change in internal variability of annual mean near surface temperature in response to human influences. Much of this response is in polar regions and a plausible consequence of the loss of sea ice. In addition, changes in internal variability impact the assumptions underlying detection and attribution studies. We plan to continue our analysis to better address the sensitivity of our results to the choice of EOF truncation and the problem of overfitting when applying discriminant analysis. We are also working on applying our analysis to the 21st century to determine if changes in internal variability continue in the future.
Figure 1. Log of noise-to-control ratios for three case studies; before detrending the control run (red), after detrending the control run (green), and masking out the poles to analyze only the domain from 30S to 50N (blue). Black lines indicate significance at the 1% level. We find that most of the differences in internal availability happen in polar regions.
Figure 2. Variates for the optimized noise-to-control ratios. Control period is in blue. One ensemble member from the historical period is in red. For the maximized cases notice the variance appears to increase in the 20th century. Conversely, in the minimized cases the variance decreases.
0 5 10 15 20 25 30
−0.2
0.2
0.6
1.0
log
ratio
NOAA−GFDL.GFDL−CM3 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
NOAA−GFDL.GFDL−CM3 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
ampl
itude
NOAA−GFDL.GFDL−CM3 Max NtC Discriminant for 30 EOFs: Ratio= 2.71
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
ampl
itude
NOAA−GFDL.GFDL−CM3 Min NtC Discriminant for 30 EOFs: Ratio= 0.412
0 50 100 150 200 250 300 350
−50
050
−1.0
−0.5
0.0
0.5
1.0
NOAA−GFDL.GFDL−CM3 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.4
−0.2
0.0
0.2
0.4
NOAA−GFDL.GFDL−CM3 Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
0.0
0.5
1.0
1.5
2.0
log
ratio
MIROC.MIROC5 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
MIROC.MIROC5 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−6−2
02
4
ampl
itude
MIROC.MIROC5 Max NtC Discriminant for 30 EOFs: Ratio= 8.26
1650 1700 1750 1800 1850 1900 1950 2000
−6−2
02
4
ampl
itude
MIROC.MIROC5 Min NtC Discriminant for 30 EOFs: Ratio= 0.443
0 50 100 150 200 250 300 350
−50
050
−3
−2
−1
0
1
MIROC.MIROC5 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.2
−0.1
0.0
0.1
0.2
0.3
MIROC.MIROC5 Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
MRI.MRI−CGCM3 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
MRI.MRI−CGCM3 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−20
24
ampl
itude
MRI.MRI−CGCM3 Max NtC Discriminant for 30 EOFs: Ratio= 1.91
1650 1700 1750 1800 1850 1900 1950 2000
−20
24
ampl
itude
MRI.MRI−CGCM3 Min NtC Discriminant for 30 EOFs: Ratio= 0.351
0 50 100 150 200 250 300 350
−50
050
−1.0
−0.5
0.0
0.5
MRI.MRI−CGCM3 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.4
−0.3
−0.2
−0.1
0.0
0.1
MRI.MRI−CGCM3 Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
IPSL.IPSL−CM5A−LR tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.5
0.0
log
ratio
IPSL.IPSL−CM5A−LR tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
IPSL.IPSL−CM5A−LR Max NtC Discriminant for 30 EOFs: Ratio= 2.33
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
IPSL.IPSL−CM5A−LR Min NtC Discriminant for 30 EOFs: Ratio= 0.283
0 50 100 150 200 250 300 350
−50
050
−1.0
−0.5
0.0
0.5
IPSL.IPSL−CM5A−LR Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.6
−0.4
−0.2
0.0
IPSL.IPSL−CM5A−LR Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
MIROC.MIROC−ESM tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
MIROC.MIROC−ESM tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
MIROC.MIROC−ESM Max NtC Discriminant for 30 EOFs: Ratio= 2.09
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
MIROC.MIROC−ESM Min NtC Discriminant for 30 EOFs: Ratio= 0.345
0 50 100 150 200 250 300 350
−50
050
−0.5
0.0
0.5
1.0
MIROC.MIROC−ESM Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.2
−0.1
0.0
0.1
MIROC.MIROC−ESM Min NtC Pattern for 30 EOFs
0 5 10 15 20 25 30
−0.2
0.2
0.6
log
ratio
CNRM−CERFACS.CNRM−CM5 tas Log of Leading Ratios NtDetrendedCTR_3ens
0 5 10 15 20 25 30
−1.0
−0.6
−0.2
0.2
log
ratio
CNRM−CERFACS.CNRM−CM5 tas Log of Trailing Ratios NtDetrendedCTR_3ens
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
CNRM−CERFACS.CNRM−CM5 Max NtC Discriminant for 30 EOFs: Ratio= 2.22
1650 1700 1750 1800 1850 1900 1950 2000
−4−2
02
4
ampl
itude
CNRM−CERFACS.CNRM−CM5 Min NtC Discriminant for 30 EOFs: Ratio= 0.323
0 50 100 150 200 250 300 350
−50
050
−0.20.00.20.40.60.81.0
CNRM−CERFACS.CNRM−CM5 Max NtC Pattern for 30 EOFs
0 50 100 150 200 250 300 350
−50
050
−0.2
0.0
0.2
0.4
0.6
CNRM−CERFACS.CNRM−CM5 Min NtC Pattern for 30 EOFs
Minimized
Maximized GFDL-CM3 MIROC-5
MRI-CGCM3
MIROC-ESM
IPSL-CM5A-LR
CNRM-CM5
Figure 3. The spatial pattern that corresponds to the variates in Fig. 2.
We apply DA to annual mean, near-surface (2m) temperature from 14 CMIP5 climate models that simulated at least 500 pre-industrial control years and had a 3-member historical ensemble. The data was interpolated onto a common 5x5 grid and climatology was removed. DA finds the linear combination of variables that maximize (and minimize) the ratio of variances. The resulting ratio is called the noise-to-control ratio and given by Internal variability in the 20th century is estimated by subtracting out the ensemble mean. Optimization is performed on the leading 30 EOFs. The upper panel in Fig. 1 shows the log of the maximized ratios for three separate applications of DA and the horizontal black line provides the 99% confidence level. In this figure, a significant maximized ratio means that an INCREASE in 20th century internal variability was detected. Similarly, the lower panel shows the results for the log of the minimized ratios and the horizontal black line there indicates the 1% confidence level. Bars that stay between these two black lines indicate no changes in 20th century internal variability are detected. We found that 2 models had significant maximized ratios and 4 models had significant minimized ratios. However, we noted trends in the control runs, so, we detrended all the control runs and repeated our analysis. The green bars indicate detrended data; detrending substantially reduced the change in variability for two models.
Relation Between AMO and AMOC
MOTIVATION: Research suggests that the Atlantic Multi-decadal Oscillation (AMO) is the most predictable mode of temperature variability. This predictability is often attributed to variations in the Atlantic Meridional Overturning Circulation (AMOC). Here we ask the following questions: 1. Is there a relation between the AMOC and AMO? 2. Is the relation consistent across climate models? 3. Is the maximum streamfunction the best variable for
assessing the AMOC-AMO relation?
Methodology
Results
Future Work
DATA: Analysis is performed on 450 years of annual anomalies of sea surface temperature (SST) and Atlantic mass overturning circulation (AMOC) data from the following pre-industrial control runs: An index for the Atlantic Multi-decadal Oscillation (AMO) is defined as the area-weighted annual SSTA over the north Atlantic from 0-60oN. The maximum strength of the AMOC is defined as the maximum in the AMOC between 30oN-60oN, and denoted as Ψmax. Control runs are split into two parts: a training part (225 years) for building the model and a verification part (225 years) for testing the model. OPTIMIZATION: The relation between the AMO and AMOC is diagnosed by finding the linear combination of AMOC principal components (PCs) that are most highly correlated with AMO in an integral sense (Jia and Del- Sole, 2011). This integral is called Average Predictability Time (APT):
1. AMOC MOST RELATED TO AMO 2. MODEL COMPARISION 3. AMOC and Ψmax
• Only 3 APT are found to be significant across all models
• No robust relation was found on decadal-to-multidecadal time scales
Figure 1: The leading three components of the AMOC that are most related to the AMO, in a multi-model sense. Structures are for the training period.
Differences in cross model AMOC-AMO relation are quantified by fitting an empirical model: Αn F-test is used to test equality of all models
Figure 2: Diagram showing the clustering of the empirical models for 5 CMIP5 climate models. Clustering is based on F-Statistic for pairwise comparison of empirical models. Lines connecting models denote statistically equivalent models at α=0.05. The distance between unconnected models is proportional to F.
• AMOC-AMO relations are statistically indistinguishable for MRI/MPI/NCC
• CCC and NCAR are significantly different from all other models except MRI and NCC
• APT patterns have larger correlation with AMO compared to Ψmax
• Regressing out APT from Ψmax removes significant correlations with AMO
Figure 3: Squared correlation for 3 leading APT-AMO (black), ψmax-AMO (red), and residual ψmax -AMO (blue). Residual ψmax is recovered after the 3 leading APT has been regressed out of ψmax. Negative lags indicate APT and ψmax lead AMO.
• Extend optimization to include sea level pressure over the North Atlantic • Construct simple dynamical model to describe AMO-AMOC mechanism • Attribute differences in AMO variability to differences in regression model and
AMOC variability
Summary
