cc: p.jones@uea.ac.uk,K.Briffa@uea.ac.uk date: Wed, 06 Jun 2001 10:38:37 +0100 from: Tim Osborn subject: Re: volcanoes to: tom crowley Tom, thanks for the recent message (and the previous one which I hadn't had time to answer till now!). Here are my comments/thoughts: At 10:18 31/05/01 -0500, tom crowley wrote: >we have just completed a comparison a seasonal model simulation of climate >change over the last 1000 years with the Briffa et al data. the seasonal >ebm allows us to sample the model field over the area (20-90N land) and >time (April-Sept) of the Briffa et al. 2001 JGR paper. Correlations of >decadally smoothed model and observations is 0.76 for the whole record and >0.82 over the post-1500 interval. The amplitudeand shapte of the very >large decadal oscillationis in the 17th century are generally very well >simulated. > >We are now comparing the model results with the Briffa et al 1998 volcano >paper for individual years. Inspection of the annual model time series >shows many of the same peaks you identify (we force our volcano record with >a composite of ice core time series from Crete and Gisp2). > >BUT I am still a little unsure about how we should average our data, since >your 1998 paper - if I understand your paper correctly you are not >averaging over the same area as the 2001 paper - I read your 1998 paper as >indicating all NH April-Sept. > >Is this correct? Please verify so we can analyze our model fields in an >identical manner. The NH time series in the 1998 volcano paper was not calibrated in a particularly rigorous manner, though you are correct to say that we did calibrate it against the full NH mean (i.e., land AND marine, everything north of the equator) for April-September, subject of course to the incomplete coverage of observed temperatures over certain regions during the 1881-1960 period. The data used in the 1998 paper were the standardised tree-ring densities (i.e., dependence of density on tree-age was removed by fitting a Hugershoff function to each tree-core series), which have less low-frequency variance than the age-banded approach used in our 2001 JGR paper. We have also used the "Hugershoff standardised" data set to reconstruct the April-Sept land north of 20N series (i.e., the same series as we reconstructed in the 2001 JGR paper using the age-banded version). This improved quasi-NH reconstruction is described in a paper (Briffa et al.) that has been provisionally accepted for a special issue of The Holocene, which will probably appear in 2002. If you're interested in a series with the short-timescale variations but not the long-timescale variations, then it is probably better to use the 2002 Holocene calibrated reconstruction rather than the 1998 volcano series (they're obviously similar, but not identical). I'm not sure exactly when we'll release that series, however, since the paper won't appear for some time - I'll have to check with Keith about that. >Thanks for any help and of course I will send you a preprint of the paper >when it is written up. We would, indeed, be keen to see the preprint - and also to make a contribution as co-authors if that is appropriate or necessary. At 14:55 25/04/01 -0500, tom crowley wrote: >we are in the preliminary stages of comparing our 2d ebm simulations >(April-Sept, >30 N land) with the Briffa et al time series you sent me. We >simulate a larger early 20th century temp rise than you reconstruct, but >yours is also smaller than the Jones et al observations you illustrate in >your JGR paper. > >do you have any idea why you are not reconstructing the full magnitude of >the 20th century rise? it certainly can not be attributable to low >sensitivity of your index, for it is very "excitable" in the earlier part >of the record. We have thought about this in some detail, and also tested a few things out. The correlation between the tree-density series and the observed temperature series is stronger at the longer timescales (e.g. for decadally smoothed data) than for shorter timescales (sub-decadal). Since the regression slope (and hence the variance of the reconstruction) is always reduced in situations where the correlation is poor, it seems that the poorer performance at the shorter timescales has degraded the overall variance of our reconstruction and hence resulted in a smaller magnitude of early 20th century warming in the reconstruction. We have attempted to create an alternative reconstruction, where we calibrate separately on the high and low frequencies, and this results in a noticeably greater early 20th century warming. However, the degrees of freedom for the decadal calibration are rather few, due to the smoothing, and therefore it is very easy to overfit the decadal component of the reconstruction. We have a paper describing this study coming out soon in Dendrochronologia - let me know if you'd like a preprint/reprint. >also can you send me the unsmoothed Jones et al NH land Ap-Sept time series >you used so I can be sure I am using the same observations you are? The file is attached (Apr-Sep land north of 20N), as anomalies in degrees C from the 1961-1990 reference period mean. We only use 1881-1960 for our final calibration. Best regards Tim 138 1860 -0.14 1861 -0.16 1862 -0.18 1863 0.03 1864 -0.28 1865 0.28 1866 -0.05 1867 -0.42 1868 0.38 1869 0.09 1870 0.19 1871 -0.22 1872 0.31 1873 -0.25 1874 -0.05 1875 -0.31 1876 -0.11 1877 -0.27 1878 0.26 1879 -0.28 1880 -0.10 1881 -0.14 1882 -0.38 1883 -0.39 1884 -0.71 1885 -0.59 1886 -0.25 1887 -0.22 1888 -0.45 1889 -0.21 1890 -0.27 1891 -0.42 1892 -0.31 1893 -0.35 1894 -0.21 1895 -0.22 1896 -0.20 1897 0.02 1898 -0.20 1899 -0.22 1900 -0.09 1901 0.02 1902 -0.63 1903 -0.60 1904 -0.57 1905 -0.31 1906 -0.06 1907 -0.80 1908 -0.42 1909 -0.48 1910 -0.35 1911 -0.28 1912 -0.70 1913 -0.54 1914 -0.31 1915 -0.27 1916 -0.40 1917 -0.46 1918 -0.49 1919 -0.23 1920 -0.14 1921 0.09 1922 -0.03 1923 -0.30 1924 -0.19 1925 -0.02 1926 -0.28 1927 -0.07 1928 -0.28 1929 -0.23 1930 0.05 1931 0.04 1932 0.15 1933 -0.06 1934 0.05 1935 -0.12 1936 0.17 1937 0.27 1938 0.38 1939 0.19 1940 0.08 1941 -0.05 1942 0.00 1943 0.23 1944 0.19 1945 0.01 1946 0.14 1947 0.12 1948 0.21 1949 0.02 1950 -0.09 1951 0.14 1952 0.20 1953 0.33 1954 -0.01 1955 0.06 1956 -0.33 1957 -0.07 1958 -0.07 1959 0.11 1960 -0.02 1961 0.15 1962 0.05 1963 0.02 1964 -0.23 1965 -0.35 1966 -0.06 1967 -0.01 1968 -0.18 1969 -0.20 1970 -0.02 1971 -0.14 1972 -0.30 1973 0.05 1974 -0.10 1975 0.19 1976 -0.23 1977 0.23 1978 -0.13 1979 -0.08 1980 0.07 1981 0.32 1982 -0.00 1983 0.18 1984 0.03 1985 0.01 1986 0.07 1987 0.11 1988 0.50 1989 0.37 1990 0.50 1991 0.49 1992 -0.19 1993 -0.00 1994 0.60 1995 0.54 1996 0.11 1997 0.42 Dr Timothy J Osborn | phone: +44 1603 592089 Senior Research Associate | fax: +44 1603 507784 Climatic Research Unit | e-mail: t.osborn@uea.ac.uk School of Environmental Sciences | web-site: University of East Anglia __________| http://www.cru.uea.ac.uk/~timo/ Norwich NR4 7TJ | sunclock: UK | http://www.cru.uea.ac.uk/~timo/sunclock.htm