How Many Edge Trees Do You Have?
A friend of mine brought up a good point a few days ago. He wondered if the “walkthrough method” was working well enough. It’s a good edge effect correction, and an excellent way to properly balance the selection of trees correctly (something half plots and other techniques do not do well).
The question boiled down to “are people doing the field work well enough – and how can we check?” It can be a pain to check the distance on the other side of that tree “way out there” and see if it lands on a road or across some boundary. Is it getting done?
Good question. No technique, no matter how clever or correct will be a success if the field work is not carried out with enough care and attention. This brought out an interesting question. How many “border trees” occur in your cruising right now?
How many duplicate trees occur when you do half plots? Shouldn’t a roughly equal number occur with any newer system? Probably so (although not exactly, for several technical reasons). You at least ought to know the difference. In the course of computing your cruises are you seeing these kinds of answers? If not, why not? This kind of practical, consistent effort in quality control is really a good idea, and perhaps if you ask for this kind of thing often enough you will eventually get it.
Computer output should be telling you the kind of numbers that allow you to track the typical situation, and let you know when they start to change. The reason that compilers often do not provide these answers is that nobody asks for them persistently enough. I hear all the time that “Variable Plot samples are not good as good for computing the number of trees in stands as fixed plots are.” A quick check showed me that for cruising in my part of the world the number of trees was as well known as the basal area (which everyone thinks VP sampling is quite good for).
None of these folks had ever checked. They just repeated the folklore they had heard. Now it took some work to check, so maybe it is not surprising that people had not looked at this. The question might well be “if you are computing numbers of trees – why aren’t you reporting the sampling error for that answer?”
Compilation packages and other systems should be telling us this sort of thing every time we see a result. How else are we going to know if changes in procedure are working as expected or if systems are working “well enough.” We are awash in data, and yet are not getting the simple messages the data are trying to give us. If the tree count correction for edge effect is not about the same as the last technique I used, I want to know why – or why not.
So – what should you be tracking in your data?
Originally published February 2007
Regular Article Index