If You Don’t Like Measuring Borderline Trees, What Can You Do?

 Checking borderlines does take time.  A friend in California (Kenny Lucas) was frustrated because he was spending too much time and client money measuring borderline trees.  He didn’t think he was missing too many trees, and decided to prove that it was not necessary – not just once, but on an ongoing basis.  It’s a great idea, and you might consider it.  His website is:  http://www.tsiwood.com (click: timber cruising / P-3-P cruising / PDF file). 

As those of you who have been to the short course at OSU might remember, we recommend that you either measure borderline trees (using a plot radius factor times the DBH) or do something to verify the extent of the bias you get from any counting errors.  If the error rate is shown to be small enough, you might choose to ignore that bias, as we do with small errors in volume tables and other field processes.  If the error is large enough, you should correct for it.  Ken reports that he makes a correction about half the time.  We have not checked the computer programs he is using, but we certainly think the approach makes sense. 

In Ken’s case, he is faced with brush problems and visibility difficulties from young redwoods that branch nearly to the ground.  He takes a sample to check the number of trees that are actually in the plot compared to the original call the cruiser has just made.  This quantifies the correction that must be made (or the size of the bias that is being ignored).  It sure would be nice to see what that comparison is for a year or more of work by a cruiser.  

To select sample plots, you can use several methods.  One is a 3P process, where you select plots proportional to the estimated number of “in” trees the cruiser calls.  One cruiser reports this over a two‑way radio, and another cruiser tells him whether to carefully check the trees.  You have to do something about checking the “zero tree” plots, since a 3P process would never select those.  The simple way is to make zero tree counts into a different strata and check one in 5 or some other fraction.  A simple PDA using an EXCEL spreadsheet can be used for this process, but a list of the numbers can also be printed and taken to the field (no water damage to the PDA that way).  The correction to stand basal area is the average ratio of each individual answer [true count ÷ estimated count], which is called the “average of means” approach in statistics books (or words to that effect).  Because low tree counts can give a very large correction, you might make zero or 1‑tree counts a separate strata or check them every time (since they are not much effort to check).   

A second and simpler option is to use a random selection of 1 plot in “X”.  1 in 7, for instance, might be checked.  Create a random number between 1 and 7.0, and (after the initial count is made) the plot is selected if the random number is between 0 and 1.  In EXCEL, the equation [ Rand( )*7 ] will produce these random numbers.  You can also print a list to take to the field.  The final simple correction for stand basal area is the sum of all carefully measured counts divided by the sum of the estimated trees.  This is called the “mean of averages” approach in statistics books.  No need for special procedures on the zero count plots with this method.   

Ken, however, correctly takes the comparison one step further, and compares the complete compiled answer using the carefully checked trees vs. the answer with the initial count.  This finds out if there is any meaningful difference in the VBARs, grade percentages, or any other changes from the quick tree selection rather than the more careful check of borderlines.  This gives you a better correction, and we recommend it.  It’s a good precaution, and is the same process as comparing check cruiser results vs. the initial cruiser results.  In many practical cases, differences are small enough to ignore.  It may be that when you do a full comparison a few times you become convinced that the simple correction for tree count is adequate.  Any tree count difference will account for most of the difference in stand volumes or other values.   

This is a good example of a practical approach that is done carefully and consistently.  We appreciate Ken Lucas, Thomas Blair, and Doug Maxey presenting us with their process and results.  They report considerable increases in plots per day, and are happy with the procedure.  It is a good example of accepting a possible bias when it can be proven to be small rather than working very hard to get the tree count exactly.  The “proven” part of that last sentence is the difference between this approach and an arrogant assumption that can get you into trouble.  Lots of people claim that they never miss trees or are “sure they cancel out”, but not many can show it.   

Kim Iles