Guest Editorial:

Production in Forest Inventory – What is it?

by Kim Iles 

You hear comments about “production” in the inventory business a lot.  There has to be a “product” in “production” – so what is it?  I would submit that it is NOT the number of plots you installed.  It is not hours worked, or pain and suffering, or the faithfully consistent application of methods that are not helping.   

The product of an inventory could well be considered to be:

1. an unbiased answer with

2. the smallest sampling error for the amount of money (or perhaps time) spent. 

When you get an unbiased answer with a smaller sampling error by any means, you are more productive.   

That might seem obvious to a few of you, but many folks seem to have missed that insight.  They insist on pounding in plots that do little good, and often at the expense of making correct measurements that would make the answer unbiased.  When there is too little time to get quantity done, the quality of work often suffers.  What is the point of guessing 200 form factors and getting a wrong answer by -3% with ±0.3% sampling error?  You might be far better off measuring just a few of them.

By “unbiased,” I mean that the average of all possible samples would be the right answer.  This normally requires correct measurements on a correct sample, with correct compilations and correct tables or factors applied.  At the very least, the field crew has the opportunity to put in a correct sample with correct measurements. 

In general, I think it makes sense to try living with a larger sampling error than to mess up the field measurements in the pursuit of a larger sample size.  A larger sample size will only get you a more final answer; doing the work right gets you to a more correct answer.   

If there is any way to measure production, it has to involve looking at the sampling error you can get when you do the work correctly.  A better sample design is one option to improve the sampling error.  What are you doing that is not improving the answer?  What is really essential in the field?  How can you get a better sampling error with less work?  What is it that you really need the sampling error for – volume?, value?, stumpage cost?, or what?

 The correct answer is the product, and the sampling error is the way to measure its precision.  Pounding in more plots is only one way to reduce that sampling error, and only helps when you can do those plots correctly (or prove that any biases are small).  If you think that numbers of plots IS production, you might be looking in the wrong direction.