A Rare Species Problem

A Canadian client encounters an occasional large cedar - about 10% of the stand basal area, and fewer in number. He was worried about that. He needed a good estimate of the volume in the two premium grades of the butt logs for marketing purposes.

It was a classic situation. The answer was well known - the difference was that this client was willing to apply that simple answer. So many times you just hear the standard reply of "well, we have never done that, and we don't want to change." Strangely, such folks never seem to improve their results.

The trees averaged about 160 cm in DBH (63 inches) and were about 43 meters tall (140 feet). Remember when we cruised that kind of timber in the U.S.? The principles apply to other rare items too, so perhaps you should read on …

I suggested that they stop every 25 meters (yards) and put in a count only for large cedar with premium grades in the butt log. Ignore all the other trees. That came out to about 14 "Big cedar count plots" per measured plot.

Could they identify these high grade trees quickly and with fair accuracy? Yes, they could, they assured me - although they were quite worried that many times they would get a zero tree count. And what about all the big trees that they would only count and never measure. Would that be a problem?

Most of the time they did get a zero count for large premium cedar even when the average tree count was about 6. Once in a while they would land in a clump. The tree count would have plenty of zeros and a high variability. Luckily, that variability was on something cheap and easy to do, so they were willing to do it. In addition, the count could be done with a large BAF to insure that these big cedar were not likely to be missed.

By putting in many counts, the basal area of that type of tree would be well known. Getting a good basal area is the big hurdle with any rare species (or rare anything - not just a species).

Off they went into the woods to do the work. When they came back they processed the results. You can do that as an entirely different cruise with just the large cedar data. Add that volume to the a cruise that processed all the other trees (ignoring the large cedar) to get the total for the tract. The main difference is that the regular cruise will have many fewer plots This will make the processing fairly simple, and it will work with any cruise compiler.

[Hint - if you pay for compilation by the plot, you can simply combine several of the plots. If you combine the data from 4 plots and record the BAF as 4 times smaller than it was - a 10 BAF rather than the 40 BAF that you actually used, for instance]

The results were classic. The client ran the results through the STAR_BAR spreadsheet to get the optimum mix of counts and measured plots. "Does it make sense that you would tally about 25 counts for every measured tree?", he asked. Wow, that seemed pretty extreme. It was true.

{You can get this spreadsheet from Kim Iles at kiles@island.net }

It turned out that the CV of the tree count was 180%. Not unusual for anything that is a bit rare. What it means, of course, is that you will need plenty of counts to get a good answer. If the basal area is off, so is everything else - value, grade volume, weight, etc, etc.

The VBAR (the only reason to measure trees) was less variable. "Less variable" hardly describes it, actually, the CV was 7.5% for gross volume. By the time you took off defect, waste, breakage, etc. it jumped up to 11.5%. That means that you need to measure about 12 trees for every 180 count plots you establish. One measured tree for every 15 counts.

"Why was the CV of the VBAR so consistent?", you might ask. Simple. The tree heights only varied about 9%, and that is the major component of VBAR variability. The same height implies about the same gross VBAR (give or take a bit). Graph it up for yourself if you have any doubt.

This was using the standard British Columbia Ministry of Forests "valuation" approach, which is not particularly astute about decay. They put trees into large categories and give the entire category the same percent of defect. In the U.S. (and hopefully in British Columbia in the near future) we adjust for the individual defect and its location, so we have a better estimate for each individual tree.

Would another way of doing the defect give more variability in the VBAR? Well, that would be the next thing to check. For second growth it is probably about right, but it is easy to find out. Simply divide each tree volume by its basal area, and find out how variable that VBAR is (an Excel spreadsheet will easily do this). Once you have that CV, it again gives your appropriate ratio of Count plots to measured trees.

Did they get enough trees? They measured 8 trees on 11 measure plots, with 152 count plots, and the optimum number was actually 6 trees for gross volume and about 10 trees for net volume. Looks like they hit it about right. If the values for the two grades were similar, this sample size is also correct for dollar value.

If your own data is saying the same thing, maybe you ought to believe it. At any rate, you ought to know the situation. How do you find out? Simply pick out some "rare" situation from one of your old cruises, and run only that count and measured data as a special compilation. That will tell you the CV for the tree count (it might be labeled as the CV of the basal area - but they are the same). You might have to work out the CV of the VBAR or $BAR by hand. Better yet, get after your compiler (perhaps in a nice way) to give you that information on a regular basis with every compilation.

For at least one company, this simple comparison was a big cost saver.