Ecological Archives A017-002-A1

K. Norman Johnson, Pete Bettinger, Jeff D. Kline, Thomas A. Spies, Marie Lennette, Gary Lettman, Brian Garber-Yonts, and Tad Larsen. 2007. Simulating forest structure, timber production, and socioeconomic effects in a multi-owner province. Ecological Applications 17:34–47.

Appendix A. The effect of random variables on the LAMPS simulations.

The LAMPS simulations utilize many random variables in each simulation (Table A1). However, we believe that a single simulation gives a reliable estimate of the implications of any set of policies. In this appendix, we provide evidence for that conclusion.

We divide the Coast Range into six “megasheds” for simulation. Each megashed contains approximately 2 million to 4 million basic simulation units (BSUs), ranging in size from 0.06 to 1.94 ha, with an average size less than 0.5 ha. They are the fundamental building blocks of the simulation. Spatially contiguous BSUs are combined into parcels for management with parcels averaging approximately 6 ha in size giving approximately 50,000–100,000 parcels per megashed.

In terms of use, the random variables fall into three categories: (1) Those applied to BSUs at the beginning of the simulation to fill a missing stand characteristic such as site class. Less than 5% of the BSUs need these values so we would not expect them to affect the results much. (2) Those applied to BSUs each period. Given the large number of BSUs, we hypothesized that each of these random variables would produce approximately the same distribution in each period. (3) Those applied to parcels each period. Again, given the large number of parcels, we hypothesized that the random variables would produce about the roughly the same distribution in each period.

To test these hypotheses that the results will not change much from run to run, we ran 10 simulations for the Mideast megashed and ten simulations for the Midwest megashed, using a different seed to start each simulation. Both of these megasheds have significant amounts of forest industry and nonindustrial private forest and low amounts of State forest (Table A2). Federal forests are heavily represented in the Midwest megashed, but not in the Mideast .

We then compared five outcomes for each owner: (1) hectares clearcut by period, (2) hectares thinned by period, (3) hectares by age-class of softwood stands in period 20 (last period), (4) hectares by age-class of mixed stands in period 20, and (5) hectares by age-class of hardwood stands in period 20.

We evaluated the variability from run to run using the coefficient of variation (CV) (Table A3, Table A4). For federal and forest industry ownerships, the median percent CV for the different outcomes is less than five percent, except for cases where a relatively small number of hectares are involved (mixed area by age class on federal lands in period 20 in the Midwest).

The nonindustrial private ownership show more variability, though even here, the percent CV is generally less than 10 percent. The highest variability for nonindustrial lands is associated with area thinned and the hardwood age-class inventory in period 20. Whether an area is thinned is determined at the parcel (group of BSUs) level and involves both the probability function and an analysis of whether the parcel has enough value to thin. All of these factors contribute to a higher variability than other owners. The relatively high CV for nonindustrial hardwoods in period 20 in the Midwest can be attributed to the presence of relatively low amounts of hardwoods by that period.

State lands cover a very small area in each megashed (Table A2). Therefore, relatively few observations are drawn each period to determine actions and effects there and the higher percent CVs is not surprising.

Overall, the coefficient of variation was relatively low in most cases, except where an owner group had very few hectares in the category being analyzed. From this analysis, we concluded that recognition of randomness in the spatial location of actions and impacts does not significantly affect the aggregate results of our analysis.

TABLE A1. Random variables utilized in a LAMPS simulation.

All landowners:
Assigning vegetative class to BSU in cases where it is missing.
The site index for a BSU in cases where it is missing.
BSUs for riparian stochastic disturbance.
BSUs for upland stochastic disturbance.
Regeneration condition of the BSU after a stochastic disturbance.
Regeneration condition of a BSU after a regeneration harvest.
 
Forest industry:
Timing of thinning for each parcel.
New management intensity for each clearcut parcel.
Vegetative condition of each regenerated BSU.
 
Non industrial private forests:
Timing of thinning for each parcel.
Whether a parcel gets clearcut, partial cut, or left unharvested.
 
State:
Size distribution of interior habitat patches.
 
Federal:
Timing of thinning (choosing a RX) for each parcel.
Parcel to schedule for harvest.

 

TABLE A2. Initial extent of forests in each ownership group in each megashed.

Megashed Forest Industry Nonindustrial Federal State
         
Midwest 213660 (ha) 87825 (ha) 248983 (ha) 19285 (ha)
Mideast 72724 (ha) 231728 (ha) 21936 (ha) 12648 (ha)

 

TABLE A3. Median percentage CV values for types of harvest for all randomized runs in the Mideast and Midwest megasheds.

   

Mideast
total (ha)

 

CV (%)

 

Midwest
total (ha)

 

CV (%)

Clearcut

 
 
 
 
 

Federal

1603.97

 

3.29

 

2118.93

 

2.57

 

Industry

110113.95

 

0.75

 

398335.92

 

0.37

 

Nonindustrial

112741.78

 

4.95

 

75217.50

 

6.54

 

State

1205.76

 

13.80

 

3658.56

 

6.64

Thinning

 
 
 
 
 

Federal

10137.58

 

3.59

 

53948.24

 

2.07

 

Industry

49545.26

 

2.45

 

125481.28

 

1.70

 

Nonindustrial

19518.80

 

9.64

 

11684.69

 

10.48

 

State

6745.10

 

6.27

 

18498.18

 

3.67

Combined

 
 
 
 
 

Federal

11746.20

 

2.76

 

56084.37

 

1.59

 

Industry

159567.75

 

0.79

 

523474.03

 

0.50

 

Nonindustrial

132236.50

 

4.51

 

86915.75

 

5.56

 

State

7965.02

 

5.67

 

22201.66

 

2.80


 

TABLE A4: Median %CV values of inventory across all age distributions for periods 1, 10, and 20 in Mideast and Midwest megasheds.

   
Mideast
Midwest
   

   
Period 1
Period 10
Period 20
Period 1
Period 10
Period 20
Total  
  Federal
0.08
0.93
1.99
0.04
0.12
0.26
  Industry
0.56
1.61
2.09
0.73
2.09
1.67
  Nonind.
1.21
3.14
4.56
2.35
3.75
2.68
  State
0.63
2.72
4.38
0.53
1.80
2.69
Softwood  
  Federal
0.12
1.55
3.11
0.06
0.14
0.45
  Industry
0.87
1.79
2.67
1.16
2.89
1.99
  Nonind.
1.83
5.00
4.96
3.92
6.34
6.06
  State
1.01
3.43
5.00
0.66
2.17
3.24
Mixed  
  Federal
0.00
0.22
2.37
0.05
0.25
7.62
  Industry
0.29
2.00
3.81
0.49
2.54
2.28
  Nonind.
0.00
4.20
5.82
1.61
5.01
6.74
  State
0.00
7.04
19.87
0.79
3.63
12.72
Hardwood  
  Federal
0.00
0.00
0.00
0.03
0.14
0.00
  Industry
0.00
0.00
0.00
0.08
1.70
0.00
  Nonind.
0.00
1.66
7.95
0.03
5.35
13.42
  State
0.00
0.00
0.00
0.00
0.00
0.00



[Back to A017-002]