..‘.|_vPL..(.?.. .. .|$_ll 7.4.24 2
.1 . . .
. .vkfifiu.‘
ht»...

. .
Emu“... v..- a}

1,. ..N...».. 0 5n uh... .
». ...~.V...7.-.$¢
. {o
J?-J»t!.dfld..fl.... .

’fi.

.....\..vbl

At. Huang z
. 0. . ,VHE‘lofl-IL’I

I ’1
ii ':5'
L 35'?

. ‘f-
, E};
'1'
."r

:u
‘g
h.

D
|
I

{i
'1

§

X.‘ul.p.0tot
. ‘ln'..\.tt\!l

. .

0.3'9110It1r.
.t.‘.-.Iv-- n Iv
. 1-. 'auoa. vu‘
.0 ‘VIuc

 

\o. .I... NI.» 5
..s)tiI|‘-u"‘c
.‘all>v. ....‘
4 | II
9!. .V“?¢v|.lu
vl.v Avv. t; v:
'0

0":

‘ ‘ "in“.hur y
.. . . . . . .vautll n...
A y .. .| (I ,
‘ .13 I.
v vl'L‘ . v
. . ulln‘.lv.
v:|.u'.. . .... . .. . . y. . ‘ ..
ncnlbt. . . ‘ y . . . . l... 4 .

4,; '

aw '1'
. .
"3"? . .
b

H

..

g

\

i

D

. . Int...
...vl.l. 9|--nnf
V Iv\\ uc.o'|ob!“|t|t¢vi

I

.‘7“ I‘-
‘ .

'J"°l~: :

I II! I.
II:-
nil"

\' I M"

.> ‘ 'Iv1 »
ft‘t.l.vo-vt.l'|l
..l\. 3L,|‘...pbv!

 

1 'Au. ‘ v
a» ,...I '10..
I...o'6.li . .
. v 23.1}

‘

 

 

THESlS

L

MICHIGAN STATE UNIVERSITY UBRAR

Willi/Ill”)!!!Ill/HUN!!!){Ill/Ill!!!Willi/ml!

' 3 1293 01718 1086

This is to certify that the

thesis entitled

Projecting Michigan's Aspen Timber Resource ,

presented by

Heidi Cherry

has been accepted towards fulfillment
of the requirements for

M.S. Forestry

degree in

 

 

 

Major professor

Date 2/17/Q8

0-7 639 MS U is an Affirmative Action/Equal Opportunily Institution

 

 

 

LIBRARY

Mlchlgan State
University

 

 

 

PLACE IN RETURN BOX
to remove this checkout from your record.
TO AVOID FINE-S return on or before date due.

 

 

DATE DUE um: DUE I DATE our.
2004

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

use mus-m4

 

PROJECTING MICHIGAN’S ASPEN TIMBER RESOURCE
By

Heidi Ruth Cherry

A THESIS
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

MASTER OF SCIENCE

Department of Forestry

1998

ABSTRACT

PROJECTING MICHIGAN’S ASPEN TIMBER RESOURCE
By

Heidi Ruth Cherry

Remeasured plots from 1980 and 1993 Michigan forest inventories were used to
develop harvest probability equations based on stand characteristics such as ownership,
stocking, and location. Michigan’s 1993 inventory plots with aspen were projected thirty
years using a modified version of STEMS85 with management based on these equations.
Michigan Specific diameter growth and mortality correction factors were included. The
projected net loss between 1993-2023 in aspen growing stock volume is 60%, from 40
million to 16 million cords. The greatest proportional loss of aspen is projected to occur
on private ownerships and in the Upper Peninsula while public lands and the Lower
Peninsula increase their proportions. The average annual harvest is projected to increase
from a 1983-1993 level of 122 million cubic feet to 128 million cubic feet between 1993-
2003. However, projections Show a decrease to 97 million cubic feet/year between 2003-

2013 and 74 million cubic feet/year from 2013-2023.

ACKNOWLEDGMENTS

I wish to acknowledge the financial support given by the USDA Forest Service
North Central Forest Experiment Station (NCFES) for this study. They supported not
only the research itself but also my degree studies. Mark Hansen and Pat Miles from
NCFES Forest Inventory and Analysis provided support by being a resource for questions
about the inventory data and growth and mortality correction factors developed from it.
Gary Brand from NCFES Mathematical Models in Forest Dynamics for the North Central
Region provided support in getting STEMS prOperly compiled and functioning and
assisted in my understanding of some of its basic components. Jeff Bowles from the
NCFES Economics Units in East Lansing, MI developed the program that properly
formatted the inventory data into tree lists readable by the STEMS program and did the
compiling of the original STEMS code into a properly functioning program. I also wish
to thank my committee for their support, time and effort in helping me to develop and
finish this research. They are Dr. Donald Dickmann and Dr. Karen Potter-Witter from
Michigan State University Department of Forestry, Dr. Daniel Stynes from Michigan
State University Department of Parks and Recreation, and Dr. J. Michael Vasievich from

NCFES Economics Research Unit in East Lansing, MI.

iii

TABLE OF CONTENTS

LIST OF TABLES ............................................................................................................. vi
LIST OF FIGURES ........................................................................................................... vii
INTRODUCTION ................................................................................................................ 1
1, THE BIOLOGY OF ASPEN ........................................................................................... 2
11, THE ASPEN RESOURCE ............................................................................................. 5
Current Timber RCSOUI'CC ................................................................................................. 8
Management Concerns ................................................................................................... 10
Demand Concerns ........................................................................................................... 11
Structure Concerns ......................................................................................................... 14
Ownership Concerns ....................................................................................................... 15
111. BACKGROUND ......................................................................................................... 17
Past Projections .............................................................................................................. 17
Recent Projections .......................................................................................................... 19
Study Objectives ............................................................................................................. 23
IV. STUDY METHODS ................................................................................................... 2 5
Data Collection and Resource Characterization ............................................................. 2 5
Growth Model ................................................................................................................. 29

iv

STEMSSS Modifications ................................................................................................ 30

Regeneration ................................................................................................................... 35
Past Trends Management Scenario ................................................................................. 37
V. RESULTS AND DISCUSSION .................................................................................. 49
Findings .......................................................................................................................... 49
Analysis and Future Research ......................................................................................... 56
Limitations ...................................................................................................................... 60
Conclusion ...................................................................................................................... 62
APPENDIX A .................................................................................................................... 64
Distribution of Michigan’s Aspen Resource According to 1993 Survey Data --------------- 64
APPENDIX B .................................................................................................................... 73

Comparisons of 1980 Survey Plots Projected Using the Initial Logistic Regression
Results to be Correctly and Incorrectly Cut as Determined by 1993 Survey

Remeasurements ............................................................................................................. 73
APPENDIX C .................................................................................................................... 79
Study Projection Data for 1993-2023 ............................................................................. 79
LIST OF REFERENCES ................................................................................................... 84

LIST OF TABLES

Table 1: Average DBH differences in inches between remeasured data and projections

including the modified growth and mortality equations in STEMSSS. -------------------- 31
Table 2: Distributions of survival rate differences in proportion of trees. [Mean, Standard
Dev.] (number of observations). ................................................................................ 33
Table 3: Remeasured minus projected volume before and after mortality correction. ----- 33
Table 4: Logistic regression estimates for stand level harvesting. ------------------------------------ 39
Table 5: Logistic regression estimates for tree level harvest. ------------------------------------------- 40
Table 6: Hosmer and Lemeshow Goodness-of-Fit Test ................................................... 41
Table 7: Stand level harvest event classification table for model dataset with cutoff point
of .3. ........................................................................................................................... 42
Table 8: Tree level harvest event classification table for model dataset with cutoff point
of .44. ......................................................................................................................... 42
Table 9: Stand level harvest event classification table for validation dataset with cutoff
point 0f .3. .................................................................................................................. 43
Table 10: Tree level harvest event classification table for validation dataset with cutoff
point of .44. ................................................................................................................ 43
Table 11: Comparison of 1980-1993 projected and measured harvest and stocking
volumes. ..................................................................................................................... 47
Table 12: Comparison of 1980-1993 projected and measured harvest levels by
proportional distribution across ownership and location. ------------------------------------------ 47
Table 13: Averages for projections of Michigan’s aspen resource by timber class, 1993-
2023 [std]. .................................................................................................................. 79

Table 14: Averages for projections of Michigan’s aspen resource,1993-2023[min,max]80
Table 15: Averages for projections of Michigan’s Federal aspen resource by timber class,

1993-2023 [std]. ......................................................................................................... 31
Table 16: Averages for projections of Michigan’s Private aspen resource by timber class,
1993-2023 [std]. ......................................................................................................... 81
Table 17: Averages for projections of Michigan’s State/Local aspen resource by timber
class, 1993-2023 [std]. ............................................................................................... 32
Table 18: Averages for projections of Michigan’s Upper Peninsula aspen resource by
timber class, 1993,2023 [std]. .................................................................................... 32
Table 19: Averages for projections of Michigan’s Lower Peninsula aspen resource by
timber class, 1993,2023 [std]. .................................................................................... 83

vi

LIST OF FIGURES

Figure 1: Aspen sawlog production and receipts in Michigan, 1969-1994. ----------------------- 6
Figure 2: Aspen pulpwood production and receipts in Michigan, 1969-1994. ------------------- 6
Figure 3: Aspen industrial roundwood production in Michigan by product type, 1984-
1994. ............................................................................................................................. 7
Figure 4: Percent of harvested aspen sawtimber used in sawlog or veneer production,
1972,1994. ................................................................................................................... 7
Figure 5: Acreage in aspen covertype by survey and FIA unit, 1935-1993. ------------------------ 9
Figure 6: Aspen volume by survey and F1 A unit, 193 5.1993_ .......................................... 10
Figure 7: Aspen growing stock & non-growing stock volume production (Removals
minus residuals & slash), 1975,1993 ........................................................................ 12
Figure 8: Average aspen pulpwood stumpage price per cord in constant 1982 dollars,
1954,1994. ................................................................................................................. 13
Figure 9: Average aspen sawtimber stumpage price per mbf in constant 1982 dollars,
1954-1994. ................................................................................................................. l 3
Figure 10: Area of aspen covertype on timberland by age-class in MI, 1993. ------------------- 14
Figure 11: Aspen covertype acreage and aspen volume on timberland by ownership. 15
Figure 12: Michigan county aspen growth per acre by ownership class, 1993. ---------------- 27
Figure 13: Michigan county aspen volume distribution by stand covertype, 1993. ---------- 28
Figure 14: Stand and tree harvest flowchart ..................................................................... 44
Figure 15: Comparison of 1980-1993 projected cut plots’ cubic foot volume/acre,
1930. ........................................................................................................................... 46
Figure 16: Michigan’s aspen resource from 1935-2023 based on published inventories
(1935-1993) and study projections (2003,2023). ....................................................... 51
Figure 17: Michigan’s aspen growing stock volume by ownership and peninsula, 1993-
2023 from published 1993 inventory and study projections (2003-2023). ---------------- 51
Figure 18: Michigan’s aspen resource cut and mortality from published reports (1983-
1993) and study projections (1993,2023). ................................................................. 52

Figure 19: Michigan’s aspen harvest volume by ownership and peninsula from published
inventory and timber product output reports (1983-1993) and study projections

(1993,2023). ............................................................................................................... 54
Figure 20: Comparison of Michigan’s 1993 and 2023 aspen covertype age class
distributions from 1993 inventory and study projections for 2023. --------------------------- 56
Figure 2]; 1993 surveyed aspen covet-type plot locations. ............................................... 64
Figure 22: Distribution of aspen cubic foot volume cut, 1980-1993. ------------------------------ 65
Figure 23: Distribution of aspen cubic foot volume cut by ownership, 1980-1993. -------- 66

vii

Figure 24: Distribution of aspen covertype acreage by ownership class, 1993. --------------- 67

Figure 25: Distribution of aspen covertype acreage lost by ownership, 1980-1993. -------- 68
Figure 26: Distribution of standing aspen cubic foot volume, 1993. -------------------------------- 69
Figure 27: Distribution of aspen cubic foot volume by ownership, 1993. ------------------------ 70
Figure 28: Distribution of average aspen cubic foot volume/acre on all acreage with

aspen volume, 1993. ................................................................................................... 71
Figure 29: Distribution of aspen cubic foot volume by stand covertype classification,

1993. ........................................................................................................................... 72
Figure 30: Comparison of 1980-1993 projected cut plots’ aspen square feet/acre,

1980. ........................................................................................................................... 73
Figure 31: Comparison of 1980-1993 projected cut plots’ aspen cubic foot volume,

1930. ........................................................................................................................... 74
Figure 32: Comparison of 1980-1993 projected cut plots’ average dbh, 1980. ---------------- 74
Figure 33: Comparison of 1980-1993 projected cut plots’ cull in square feet/acre,

1930_ ........................................................................................................................... 75
Figure 34: Comparison of 1980-1993 projected cut plots’ aspen mortality in square

feet/acre, 1980. ........................................................................................................... 75
Figure 35: Comparison of 1980-1993 projected cut plots’ mortality in square feet/acre,

1930. ........................................................................................................................... 76
Figure 36: Comparison of 1980-1993 projected cut plots’ number of live trees per acre,

1980. ........................................................................................................................... 76
Figure 37: Comparison of 1980-1993 projected cut plots’ location, 1980. ---------------------- 77
Figure 38: Comparison of 1980-1993 projected cut plots’ ownership class, 1980. ---------- 77

Figure 39: Comparison of 1980-1993 projected cut plots’ live square feet/acre, 1980. ---78

viii

INTRODUCTION

The sustainability of an aspen supply into the twenty-first century has been the
subject of research from before the 19703 to the present. The extensive land clearing
within Michigan in the late 18005 and early 19005 and subsequent desertion of that land
provided an optimum environment for the establishment of an extensive aspen resource
that was becoming mature during this time period. As technology improved, a market for
the resource was eventually created. However, the maintenance of the resource is highly
subject to the management practices applied. As the resource continues to mature and
decline, the question of sustaining current output levels is becoming of increasingly
critical importance. This research projected the resource as of 1993 for thirty years based
on harvesting practices between 1980 and 1993. The goal being to characterize harvest

and stocking volume trends during that time period.

I. THE BIOLOGY OF ASPEN

Quaking aspen (Populus tremuloides Michx.) and Bigtooth aspen (Populus
grandidentata Michx.) comprise an early successional or pioneer forest type commonly
referred to as aspen. Its natural growth—pattem is to overtake areas that have experienced
a large disturbance such as wildfire, tornado or timber harvest. Growth characteristics
vary across its natural range which is transcontinental along the northern border of the
United States and into Canada. The expected lifetime of aspen increases as the average
annual temperature decreases (Perala et. a1. 1995, Shields & Bockheim 1981). In the
Lake States of Michigan, Minnesota and Wisconsin, it is generally considered a fast-
growing and Short-lived species that develops on a variety of sites. Good aspen sites are
characterized by well-drained loamy soil with a good supply of nutrients to support the
fast growing species. While stands in the Rocky Mountains may live for 200 years and
allow a rotation age of 125 years, aspen stands in Michigan are often managed on a 35-50
year rotation and will typically begin to break-up after 60 years.

Breaking-up is a process in which small gaps created by mortality in the stand are
not replaced by new aspen stems (Perala 1991). The shade of older, more mature trees on
the site does not promote successful aspen regeneration. More shade tolerant tree species
invade the stand and outcompete any naturally regenerating aspen stems. Contributing
factors in stand decay include mean annual temperature, depth to water table and

exchangeable potassium (Shields & Bockheim 1981). For modeling purposes, Perala et.

3
a1. (1996) identified the decay triggers to be a threshold stand density of 147 trees per

hectare and a stand age equal to 10 plus the site index.

Natural seed regeneration is not considered a reliable or optimal way to fully
regenerate an aspen stand. The aspen seed is very fragile, short-lived and specific in its
seedbed requirements. However, stand regeneration through root suckers is relatively
inexpensive and highly reliable if a few critical conditions are met.

The critical factors in successful aspen regeneration are the parent stand of aspen,
the site quality, the amount of direct sunlight, the type of harvest activities, and the season
of harvest. Maximum aspen sucker regeneration occurs on well-drained and aerated sites
where at least 50 aspen trees or 20 square feet of basal area per acre were cut (Perala
1977). Root suckers originate from dormant buds kept in check during the lifetime of the
parent tree through chemical suppression called apical dominance. When the parent tree
is cut or top killed, this chemical suppression ceases. Root suckers can initiate growth
underneath an overstory of older trees but cannot continue to grow and mature. “AS little \
as 10 to 15 square feet of basal area of residual overstory will slow sucker growth by 35
to 40 percent” (Perala 1977). Direct sunlight allows aspen to take full advantage of its
capacity for fast growth when invading newly disturbed areas. On open sites, it can
outcompete any later establishing vegetation.

Additional regeneration success factors include the age and health of the parent
stand and the type and season of harvest. As the parent stand gets older and more
diseased, its regenerative ability decreases. A Silvicultural clearcut in which all stems

over 1 inch in diameter at 4.5 feet above the ground(dbh) are removed through whole tree

harvesting will provide an adequate site preparation through soil scarification and

4

vegetation removal. However, excessive soil compaction by harvesting equipment or
extensive disking operations can result in decreased regeneration.

The season of harvest effect is related to the root reserves. Aspen stand
regeneration is at a maximum when harvesting occurs from summer through early Spring.
Harvesting during late spring and early summer decreases initial regeneration capacity.
The available food reserves important for sucker growth have been diminished by the
parent tree’s first growth of the season. A study in Wisconsin Showed that the number of
root suckers after summer harvesting was only 74 to 78 percent of the number of root
suckers after fall and early spring harvesting (Stoeckeler & Macon 1956). Inadequate

regeneration can allow competing vegetation to take over the site.

II. THE ASPEN RESOURCE

The aspen covertype is an important natural resource in the state of Michigan. It
is valued as a commercial timber species, important wildlife habitat, a resilient pioneer
within the ecosystem and an aesthetic part of the landscape. In 1994, there were 154
primary and 148 secondary manufacturers in Michigan using aspen as a raw material

(Bertsch & Weatherspoon 1994). All types of aspen production and the proportion of it

 

kept in the state have increased Since the 19705 as shown in Figures 1 and 2.’ Aspen is
primarily used as pulpwood by the timber industry (Figure 3). Over half of the sawtimber
harvested is either left on the site or used for products other than sawlogs. Figure 4

shows the proportion of harvested aspen sawtimber used for sawlog production. This
could indicate a potential for competition between the pulpwood and sawlog industries as
the resource matures. Aspen provides the timber industry with a short-rotation fiber
supply that can be easily and economically regenerated if the parent stand was healthy,

not overmature, and effectively harvested to promote root suckering.

| Lake States’ timber production estimates by species, product and location are compiled and published
PCYIOdically by the USDA Forest Service, North Central Forest Experiment Station in cooperation with
individual states. The most recent published statistics are presented by May & Pilon (1995) and Piva
(1997). 1994 sawtimber statistics are based on preliminary data at the time of this research.

5

120

 

TSEMTOEPTOBDEién l
1 I Kept ln-State
100 .. L13 Sawmill Reoeipts

   

Volume (mmbf)
8

20.

fir

 

 

 

 

 

 

 

 

 

 

 

 

 

f

1969 1972 1975 1977 1984 1988 1990 1992 1994
Year

Figure 1: Aspen sawlog production and receipts in Michigan, 1969-1994.

 

1200 .

1000
’4? (”i PUIprdddion
g 800 _ l .Kept In-State !
0 I Own Receipts i
a __.i_____.____ __ __- _
C
E 600 i.
o
i
E 400-.
3
>

200 -.

0 - i #1 i 4

 

 

 

 

 

 

 

 

 

 

 

 

1969 1972 1975 1977 1984 1988 1990 1992 1994
Year

Figure 2: Aspen pulpwood production and receipts in Michigan, 1969-1994.

120

 

100 ..

Volume (mmcf)
8

40 47

20.

 

0
1984

 

fiber WOOdl

   
 

{#361638 I

 

1988 1990 1992 1994
Year

Figure 3: Aspen industrial roundwood production in Michigan by product type, 1984-

1994.

70%

 

 

 

 

1972 1977 1988 1990 1992 1994

Figure 4: Percent of harvested aspen sawtimber used in sawlog or veneer production,

1972-1994.

8
Aspen ecosystems also provide habitat to 23 species of mammals, 31 species of

birds, 3 Species of reptiles and 3 species of amphibians (Ottawa N. F. 1995). These
species represent an “important component of the total number of species on the forest”
(Ottawa N. F. 1995). In addition, due to the invasive nature of the aspen Species in
disturbed areas, it helps “stabilize the water regime of streams and lakes” (Brinkman &

Roe 1975).

Current Timber Resource -1

According to the 1993 Michigan forest inventory, there was approximately 3.1
billion cubic feet of live standing aspen timber on 8.4 million acres of timberland
(Leatherberry & Spencer 1996).2 Approximately 5% of the total aspen resource, 173.6
million cubic feet, was standing dead wood. In stands classified as aspen covertype on
timberland, there was 1.8 billion cubic feet of aspen timber on 2.6 million acres. Thus,
approximately 42% of the aspen timber resource is in stands not typed as aspen.

The 1935 forest inventory reported 5 million acres of aspen-birch covertype and
12 million cords, 948 million cubic feet, of aspen growing stock. The 1955 inventory
reported 4.8 million acres of aspen-birch covertype and 22 million cords, 1.8 billion cubic
feet, of aspen growing stock. The 1966 inventory reported 4.7 million acres of aspen-
birch covertype, of which 4.3 million acres were aspen, and 28.6 million cords, 2.2 billion
cubic feet, of aspen growing stock. Thus, for thirty years the aspen acreage remained

fairly constant while the aspen stems matured. However, from 1966 to 1980, the aspen

2 Actual data were summarized from the inventory database developed and maintained by the USDA Forest
Service Forest Inventory & Analysis Unit, St. Paul, MN. Survey procedures and data collection are
outlined in Field Manual Michigan (1991).

9
covertype declined by 20% (Potter-Witter & Ramm 1992). The 1980 inventory reported

3.4 million acres of aspen covertype and 32.9 million cords, 2.6 billion cubic feet, of
aspen growing stock. The 1993 inventory reported 2.6 million acres of aspen covertype
and 40.5 million cords, 3.2 billion cubic feet, of aspen growing stock. This is a net
decline of 400 thousand acres, approximately 13% of the 1980 resource. This trend
represents the invasion of lands cutover or burned at the turn of this century in Michigan
by aspen, and its maturation and successional nature given a decreased level of

disturbance. Figures 5 and 6 display this trend by survey unit.3

 

 

 

 

 

 

 

 

 

 

 

3.000
—— " .‘WTUP
2.500 .5. up
‘ DN. LP ‘
_ ‘ .8. LP 1
2.000 __
§ _
<
'2 1.500 1.
g ___
O
.C
I-
1.000 1.
500 ..
O .

 

 

 

1935 1955 1966 1980 1993

Figure 5: Acreage in aspen covertype by survey and FIA unit, 193 5-1993.

3 The 1935 and 1955 surveys represent acreage in the aspen-birch covertype. The other surveys represent
acreage in only the aspen covertype. Lake States’ forest statistics are compiled and published periodically
by the USDA Forest Service North Central Forest Experiment Station in cooperation with individual states.
The most recent statistics for Michigan are presented in Leatherberry & Specner (1996).

10

 

 

 

 

 

 

 

 

 

25.000
PivvfuP ;
l ‘ —‘
.E. UP
20.000 ] 0N“,
! .3. LP
a __
2 15.000 .
O
o —1
t
c
N
0
3
.1: 10.000.
1-
5.000
O 1

 

 

 

1935 1955 1966 1980 1993

Figure 6: Aspen volume by survey and FIA unit, 1935-1993.

Management Concerns

The Silvicultural system used to manage an aspen stand is one of the most
important factors in sustaining aspen as a resource. Success in regenerating a vigorous
aspen stand is highest with even-aged management practices. With uneven-aged
practices such as cutting aspen out of a stand composed predominately of other species
limits the opportunity for dense aspen regeneration (Ottawa N. F. 1995). Forty-four
percent of the annual net aspen harvest from 1980-1993 occurred in stands not classed as
aspen covertype in 1993. These stands supplied 32% of the annual net aspen growth
between 1980 and 1993 on stands that incurred harvesting. Of the 956 thousand acres not

classed as aspen in 1980 but that did have aspen stems harvested between 1980 and 1993,

11

only 26% or 248 thousand acres were classed as aspen in 1993. Practices such as this
contribute to further aspen resource declines.

Another potential loss of aspen occurs in cutting aspen too late in its life cycle or
after it has been grossly infected with diseases such as white trunk rot. Overrnaturity and
decreased vigor and nutritional reserves inhibit the root suckering needed to fully restock
the stand with aspen. “Aspen stands begin to deteriorate rapidly when they reach 50 to 60
years of age” (Ottawa N. F. 1995). When disturbance does not occur at an appropriate
time, age or scale, aspen will be replaced by later successional covertypes such as shade-
tolerant conifers and northern hardwoods.

The Michigan forest inventory indicates that over 600,000 acres lost classification
as aspen covertype between 1980 and 1993. Across ownerships, only 4% was converted
by planting or seeding, most on public lands. The remaining 96% is classified as natural
conversion. Forty-four percent of the converted lands is now predominately maple-

beech-birch, a late successional covertype.

Demand Concerns

Overutilization is a concern reflected by a 1.02 aspen growth to removal ratio in
1992. The net annual aspen growth was 111.5 million cubic feet while timber production
output was 108.9 million cubic feet (Leatherberry & Spencer 1996). “Removals did
exceed projected growth for the northern Lower Peninsula for the 7-year period between
1980-1987” (Potter-Witter & Ramm 1992). Figure 7 shows an increasing trend in aspen

removals.

I
i

12
120

 

110 7

100

Volume (mmcf)

50

 

 

 

1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995

Year

Figure 7: Aspen growing stock & non-growing stock volume production (Removals
minus residuals & slash), 1975-1995.4

The price per cord for aspen has increased sharply from 1987 to 1994. A 1995
USFS aspen ecosystem report states that the stumpage price in constant 1994 dollars has
gone from $3.76 to $20.48 per cord. Figures 8 and 9 Show this increasing price trend
based on stumpage price data collected on Michigan State Forest sales and data reported
by TimberMart North. This increase reflects both a rising demand for the resource and

uncertainty about its future supply from national forests (Ottawa N. F. 1995).

’ Growing stock volume is the central stem volume ofpoletimber and sawtimber size trees. Non-growing
stock volume is limbwood, saplings, cull and dead volumes. Logging residual ismerchantable volume not
used for roundwood production. Logging slash is the unmerchantable volume not used.

20

 

f

18.

16

14.

121

10-

1 982 DollanlCord

 

 

 

Figure 8: Average aspen pulpwood stumpage price per cord in constant 1982 dollars,
1954-1994.

 

70

1982 Dollars/mbf

20.

10.

 

 

0 + , i g . .
54 59 64 69 74 79 84 89 94

 

Figure 9: Average aspen sawtimber stumpage price per mbf in constant 1982 dollars,
1954-1994.

1 4
Structure Concerns

Another concern about the aspen resource is an imbalance in the age classes. A
supply shortage is predicted when the current 35 year age-class aspen matures. This
shortage is related to the imbalanced aspen age-class structure shown in Figure 10.
Relating to this, Potter-Witter and Ramm 1992 stated that under current conditions, the
price for aspen will continue to rise steadily until 2000 and then it will rise
“dramatically.” However, because much of the aspen resource is in mixed species and
multi-story stands, the stand age collected during the inventory may not be representative

of the aspen resource.

600

 

500 ._

A
8

Acres (10008)

200 7

100-.

 

 

5 15 25 35 45 55 65 75 85 95 100
10-Year Age Class

Figure 10: Area of aspen covertype on timberland by age-class in MI, 1993.

The younger age classes are presently more abundant than the current mature age classes.
If these lands aren’t converted or lost, early harvesting may supplement a shortage if they

are productive enough.

l 5
Ownership Concerns

A critical factor in future availability of an aspen timber resource is ownership.
Landowners have a variety of management goals and take a variety of actions to achieve
them. These management actions may not be directed at providing for a future aspen
resource. Therefore, the aspen resource distribution across ownerships should be
considered in projections. Currently, 57% of Michigan’s timberland acreage in aspen
type, 1.5 million acres, is held by non-industrial private landowners. Forest industry
owns only 5% of the current aspen timberland acreage, 123 thousand acres, in Michigan.
Approximately 38% of the aspen timberland acreage, 990 thousand acres, is in public
ownership. As seen in Figure 11, the privately held lands have a greater amount of aspen

in aspen typed stands that are either better stocked with aspen or more mature.

1600 900

 

14001- 'H-Akcrre'agefi T " ” I .- 800
‘ .Average Volume/Acre 1

, 700

1 600
1000 1
. 500

800 ,
400

Thousand Acres
Cubic Feet/Acre

600 -
300

400
200

200 1 100

 

 

 

 

 

Federal State Private

Figure 11: Aspen covertype acreage and aspen volume on timberland by ownership.

 

16

Management goals may place wildlife management, old-growth preservation,
aesthetics and recreation above maintaining aspen as a timber resource. Additionally, the
value of the aspen as a timber resource may not be recognized due to historically lower
aspen prices. It may take time for the market supply to adjust to the increasing demand
level and for an equilibrium point to be established. Considering these factors, there is a
potential for approximately 95% of the current aspen resource not to receive the specific
management practices needed to sustain it. Individual ownerships should be aware of the
contribution they can have in their area. A landscape view of what and where the future
aspen resource will be can facilitate stand management decisions and is important in

making public forest planning responsive to all.

III. BACKGROUND

There has been increasing concern within the forest products industry and other
organizations about maintaining an aspen timber resource and sustaining harvest levels.
In the past, this need for timber supply projections has resulted in studies that project
available resource data under developed assumptions or identify and analyze changes in

the resource over time.

Past Projections

Leuschner (1972) projected the Lake State’s aspen resource thirty years using the
forest inventory data collected in the 19605 and updated to a base year, 1968. His model
incorporated changes in commercial forest area, cut allocations and shifts, growth and in-
growth in three regions. These regions were northern Wisconsin and Michigan’s upper
peninsula, Michigan’s northern lower peninsula, and northern Minnesota. Growth was
based on trends in net annual growth. Total cut was an estimate of industry demand
adjusted for availability. Each survey unit was allocated a portion of the total cut based
on past production.

Projection assumptions included past resource trends, loss due to overmaturity
and succession, and positive, pro-active management practices. The trend assumption
was based on the previous inventory cycle. Overmaturity and succession were modeled
by controlling resource availability and in-growth. Positive practices were modeled by

controlling harvest type, allocation and in-growth.

17

 

18
This study is an update of Leuschner’s aspen projections based on recent trends.

Data for the past two survey cycles are available in electronic format which facilitates the
modeling process. This projection focuses on the entire state of Michigan. As Leuschner
did, the recent trends assumption is based on an inventory cycle, 1980 to 1993.

Leuschner projected that the cut would diminish in Michigan’s upper peninsula by
1990 or by 1995 if industry accepted smaller material. By 1998, the cut was projected to
be coming from the 6-8 inch dbh classes. In 1990, the projected cut for the Upper
Peninsula totaled 67.2 million cubic feet—26.9 million cubic feet from the eastern half
and 40.3 million cubic feet from the western half. The actual 1992 harvest levels were
16.8 and 31.7 million cubic feet respectively. The overestimate could be related to
industry increasing their proportional use of other Species. Leuschner projected a
continuous decline in growing stock throughout the projection but more sharply in the
western half than in the eastern half. The projected growing stock levels were 340
million cubic feet in the eastern half and 350 million cubic feet in the western half of the
Upper Peninsula. The growing stock volumes in 1992 were 469 and 803 million cubic
feet respectively.

In the northern Lower Peninsula, the projected cut increased until after 1995 and
was projected to come from the 6-8 inch dbh classes. A sharp downward trend in
growing stock was projected due to a perceived lower level of growth capacity. The
projected cut and growing stock level for the northern Lower Peninsula in 1990 was 59.5

and 550 million cubic feet respectively. In 1992, the actual cut level was 53.6 million

cubic feet and growing stock was at 1.7 billion cubic feet. This large discrepancy may be

19
related to the assumption of lower growth levels in the northern Lower Peninsula, lower

levels of harvesting than assumed, or the structure of the analysis regions.

Leuschner projected that most of the available resource in 2000 would be in the
12 inch dbh and smaller classes. In fact, 97% or more of northern Michigan’s aspen
stems in 1992 were in this range. Across Leuschner’s projection assumptions, the
projected total cut ranged from 104 to 127 million cubic feet for Michigan in 1990. In
1992, aspen production was at 102.7 million cubic feet. Leuschner’s projected growing
stock in 1990 ranged from 730 to 1,270 million cubic feet. In 1992, aspen growing stock

was close to 3 billion cubic feet according to the inventory data.

Recent Projections

Liggett and Leefers (1990) modeled Michigan’s aspen resource from the 1980
forest inventory data and projected a significant decline in the aspen resource over the
next 50 years due to its early successional nature. They analyzed current management
plans and recent trends in the resource to develop their model assumptions. The National
Forest management scenario was derived from the 1986 forest plans. The State Forest
management regime was extrapolated from draft plans for two of the forests. They cited
a 1985 study by Carpenter and Hanson that estimated 23% of non-industrial private
landowners would harvest in the next 10 years, and 21% would never harvest. Using
these numbers, they modeled the area of the resource that would likely be available for
harvest.

Liggett and Leefers also stated that natural succession should be modeled and is

dependent on the “extent and tolerance of other species.” They referred to a study by

20
Shields and Bockheim (1981) which reported 37% mortality in declining asan stands

with a mean age of 54. The Hiawatha National Forest Plan assumed aspen would convert
to other forest types at age 90 unless harvested. Liggett and Leefers assumed Significant
deterioration at ages greater than 70--aspen regeneration was not assured when these
stands were harvested, and natural succession at age 90.

Liggett and Leefers used F ORSOM to project the managed resource. This model
requires that expected yields for the projection period be entered by the user. The total
aspen acreage in the year 2030 was projected to range from 2.7 million acres to 2.4
million acres depending on harvesting intensities on private land. This is down 634 to
924 thousand acres from the 3.4 million acres reported in the 1980 forest inventory.
According to the 1993 inventory data, this benchmark has already been reached with 2.6
million acres reported. The authors felt that decreased logging and interference in natural
fire cycles has opened the door to a continuing loss of acreage for the pioneering aspen
species.

A Generic Environmental Impact Statement (GEIS) was done in 1992 on
maintaining productivity and the forest resource base for Minnesota (J aakko Poyry
Consulting, Inc. 1992). The study was based on the 1990 forest inventory data collected
in Minnesota. The scope of the GEIS is more expansive and all-encompassing than this
proposed study and applies only to Minnesota. However, the techniques used to model
the resource provide a good framework.

The GEIS used the GROW program--a scaled-down version of STEMS much

more dependent on user programming. The GEIS formulated three harvesting levels: a

base level related to current demand, a medium level related to increased demand by

21

proposed industry expansion and a high level which took advantage of all available but
unharvested land in the medium harvest scenario.

Management scenarios used a treatment option model and a harvest scheduling
model. The harvest model projected the resource for 50 years with the goal of Optimizing
the marginal cost of production. Prices included harvesting costs based on quality,
stocking and distance from designated market centers. Optimal solutions were found for
each harvest target level.

The GEIS modeled regeneration by adding a typical tree list into the growth
model 15 years after harvest. This tree list was derived from FIA data for regenerated
stands ranging in age from 10 to 20 years. These stands were screened from the database
based on age, visual verification, commercial forest use and seedling/sapling
classification. Stands were aggregated by site quality, dbh and crown ratio. More precise
models were not found and researchers such as Belli and Bk (1988) suggest a precise
model “is still very distant.” All trees greater than l-inch dbh on the FIA plots were
included to provide for in-growth.

The study analysts assumed that lower competition levels would favor lower
mortality and larger trees and compensate for lack of in-growth in the model. Studies
showed that over 50 years, the difference between projected and measured aspen stands
could range from -61 to -51 trees per acre, -15 to -5 square feet per acre, and -l .2 to 1.4
cubic feet per acre for aspen less than 5 inches dbh.

The GEIS authors felt STEMS85 construction underrepresented Minnesota

stands, didn’t account for catastrophic losses and overestimated growth. Concerns

relevant in Michigan as well. The authors studied stands undisturbed by human actions

22
to try to correct for catastrophic events but did not find a universal solution. The authors

were also concerned that some F IA stand data may actually represent two different stands,
but no solution was found to correct this problem.

In the first model run, ownership was not a consideration in projecting the
resource. The second runs included updates on growth and yield coefficients, changes in
forest area, changes in covertype, and timberland availability based on ownership
constraints. Other factors affecting timber production included policy, silviculture,
rotation length, management alternatives, old-growth, best management practices, buffers
and sensitive areas. Allowable cut for long-term sustained yields was modeled using the
ACES (Rose 1992) model and the 1990 and projected 2040 age class distribution. The
growth mOdel in ACES is very conservative so the allowable cuts derived were
considered to represent the lower bounds.

In the first model run, all three harvest levels were met in all market centers. The
aspen covertype experienced the heaviest level of cutting and represented 40-60% of the
harvested area across the state. Rotation age ranged from 40 to 90 years. High increases
in demand could only be met if privately held aspen was available. Long-term trends
suggested a “slight decline” in aspen acreage over the projection period. It also showed a
short-term deficit in 20-30 years because of an imbalanced age class distribution—a
similar concern in Michigan.

The second runs showed infeasibilities in meeting target harvest levels for species
such as aspen due to constraints such as longer rotation lengths and restrictions on

clearcutting. The only solution for the GEIS was to lower the base target level for aspen

23

or to relax some of the constraints. The rising marginal price for aspen confirmed
concerns for its future availability.

Potter-Witter and Ramm (1992) analyzed the aspen supply in Michigan based on
forest inventory data, Michigan Department of Natural Resources data, stumpage prices
and industry data. Through comparison and trend analysis of the data from past decades,
they reported a declining aspen acreage. They referred to a study by Spencer et al.(1990)
that looked at the aspen resource from the mid 1960’s to 1987 for the Lake States. It
reported that Michigan had the greatest decline in aspen acreage, 19.9%. Annual growth
also declined by 6.7 million cubic feet from 1980-1986. Scarcity is reflected in the 4-5%
real rate of increase for stumpage prices between 1980 and 1987. Potter-Witter and
Ramm projected that by the year 2000, there would be only 143,000 acres or less of
mature aspen.

Einspahr and Wyckoff (1990) found shortages in the 1 1-20, 21-30 and 31-40
aspen age classes with large volumes of overrnature aspen. They predicted shortages in

2000-2020 related to increased demand and age class imbalances.

Study Objectives

The timber industry in Michigan currently relies on the future availability of the
aspen timber resource for pulp, manufactured panel and saw mills. While industry is
capable of changing to other timber species to supply mills, this can be costly. It would
also require the existence of an alternative supply equal to aspen’s growth potential and
economically efficient regenerative capacity. Industry needs reliable estimates of future

timber supplies to plan manufacturing operations. A decrease in species diversity and

24
current resource status due to changing harvest practices and attitudes about timber

harvest is also of interest to ecologists, wildlife and biodiversity proponents.

The objective of this study is to project the current Michigan aspen resource for
thirty years. This projection is based on past harvesting trends developed from an
analysis of the 1980 and 1993 remeasured forest inventory plots. The goal is to describe
the potential aspen resource distribution across the state of Michigan as characterized by
ownership, age class, standing timber volume, and harvest yields assuming current
management trends continue.

Timber production is not the only consideration related to the outputs from an
aspen forest ecosystem. However, it is beyond the scope of this study to analyze the
interrelationships between the various market and non-market forest values and the
tradeoffs for the aspen timber resource. This study is not a prediction. There are many
assumptions inherent in this type of projection. It is meant to present a generalized base
description of one potential future resource status. From which, one can decide upon
alternative management strategies related to how they might wish to alter this potential

result.

IV. STUDY METHODS
Data Collection and Resource Characterization

The initial step was the compilation of aspen resource data from the 1980 and
1993 forest inventories. These forest inventories were conducted by the USDA Forest
Service, North Central Forest Experiment Station, Forest Inventory & Analysis Unit
(FIA). A description of the survey procedures used in the inventory can be found in
Hansen et. a1. (1992) and the Field Manual Michigan (1992). The inventory data were
screened for plots containing any stems of aspen qualifying as live growing stock and
greater than or equal to 1 inch dbh. The procedures used for determining stocking and
covertype on the plots can be found in Hansen & Hahn (1992).

Individual FIA survey plots were mapped, and areas related to ownership,
stocking and harvest levels were delineated in each county. This characterization and
mapping provides a good overview of the current resource, possibly identifying areas of
highest impact due to changes in the resource status, and providing a good base level
from which to work. Figures 12 and 13 Show that the greatest aspen growth per acre
occurs in counties where the majority of the aspen resource is in stands typed as aspen.
They also show that this growth is occurring across all ownerships. Additional maps are
in Appendix A. The size of the circles on the maps are representative of the proportion of
the statewide resource within each county. The subdivision of these circles represent

proportions of the county resource. Depictions of the survey data at the county level are

25

26
subject to high levels of error. However, they are useful in generalizing resource

characteristics.

The individual tree records on these plots were used as input into the projection
model. Each survey plot represented a separate stand with its own characteristics. Each
measured tree on the plot has an expansion factor to a trees per acre basis. In addition,
each plot has an expansion factor to acres of land within the respective county. A
computer program was developed to facilitate the organization of the survey data into
properly formatted tree lists with appropriate stand information header lines.

Data from plots measured in 1980 and remeasured in 1993 were used to develop
the past trends management scenario. The selected plots had measured aspen stems in
1980, were classified as timberland in 1993, and did not change ownership between 1980
and 1993. Ownership was classified as federal, state or private. Private ownership
includes industrial, non-industrial and Indian stands. These criteria allowed an analysis
of available timber lands, with an aspen component, that included ownership and stocking
as possible factors in predicting future management actions.

There were 3,730 remeasured plots. Four hundred and thirty-six plots were
deleted due to a change in ownership. Of the remaining plots, 1,547 plots had at least one
live and measured aspen stem. Those 1,547 plots had 66,787 measured trees in the 1980
survey, 9,350 of which were aspen. There were 89 plots with evidence of harvest activity
prior to the 1980 survey. Between the 1980 and 1993 survey, there were 30,673
remeasured trees and 481 plots with evidence of cutting. On those 481 plots, 10,770 trees

were measured in 1993, 4,929 of which were cut.

27

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Federal

 

Aspen Growth/Acre by Ownership

e'~'-.I.h.1' o
I‘I."U.Olc"
2:33;. State

 

 

l

 

Figure 12: Michigan county aspen grth per acre by ownership class, 1993.

28

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Volume Distribution by Covertype
- Other Covertype
I: Aspen Covertype

 

Aspen

l

 

 

 

 

 

 

 

 

Figure 13: Michigan county aspen volume distribution by stand covertype, 1993.

29
The 1993 survey data to be projected was collected from stands with at least one

live, measurable aspen stem with a dbh of at least 1 inch. The stand information required
for the projections includes the state, plot name, forest type, stand age, site index species,
stand site index, survey date, stand origin (planted/natural), plot area expansion factor,
FIA unit in Michigan, distance from nearest road in chains, and ownership group. The
individual tree information includes species, dbh, crown ratio, tree expansion factor to a
per acre basis, tree site index, tree history of 1 if alive, 2 if cut, and 3 if dead, and tree
classification as acceptable, rough or rotten. All tree site indexes were entered into
STEMS85 as 0. This prompts STEMS85 to calculate the individual tree site index based

on the stand site index.

Growth Model

The Stand and Tree Evaluation and Modeling System or STEMS85 was used for
the projections (Brand et. al. 1988). This is an individual-tree, distance-independent
growth and yield model developed for the Lake States. Growth and mortality were
projected on a yearly basis for 30 years with potential management actions applied at 10
year intervals. STEMS85 allows the user to modify management decision trees or select
management options, perform batch or interactive processing, modify output reports, and
provides a regeneration algorithm for clearcuts. STEMS85 incorporates a diameter
adjustment to the growth component and a modified mortality model as compared to the
original STEMS (Holdaway & Brand 1986). STEMS85 was originally in FORTRAN

Computing language. The program was compiled into an MS-DOS executable

30
application with Microsoft Fortran Powerstation Version 4.0. This facilitated easier

alteration to the programming code and more control over the program elements.

A model validation was done by Holdaway & Brand (1986). Mean dbh errors
over a 10 year period showed an average overprediction of 0.09 inches for bigtooth and
quaking aspen. Mean number of trees error for aspen was +5 trees/acre. Mean basal area

error for aspen was +4 sq.ft./acre. Caution is recommended in utilizing the model beyond

30 years (Holdaway & Brand 1986).

S TEMS85 Modifications

Extensive modifications were made to STEMS85 to facilitate this study. The
Lake States’ growth and mortality equations were modified; the individual tree survival
rates were modified; the harvest methods were modified; the model outputs were
modified.

Michigan specific correction factors were added to modify the diameter growth
and mortality equations. These correction factors are applied according to tree species,
dbh and location within either the upper or lower peninsula of Michigan. They were
developed by the FIA Unit of the North Central Forest Experiment Station using data
collected on remeasured plots in the 1993 survey. The correction factors were calculated
based on the procedure outlined by Smith (1983). The diameter growth factors are
multiplicative coefficients that are used to modify the annual growth projected by
STEMS85. The mortality factors are multiplicative coefficients that are used to modify

the probability that a tree will die over the course of a year.

31
There were 1,635 trees remeasured on plots that were classified as undisturbed

between 1980 and 1993 with a remeasurement period of 9 to 10 years. After projecting
the 1980 tree data using the corrective growth and mortality coefficients, the projected
and remeasured survey data were compared. When all remeasured trees were considered,
the average difference between remeasured and projected diameters for some species such
as quaking aspen were high. However, there were 1,387 trees that survived the
remeasurement period. The average differences between projected and remeasured
diameters for these trees are more accurate, especially for aspen, as shown in Table 1.
Thus, the corrected mortality equations still did not fit well and were introducing a high

degree of error into the projections.

Table 1: Average DBH differences in inches between remeasured data and projections
including the modified growth and mortality equations in STEMS85.

SPECIES Number of Trees AVG. DBH DIF (in.) AVG. DBH DIF (in.)
Observed-Pro ected 0

ALL 1635 -.20 -.10
Jack ' 333 -.29 -.13
Balsam fir 35 -.23 .01
Red ine 289 -.24 -.22
E. white ' 36 .19 .25
N. white cedar 175 -.03 -.03

Red le 316 -.10 -.09
S le 318 -.08 -.08
375 -.10 -.01

Bi th 161 -.08 .03
' 130 -.40 .01
White oak 26 .09 .09
N. red oak 90 -.01 .03
Black oak -.44 -.40

 

32
Additional mortality correction factors were developed. For each species on all

remeasured and undisturbed plots, the annual survival rate was calculated for both the
remeasured 1993 data and the data projections of the 1980 survey. These calculations
were based on the methods outlined in Buchman (1983). The survival rates were
calculated for 4 dbh classes. The classes were Class 1: <5 inch, Class 2: >49 and <10
inch, Class 3 :>9.9 and <15 inch, and Class 4 :>14.9 inch dbh. The mean and standard
deviation of the difference between the projected survival rate and the remeasured
survival rate were determined for each species and dbh class with at least 50 observations.
The means and standard deviations are shown in Table 2. The number of trees on which
the distributions are based is shown in parentheses. Correction factors randomly selected
from a normal dataset based on these distributions were applied to the appropriate trees in
STEMSSS according to species and dbh class.

After the annual probability of mortality is calculated by STEMS85 for each tree,
two random variates are drawn from a uniform distribution (0,1). These variates are then
used to calculate a normal deviate using the Box-Muller transformation. The deviate is
then transformed to be representative of a distribution with a mean and variance equal to
that above for a specific species and dbh class combination. This value modifies the
annual probability of mortality for the tree. With this correction in place, the projected
volumes become more precise and accurate as compared to the remeasured volumes.

This is shown in Table 3 with the difference representing remeasured volume minus

projected volume.

33

Table 2: Distributions of survival rate differences in proportion of trees. [Mean, Standard
Dev.] (number of observations).

 

SPECIES

CLASS 1

CLASS 2

CLASS 3

CLASS 4

 

lack pine

[a026,0131(294)

1101,0051(845)

[n007,011(159)

 

Red pine

[,004,0041(153)

[1001.000031(7i3)

1-.002,.9E-g (204)

[2001,00011(105)

 

White pine

[a006,003](54)

[1005,00021(119)

[1003.00031(127)

[o009,0051(231)

 

White spruce

L025,071](§2)

[n012,0101(i37)

[2010,018](114)

[-.015,.0311 (62L

 

Balsam fir

[4019,013](678)

[a035,016](782)

[-.068,.0291(136)

 

Black spruce

I006,0161(231)

_L009,019](481)

 

N. white cedar

[.13E-3,.O20] (532)

[1003,002](1529)

[1004,0021(533)

1005,0131(105)

 

Hemlock

L0001,005](46)

[4003,00071(133)

110031001](208)

[10004.0021(179)

 

Black-Green ash

[1007,0121(307)

i002,014](311)

1007,015](120)

 

Red maple

1040,0561(710)

(0009,0051(1201)

10005,003](§34)

[,003,0011(336)

 

Elm

L014,040](200)

[1084,0031(201)

[1144,004](84)

 

Yellow birch

[1011,013](62)

[1005,009](239)

[a002,0021(i99)

L005.025](234)

 

Basswood

1057,0681(58)

[1004.0021(255)

[1004,00061(190)

[1012,0021(77)

 

Hard maple

1003,0201(718)

10002,0011(9s§)

[2001V001](484)

[1003,0007](326)

 

White ash

1046,060](71)

[1005,0041(150)

[a006,003](84)

1024,07i](79)

 

White oak

[a026r0009](86)

[a004,00011(219)

[a002,0001](185)

[~.OOO4,2E-6] Q93)

 

N. red oak

1156,101(333)

1008,0171(357)

[a0005,0011(224)

1002,0031(224)

 

Other red oak

L014,0321(222)

10003,0011(199)

L008,028](121)

 

Bigtooth aspen

[.ll,.126] (183)

[;006,004](322)

[4004,003](264)g,

U1005,005](74)

 

Quaking aspen

[100910451(530)

[5023,008](79l)

[a016v015](646)

[3011,012](l89)

 

 

Paper birch

 

[1044,0051(111)

 

[4017,003](461)

 

[1017,005](223)

 

 

Table 3: Remeasured minus projected volume before and after mortality correction.

 

 

 

 

 

 

 

Volume Difference Before Volume Difference After
Total Growing Stock -1% +4%
Total Growing Stock-Aspen -9% +6%
Total Aspen Sawtimber -4% +9%
Total Aspen Pulpwood -13% +5%

 

 

 

 

The volume equations used in STEMSSS are based on data collected in the upper

peninsula of Michigan as described in Raile et. al. (1982). However, the equations used

to calculate volumes in the forest inventory database are based on data collected across

the Lake States as described in Hahn (1984). Additionally, there are three modifications

34

incorporated into these equations by the forest inventory unit. If stand basal area is less
than 50 square feet per acre, stand basal area is set equal to 50 square feet per acre. If the
tree site index is less than 20 feet or greater than 120 feet, the tree site index is set equal
to 50 feet. Finally, if the current basal area is less than the previously measured basal
area, the stand basal area used in the volume equations is set equal to the greater of the
two (Miles 1997). Calculation of the total cubic foot volume for the 1993 inventory trees
shows an average 10% difference between the results of these two sets of volume
equations. Consideration should be given to this difference when comparing base
STEMS85 projected volumes and remeasured volumes from the survey data.

STEMS85 was modified to allow the input of additional stand information
variables such as plot location in the upper or lower peninsula of Michigan, distance of
the plot from the nearest road, stand expansion factor, stand ownership classification and
history of previous cutting on the stand. The original random number generator relied on
the user to specify a new seed value for each run. This was replaced with a seed value
generator tied into the computer’s internal time clock. For each stand projection, a new
set of uniform random numbers between 0 and 1 were generated. These random numbers
were used in correcting the annual probability of mortality for each tree, in determining
probabilistic mortality, and in determining stand harvesting during the projection.

Mortality can be modeled in a deterministic or probabilistic nature. For large
datasets, the differences should be minimized. For these projections, mortality was
probabilistic. Annual mortality was determined by comparing a random number and
calculated annual probability of mortality in STEMSSS. If the calculated probability is

greater than the random number, the measured tree and all those it represents are killed

35

during that year’s Simulation. The use of random numbers in determining stand
harvesting is described in a later section.

The STEMS85 batchrun capability was modified so that the user can set the
number of simulations a group of stands will go through. The subsequent output values
can be averaged, and the effects of the stochastic elements in the projection can be
determined by examining the distribution of the outputs.

The program output of STEMS85 was also modified in order to facilitate its use.
This required the addition of tracking and summary variables in the program code. A
subroutine to calculate mortality volume was also added.

STEMS85 requires several assumptions: no change in tree quality such as cull
status over the length of the projection; no undetected disturbance in the survey plots
affects the projection; no catastrophic events or insect or disease outbreaks occur; high
variability occurs in the projection of seedling/sapling size stands. Management activities
occurred in the middle of the projection cycle so that variations in growth and mortality

from cutting stands early in the cycle or late in the cycle would be offset.

Regeneration

The possibility that incorporation of a more accurate representation of those 1993
stands that were regenerating aspen would enhance the declining values was explored.
The regenerating stands were identified as 1993 stands less than 20 years old with an
overstory of less than 40 square feet/acre and either greater than 20 square feet/acre of
aspen basal area cut between 1980 and 1993 or seedling sized aspen less than 1 inch dbh

recorded on the plot. For the first decade, all stands were projected with the modified

36
STEMS85 based on the plot data taken from the 1993 inventory. Aspen regeneration

stocking levels were generated for the regenerating stands using the equations described
in Ek and Brodie (1975).

For those stands that met the minimum aspen harvest requirement, the stocking
level was calculated for a stand age of two and then projected for an additional eight
years. For the sucker/seedling stands, each recorded seedling size aspen was given a plot
expansion factor of 30 based on ten 30ft. radius plots. The sum of these stems was set as
the age two stocking and projected for eight years using the Ek and Brodie equations.

Afier the first decade of the modified STEMS85 projections, the Ek and Brodie
regeneration was added to the respective stand tree lists. These individual tree entries
were determined by the application of the integral of Gaussian probability function from
Bevington (1969) to the number of seedlings and their average dbh from the regeneration
projections. This process develops a tree list based on a normal distribution. Each
regenerating stand was given an age of 10. The compiled tree lists were then projected
for the remaining twenty years using the modified STEMS85. Because the regeneration
was added to tree lists that had already been projected for 10 years, it was not possible to
model mortality probabilistically. Therefore, the tree lists at the end of the first 10 years
were the result of a single simulation with deterministic mortality. It is possible that this
approach biased the status of these regenerating stands at 10 years. However, there were
only 202 regenerating plots identified, and so any bias should have minimal impact on the

overall projections.

3 7
Past Trends Management Scenario

Equations to calculate the probability that a stand and then a single tree within that
stand will be harvested were developed using the logistic regression routine in SAS.
Logistic regression models the association between a binary dependent variable and a set
of binary and/or continuous independent variables. Because of the binary explanatory
variables, discriminatory analysis could not be used. Logistic regression analysis
provides coefficients for the independent variables to calculate the probability of an
event, designated by the binary variable. This type of analysis has been used in predicting
timber harvest and supply and postfire mortality (Ryan & Reinhardt 1988, Jamnick &
Beckett 1988, Connaughton & Campbell 1991, Bell & Eriksson 1991). The regression
results are easy to apply and not biased by the variables’ distributions.

For this study, a two-fold approach to harvesting was used. The first logistic
regression was to determine the probability that a stand will incur some degree of
harvesting over a 10 year period. The second logistic regression was to determine the
probability that an individual tree would be cut. Although the surveys are classified as
1980 and 1993, the remeasurement period for the stands actually ranged from 6 to 17
years. The stands were not adjusted to one base year. Therefore, the logistic regression
results represent the probability of harvest for a varying cycle length. Most of the
remeasurement periods were between 9 and 13 years. So, for these projections, the
logistic regression results will be used to determine the probability of harvest for a 10-
year interval.

The independent variables considered for this regression included ownership,

cubic foot and board foot volume (stand, aspen, hardwood, softwood, cull), trees per acre

38
(live, greater than 5 inch dbh), basal area per acre (live, cull, dead, aspen), age, Site index,

stand distance from nearest road, location in Michigan, and past history of harvesting.
These variables were selected because they are indicative of the stand’s life stage,
potential growth, timber value, availability, and possible management intentions and
possible conflicts. Correlation matrices were developed so regression results would not
be biased by the inclusion of highly correlated variables. If two variables had a Pearson
Correlation Coefiicient greater than .4, only one was included in the regression.

Stand variable information was taken from the 1980 survey data for plots
remeasured in the 1993 survey. The history of cut was a binary variable of 0 if the 1980
plot record indicated no harvested trees, and 1 if it did. The ownership factor was
included with two binary variables. If the ownership was federal, the first variable was 1
and the second variable was 0. If the ownership was state or local government, the
opposite applied. If the ownership was private, both variables were 0. The 1993 plot data
were used to determine the value of the dependent variable. If any remeasured tree on the
plot was harvested, the dependent harvest variable was 1. If not, the variable was 0.

The total dataset was subdivided into two subdatasets. Afier randomly ordering
the observations, approximately the first two-thirds were put into a model dataset. The
remaining observations were put into a validation dataset. The model dataset was used in
the regression. An alpha-value of 0.1 was used to determine the significance of each
variable included in the regression.

The variables that were significant in the regression are shown in Table 4. Own]
designates federal ownership, Unit designates the upper peninsula of Michigan, DRFT is

the distance from the road in chains, Live5 is the number of trees per acre greater than 5

39
inch dbh, GSBDVH is the board foot volume per acre of hardwood species, Numlive is

the number of live trees per acre, Aspvol is the cubic foot volume per acre in aspen, and
Volratio is a ratio of board foot volume per acre to cubic foot volume per acre. The odds
ratio represents the increase in the odds of an event given a one unit increase in the
respective variable. The r-square value, rescaled to account for the number of
observations, was .143. The probability of harvest is calculated by l/(l+exp(-b - BX)
where b is the intercept parameter estimate, B is the vector of slope parameter estimates,

and X is the vector of explanatory variables (SAS Institute Inc. 1995).

Table 4: Logistic regression estimates for stand level harvesting.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Estimate Std. Error Pr > Chi-square Std.Estimate Odds Ratio
Intercept -1.6127 .2166 .0001

Ownl -.8477 .2745 .0020 -.14543 1 .428

Unit .3549 .1520 .0196 .097870 1.426
DRFT -.0327 .0125 .0086 -.150315 .968
Live5 .00438 .000943 .0001 .216829 1.004
GSBDVH .0001 12 .000056 .0454 .099986 1.000
Numlive -.00032 .000142 .0242 -.1 103 84 1.000
Aspvol .000654 .000255 .0103 .1 19195 1.001
Volratio .123 .0674 .0677 .089934 1.131

 

 

The individual tree harvest probability regression was performed in a similar
manner. The data were collected from remeasured stands that incurred harvesting
between 1980 and 1993 and had at least one live, measurable aspen stem during the 1980
survey. Again, the dataset was subdivided similar to the stand level dataset. This model
calculated the probability that each tree on the plot, and the trees it represents, would be

cut during each 10 year period. Some independent variables to be considered were dbh,

40
net cubic foot volume, relative basal area of the trees within the stand, and some of the

same stand characteristics used in the stand level logistic regression. Variables
considered in addition to those for the stand level regression were the binary variables
SW-- equal to 1 if the tree was a softwood, Aspen» equal to 1 if the tree was an aspen,
and Own2—equal to 1 if the plot was owned by state or local government. Netcuvl is the
cubic foot volume of the measured tree times 10. BA is the square foot per acre of the
measured tree plus the trees it represents. Volasp is a ratio of the total aspen cubic foot
volume on the acre to the total cubic foot volume on the acre. Those variables in the

model dataset found to be significant are shown in Table 5. The rescaled r-square value

was .159.

Table 5: Logistic regression estimates for tree level harvest.

 

 

 

 

 

 

 

 

 

 

 

 

 

Variable Estimate Std. Error Pr>Chi-square Std. Estimate Odds Ratio
Intercept -.9868 .1086 .0001

Own2 .4933 .0578 .0001 .114072 1.638
Unit .4754 .0517 .0001 . 129908 1.609
SW .3052 .0640 .0001 .067543 1.357
Numlive -.00065 .000053 .0001 -. 1 77494 .999
Volratio -.1599 .0243 .0001 -.098227 .852
DRFT .0227 .00391 .0001 .078371 1.023
Aspen .8364 .0614 .0001 .209024 2.308
BA .1083 .0154 .0001 .093798 1.114
Netcuvl .0021 1 .000173 .0001 . 190083 1.002
Volasp .7898 .1022 .0001 .120238 2.203

 

 

 

 

 

 

 

The accuracy and precision of the logistic regression results were evaluated in
three ways. The Hosmer and Lemeshow goodness-of-fit test was applied to the model

dataset. This test is considered conservative, biased towards specific types of lack of fit

41
and dependent on the grouping of the observations (SAS Institute Inc. 1995). This

statistic compares a value based on observed and expected event frequencies in each
group to a chi-square distribution. It was significant for the stand level regression but not

for the tree level regression as shown in Table 6.

Table 6: Hosmer and Lemeshow Goodness-of-F it Test

Lo istic Re ression I Goodness-of-Fit Statistic [Degrees of Freedom I p-value I
I Stand Level | 1.6082 | 8 I .9908 |

ITree Level | 46.99 [8 I .0001 l

 

 

The second test was development of a classification table. The logistic procedure
in SAS uses a one-step approximation of a typical jackknifing method. The model is fit
using all observations but one. The resulting coefficients are applied to this one
observation to calculate its event probability. The event probability is then compared to a
predetermined cutoff point that ranges from 0 to 1. If the calculated event probability is
greater than the cutoff point, then an event occurs. If not, the event does not occur.

For this study, the cutoff point was determined by examining a table of all
possible cutoff points and their resulting sensitivity and specificity. Sensitivity is a ratio
of the number of correctly classified events over the total number of events (SAS Institute
Inc. 1995). Specificity is a ratio of the number of correctly classified nonevents over the
total number of nonevents (SAS Institute Inc. 1995). The cutoff point was set where
these two ratios were approximately equal. This assures that the regression results are not
biased toward more accurately predicting events or non-events. There is an equal

probability that an event or non-event is accurately predicted.

42
For the stand level regression, this cutoff point was .3 and produced the

classification table shown in Table 7. Approximately 66% of the observations were
correctly classified. For the tree level regression, this cutoff point was .44 and produced
the classification table shown in Table 8. Approximately 64% of the observations were

correctly classified.

Table 7: Stand level harvest event classification table for model dataset with cutoff point
of .3.

 

 

 

Predicted
Actual NOT HARVEST HARVEST | Total
NOT HARVEST 451 234 J 685

 

 

 

HARVEST 106 195 301
I Total I 557 429 986

 

 

 

Table 8: Tree level harvest event classification table for model dataset with cutoff point
of .44.

 

 

 

 

 

 

 

 

 

Predicted
Actual NOT HARVEST [HARVEST [ Total
NOT HARVEST 2661 I 1629 | 4290
HARVEST 1266 [2430 I 3696
Total [3927 [4059 [ 7986

 

The final test was the application of the regression results to the independent
validation datasets. Only the model dataset was used to determine the coefficients for the
independent variables. These coefficients were then used with the validation dataset to
calculate the event probabilities. The same cutoff point from the model dataset
classification table was used to develop a classification table for the validation dataset.

The percent of accurately predicted events and non-events was examined. The

43
classification table for the stand level validation dataset is shown in Table 9

Approximately 67% of the observations were correctly classified. The classification table
for the tree level validation dataset is shown in Table 10. Approximately 66% of the
observations were correctly classified.

Table 9: Stand level harvest event classification table for validation dataset with cutoff
point of .3.

 

 

  
 
    
 

 

 

 

Predicted
Actual NOT HARVEST [HARVEST [ Total
NOT HARVEST 249 [ 119
HARVEST 59

 

 

   

 

Table 10: Tree level harvest event classification table for validation dataset with cutoff
point of .44.

 

 

 

 

 

 

Predicted
Actual NOT HARVEST HARVEST [ Total
NOT HARVEST 994 557 [ 1551
HARVEST 377 856 1233
Total 1371 1413 2784

 

 

 

The stand level model was selected because of its Significant goodness-of-fit test
and its approximately equal classification error rate for both the model and validation
datasets. The tree level model was selected because of its approximately equal
classification error rate for both datasets, and it outperformed any other combination of
variables in respect to all three tests.

These harvest probability models were incorporated into the STEMSSS

management routine. During each management activity, the stand and tree data values

44
from the beginning of the projection period were used as input into the harvest probability

equations. The cutoff probability values set above were compared with the calculated
harvest probability for each stand and each tree. If the cutoff probability was less than the
calculated harvest probability, the stand or the tree was harvested. Figure 14 illustrates

this procedure.

 

A nnual Grow 1h

4; 8: Mortality Loop ‘

 

 

 

 

 

V

Management
Loop

 
  

 

NO

  
   

(‘uloffpr <
Stand Harvest

 

 

 

 

Stand lncurs
H arvesting

I YES NO Cutoffpt<
Tree Harvest
Prob.

 

 

 

 

Tree is Cut

 

Figure 14: Stand and tree harvest flowchart

The assumptions necessary for this model are that all survey plots on timberland
are available and accessible for harvest; harvesting can be modeled at an individual tree
level; there are no constraints on supply; there are no effects due to stumpage price; the
F IA survey plots are representative of the actual resource.

A discrepancy between volume equations used by STEMS85 and by the F IA unit
was described earlier. The logistic regression coefficients were based on tree volumes
calculated by the F IA unit’s equations. In addition, the survey database only includes tree

volumes for live, acceptable growing stock trees. The volume equations in STEMSSS

45

calculate volume for all trees of all classes. In order to accurately utilize the regression
models, STEMS85 was modified so that the volumes used as independent variables in the
harvest probability equations are calculated using the F IA unit’s equations. In addition,
the stand and individual tree volumes were based on only those trees with a tree class of
20 or acceptable.

This modified version of STEMSSS was tested by projecting the 1980 plots with
aspen that were remeasured in 1993. Mortality was deterministic. Management was
implemented according to the logistic regression equations. The projection consisted of
12 one year periods with management in year 6. More than twice as much volume was
harvested by STEMS85 than was measured as harvested in the 1993 survey. This
discrepancy can be explained by the classification tables developed in testing the stand
level model. The cutoff point was chosen so that the probability of misclassifying an
event was equal to that of misclassifying a non-event. This allows for a more realistic
representation of the projected timber flow and resource status across the State and
ownerships. However, it results in a 59% false positive rate and a 20% false negative
rate. Thus, 59% of the plots harvested by the modified STEMSSS were not remeasured
as having had any cutting by the 1993 survey, and 20% of the plots not harvested by the
modified STEMS85 were remeasured as having had some cutting by the 1993 survey.
This discrepancy could be a function of the variables selected for the regression, the
exclusion of other factors such as markets and accessibility, and possible missed harvest
events on plots classified as undisturbed in the survey data. Variations in the cutoff point
from .36 to .4 and inclusion of an age limit for harvesting a stand only slightly improved

the results.

46

An analysis of the plots that were correctly and incorrectly classified showed no

significant difference between the two groups in the distribution of stand values such as

aspen volume, quality, size and species mix, etc.. Only stand stocking characteristics

such as total cubic foot volume and basal area showed a clear variation between the two

groups. As Figure 15 shows, the stands that were wrongly cut by the modified STEMSSS

tended to have lower stocking levels than those correctly cut. Additional comparisons are

in Appendix B.

Frequency

3

A
O

 

 

 

I ‘ + Correctly Cut PIOISAE

.. a - Wrongly Cut Plots .
t , I .- ”-2-

 

 

500 1000 1500 2000 2500 3000 3500 4000 4500
Plot Cubic Foot Volume

Figure 15: Comparison of 1980-1993 projected cut plots’ cubic foot volume/acre,
1980.

Since there was no bias in terms of levels of aspen stocking on the plots, 59% of

the plots projected to be harvested were randomly uncut, and 20% of the plots projected

to not be harvested were randomly cut. After the harvest probability for each stand was

47

calculated in the modified STEMS85, a random number was selected from a uniform
distribution (0,1). For those stands whose calculated harvest probability was greater than
or equal to 0.3 and the random number was less than 0.59, the stand would not be cut and
would return to the growth and mortality loop. If the harvest probability was less than 0.3
and the random number was less than 0.2, the stand would be cut. The total volume cut
between 1980 and 1993 with and without the random harvest correction were then
compared. Tables 11 and 12 show a comparison between the 1993 remeasured data, the
base projection data and the corrected projection data. The harvest correction produces
more accurate projections of the remeasured plots while still maintaining an

approximately equal distribution of harvest values across ownerships and units.

Table 11: Comparison of 1980-1993 projected and measured harvest and stocking
volumes.

 

 

 

 

Remeasured Data Base Projection Corrected Projection
Cubic feet aspen cut, 183.8 456.3 216.7
1980-1993
Cubic feet aspen stocking, 465.6 187.6 399.8
1993

 

 

 

 

Table 12: Comparison of 1980-1993 projected and measured harvest levels by
proportional distribution across ownership and location.

 

 

 

 

 

 

 

1980-1993 percent total cut Remeasured Data Corrected Projection
Federal aspen cut 1 1% 12%
State asan cut 28% 26%
Private aspen cut 61% 61%
Upper Peninsula aspen cut 55% 51%
Lower Peninsula aspen cut 45% 49%

 

 

 

 

 

48
In projecting the 1993 data, the modified STEMS85 was run 55 times for each

plot in order to provide data to estimate the variance in the projected outcomes. No
provision was made for regenerating stands harvested during the projection. I did not feel
that the simulations accurately reflected harvest activities on an individual tree basis, and
so it would be difficult to estimate the possible effect of residual overstory on any
regenerating aspen.

Successional processes are not specifically addressed in the model because the age
data for the projected plots is unreliable. The stands continue to grow and die throughout
the projection period. Because aspen is intolerant, it was assumed that the ingrowth on
these plots did not consist of a large aspen component. Those stands that did incur
harvesting were tracked and summarized by acreage and residual volume in order to

provide a generalization of the standing resource at the end of the projection.

V. RESULTS AND DISCUSSION

Michigan’s aspen resource continued to increase in volume prior to 1993 as the
lands cleared and abandoned at the turn of the century reverted to forestland and the
pioneering aspen became established and matured. However, these projections indicate
this trend is likely to reverse given continued harvest levels, management practices and
natural succession. From 1935 to 1993, the acreage in aspen covertype decreased by 48%
as natural succession, development and aging processes continued and harvest rates
increased 139% between 1975 and 1994. Possible shortages in aspen inventories have
been the subject of much research since the 19705 (Leuschner 1972, Liggett & Leefers
1990, Potter-Witter & Ramm 1992). This potential Shortage has been linked to increased
harvest levels, imbalanced age classes, natural succession due to lower levels of
disturbance, increasing resource maturity, and management practices that don’t promote
natural aspen regeneration. These projections suggest that aspen~volumes will start to
decline between 1993-2003 as the average annual harvest rate is almost sustained at 1994
levels. After 2003, recent harvest rates cannot be maintained by the projected aspen

inventory.

Findings
These results represent the average of 55 projections of Michigan’s aspen resource
from 1993-2023 simulated with a modified version of STEMS85. This program

projected growth, mortality and harvest of Michigan’s aspen resource based on the 1993

49

50

statewide forest inventory. The growth, mortality and harvest probability equations were
fit to data collected in Michigan on remeasured plots from the 1980 and 1993 inventories.
These results represent one possible future condition based on the assumption that recent
trends in growth, mortality and harvesting continue. All error bars represent 3 standard
deviations from the mean based on the averaging of the simulations.

Projections show that standing aspen volume declines by approximately 25% over
each of the first two decades from 1993-2013 and 30% over the third decade from 2013-
2023. Figure 16 demonstrates the projection of an ovennature resource that has been
significantly influenced by past and current harvesting patterns, natural successional
losses, and acreage conversions. The projected net 1055 between 1993-2023 in the aspen
growing stock volume is 60%, from a 1993 level of 40 million cords to a 2023 level of 16
million cords. This decline occurs across all ownerships and both peninsulas in Michigan
as shown in Figure 17.

The greatest proportional loss of aspen from 1993-2023 is projected to occur on
private ownerships (67%) and in the Upper Peninsula (64%). Federal and state/local
ownerships and the Lower Peninsula increase their proportion of the total state aspen
resource while private ownerships and the Upper Peninsula decrease their proportions by
2023 as compared to 1993. This decline in the aspen resource is related to the
continuation of increased aspen harvest levels from the past two decades into the first
decade of the projections, and the effect of minimal acres and stocking in younger age
classes due to the increasing proportion of the resource progressing undisturbed into

ovennaturity.

45

 

4710151 Asia—n Volumé (cords) T W l
40 1. n. . .Aspen Pulpwood Volume (cords)
q—Aspen Sawtimber Volume (cords)

351

25
20 .

15 ,,

Million Cords of Aspen

10

 

5
. /
/

"—— .

 

 

o I‘—

1930 1940 1950 1960 1970 1980 1990 2000 2010 2020 2030

 

Figure 16: Michigan’s aspen resource from 1935-2023 based on published inventories
(1935-1993) and study projections (2003.2023).5

 

2,500

2.000

1,500 -

1,000

 

Million Cubic Feet Aspen Growing Stoc

 

 

 

 

 

 

 

 

 

 

Federal State/Local Private Upper Lower
Peninsula Peninsula

Figure 17: Michigan’s aspen growing stock volume by ownership and peninsula, 1993-
2023 from published 1993 inventory and study projections (2003-2023).

5 Sawtimber volumes were converted using 1 mbf=l60 cubic feet and 79 cubic feet=l cord.

52
Total projected aspen removals for 1993-2003 were only 63 million cubic feet

greater than removals for 1983-1993, an increase of 5%. However, the period of 1993-
2003 had the greatest amount of mortality within the time frame of the projections as
shown in Figure 18. This could be related to a large portion of the aspen resource in 1993
being mature to overmature and subsequent decades’ mortality coming from a decreasing

aspen base.

1,600

 

1-400 1‘ a19831993
[1:1 1993 2003
1.200 —- 132003-2013;

111203-2023
1,000 .,

800 -

  

600*

Million Cubic Fee

400 -

l

    

-11

t 1' ‘ .1 1'1: l[lllliillllll il l l

 

 

 

 

 

Aspen Cut Aspen Mortality

 

Figure 18: Michigan’s aspen resource cut and mortality from published reports (1983-
1993) and study projections (1993-2023).6

Based on these projections, this increased aspen harvest level is not sustainable.

The resource becomes overmature as the increased mortality indicates, and the maturing

6 Aspen removals for the period 1983-1993 (Blyth et. A1. 1988, Smith et. A1. 1990, Hackett & Pilon 1993,
May & Pilon 1995). Aspen mortality for the period 1983-1993 is based on average annual mortality
(Leatherberry & Spencer 1996).

53
resource volumes are not sufficient to maintain the 1994 harvest level. The average

annual harvest level between 1983-1993 was 122 million cubic feet. This is projected to
increase to 128 million cubic feet between 1993-2003. The projected average annual
harvest then falls to 97 million cubic feet between 2003-2013 and 74 million cubic feet
between 2013-2023. Based on the timber product output publications referenced in
Figure 18, approximately 16% of the harvested volume does not make it into production.
Assuming this continues to apply to the projections, only 108 million cubic feet per year
between 1993-2003 will be available for industrial roundwood production. This is 5%
less than the 1994 industrial roundwood production level of 114 million cubic feet.
Despite a decreasing harvest level on federal and state/local ownerships, the 1983-
1993 harvest level is projected to be maintained between 1993-2003 by an increased
harvest level on private ownerships. The 1983-1993 ownership and geographic harvest
distribution was calculated by applying the respective removal proportion based on the
1993 Michigan inventory remeasured plots to the aspen volume cut based on the timber
production output publications. Private aspen harvest volumes go from 708 million cubic
feet between 1983-1993 to a projected 881 million cubic feet between 1993-2003, an
increase of 24%. The projected harvest levels by ownership and peninsula are shown in

Figure 19.

54

 

1000 I
900 ..
1.19831993I
800 I01993-2003
‘“ 32003 2013
.2013 2023

700 1-

600 1.

300 1-

Million Cubic Feet Aspen Cut

  

 
 
 
 

 

 

 

 

 

 

 

 

 

 

 

 

200 - ,
[I lmll
100 1- lllllI
‘ ll
I lll
0 1 . . .- illll . - 1 _.-
Federal State/Local Private Upper Lower
Peninsula Peninsula

Figure 19: Michigan’s aspen harvest volume by ownership and peninsula from published
inventory and timber product output reports (1983-1993) and study projections (1993-
2023)

After the initial projection period of 1993-2003, the private ownership and the
Upper Peninsula show the greatest decrease in aspen harvest levels which coincide with
the large decreases in aspen growing stock shown in Figure 17. These projected 27%-
28% volume reductions result from maintaining the harvest level during the first decade
from a resource not capable of replacing that volume in its current status. From 1983-
1993, aspen harvests on federal and state/local ownerships and in the Upper Peninsula
were proportionally greater than their share of the state’s aspen growing stock. While the
federal and state/local ownerships and the Upper Peninsula had 10.5%, 21.5% and 39.6%
of the state’s aspen growing stock, the harvest levels were 12.4%, 29.3% and 44.3% of

the state’s aSpen harvest, respectively. After 1993, the federal and state/local ownerships’

55
harvest proportions are projected to be less than their statewide growing stock proportions

which increase by 4% and 6%, respectively, between 1993-2023. Private ownerships are
projected to increase their 1993-2003 proportional harvest levels by almost 11% more
than their 1983-1993 proportion. The Lower Peninsula is projected to increase both its
proportional harvest level and standing aspen volume between 1993-2003. This may
indicate better growing sites in the Lower Peninsula and/or the effect that increased
harvest/disturbance levels can have in terms of future production.

The age class distribution projected for 2023 shows a large increase from 1993 in
the number of aspen covertype acres in the overmature age classes of 75 years and older.
This acreage does not necessarily represent aspen stands. It represents aspen stands in
1993 in which the aspen is not harvested during the projections to 2023. However, it is
unlikely this aspen will survive into these older age classes. Perala et a1. (1996) identified
one of the decay triggers to be a stand age equal to 10 plus the site index. It is likely that
these stands represent aspen acreage lost through conversion to a different forest type if
ingrowth and natural successional processes were incorporated into the projections. The
projected acreage in the mature age classes of 55 and 65 is 32% and 65% less than the
acreage in 1993, respectively. The age classes of 35 and 45 do contain a greater number
of acres in 2023 than in 1993. It is possible that these maturing acres could provide for an
increased aspen harvest level in succeeding periods. However, these periods are beyond
the scope of these projections due to the time period and the fact that regeneration on
stands cut during the projection was not included. Figure 20 shows the acreage

distribution by age class.

56
350

 

30° .. "ire—m near—bunch

__ 02023 Distribution

 

250

200.

 

150 -

100

Thousands of Aspen Covertype Acre

 

 

 

 

 

 

 

 

 

 

 

 

 

 

35 45 55 65 75 85 95 105+
10-Year Age Class

Figure 20: Comparison of Michigan’s 1993 and 2023 aspen covertype age class
distributions from 1993 inventory and study projections for 2023.7

Analysis and Future Research

Aspen regeneration will have a large impact on future resource levels, and there is
significant need to consider the success of aspen regeneration in any long-term resource
projection. However, within the short time span of these projections, the additional
regeneration is not successful in preventing a loss of the current resource.

There were 202 plots in the 1993 inventory that could be considered to be in the
process of regenerating at the time measurements were taken. These plots were less than

20 years old with an overstory of less than 40 square feet of basal area per acre and either

7 Dashed borders represent aspen acreage in 1993 with aspen volume not projected to be harvested by
2023. These acres in 2023 are not likely to still be in aspen but rather, converted to another forest type.

57
greater than 20 square feet of basal area per acre of aspen cut between 1980 and 1993 or

aspen seedlings/suckers measured on the plot. These parameters identify plots with
potential future aspen volume resulting from regeneration not overly suppressed by a
more mature overstory.

The effect of these stands with enhanced aspen stocking on both the projected
harvest and residual stocking levels was minimal due to the short-term nature of the
modified STEMSSS projections and the proportionately small number of these stands.
The average projected output and growing stock values were within the range of the
original projections that did not include the enhanced regeneration. At most, the
projections with the regenerating aspen stocking increased average period ending
stocking levels by less than 1%.

A large portion of the aspen resource is projected to progress undisturbed into the
overmature age classes given the past, and projected, harvesting patterns. Over 612,000
acres of aspen covertype are projected to be over 75 years old in 2023. This is 23% of the
acreage in aspen covertype in 1993. The possibility of salvaging some of the aspen
stands that progressed into overmature age classes or were lost to mortality is highly
dependent on the age and quality of the resource being harvested in the projections. If the
harvests are coming from newly mature stands in which volume would not be lost by
delaying harvests a decade, perhaps harvests could be shifted to older, overmature stands
during the first decade and inventory sustained for harvest in succeeding decades. This
could possibly maintain 1994 harvest levels into the succeeding decades. Because the
mortality level is at its highest during the first decade, much of the resource may be too

overmature and diseased to be utilized. If it isn’t, it may not be possible to fully utilize

58
the older timber resource without increasing overall harvest levels in the initial projection

period. Since the first decade projections are relatively close to meeting a continuation of
the 1994 demand level, the market may not provide an opportunity for salvage of the
older resource.

The possibility of salvaging the older aspen resource would require a more
detailed examination of the resource being harvested in each decade in terms of the
mortality and growth rates, distribution across ownerships and where changes in harvest
potentials could occur. A more sensitive modeling approach would be required to
estimate any potential effects of specific management strategies to maintain the aspen
resource. This would require a shorter projection cycle and more specific models of
harvest probabilities.

The projected harvest does not maintain the 1994 harvest level throughout the
next thirty years. Maintaining 1994 harvest levels would require an unparalleled effort at
statewide management of the aspen resource. It is possible that harvest levels could rise
after the initial lull between 2003-2023, but this is highly dependent on the management
practices implemented over the next thirty years. It may require the clearcutting of older
stands not past their prime but that are not necessarily marketable or stands with a
minimal aspen component that are in later successional stages. Maintaining an aspen
timber resource will require a concerted effort to facilitate active regeneration in
management practices and avoid further successional losses on large ownership blocks
such as publicly held lands or to increase industry’s aspen holdings.

It appears the Lower Peninsula and publicly held lands are potential sources for

increased resource output in the future. After 1993, their‘projected proportion of standing

59

resource and timber supply increase. It is possible that an increase in their output levels
could offset some of the harvest volume decline in the next thirty years without impairing
future contributions. A closer look at the resource on these ownerships in terms of
maturity, stocking, and availability would be required to determine if these lands could
contribute greater volumes over the next thirty years. However, this possibility could be
negated if the management practice of clearcutting is legally restricted from being
applied. The effect of restricting clearcutting on federal lands is not easily determined in
this analysis. It would likely decrease harvest levels over the next thirty years. The
degree to which it would decrease would depend on the amount of aspen harvested using
a selection management approach. If no aspen harvesting occurred on federal lands
between 1993-2003, Michigan’s projected average annual harvest level would be 116
million cubic feet, 2% greater than the 1994 harvest level of 114 million cubic feet if the
total harvest volume was utilized for roundwood. The greater effect of not allowing
clearcutting of aspen would come into play beyond the time period of these projections as
the lack of regeneration in these stands decreased the possible future mature aspen
resource growing stock levels.

A closer examination of the aspen resource being harvested would be required to
determine if shifts in harvesting practices could increase aspen inventories and sustain
harvests. While private and Upper Peninsula aspen resources are projected to have a
proportionately greater drain over the next 30 years as compared to that on public and
Lower Peninsula lands, this might be a factor of the aspen resource age and quality
distribution across these ownerships and peninsulas. Further classification of the aspen

resource status and outputs on these ownerships and peninsulas at shorter projection

60
cycles could provide a clearer picture of the options available to aspen resource managers.

It might facilitate the development of management alternatives that could alter the

projected outcomes for the resource.

Limitations

There are five main factors limiting this modeling approach. These include
assumptions related to the inclusion and application of regeneration, natural succession,
harvest probabilities based on logistic regressions, 10 year projection periods, and survey
data as representative of the aspen resource. Each is addressed below.

It is possible that modeling the first period mortality for the 202 regenerating
stands deterministically biased the status of these stands at 10 years. With only 202
regenerating plots identified out of over 4500, any specific bias should have limited
effects on projection results. Any possible regeneration from stands harvested during the
projections would not produce measurable aspen volume before 2023. So, the exclusion
of this regeneration process is not likely to affect projected inventory volumes.

Another potential source for error in describing the future aspen resource is that
plots converting to aspen through disturbance or natural successional processes were not
modeled. Between 1980 and 1993, there was a net loss of approximately 400,000 aspen
acres. Included in this net loss were approximately 200,000 acres which changed
classification to aspen covertype. It would be reasonable to expect that some acres would
gain an aspen component during the projections. These new aspen components could be
added to the projections if the parameters capable of defining the probability of their

future occurrence were identified and an average tree list was defined.

61
Natural succession and stand breakup were not specifically modeled. These

processes were generally included through the modified mortality rates, harvest
probability equations, and recalculation of the stand covertype after each projection cycle.
A more specific approach would be to remove any stands that reach a specific age from
harvest consideration and inclusion in standing volume totals. However, the age of the
aspen resource would have to be more accurately determined. Modeling natural
succession would require the addition of ingrowth tree lists over the course of the
projection and a more accurate model for determining the specific plot harvest patterns.
Natural succession is likely to affect the growth of aspen originally in a stand but not add
new aspen stems due to the shade intolerance of aspen.

The reason for the logistic regression resulting in overcutting stands with lower
stocking levels than the plots actually harvested between 1980 and 1993 has not been
specifically defined. It is possible that refinement of the harvesting equations could
result in different harvest volume projections. The choice to assume harvests occur in the
middle of the projection period may bias the projections. Mid-cycle harvesting was
chosen in order to balance the volume lost on stands cut early with the volume gained on
stands cut late. More specific knowledge of harvest timing and conditions at that time are
necessary to improve the modeling of harvest patterns.

The use of data collected in 1980 and 1993 to develop the harvest probability
equations limits this analysis by locking the projections into a 10 year cycle. This limits
its usefulness for those who plan on a shorter time horizon. However, it was not possible
to calculate probabilities for a shorter projection cycle without knowing when the stands

were cut between 1980 and 1993. It is possible that the total harvest level over a decade

62
could be the result of a widely varying annual harvest rate. The average annual harvest

levels were calculated by dividing the period harvest by 10 years.

A more comprehensive harvesting model might be developed with more detailed
knowledge of financial considerations, mill location, specific harvest levels and timing,
bidding and contract considerations, and market fluctuations. Harvest probabilities were
based on past management related to stand-specific variables. It may be possible to
include additional variables to increase the efficacy of the regressions. Additional
research is required to improve understanding of harvesting patterns and to implement
harvesting behavior models within a projection framework.

These results are meant to provide information to those interested in the status of
the aspen resource and its outputs, and those who are in a position to affect the outcomes.
Improvements in projection methods and especially models of harvesting patterns would

likely produce more detailed insights into the dynamics of this important forest resource.

Conclusion

Unless recent management practices change, the aspen resource is projected to
continually decline over the period of 1993-2023. The 1994 harvest level cannot be
maintained throughout the first decade, and subsequent decade harvest levels are
projected to continue declining. Public ownerships increase their statewide proportional
aspen resource due to their decreasing harvest levels from 1993-2023 and an increased
harvest level on private ownerships which greatly reduces their standing aspen volume.
In addition, a large portion of the aspen resource is projected to progress into overmature

age classes which decreases its utility and ability to regenerate and represents losses

63
through natural successional processes. Increased harvesting of the aspen resource in

mature age classes in 1993 would be required to salvage this loss, possibly increase
harvest levels before 2023, and provide a potential future aspen resource stemming from
the regeneration on these stands. It is clear that if recent management practices continue,

the current aspen inventory and harvest levels cannot be maintained between 1993-2023.

APPENDICES

APPENDIX A

APPENDIX A

Distribution of Michigan ’s Aspen Resource According to I 993 Survey Data

 

 

 

 

 

 

 

 

 

 

 

Figure 21: 1993 surveyed aspen covertype plot locations.

64

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Figure 22: Distribution of aspen cubic foot volume cut, 1980-1993.

66

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

.Federal
.srae
[jpn'vate

pen Volume Cut by Ownership

 

I

 

Figure 23: Distribution of aspen cubic foot volume cut by ownership, 1980-1993.

67

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

AspenCovertypebyOwnersh'p

 

 

 

 

- Federal

 

Figure 24: Distribution of aspen covertype acreage by ownership class, 1993.

68

 

 

- Federal
[:I State
[:] Private

Aspen Acreage Lost by Ownership

 

I

 

Figure 25: Distribution of aspen covertype acreage lost by ownership, 1980-1993.

69

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

'\ 'I .7 . ." I.I‘. '1 ."""'. TI. 17.3: ”I .
Aspen Cubic Foot Volume x a ‘ ~ "=1”
- '.: ' film-I . ”3., "2.17"”: . '~
Each Dot = 200,000 cubic feet . . ,4, , . ~~ gym.- --' »
, 7 . {q‘rr’l :i'; r H A-

 

 

 

 

 

I
f ‘

Figure 26: Distribution of standing aspen cubic foot volume, 1993.

 

+4

7O

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I Aspen Cubic Foot Volume by Ownership I
- Federal I ®
/’
I. 3...... . .>‘ e
I - fl Private l i G
T *7 “I 6
a G G
.- G
a G v
6/
G
G O G \r"
G G G
Q l G G
A _# _,

Figure 27: Distribution of aspen cubic foot volume by ownership, 1993.

71

 

 

 

‘

AspenCubicFoot/Acre
D Oto 100

 

r

UH] 100t0200

200to300

300lo4oo

m 400t0500

m 500t0600

- 600to700

L

 

Figure 28: Distribution of average aspen cubic foot volume/acre on all acreage with aspen

volume, 1993.

72

O Q o O Q
Q L
e QOIQIO f
O O Q Q (56,
00.00.09 .
[Aspen Volume Distribution by Covertype e 0 e G O O I ”I, I

 

- Other Covertype e G
E] Aspen Covertype O W a I . . I . I I

Gee-colic,
c e e o oIe_
e‘e e‘e .7.
9019307. .‘.'4

 

Figure 29: Distribution of aspen cubic foot volume by stand covertype classification,
1993.

 

APPENDIX B

APPENDIX B

Comparisons of 1980 Survey Plots Projected Using the Initial Logistic Regression
Results to be Correctly and Incorrectly Cut as Determined by 1993 Survey
Remeasurernents

180

 

160 ii '
i -o¥- Correctly??? Plots 1
140 er ' +Wrongly Cut Plots ‘

120 ,,

 

 

100 »

Frequency
8

60m

404

20 so

 

 

 

 

 

0 20 40 60 80 100 120 140 160
Flat Aspen Square FeetlAcre

Figure 30: Comparison of 1980-1993 projected cut plots’ asan square feet/acre,
1980.

73

74
180

 

160 4» ‘+ Correctly Cit Plots
+ Wrongly Cut Plots
140 W P i

120 -r

 

100 4.

Frequency

 

 

 

 

A A ‘ A A A
V

 

 

f‘

0 500 1000 1500 2000 2500 3000 3500 4000
Plot Aspen Cubic Feet

Figure 31: Comparison of 1980-1993 projected cut plots’ aspen cubic foot volume,
1980.

90

 

80 _ _ _.
+ Correctly Cut Plots ‘

+ Wrongly Cut Plots
70 i , ~-~ ,

60

50

Frequency

40

30

20

1O

 

 

 

0 2 4 6 8 10 12 14 16
Plot Average DBH

Figure 32: Comparison of 1980-1993 projected cut plots’ average dbh, 1980.

 

75

 

 

 

 

140
120
4: CoFectly CUFF—dots '
100 r ‘fwfll‘g'lcufpts ‘
5‘ 80 j
r:
o
E
u. 60
4o
20 5‘10‘“ ”‘5’
o \ 1“?3¢:A
35 40 45

Flat Cull 202Square FeetIAcre

Figure 33: Comparison of 1980-1993 projected cut plots’ cull in square feet/acre,
1980.

 

70

60 .i

50 + Correctly cut my
' ' +Wrongly Cut Plots ‘

Frequency

 

 

 

 

25 30

5 1O 15 20
Plot Aspen Mortality Square Feet/Acre

Figure 34: Comparison of 1980-1993 projected cut plots’ aspen mortality in square
feet/acre, 1980.

100 ,____. - _ , , i_.__n__.___i_ _c_ t , Ms M is, a a

80.

+ Correctly CCU! Plots
70 - +Wrongly Cut Plots

Frequency
8

0 - 7 . w- .__.-_—o e 77 r +7 ‘r - ‘7

0 5 10 15 20 25 30 35 40 45 50
Plot Mortality Square FeetIAcre

_1_,___-i ' _A.___ -a———

 

Figure 35: Comparison of 1980-1993 projected cut plots’ mortality in square feet/acre,
1980.

 

80

t + Correctly Cat Flori;

L... Wrongly Cut Plots 1

Frequency

 

 

 

 

0 100 200 300 400 500 600 700 800
Number of Live Trees on Plot

Figure 36: Comparison of 1980-1993 projected cut plots’ number of live trees per acre,
1980.

77
70%

 

60% 4, 'IVCErect’ly CEt'PlotsL

lDWrongly Cut Plots
50% ,,

40% ._

30% ,7

Percent of Plots

20% ,_

10% ,

 

 

 

 

 

0%.

 

 

 

Upper Peninsula Lower Peninsula

Figure 37: Comparison of 1980-1993 projected cut plots’ location, 1980.

70%

 

' 'Correctly’Cut Ploy
60°/ 7 ' 1
o lgWrongly Cut Plots
50%

40% .,

30%

Percent of Plots

20% i,

10% ,_

 

 

0%-

 

 

 

Federal State/Local Private

Figure 38: Comparison of 1980-1993 projected cut plots’ ownership class, 1980.

78
7o

 

\ " _._ CorrectTy ‘cm Plots 1
fl '\ l— g - Wrongly Cut Plots ‘
I A , ‘-_ A” , .
l
l

8 8

Frequency
8

 

 

 

 

0 20 40 60 80 100 120 140 160 180 200
Plot Live Square FeetlAcre

Figure 39: Comparison of 1980-1993 projected cut plots’ live square feet/acre, 1980.

 

APPENDIX C

APPENDIX C

Study Projection Data for 1993-2023

Table 13: Averages for projections of Michigan’s aspen resource by timber class, 1993-

2023 [std].

 

 

 

 

 

 

 

 

OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen Pulpwood (met) 1225213 897820 588306
[23892] [3 1 174] [33089]
Aspen Sawtimber (mbf) 7376350 5723956 4170660
[158239] [227207] [231465]
Aspen Pulpwood Cut (mcf) 622764 462645 347317
[24805] [28711] [28454]
Aspen Sawtimber Cut (mbf) 4042981 31 14614 2398906
[170760] [164452] [189717]
Aspen Pulpwood Mortality 441198 351562 247970
(met) [10471] [14291] [15038]
Aspen Sawtimber Mortality 2274411 1765008 1232581
(mbf) [75414] [80406] [79660]

 

 

 

 

79

 

80

Table 14: Averages for projections of Michigan’s aspen resource, 1993-2023

 

 

 

 

 

 

 

 

 

 

 

[min,max].
OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen (met) 2421244 1826542 1264725
[2325587,2504435] [1725180,1956163] [1138522,1443844]
Aspen Cut (met) 1278201 968373 736936
[1 198083,]366514] [869337,1058887] [598829,874747]
Aspen Mortality (met) 808284 636916 447261
[766965,84171 1] [582156,673399] [402660,505149]
Federal Aspen (met) 292500 247831 188356
[255719,322958] [199079,288487] [141 162,268568]
Federal Aspen Cut (met) 1 17101 95609 83186
[83656,165360] [61694,143420] [54102,] 18142]
Federal Aspen Mortality 1 12389 94048 74651
(met) [96470,132224] [68884,114265] [55358,101330]
Lower Pen. Aspen (met) 1517478 1 167774 809097
[1459690,1587256] [1066958,1246004] [730255,885204]
Lower Pen. Aspen Cut 741874 622952 487747
(met) [666846,81 1629] [555450,719307] [361886,585788]
Lower Pen. Aspen 435786 366872 262906
Mortality (met) [397664,461910] [334685,397873] [227193,300997]
Private Aspen (met) 1544063 1098602 730982

 

 

 

 

 

 

 

 

 

 

 

 

[1468254,1607938] [1014200,] 183137] [629976,821201]
Private Aspen Cut (met) 881 142 634625 462018
[817006,987490] [539781,709345] 394598,521467]
Private Aspen Mortality 514449 383847 251597
(met) [474860,539391] [350596,422406] [216407,287167]
State Aspen (met) 584681 480109 345388
[544643,6223 26] [424496,529304] [292033,41 1072]
State Aspen Cut (met) 279957 238139 191733
[238575,32071 1] [197383,277869] [137992,249513]
State Aspen Mortality 181447 159021 121012
(met) [167289,197945] [135819,180093] [95948,145561]
Upper Pen. Aspen (met) 903767 658768 455628
[853667,970351] [601038,729984] [381321,558641]
Upper Pen. Aspen Cut 536326 345420 249189
(met) [456547,598388] [305438,396190] [206607,304835]
Upper Pen. Aspen 372499 270044 184354
Mortality (mcfl [347984,401995] [241 1 1 1,302079] [156906,216507]

 

 

Table 15: Averages for projections of Michigan’s Federal aspen resource by timber class,

1993-2023 [std].

 

 

 

 

 

 

 

 

 

 

 

OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen Pulpwood (met) 171838 148175 107931
[10116] [13481] [16162]
Aspen Sawtimber (mbt) 745651 615719 496981
[43 821] [63446] [66540]
Aspen Pulpwood Cut (met) 65058 54664 49626
[10353] [11090] [13003]
Aspen Sawtimber Cut (mbt) 321557 252470 207099
[55273] [48260] [42339]
Aspen Pulpwood Mortality 64442 58852 47942
(met) [4046] [7259] [7852]
Aspen Sawtimber Mortality 297540 217960 165480
(mbf) [3 1 194] [26269] [26679]

 

Table 16: Averages for projections of Michigan’s Private aspen resource by timber class,

1993-2023 [std].

 

 

 

 

 

 

 

 

 

 

 

OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen Pulpwood (met) 731754 47621 1 291689
[18068] [20150] [21700]
Aspen Sawtimber (mbt) 5005983 3832065 2705435
[126648] [202107] [194726]
Aspen Pulpwood Cut (met) 414916 277971 1891 16
[20920] [22861] [15362]
Aspen Sawtimber Cut (mbt) 2874294 2195269 1678578
[130254] [145076] [147039]
Aspen Pulpwood Mortality 274082 195788 124083
met) [943 5] [8693] [92 15]
Aspen Sawtimber Mortality 1488205 1 162336 787810
mbf) [59610] [60069] [64979]

 

 

 

Table 17: Averages for projections of Michigan’s State/Local aspen resource by timber

class, 1993-2023 [std].

 

 

 

 

 

 

 

 

 

 

 

OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen Pulpwood (met) 321621 273435 188686
[11052] [15717] [14728]
Aspen Sawtimber (mbt) 1624714 1276173 968244
[78045] [94396] [89857]
Aspen Pulpwood Cut (met) 142790 130010 108575
[10512] [14238] [15090]
Aspen Sawtimber Cut (mbt) 847131 666874 513228
[76623] [71288] [76590]
Aspen Pulpwood Mortality 102673 96923 75945
(met) [4735] [6686] [7493]
Aspen Sawtimber Mortality 488666 384712 279290
(mbt) [25580] [33296] [27694]

 

Table 18: Averages for projections of Michigan’s Upper Peninsula aspen resource by
timber class, 1993-2023 [std].

 

 

 

 

 

 

 

 

 

OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen Pulpwood (met) 440291 327490 221521
[12985] [15680] [20250]
Aspen Sawtimber (mbt) 2868791 2049576 1449515
[84057] [87151L [96340]
Aspen Pulpwood Cut (met) 254848 159773 121 177
[14044] [1 1726] [16502]
Aspen Sawtimber Cut (mbt) 1742375 1 147354 791587
[96038] [791 15] [67187]
Aspen Pulpwood Mortality 194132 146087 104239
(met) [645'L [8335] [9946]
Aspen Sawtimber Mortality 1 108094 768731 496836
(m bt) [49769] [40515] [38427]

 

 

 

 

 

Table 19: Averages for projections of Michigan’s Lower Peninsula aspen resource by

timber class, 1993-2023 [std].

 

 

 

 

 

 

 

 

 

 

 

OUTPUT Period 1 Period 2 Period 3
1993-2003 2003-2013 2013-2023
Aspen Pulpwood (met) 784922 570330 366785
[20125] [24956] [21648]
Aspen Sawtimber (mbt) 4507558 3674381 2721 144
[132819] [187001] [190599]
Aspen Pulpwood Cut (met) 367917 302873 226140
[21 188L [24288] [24105]
Aspen Sawtimber Cut (mbt) 2300608 1967260 1607319
[139472] [136682] [163539]
Aspen Pulpwood Mortality 247066 205475 143731
(met) [9329] [9847] [1 1262]
Aspen Sawtimber Mortality 1 166317 996277 735745
(mbf) [5 7607] [61780] [64926]

 

 

LIST OF REFERENCES

LIST OF REFERENCES

Beleher, D. M. 1981. The user’s guide to STEMS. General Technical Report NC-70. St.
Paul, MN: USDA Forest Service. 16 pp.

Beleher, D. M., Holdaway, M. R., and Brand, G. J. 1982. A description of STEMS.
General Technical Report NC-79. St. Paul, MN: USDA Forest Service. 18 pp.

Bell, K., and Eriksson, M. 1991. Nonindustrial timber supply in the south: A
probabilistic evaluation for east Texas. Proceedings of the 1991 Symposium on
Systems Analysis in Forest Resources, p.337-341. General Technical Report SE-
B. Asheville, NC: USDA Forest Service.

Belli, K. L., and Ek, A. R. 1988. Growth and survival modelling for planted conifers in
the Great Lakes region. Forest Science 34:458-473.

Bertseh, R., and Weatherspoon, A. 1994. Wood products in Michigan. A directog of
mills & manufacturers. Edited by Vi Burcharn. Lansing, MI: Michigan
Department of Natural Resources Forest Management Division.

Bevington, P. R. 1969. Data reduction and error analysis for the physical sciences, p. 48.
New York: MeGraw-Hill.

Blyth, J. E., Weatherspoon, A. K., and Smith, W. B. 1988. Michigan timber industry- an
assessment of timber product output and use. Resource Bulleting NC-109. St.
Paul, MN: USDA Forest Service. 55 pp.

Brand, G. J., Holdaway, M. R., and Shifley, S. R. 1988. A description of the TWIGS and
STEMS individual-tree-based growth simulation models and their applications.
Forest growth modelling and prediction. General Technical Report NC-120. St.
Paul, MN: USDA Forest Service.

Brinkman, K. A., and Roe, E. I. 1975. Quaking aspen: Silvics and management in the
Lake States. Agriculture Handbofiok No. 486. Washington, DC: USDA Forest
Service. 50 pp.

84

85

Buchman, R. G. 1983. Survival predictions for major Lake States tree species. Research
Paper NC-233. St. Paul, MN: USDA Forest Service. 7 pp.

Connaughton, K. P., and Campbell, C. D. 1991. Comparing industrial and nonindustrial
harvesting behavior: A probabilistic approach. Proceedings of the 1991
Symposium on Systems Analysis in Forest Resources, p.342-347. General
Technical Report SE-74. Asheville, NC: USDA Forest Service.

Einspahr, D. W., and Wyckoff, G. W. 1990. North American aspen: Timber supply,
utilization, and research. North. J. Appl. For. 7(4):168-171.

Ek,‘A. R., and Brodie, J. D. 1975. A preliminary analysis of short-rotation aspen
management. Can. J. For. Res. 5:245-258.

Field Manual Michigan. St. Paul, MN: USDA Forest Service North Central Forest
Experiment Station Forest Inventory and Analysis, [1992].

Hackett, R. L., and Pilon, J. 1993. Michigan timber industry- an assessment of timber
product output and use, 1990. Resource Bulletin NC-144. St. Paul, MN: USDA
Forest Service. 56 pp.

Hahn, J. T. 1984. Tree volume and biomass equations for the Lake States. Research Paper
NC—250. St. Paul, MN: USDA Forest Service. 10 pp.

Hansen, M. H., and Hahn, J. T. 1992. Determining stocking, forest type, and stand-size
class from forest inventory data. North. J. Appl. For. 9(3):82-89.

Hansen, M. H., Frieswyk, T., Glover, J. F., and Kelly, J. F. 1992. The eastwide forest
inventory data base: Users manual. General Technical Report NC-151. St. Paul,
MN: USDA Forest Service. 48 pp.

Holdaway, M. R., and Brand, G. J. 1986. An evaluation of Lake States STEMS85.
Research Paper NC-269. St. Paul, MN: USDA Forest Service. 10 pp.

Jaakko Poyry Consulting, Inc. 1992. Maintaining productivity and the forest resource
base. mal generic environmental impact statement study on timber harvesthg
and forest magagement in Minnesota. Prepared for Minnesota Environmental
Quality Board. Tarrytown, NY: Jaakko Poyry Consulting, Inc.

Jamnick, M. S., and Beckett, D. R. 1988. A logit analysis of private woodlot owner’s
harvesting decisions in New Brunswick. Can. J. For. Res. 18:330-336.

Leuschner, W. A. 1972. Projecting the aspen resource in the Lake States. W
NC-81. St. Paul, MN: USDA Forest Service.

86

Leatherberry, E. C., and Spencer, J. S. Jr. 1996. Michigan forest statistics, 1993. Resource
Bulletin NC-170. St. Paul, MN: USDA Forest Service. 144 pp.

Liggett, C., and Leefers, L. A. 1990. Temporal aspects of ecosystem biodiversity: a case
study of the aspen type in Michigan. Michigan Academician 22(4):371-379.

May, D. M., and Pilon, J. 1995. Michigan timber industry - an assessment of timber
product output and use, 1992. Resource Bulletin NC-162. St. Paul, MN: USDA
Forest Service. 64 pp.

Miles, Pat. North Central Forest Experiment Station Forest Inventory & Analysis, St.
Paul, MN. Personal Communication, July 1997.

Ottawa N. F. “Sustaining the aspen ecosystem on the Ottawa National Forest.” 11 July
1995. Ironwood, MI: USDA Forest Service.

Perala, D. A. 1977. Manager’s handbook for aspen in the north-central states. General
Technical Report NC-36. St. Paul, MN: USDA Forest Service. 30 pp.

Perala, D. A. 1991. Renewing decadent aspen stands. Aspen management for the 21St
centm. Proceedings of a symposium held November 20-21, 1990, Edmonton,
Albertaz77-82. Edited by S. Navratil and PB. Chapman. For. Cen., Northwest
Reg., North. For. Cent. And Poplar Counc. Can., Edmonton, Alberta:

Perala, D. A., Host, G. E., Jordan, J. K., and Cieszewski, C. J. 1996. A multiproduct
growth and yield model for the circumboreal aspens. North. J. Appl. For.
13(4):164-170.

Perala, D. A., Leary, R. A., and Cieszewski, C. J. 1995. Stockability, growth, and yield of
the circumboreal aspens (Populus tremuloides Michx., P. tremula L.). Research
Paper NC-321. St. Paul, MN: USDA Forest Service. 24 pp.

Piva, R. J. 1997. Pulpwood production in the North-Central region, 1995. Resource
Bulletin NC-180. St. Paul, MN: USDA Forest Service. 39 pp.

Potter-Witter, K., and Ramm, C. W. 1992. Aspen supply, use, and demand in Michigan.
North. J. Appl. For. 9(1):]0-14.

Raile, G. K., Smith, W. B., and Weist, C. A. 1982. A net volume equation for Michigan’s
upper and lower peninsulas. General Technical Report NC-80. St. Paul, MN:
USDA Forest Service. 12 pp.

Rose, D. W. 1991. ACES version 3.1 - An allowable cut evaluation simulation program:
User’s manual. Staff Paper no. 76. St. Paul, MN: University of Minnesota,
Department of Forest Resources.

87

Ryan, K. C., and Reinhardt, E. D. 1988. Predicting posttire mortality of seven western
conifers. Can. J. For. Res. 18:1291-1297.

SAS Institute Inc. 1995. Logistic regression examples using the SAS system, version 6,
First Edition. Cary, NC: SAS Institute Inc. 163 pp.

Shields, W. J ., Jr., and Bockheim, J. G. 1981. Deterioration of trembling aspen clones in
the Great Lakes region. Can. J. For. Res. 11:530-537.

Smith, W. B. 1983. Adjusting the STEMS regional forest growth model to improve local
predictions. Research Note NC-297. St. Paul, MN: USDA Forest Service. 5 pp.

Smith, W. B., Weatherspoon, A. K., and Pilon, J. 1990. Michigan timber industry- an
assessment of timber product output and use, 1988. Resource Bulletin NC-121. St.
Paul, MN: USDA Forest Service. 67 pp.

Spencer, J. S., Leatherberry, E. C., and Kingsley, N. P. 1990. The Lake States’ aspen
resource revisited: Mid-196OS-1987. Aspen symposium ’89 proceedings, pp. 243-
252. General Technical Report NC-140. St. Paul, MN: USDA Forest Service.

Stoeekeler, J. H., and Macon, J. W. 1956. Regeneration of aspen cutover areas in northern
Wisconsin. J. For. 54(1):]3-16.

 

"1111111]llllllllr