Why monitor revegetation?
Revegetation can be expensive. You need to prepare the revegetation site, purchase plants from the nursery, have them installed, install protection like tree guards (if required), and maintain weeds around them until they are past their establishment period. Let’s assume that whole process is $10 per plant where it is $1 to prep the site, $2 to buy the plant, $2 to install, $2 for protection materials like stakes and guards and $3 to maintain for 12 months. It’s probably not accurate for your exact situation, but it’s a ballpark, and it’s a nice round number to use.
Now let’s assume you installed 20,000 plants to restore a 10 hectare site which includes a full suite of life forms from canopy trees, to understorey shrubs to ground storey like herbs, forbs and graminoids. Using the rate listed above, that equates to an investment of $200,000.
Though, for a range of reasons, sometimes some plants just fail to thrive, and losses occur. It is part of the revegetation industry. But at what point do you consider losses unsatisfactory requiring replanting? 5%? 10%? 20%? More? Losses of 30% would significantly impact the outcome of the revegetation program as by definition, almost a third of the plants are gone, so would need to be replanted. In our example, this would cost an additional $60,000 and take another 12 months to complete. Conversely, losses of 5% would be far more inconsequential and the decision may be made to forego replanting.
So, it is clear that an accurate estimation of your losses is important, not only to know your replanting rates, but also to intervene early if losses are beginning to occur or to be comfortable a planting campaign was a success.
Let’s set an arbitrary goal of detecting losses of 10%.
The pitfalls of revegetation monitoring
So, you understand revegetation monitoring is important and throw out a few quadrats. A 1 m x 1 m quadrat is clearly too small for your planting density as you’d need to average about five quadrats to detect a single plant (0.2 plants /m2). A 5 m x 5 m quadrat only averages a count of five plants, so you opt for 10 m x 10 m which averages 20. That sounds ok, right?
You don’t want to be there forever, so you decide on 10 of these quadrats. That means you’re sampling 1,000 m2 and if there are no losses, you are counting 20 plants in each (on average), or 200 overall. If there are less than 180 plants counted in total, you are below your 10% objective.
That all sounds reasonable on paper, but consider the complexity of repeatedly, and accurately laying out 10 m x 10 m quadrats in a natural landscape. Then searching and counting up to 20 plants accurately within the existing ground layer and emerging weeds. The job would certainly take the better part of a day so isn’t very efficient.
This is what many people think of when they think of a revegetation site. Mulch on the ground, very low-lying weeds (if at all present), clearly visible revegetated plants.
This is the same revegetation site a year later (though not the exact same location). The grasses are native and have been allowed to grow, but there is also a plethora of weeds. Detecting the revegetated plants is now a time consuming process.
And would this approach detect the plant losses of 10%? The short answer is no.
We use the formula n = SE s2 / d2, where n is the number of samples required, SE is the standard error, s is the variance of the data, and d is the percentage change to detect.
We put in an example range of values where 25% losses were observed (mean value of 15), SE = 1.5, v = 30 and d = 0.1 (10%) and we determine that we would need 20 quadrats to detect that change repeatedly … We have 10.
How can we improve our revegetation monitoring?
Let’s use the example above, but use a 5 m x 5 m quadrat. Now we expect 5 plants per quadrat. Inputting an example range where the mean is 3.75 (25% losses), SE = 0.375 and v = 7.5, we determine that we still only need 20 quadrats. Yet our survey area has dropped by a quarter from 100 m2 to 25 m2. This also means the time taken to find those plants should reduce by about a quarter too.
If we scale the range of values to match our example with the 10 m x 10 m quadrats, it is even lower, with 8 quadrats required.
This is all to say, what matters is the variation in your range of values. If you have uniform losses across your planting site, you will have a low variance and require a small number of quadrats. This makes sense as losses will occur evenly everywhere, but if variance is high (that is, you suffer losses in discrete clumps), then you will need more quadrats to detect that change.
Stratifying your site
If you suspect that you might not have a uniform planting area, then you might consider breaking it up into sections and assessing each one independently. In practice, this might occur if you have different vegetation communities and you suspect one might have more trouble establishing.
For example, if our 10 hectare planting site was nine hectares of forest, and one hectare of scrub, we might not be able to detect a change in the scrub unless we anticipated a high variance and had a high number of quadrats. We may need 30 to 40 quadrats overall. But if we treat them independently, we may need 10 quadrats in each. Sure, it is a high quadrat load in the scrub, but in total it is less (20) than trying to account for it across the whole project area (30-40).
Adjusting the detection rate
So far, we’re been using a detection rate of 10%. But does that meet the outcomes of your project? Are you implementing action at 10% losses? Or is it more severe likely 20% losses will trigger a replanting event? If so, lets adjust that. Using the values of our 10 m x 10 m quadrat before, if we drop the % change to detect from 10% to 20%, we only need five quadrats. Recall that previously it was 20 quadrats so we’ve reduced survey effort by 25% by relaxing our detection parameter and also reduced the survey size by 25% making each quadrat faster to survey.
Conversely, if we want to detect a 5% or 1% change, we would need 80 quadrats and 2000 quadrats, respectively.
Monitoring species or life forms
To this point, the discussion has assumed that plants are plants. Trees, herbs, graminoids. They are all the same and lumped into one bucket so if you detected 20% losses, you wouldn’t know if it was trees, graminoids, or what life form.
To add this resolution to your program, you may consider treating each life form (or species) like the vegetation types mentioned above. That is, have an independent test for each one.
The input values will all be changed because the density of each life form will be different, but the samples can be the same. So, quadrat 1 might have an independent count for each life form essentially counting as three independent tests. E.g. canopy trees = 2, shrubs = 8, ground storey = 10. So you’re on ground survey effort is the same, for a higher resolution survey (noting that you might need to use a high quadrat size to detect the sparser life forms).
A more sophisticated approach might be to implement a chi squared (x2) test which is worth mentioning, but beyond the scope of this article to discuss.
Revegetation monitoring frequency
This just comes down to how soon you want to detect the change. If you hope to catch change as it is occurring so that you can potentially intervene with a change in management techniques, then you would opt for a higher frequency of monitoring. But if all you want is to know how many survived at the end of the establishment phase and that’s it, then you might perform monitoring once. As a good balance, it can be useful to tie monitoring to other management such as weed control so that your resources are already deployed on site.
Sample type
We jumped straight in and started talking about quadrats merely because it is the go-to option for many people. But there are an assortment of options out there such as belt transects, wandering transects and more that may be faster to implement in the field. Transects also have the benefit of being able to use measuring wheels when the landscape allows for it which reduce set up time considerably.
Hopefully we’ve demonstrated that be investing a little bit of desktop time in the design of your revegetation monitoring program, you can save yourself a lot of time counting plants and improve the reliability of your results.
For an automated solution to revegetation monitoring, consider using a STA logger while planting and spraying.
If all of this is a bit much, then contact us for a quick chat about your revegetation monitoring program.