by Kim Iles

One of the people at the recent OSU short course was not a regular cruiser, and asked “what is the sequence of steps for carrying out a Variable Plot cruise”. It’s a fair question, so here is one attempt to do that.

In all cases, make sure your computer program can handle the approach (and if not, lobby to fix it - or get another program). There is usually some way to fool the computer into doing what you need, but it should not be necessary. Computer people are supposed to make things easy for you, not the other way around.

zzzzz

1. What is the area to be cruised? A large tract probably means a VP cruise, while a small one would suggest 3P sampling. With VP cruising, you must measure the area that will be sampled. Any error for stand area will cause a proportional error in the cruise.
    
2. If the area can be broken down into strata that have different characteristics (and can be recognized from the ground) you might stratify the area to be cruised. This is really doing separate cruises, so you can cruise the strata in different ways if you wish. Sometimes you stratify just so you can report strata separately. What people want to know, and how they might still change the logging boundaries, is important to consider.
   
   Taking out small blank areas and roads is quite possible with GIS systems, but the “edge” this causes can be more trouble than it’s worth. The areas removed may complicate the field sampling process too much.
    
3. Do you know how variable the stand is? If you have cruised similar stands before, they can suggest a variability (measured by CV) for both the tree count and VBAR. If you still measure every tree on every plot, there is a CV that your cruise program should show. For count/measure plots there is a similar approach called “Johnson’s method” that gives you a single CV. The most efficient way to do VP sampling is to do the counting for basal area separately from the measuring that converts basal area to other values.
   
   The CV of the Tree Counts is easy to get using a few tree counts in the area, but is often in a narrow range you can find from your past results. The average CV of VBAR is often about the same as the CV of tree heights chosen with a prism multiplied by about 1.3 (check this with your own data measurement methods).
   
   In general, if you have CVs of 50% for Tree Count and 25% for VBAR it suggests measuring 2 count plots for every measured tree. That would be about every 10th tree. You might do them more frequently to get better grade data, but this would be the right ratio for volume information.
   
   Some folks visit the area to get this CV information before formal cruising starts, and they can check access and other interesting things at the same time.
    
4. A) How much time and staff can you afford to spend on the project? B) What SE% would that level of effort give you? C) Can you live with that result?
   
   If that SE% is much too big or too small, you can also compute a sample size based on some SE% that you would like. One SE% is CV divided by the square root of the number of plots (for basal area or the number of measured trees for VBAR). The smart thing is probably to use the STAR_BAR EXCEL program to suggest the numbers to measure for each of these.
    
5. Systematic sampling is always best, so to get the “square spacing” between plots for that sample size, divide the area in square feet (acres*43,560) by the number of plots, then take the square root of that number to get the plot spacing.
    

   
   (why not round to 100¸ or 1.5 chains?)
    

6. Would you like to change the suggested sample sizes to make the process more practical or easier? If so, use the second section of the STAR_BAR program to see if that answer gets you a result you can still live with. It probably does.
    
7. Get the GPS system to choose a random point in the area, then spread out the grid from that point in each direction. No need to get too upset if that does not give you exactly the sample size you were looking for.
   
   Lines of plots are also acceptable, and sometimes cost-effective. Use the random point to tell you where the first line should pass through, and set the other lines off from that one. This is a more correct method than using a “baseline” to lay out the sample lines.
    
8. If you want to count an average of about 6 trees per point, divide an approximate basal area of the stand by 6, and that gives you the BAF to use. You can use any BAF (or your thumb) to get the approximate Basal Area of the stand. Round the calculated BAF off to a convenient number based on the instruments you have for tree counts. An average of about 5-7 is most efficient, but use one where you will not miss trees – even if you are getting an average Tree Count of 1 or 2 in horrible brush.
   
   If you want every 5th tree, you can measure them by groups on every 5th plot, or use a second BAF (5 times larger) to potentially choose sample trees on each plot (this is the “Big BAF” method – which spreads out the measurements, and is as effective as measuring about twice as many sample trees in groups).
    
9. Time to cruise. Correct for edge effect when you are near the edge of the stand, and be careful about tree heights. Correct diameters make little difference to the volume, but might affect number of trees per acre, merchantable height, and grades - and DBH is easy to do correctly.
   
   We would suggest checking trees that are “borderline” unless you have done something to prove that you can guess them really well. Tree counts matter when calculating basal area, and directly affect volume and many other results.
    
10. Some people take the diameters of all the “in” trees. They then assign approximate heights to calculate gross volume. They also get a better DBH distribution in numbers of trees and logs. Just make sure that your assignment of heights by DBH does not bias the volume too much (see an earlier article in the newsletter called “Imaginary Heights in Variable Plot Cruising”).
     
11. Now examine the results. SE% not good enough? Really? Go back to *BAR and see how many extra count plots would need to give you the result you want. Those have to be spread over the whole area when you go back, but perhaps it is worth it to you. Better yet, revisit whether you can live with the result. You can also put in the same mixture of plots you did before. An additional strip of plots to make up the additional plots is probably OK too, if it is located properly (see item 7).
     
12. Do you want answers on specific groups of trees? You can break down information in ways that the compiler does not normally report by simply dropping trees that do not apply, and rerunning the data. This also provides statistics for that computed group. When things are entangled inside a tree, like grades, this is too complicated. If you need statistics on combined grades, just call those grades the same thing and get the statistics by rerunning the data with that combined phony grade. There is no other simple way to correctly combine the sampling error for grades (or combined species, in most cases). Sometimes that information is important for marketing the stand and you want to know how accurate some combinations might be.
     
13. At some point, if the accounting system allows it, you want to check the cutout for the cruise. Bear in mind that the Sampling Error is actually less than about 1/3 of the “95% confidence interval” often stated (for random samples), and it is even lower when you use systematic sampling to intentionally spread out the plots or tree measurements. The average cutout value should be zero over time, if there is no bias involved.
    
    If you are consistently getting more error in your cutouts than the sampling error shows, then it is a bias in the measurements or calculations - or an error in the accounting system. Go looking for it. To change your cruising to meet cutouts that are bogus numbers does not make much sense. The cruise says what is in the stand, not what will be made from it. If you cannot identify the cause, there is not much you can do about that except to additionally state that correction when you report future results.
     

zzzzz

What matters in the cruise data?

If you do not know what affect an error or other decision will make, try using old data that has all the trees measured. Make the change or introduce the error, then rerun the data to see what results change. The rules for “what matters” developed for fixed plot sampling in former times are often no longer true.