Transcript Adapting a mortality model for Southeast British Columbia

Adapting a Mortality Model for
Southeast Interior British Columbia
By - Temesgen H., V. LeMay, and P.L.
Marshall
University of British Columbia
Forest Resources Management
Vancouver, BC, V6T 1Z4
The 2001 Western Mensurationists' Meeting
Klamath Falls, Oregon
June 24-26/2001
Adapting a GY model
• The Northern Idaho prognosis variant
(NI) has been adapted to the
southeast interior of BC, PrognosisBC
US Habitat Types
BC Biogeoclimatic
Ecosystem
Classification units
Adapting a GY model (cont’d)
• Different measurement units
(metric), basic functions (e.g.,
volume and taper) and standards
• Classification of US habitat type to
BEC can be subjective
• Sub-models coefficients and
model form may not fit BC data
• Insufficient ground data for some
types of stands
Adapting a GY model
Sub-model components:
• large tree diameter and height
growth
• small tree diameter and height
growth
• small and large tree crown ratio
• mortality and regeneration
• others
BACKGROUND
• Mortality is:
an essential attribute of any
stand growth projection system
frequently expressed as a
function of tree size, stand
density, individual tree
competition, and tree vigor
• In PrognosisBC, periodic mortality
rate is predicted using tree (Ra)
and stand based (Rb) mortality
functions
BACKGROUND (cont’d)
• Ra is a logistic function of tree
size taken in context of stand
structure.
• Rb operates as a convergence on
normal basal area stocking and
maximum basal area (BAMAX)
• Rb is based on the concept that:
for each stand, there is a
normal stocking density
there is a BAMAX that a site
can sustain and this maximum
varies
by site quality
Objectives
• to adapt a mortality model for
southeast interior BC
• to evaluate selected mortality
models for conifers and
hardwoods in southeast
interior BC
METHODS
•
Three approaches of adapting
mortality model were assessed, using
BC based PSPs:
1. a multiplier function (Model 1)
2. re-fit the same model form by
species/zone combination (Model
2)
3. changing variables (Models 3, 4,
and 5)
•
PSPs that were re-measured at 5 to
12 years interval and that consistently
included all trees > 2.0 cm were
METHODS (cont’d)
• For each PSP, individual tree records
were coded, as either live or dead at
each measurement period, and
variables listed in the mortality models
were extracted
ZONE
ESSF
ICH
IDF
MS
Total
# of PSPs
8
243
274
137
662
# of trees
live
dead
508
36
44991
5162
40497
3909
13456
859
99452
9966
Annual
Mort. (%)
0.71
1.15
0.97
0.64
1.00
METHODS (cont’d)
• Only species/zone combinations
with more than 30 dead trees were
selected.
• To handle the unequal remeasurement periods in the PSP
data sets, each model was weighted
by the number of years between
remeasurement periods.
•
•
The PSP data set was divided into
model (70%) and test data (30%)
sets
Observed and predicted number of
live and dead trees by species/zone
RESULTS
• Noticeable differences were found
in the % of correctly classified
trees among the five models and
the species/zone combinations
considered in this study
• Model 5 had lower Akaike
Information Criterion (AIC) and
Schwartz Criterion (SC) for most
species/zone combinations
Percent of correctly classified trees
in the ICH zone, using test data
Model 1
Model 4
Percent of correctly classified trees
100.0
90.0
Model 2
Model 5
Model 3
80.0
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
At
Bl
Cw
E
Fd Hw Lw
Tree species
Pl
Pw
Sx
Number of observed (N_OBS) and
predicted (N_Exp) dead trees by species
in the ICH zone, using Model 5 on test
data
N_OBS
N_EXP
700
Number of dead trees
600
500
400
300
200
100
0
At
Bl
Cw Ep
Fd
Hw Lw
Tree Species
Pl
Pw
Sx
Number of observed (N_obs) and
predicted (N_Exp) dead trees by diameter
class in the ICH zone, using Model 5 on
test data
Species=Douglas-fir
160
N_obs
N_EXP
140
Number of trees
120
100
80
60
40
20
0
5
10
15
20
25
30
35
40
Diameter class (cm)
45
50
55
60
Percent of correctly classified trees
in the IDF zone, using test data
Model 1
Model 4
80.0
Model 2
Model 5
Model 3
% of correctly classified trees
70.0
60.0
50.0
40.0
30.0
20.0
10.0
0.0
At
E
Fd
Lw
Pl
Tree species
Py
Sx
Number of observed (N_OBS) and
predicted (N_Exp) dead trees by species
in the IDF zone, using Model 5 on test
data
700
Number of dead trees
600
N_OBS
N_EXP
500
400
300
200
100
0
At
Ep
Fd
Lw
Tree Species
Pl
Py
Sx
Number of observed (N_obs) and
predicted (N_Exp) dead trees by diameter
class in the IDF zone, using Model 5 on
test data
Species=Douglas-fir
N_obs
300
N_exp
Number of trees
250
200
150
100
50
0
5 10 15 20 25 30 35 40 45 50 55 60
Diameter class (cm)
For species/zone combination
with little or no data
• substitution by similar species or
BEC zone is suggested.
FOR
•Bl in IDF
•Cw in IDF
•E in MS
•Fd in PP
USE
ICH
ICH
ICH
IDF
Summary
• Model 5 predicts mortality of both
conifers and hardwoods
reasonably well
• BC based BAMAX values
improved the predictive ability of
the model
• Inclusion of eco-physical factors
such as slope, aspect, and
elevation into the mortality model
might increase the predictive
ability of the model.