Monday, May 7, 2012


date: Wed, 14 Aug 1996 14:59:37 -0500 (EST)
subject: Icefields paper

Dear Keith/ Phil,

I have been working on the paper but am now in
need of some help and an outside perspective. THis is
particualrly true for the introductory and concluding sections
where some broader vision is needed to put this study in
perspective. I have attached a revised text with a list of
questions and figures. I have faxed new figures to you that you
have not seen: several of the figures you already have in the
earlier report.

Comments, suggestions and corrections are needed and welcome. I
look forward to hearing from you in the near future.



general questions/comments


check Table 1 reconstructions; Is this set up correctly? I do not
see how you can get negative anomalies if you feed in MXD
chronology data which are positive. Am I missing something here?


I have severely curtailed in arm-waving contextual stuff e.g.
importance of millennial high resolution records etc.

DO we need some of this in the introduction?

Do we need ( wish) a further elaboration of which densitometry is
etc or do we assume that the readers either know or can find out?

CALIBRATION/ nature of the climate signal.

This is unchanged from the earlier report. Do we wish to report
calibration tests or should this section be cut.

I think this calibration section needs some strengthening.


Until I plotted up Figure 2 I did not realise the proportion of
the chronology that is Abies in the 14-1600 period. Although the
picea density chronology does (as stated) correlate very well
with the combined chronology throughout this interval, the
correlation between the abies and picea chronologies themselves
is much lower (ca. 0.5-0.7 see draft diagram). Examination of the
statistics for the Abies MXD series indicates that their mean
density is about 0.1 greater than picea (which should not be a
problem because the series were indexed before averaging in the
chronology-right?) and their mean sensitivity value is about half
that of picea. i.e. there appears to be less interannual

It seems to me that this could be a serious problem and/or be
picked up as such by a referee. The obvious question would be if
we substituted the picea chronology and did the reconstruction
over the results be any different over this interval? The
justification for including the abies was to increase replication
in the snag record but I assumed the climate signal was similar
from the two species. We have no data from the Icefields to test
this- Colenutt and Luckman 1991 did develop abies and picea
chronologies from the same site at Larch Valley which show a
similar pattern of response (see Figure 4, xerox sent) but they
are clearly not identical. The question is basically do we
address this issue head on, citing these data and indicate that
reconstruction from picea alone is very similar to that produced
in the paper- and present (or have available) data to back it up?
Although I could probably find picea snags (and Fritz has cores)
to address this issue it would take a year or more to process
this material and is not a solution at this stage.

so- can you try a reconstruction using picea alone and see if
they are different. I have assumed that because the two
chronologies are very highly correlated then reconstructions from
these two chronologies (using the same transfer function) would
be equally similar. Is that a valid assumption?


Do we need this or should I just omit the tree-ring record from
the diagram?


Do you wish to maplify this with some other records and/or
diagrams etc.?



needs a broader vision?

any ideas?


Luckman, Briffa, Jones and Schweingruber


Recent studies have established the potential for reconstructing
summer temperature conditions from tree-ring densitometry of
conifers growing in boreal and subalpine sites (Schweingruber
1988a, Briffa et al etc). In particular the maximum latewood
density of each annual ring has proven to be quite sensitive to
summer conditions and has been extensively used in climate
reconstruction work (Schweingruber 1988a). In North America the
principal application of this technique has been in the
development of sample site networks that provide regional
overviews for the last 2-300 years (FHS 1988b, Briffa etc).
Ongoing dendrochronological studies have also demonstrated the
potential for millennial length chronologies based on old living
trees and snag material (Luckman etc Jacoby, Briffa et al.,).
This paper presents a quantitative reconstruction of past summer
temperatures based on densitometric and ring-width measurements
from trees that grew at sites near the Columbia Icefield in the
Canadian Rockies. The reconstruction spans most of the last
millennium and is evaluated against other paleoenvironmental data
from the same site plus similar, shorter reconstructions from a
variety of North America sources.


The history of tree-ring studies in the Canadian Rockies has been
reviewed previously (Luckman and Innes 1991, Luckman 1993).
Recent studies have built up a data base of almost 50 ringwidth
chronologies focused primarily on high elevation (treeline) sites
using Picea engelmannii (Engelmann spruce), Larix lyallii
(alpine larch) or Pinus albicaulis (whitebark pine). Almost all
chronologies exceed 300 years and millennial-length chronologies
have been developed for all three species utilising living trees
and ring series from dead (snag) material lying on the surface
close to treeline. Although the strong relationships between
maximum latewood density and summer temperatures were first
demonstrated using Picea engelmannii from a treeline site within
the Rockies (Parker and Henoch 1970) there have only been limited
subsequent studies using densitometry. Luckman et al. (1985) used
several density and ringwidth variables to reconstruct June-July
temperatures and December-March precipitation from Picea
engelmannii at Lake Louise. In the only study utilising
Psuedotsuga menziesii Robertson and Jozsa (1988) used 15
ringwidth and density chronologies from 12 cores to reconstruct
June-July precipitation and temperature data from 1800 from a
site near Banff. Both of these studies were exploratory, have
limited sample depth and only reconstruct the high frequency
climate signal.

Tree-ring investigations at the Columbia Icefield

The Columbia Icefield is the largest icemass in the Canadian
Rockies and feeds several major outlet glaciers (e.g. Athabascsa,
Dome, Saskatchewan, Columbia and Stutfield). The eastern edge of
the Icefield is easily accessible by road and this area has been
the focus of considerable paleoenvironmental work in and around
Sunwapta Pass and the terminii of Athabasca and Dome Glaciers
(Luckman 1988, 1994a, Beaudoin and King 199?, etc).

Several tree-ring investigations have been carried out in this
area over the last 15 years associated with studies of glacier
fluctuations and paleoclimate. The subalpine forest at the
Icefield consists principally of Picea engelmannii and Abies
lasiocarpa with minor components of Pinus albicaulis and Pinus
contorta (Holland and Coen, 1982?). Dendrochronological studies
have focused almost exclusively on picea because abies rarely
attains ages of >250 years.

For descriptive convenience four principal sampling areas can be
identified- the Athadome, Ancient Forest, Icefields and Sunwapta
sites (Figure 1). Initially tree-ring investigations were used to
document the Little Ice Age history of Dome and Athabasca
Glaciers using dendrogeomorphic evidence to date glacier
fluctuations (Luckman, 1986, 1988). The first ringwidth
chronology was developed for the ice-proximal Athadome site
(1660-1980) that lies between the lateral moraines of Athabasca
and Dome Glaciers. A second longer (1346-1981) ringwidth and
densitometric chronology was developed from a site north of the
Athabasca Glacier using living picea up to 680 years old (Jozsa
et al., 1983, Luckman et al 1984). This "ancient forest" site is
bordered downslope by the Little Ice Age terminal moraines and
outwash of Athabasca and Dome Glaciers and grades upslope into
the treeline ecotone.

The "Icefields site" (Figures 2 and 3) is a more open area on the
lower slopes of Mount Wilcox adjacent to and upslope of the
Little Ice Age moraines of the Athabasca Glacier. Studies of
seedling establishment and tree-line dynamics on this slope
(Kearney, 1982; Kavanagh and Luckman 1996) indicate a complex
history. Treeline is presently advancing upslope but abundant
snag material lying on the slope surface indicates that treeline
was formerly higher (Figure 3). Radiocarbon dates of 980-1160 14C
yr B.P. from these snags initially suggested that this phase of
higher treelines pre-dated the Little Ice Age (Luckman 1986) and
promised the potential to extend the living tree chronology by
several hundred years. Ring-width and densitometric data were
obtained from Forintek Canada Corporation for a limited number of
these snags but crossdating trials were only partially successful
leading to the development of a "floating" (undated) chronology
from some of these materials (see Hamilton 1987, Luckman and
Colenutt 1992). Subsequent development of an extensive ring-width
data base from snags at this site led to the crossdating of the
floating chronology within the existing living tree chronology,
demonstrating that some of the previously-obtained 14C dates were
up to 500 years too old (Luckman 1994a). Ring-width series from
these snags and living picea at the Athadome, Icefields and
Sunwapta sites were used to develop a composite site chronology
(the Athabasca chronology) extending from 1072-1987 (Luckman
1992, 1993).

During the summer of 1984, Schweingruber sampled a network of 63
tree-ring sites in western North America (Schweingruber 1988b).
Ring width and densitometric chronologies were developed for all
sites in this network and have been used to reconstruct summer
temperature patterns in western North America over the last 3-400
years (Schweingruber et al., 1991; Briffa et al, 1992). This
network includes a Picea engelmannii chronology from Sunwapta
Pass that extends from 1608-1983 A.D. This Sunwapta site is about
5km upvalley of the Icefields site and at a similar elevation
(Figure 1).

Following the development of the Athabasca Chronology additional
sampling of snag material at the Icefield site was carried out in
1990. Selected samples of cross-dated snag material from this
site were submitted to the Swiss Federal Forestry Institute
Laboratory in Birmensdorf for densitometric analysis in 1992.
These samples were selected on the basis of the record length,
dating and quality of the ringwidth series. They were not
identified by species prior to selection and actually consisted
of 2 Pinus albicaulis, 14 Abies lasiocarpa and 22 Picea
engelmannii discs (identifications by F. Scharr, Birmensdorf).
Ring-width and maximum density chronologies were initially
developed separately for both the abies and picea samples but, as
the majority of samples were picea and the combined and picea
chronologies are very highly correlated over the interval from
1400-1690 A.D. (Figure 4), both species were used to develop the
chronology presented in this paper. The two pinus samples provide
important replicates in the earliest part of the record (1138-
1315; 1179-1383) and are also included.

The densitometric chronology developed from this snag material
covers the interval from 1073-1897 but is poorly replicated after
ca 1700. Few of the snag records extend into the 19th century and
only one living tree (1778-1991) was sampled for densitometry at
the Icefields site. Therefore the densitometric series from the
snag material at the Icefields site were combined with the living
picea record developed from the Sunwapta site to produce the
chronology used in this paper (Figure 4). Note that, although the
chronology ends in 1991, there are only two radii (from a single
tree) that postdate 1983.

The ring-width chronology used in this study is the Athabasca
chronology and was developed using a larger composite sample of
snags from the Icefield site and living trees from the Sunwapta,
Icefields, Ancient Forest and Athadome Sites.

Tree-ring width measurements were either obtained using the TRIM
measurement system (McIver et al, 19xx) at The University of
Western Ontario or by densitometry at Birmensdorf. All the
densitometric data used were obtained at the Birmensdorf
facility. Technical details of sample preparation techniques are
given by Schweingruber (refs). Crossdating was verified using
standard software (Holmes, etc,). Standardization techniques used
during chronology development were very conservative. Individual
ringwidth series were standardized using either a simple negative
exponential or linear fit. Maximum density values were
standardized using a linear transformation.


In the Canadian Rockies the instrumental climate records begin
with the initial European settlement of the region that followed
the development of the transcontinental railways. The oldest
meteorological stations are associated with the Canadian Pacific
Railway, namely Calgary (1880), Banff (1887, first complete year
1890), plus Donald (1891-1901) and Golden (1902-present) in the
Rocky Mountain Trench. Further north the earliest records are
from Edmonton (1880) and along the Grand Trunk Railway beginning
in 1914 at Jasper and Cranberry Lake (the present Valemount). The
highest elevation meteorological station with a long record in
the Rockies is Lake Louise (on the CPR) at ca 1530m, commencing
observations in 1915. Higher elevation sites have only short or
seasonal records (Janz and Storr 1977) that are inadequate for
calibration purposes.

These climate data were recently reviewed and summarised by
Luckman and Seed (1995). Precipitation and temperature records
were checked for homogeneity and missing values estimated by
regression techniques from adjacent stations. Complete monthly
climate data records were developed for precipitation and
temperatures at Jasper (1916-1994), Banff (1888-1994) Lake Louise
(1915-1990), Valemount (1916-1989) and Donald-Golden (1891-1990).
The Donald and Golden records do not overlap but, as the sites
are at similar elevations and close together within the Rocky
Mountain Trench, these records were combined without adjustment
and the linking 1902 data estimated from Banff.

These meteorological stations are all located in valley floor
sites 500-1200m below and 100-150km from the Icefields site
(Figure 5). A regional composite record was developed to maximise
the length and representativeness of these records for
calibration purposes. Monthly temperature series for Banff,
Jasper, Golden/Donald and Valemount were converted to anomalies
(�C) from the 1961-90 means for each station and the four series
were averaged to provide monthly anomaly values for the Canadian
Rockies from 1890-1994. These data were preferred to the Bradley
and Jones (1991) gridded temperature records because of their
more restricted geographical coverage within the Rocky Mountains
(see Luckman, in press).


Based on previous work, climatic calibrations were attempted
using a variety of temperature combinations with both ring-width
and maximum density data. After several trials, reconstructions
were developed for April-August temperatures which provided the
best verification when compared to the regional climate record.
The April-August interval combines elements of both spring and
summer conditions and is often referred to as the "summer half-
year". Two reconstructions were carried out. The first used
maximum latewood density of the growth year and prior year. The
second reconstruction also included ring-width for the same two
years. The reconstructed chronologies are very similar but the
reconstruction using both ring-width and density data explains
slightly more of the variance during the calibration interval.
Details of the chronologies and calibration statistics are given
in Table 1.



The quality of a reconstructed climate record can be verified in
several ways. Conventionally the reconstruction is initially
calibrated using only half of the available instrumental climate
record. These relations are then tested by using them to
reconstruct the climate for the period covered by the climate
data held back from the calibration and then comparing the
reconstruction with the observed climate record for this
interval. This process is then repeated reversing the two climate
data sets. If the reconstructions are of similar quality the data
set is then combined in a final run using the entire climate data
set to develop the reconstruction.


(i) Comparison with the instrumental climate record.

The calibration results are highly statistically significant
(Table 1). Over the 1890-1982 period the correlation between the
regional temperature record and the reconstructed record is 0.651
for the annual data and 0.686 using decadal values (1890-1980).
Figure 4 shows a plot of the actual and reconstructed records.
The reconstruction captures the general rise in temperatures
between 1890-1930 but performs less well during the latter half
of the calibration period (see Table 2). Examination of the
individual station records that comprise the network used to
generate the regional series also shows greater differences
between the component sites during this interval (Luckman and
Seed, 1995). The greatest discrepancy in the decadal series
occurs during the 1960s and this period, with the 1950s, has the
most extreme interannual variability in the instrumental record
(Figure 6). This may in part explain some of these differences.
Examination of the residual values (measured temperature minus
the reconstructed values) shows a strong negative correlation
with the actual values (Table 3) which indicates that the
reconstruction provides a conservative estimate of the more
extreme years in the calibration period. Examination of both the
decadal (Table 4) and annual data (Figure 6) indicates that both
the warmest and coldest years are underestimated. Possibly this
situation could be improved with a larger sample of "warmer"
years in the calibration period: the lack of densitometric data
after 1983 severely restricts the available sample for
calibration purposes.

General comment

Figure 7 shows the reconstructed temperature record for the
Icefield site and the reconstructed decadal averages are given in
Table 4. All reconstructed and instrumental temperature data are
reported as anomalies from the April-August average for the 1961-
1990 period. During the 1961-90 interval the decadal temperature
anomalies were, respectively, -0.08�C, -0.06�C and +0.15�C. The
1961-1983 mean anomaly is -0.083�C. and the 1981-90 decade
contains half of the 14 positive years during the entire 30-year
interval. As the reconstructed record essentially terminates in
1983, it should be anticipated that the most recent part of the
reconstructed record would have a dominance of negative
anomalies. The last 30 years also contain some of the warmest
decades in the instrumental climate record (Tables 4,5) and the
mean temperature anomaly over the 1891-1990 period is -0.375�C.

The fact that almost all of the reconstructed values are negative
(Figure 7) indicates that summer temperatures at the Icefield for
most of the last 900 years were considerably lower than those
experienced during recent decades. The reconstructed mean for the
period 1073-1990 is -0.656�C and is almost 0.3�C lower than the
mean of the instrumental record from 1891-1990, indicating that
conditions during the period of the instrumental record are, on
the average, warmer than the preceding 800 years. This is not
unexpected because the greatest extent of the Athabasca Glacier
during the last 10,000 years occurred in the 1840s (Luckman

The only part of the reconstructed record that has decades with
positive temperature anomalies is the late 11th century at the
beginning of the record. As this reconstruction is based on fewer
than 5 samples prior to 1140 and these are all young trees, it is
possible that these results may be unduly influenced by juvenile
effects and should be interpreted more cautiously than the better
replicated sections of the reconstruction.

Description of the reconstruction

Figure 7 shows the annual reconstruction of summer temperature
anomalies fitted with a 10-year Gaussian filter to highlight the
overall trends. Sample depth, species and composition of the
chronology are shown in Figure 4. The reconstructed decadal
averages are given in Table 4 and the extreme years and decades
in the reconstruction are listed in Table 5.

The most obvious comment about the reconstruction is that almost
all of the record is well below the 1961-1990 mean i.e. most of
the last 900 years were significantly cooler than the last 30
years of the instrumental record. The other striking feature of
the reconstruction is the severity of conditions during the 19th
century which contains 8 of the 10 coldest years and 3 of the 10
coldest decades (Table 5). The average reconstructed temperature
anomaly over the 120 year interval from 1781-1900 is -1.04�C and
three separate six year intervals within this period have mean
anomalies <-1.6�C. (Table 5). The four coldest decades were the
1780s, 1810s, 1830s and 1870s.

In addition to the 1780-1900 period, several other cold intervals
are reconstructed particularly ca. 1190-1250, 1280-1340, 1440-
1500 and the 1690s. Temperature anomalies for the 1197-1240
interval average -1.09�C and six of the decades in the 13th
century rank in the coldest 20% of decades in the reconstruction
(Table 5). The 1445- 1500 average anomaly is -0.96�C with three
decades during this interval falling below -1.0�C. The 1690s are
unique in having six consecutive years (1696-1701) with
temperature anomalies colder than -1.5�C.

Nine of the 10 warmest decades in the reconstruction are in the
periods 1073-1110 and 1921-1980 (Tables 4 and 5). Apart these two
periods the major warmer intervals were ca. 1160-1180, 1260-1280,
1350-1440 and 1710-1770. The 16th and 17th centuries are, on the
whole, rather unremarkable. This result was unexpected given the
variability of ringwidth values during this period (see below and
Luckman 1996a) and the fact that the early 1600s are considered
to be the coolest period in the northern hemisphere during the
last 500 years (Bradley and Jones, 1993). It is possible that
some of this complacency may be attributed to the admixture of
abies samples in the snag chronology but, as the mixed chronology
is very highly correlated with the less well replicated picea
chronology throughout this interval (Figure 4), this effect is
thought to be of minor significance.

This record indicates that summer temperatures during the period
from ca 1120-1960 at the Columbia Icefield have been almost
universally cooler than those recorded during the 1961-1990
reference period. However, although summer conditions were
particularily severe in the 19th century, the presence of several
episodes of relatively warmer and cooler conditions throughout
this record does not permit the clear demarcation of a classical
"Little Ice Age" period of cooler temperatures. The
reconstruction does suggest that warmer summers did occur in the
1073-1120 interval that may be interpreted as a period when
conditons were similar or warmer than present. However, low
sample depth and possible juvenile density effects suggest these
data be interpreted with caution until complementary evidence and
or longer chronologies are available.

Comparison with other proxy climate data from
the Canadian Rockies

Previous work at the Athabasca Glacier and in adjacent areas has
provided a rich source of proxy environmental data that may be
used to evaluate the temperature reconstruction shown here
(Figure 8). The local glacial record indicates the maximum
extents of Dome and Athabasca Glaciers were in 1843/4 and 1846
respectively (Luckman 1988). There is also evidence from a tilted
tree at the Athadome site that the Athabasca Glacier had advanced
very close to its maximum position earlier in 1714 (Heusser 1956,
Luckman 1988).

Examination of the regional glacial record also shows two major
periods of glacier advance in the last 300 years (Luckman 1996a,
see also Figure 9). The first of these is a tightly dated set of
moraines between 1700 and 1725 found at about 25% of the glaciers
studied in this region. The most extensive glacier event was
usually in the nineteenth century: dates derived from trees
tilted or killed by this event at 7 different glaciers suggest
this maximum extent occurred in the 1840's (Luckman 1996a). At
many glaciers in the Rockies there are several moraines between
the LIA maximum moraine and the first known position of the
glacier margin based on human records (usually photographs).
Where dated, most of these moraines were formed between 1850 and
ca 1920. At the Athabasca Glacier there are two such moraines
between the outermost limit and the first photographed position
of the glacier in 1906 (Luckman 1986b).

Almost all of the evidence presented above strongly supports the
reconstructed temperature history from the Athabasca site (see
Figure 8). The severest conditions during the last millennium are
clearly during the first half of the 19th century and the 1840s
marks the beginning of a period of slightly milder summer
temperatures that would have led to glacier recession after the
period of coldest summers. The renewal of colder summers during
several decades of the late 19th century could clearly account
for the glacial events that produced the well defined moraine
ridges that lie between the LIA maximum limit and the known 1906
ice front position of Athabasca Glacier. As the dating of
moraines at the Athabasca parallels that found at many other
glaciers within this region this would seem to be a good
indication that;
(i) the reconstructed temperature record is very useful in
understanding the local glacier history of Athabasca and Dome
Glaciers; and,
(ii) the local glacier history is reasonably representative of
the regional glacier history of the Canadian Rockies and is
therefore a useful summary of regional climate trends over this

Most of the 1700s are reconstructed as being warmer than the
1800s and this corresponds with the evidence of wider ringwidths
from many tree-ring chronologies in this area (see Figure 9 and
Luckman, submitted). Several climate reconstructions from
northern North America also show that the late 18th century was
generally milder than the early 18th and 19th centuries (e.g.
D'Arrigo and Jacoby, 1992). However, in many of these records the
onset of severe cooling occurs after ca. 1810 (similar to the
ringwidth plot shown in Figure 9) rather than in the 1780s. This
difference may reflect the fact that the tree-ring maximum

No comments:

Post a Comment