Hemlock Age   Charlie Cogbill
  May 05, 2001

From: "Charlie Cogbill" 
Date: Sat, 5 May 2001 00:04:26 -0400

Fellow Listers,

I have kept an informal database of documented ages of eastern 
forest trees and include here an excerpt of eastern hemlock maxima 
within various states and provinces. This list is obviously limited and
any additions or comments are actively sought. Several observations are 
appropriate to the discussion of hemlock ages and their determination. 
The extreme hemlock longevity of 998 years from Pennsylvania reported by 
Hough (1960) is a singularity and more than 330 years (or 50%) older 
than the penultimate age. I have always wondered if this aberration is 
due to a reporting error, an extrapolation gone astray, simply extreme 
rarity (maybe never to be repeated), or the small sample size of 
reported ages . In any case, the frequency of "old" hemlock ages truly 
ends in the 600s and the tail of over 500 years is apparently uncommon 
while the 400s are expected maxima in most regions/conditions. 
Significantly, the Massachusetts record of 450 years is obviously 
exceeded by the 474 minimum count from Dunbar Brook, but still less that 
reports from 8 different states and provinces. 
This list's discussion has addressed several methods to determine 
tree ages lacking a pith core, a subject long pondered by 
dendrochronologists and ecologists. The accuracy of these 
extrapolations are always questionable and should be of concern when 
citing ages, especially extremes (for example, it is unclear if eastern 
hemlock can live to 600 or 900 years). Obviously, beyond actual 
observation or archival documentation, any cited age not based on a tree 
core or cookie is uncertain. Furthermore, dendrochronologists will 
argue that any non-crossdated core is to be doubted (at least has some 
uncertainty relative to calendar years). Several sources of error in 
tree age determinations have been mentioned earlier: missing pith (due 
to wood rot; short core-or too big tree; or off-center core), missed 
count (improper preparation-e.g. field read?, undercounting narrow 
rings, missing rings), coring height above germination height, and 
position of pith in relation to tree center (which affects actual 
within-bark radius). Interestingly each of these errors generally biases 
the age in the same direction (younger) and the actual ring count is a 
good starting point for a minimum actual age. Extrapolation beyond this 
minimum age is based on several assumptions: primarily radial growth 
rates, missing length--from tree center position, and amount of height 
growth. For example Bob brackets his Guilder Pond core from 231 to 401 
years. I would hesitate to add that a "pure" extrapolation of the 
recent growth rate 30.8yr/" to the "remaining 5.7" (is bark thickness 
included?) would yield 367 years. Interestingly, a linear regression of 
18 record hemlocks from the cited database with both age and diameter 
determinations gives an equation of size=3D0.0405dbh(cm) + 39.7cm ; and 
the 60.45 cm hemlock above would predict (without a core!) a 513 year 
old tree. Note that the regression y-intercept is 40 cm dbh, meaning 
that the average recent growth rate (approximated slope of the 
line=3D0.008"/yr or 125yr/") is much too slow. The relation is also 
highly variable explaining only 8% of all the variation, showing that 
age is a poor predictor of size and vice-versa. In addition, the 
average lifetime radial increment of this sample is 26.9yr/", but growth 
rates range from 11 to 84 yr/". Given these observations, I would be 
hard pressed to speculate on the growth rates of the Guilder Pond 
hemlock before1811 (=3D2002-191)--interestingly what is the error in 
this determination? In fact, that was about the end of the Little Ice 
Age and the rates were more than likely changing. We might bracket the 
extrapolation between the fastest growth rate (11yr/") times the least 
missing radius (2"?) giving 22 additional years and the slowest 
(125yr/") time the greatest diameter (5.7") giving 712 additional years! 
I suspect that Bob's "calculation" of 60 years is a little short, but 
an accurate determination is elusive. Doing a proper error analysis is 
probably not going to improve this estimate (Bob's range from 200 to
400yr) and my personal solution is to ignore all extrapolations. When 
there is no indication of the location of the pith, as seen in narrow 
arc to inner rings, one must simply cite the minimum age. Thus by any 
standard the 474 ring Dunbar Brook is old (although not the 1020 year 
from "pure" extrapolation) and the missing ~ 10 cm (is it really 1 or 
20, only pith will tell) is the tease to go back and get more cores. In 
fact, ecologically the stand age structure is probably more indicative 
of the conditions of the stand than one admittedly indeterminate age. 

While I'm lecturing, I would also observe that the growth rates of 
hemlocks are highly variable, possibly rapid during release or after 
open canopy regeneration, and conversely highly suppressed for many 
decades, especially in small advanced regeneration. It is difficult to 
deduce this growth from size or recent growth. This is also reflected 
by the relatively long time cited by Paul to reach coring height. 
Incidentally this average "age to breast height" is a classic case of 
the mean being an poor determination of actual values in a highly skewed 
or bimodal distribution. Since virtually all the cited tree ages are 
from breast height cores why add the same constant to all these ages? 
In a parallel argument to the use of an "average" growth rate to 
"correct" missing core pieces, why introduce error and increase the 
variance into the age determination. It seems better to simply indicate 
that the rings were accurately counted and the count is the "age" at 
breast height.

Charles V. Cogbill
Plainfield, VT 05667

The data:

State/Prov, Ring Count, Location, Source,

CN, 350, Colebrook, Nichols 1913

GA, 309, Clayton , Baumgras et al. 1999

IN, 535, Bell Woods, ?

MA, 450, Cold River, Leverett 1991

MA, 365, Alander Mt. , Ed Cook, p.c.

MA, 255, Wachusett Mt., Orwig,

ME, 436, Big Reed Hdwd, Chokkalingam 1998

MI, 509, , Lorimer 1985

MI, 360, Huron Mts., Witty & Coffman

NC, 530, Mt. Guyot/Smokies, Blozan,

NC, 491, Robbinsville, Baumgras et al. 1999

NH, 505, Gibbs Bk. , Cook p.c.,

NH, 400, Cambridge, Lyon 1946

NS, 467, Timber Lake, Lunenburg Co. , Johnson 1986

NY, 470, Shawngunks, Ed Cook, p.c.

PA, 998, , Hough 1960

PA, 607, Tionesta, Morey 1936

PA, 560, East Tionesta, Hough 1936

PA, 492, Hearts Content, Morey 1936 Ecology 17:255

PQ, 661, Abitibi, Frothingham 1915/ Fowles 1975

PQ, 307, Muir's Woods, Nat Areas J. 12 1992

TN, 525, Smokies, Runkle 1982 from Whittaker

TN/NC, 505, Smokies, Blozan/GSMNat. Hist. Assoc.

VT, 486, , Marsh 1864

VT, 419, Gifford Woods, Bormann & Buell 1964

WI, 530, Menominee, Stand 27, Milfred et al 1967

WI, 374, Outer Island, Tyrell & Crow 1994

WV, 375, Cannan Valley, Allard & Leonard 1952