INTRODUCTION 



Mortality rates, diameter increment rates, and 

 height increment rates are the key components of 

 all yield predictions. Mortality rates are the 

 weakest of the three components. Information on 

 diameter and height increment rates is readily 

 available from silvicultural research plots where 

 scientists have studied growth under a wide array 

 of environmental conditions and management 

 strategies. In addition, in most forest survey or 

 management planning inventories, information on 

 current diameter and height increment rates is 

 collected. Where such data are not readily avail- 

 able, diameter and height increment can be ob- 

 served in a single visit to the plot. Annual rings and 

 nodal scars provide the time scale that makes 

 these measurements possible on many species 

 growing in temperate forests. There is no com- 

 parable time scale for easily and accurately post- 

 dating mortality. Thus, data describing the occur- 

 rence of mortality is more difficult to obtain. 



Silvicultural research plots on which individual 

 trees have been measured for many years provide 

 a potential source of data for estimating relation- 

 ships between stand and tree characteristics and 

 mortality rates. However, use of such data is 

 severely weakened by the common practice of 

 deleting plots containing heavier than average 

 mortality from the experiment. Further, such data 

 are not available for all species and localities and 

 do not provide measures of current levels of 

 mortality. 



Many mortality surveys estimate the amount of 

 dead timber in a population by counting dead 

 trees, measuring the volume of dead trees, or 

 measuring the number of acres that contain a 

 substantial number of dead trees. Such informa- 

 tion is of little value in predicting future mortality. 



Sampling for mortality rates requires two types 

 of information not collected in the usual mortality 

 survey: counts of green trees and the year of death 

 for mortality trees. Mortality rates specify the 

 proportion of trees with a given set of character- 

 istics that are expected to die in a fixed time 

 interval; to know that proportion both green and 

 dead trees must be sampled. It is also necessary to 

 determine a time scale for mortality occurrence; 



this requires that the time of death be estimated 

 for each dead tree sampled. 



Once mortality rates are established, they may 

 be used for many purposes. Current mortality rate 

 models are designed to describe the occurrence 

 of mortality in a stand as the stand grows. How- 

 ever, if an estimate of number of dead trees or an 

 estimate of volume in dead trees is desired, mor- 

 tality rate models may be used in the context of a 

 stand inventory compilation system to produce 

 these estimates (much as volume equations are 

 used to estimate stand volume). 



Management planning inventory information in 

 the Northern Region of the USDA Forest Service is 

 collected on variable radius plots (basal area fac- 

 tor= 40) arranged in a 1 0-chain by 5-chain grid on 

 subcompartments of about 500 acres (200 ha) 

 selected with probability proportional to National 

 Forest acreage (Stage and Alley 1972). Mortality 

 data are collected by estimating which trees have 

 died in the past 5 years on each sample point. The 

 sampling design results in efficient estimates of 

 variables such as volume, diameter, diameter 

 growth rate, and height, but the design is ineffi- 

 cient for collecting information about mortality. 



This is true for several reasons. Under normal 

 conditions, mortality is a rare event. A rule of 

 thumb is that the expected normal mortality rate is 

 about 0.5 percent per year (that is, one tree out of 

 every 200 will die in ayear'stime). Also, mortality is 

 not uniformly distributed over the forest in either 

 time or space. An assumption of some form of 

 clustered distribution for mortality trees is prob- 

 ably more accuratethan an assumption of eithera 

 uniform or random distribution. Thus, most stand- 

 ard ground inventory systems (either variable 

 radius plot sampling designs or designs using 

 small fixed area plots) are not very efficient for 

 collecting mortality information. Finally, there is 

 some question as to the ability of field crews to 

 accurately estimate which trees have died within 

 the past 5 years. Postdating mortality on trees 

 dead for more than 2 years is difficult even for a 

 trained pathologist or entomologist because of 

 the great variability in deterioration of individual 

 trees (Miller and Keen 1960: Keen 1955). 



1 



