assured replicated coverage of important variables — habitat type series, site 

 preparation, overstory density, and level of spruce budworm defoliation. 

 Stands were randomly selected to ensure an unbiased choice among all can- 

 didate stands. Plots within stands were made as independent as possible 

 by recording plot-specific variables whenever feasible and by spacing plots 

 throughout the stand. 



We feel that unbiased stand selection procedures are crucial to developing 

 a regeneration model because results are very dependent on probabilities. 

 For example, the probabihty of stocking depends on the proportions of stocked 

 and nonstocked plots. These proportions could be easily changed if study 

 areas are not chosen in an unbiased manner. 



The importance of random stand selection does not mean that case history 

 studies of individual or small groups of stands are not needed. These studies 

 provide very useful information on silvics of species, regeneration success, 

 processes leading to successful regeneration, species composition, and so on. 

 Case history studies helped us design this study, but it is inappropriate to 

 use case history studies to develop a regeneration model because of potential 

 problems with low number of stands in the sample, narrow range of years 

 of treatment, and the possibility of unintended stand selection bias. 



Stands used to develop the regeneration model were chosen from a list of 

 stands harvested by commonly used methods in the Northern Rocky Moun- 

 tains. These stands were harvested as part of commercial timber removals 

 on Ranger Districts and State and corporate ownerships. Thus, the regen- 

 eration model should reflect what can be expected when actual silvicultural 

 prescriptions are implemented. 



Retrospective examination procedures may not be suitable for developing 

 regeneration models in all cases. Sometimes stand conditions prior to harvest 

 may be important in determining regeneration in the next rotation. For exam- 

 ple, the presence of serotinous lodgepole pine cones is important for lodgepole 

 pine regeneration. 



Retrospective examination procedures used in this study have the drawback 

 of sampling stands that are up to 20 years old. During that time, technology 

 could improve regeneration success. The same concern is true of a prospec- 

 tive study. Retrospective examinations can be conducted in a much shorter 

 time fi'ame — say 2 to 3 years to collect a sufficient amount of data. Retrospec- 

 tive studies, however, can only be done if adequate historical records have 

 been kept for each stand. 



Another important part of data collection is to sample from a range of treat- 

 ment years. This spreads out the sporadic effects of variables not explicitly 

 accounted for in the model — weather, seed crops, animals, insects, disease, 

 and so on. Data collected over a range of treatment years should result in 

 model predictions that are closer to long-term averages. 



An ecologically based land classification system was very important for 

 model development. The regeneration model makes extensive use of habitat 

 types and habitat type groups. Other alternatives would not have worked 

 as well. For example, site index does not provide the necessary information 

 to model regeneration success. Site index alone is not useful for determin- 

 ing the species that can occupy the site. 



Ecological land classifications provide a logical framework within which 

 to model regeneration success. The successional pathways to a climax for- 

 est can be modeled using an ecological classification system. Management 

 imphcations can be cataloged using the same system. The analyses and model 



38 



