Why big square vegetation plots are best


I had to write an email recently to persuade someone that using bigger permanent vegetation survey plots in savannas was better than small plots. In our field of research one of the basic field measurements we do is to demarcate an area of savanna, tag all the tree stems gt5 cm diameter, and then return every few years to track growth and mortality of those trees.

The reasons for constructing small plots are mostly practical. Small circular plots are very quick to set up, with a single metal pole sunk into the ground at the centre of the plot. Then it’s just a case of running a tape measure out to the required radius and moving round with the tape measure measuring all the trees that are within the radius. It’s also fairly easy this way to determine the spatial distribution of tree stems by using a distance and compass angle coordinate system. Big plots on the other hand are quite slow to set up, because the trees get in the way of the tape measures, and there’s more chance of the tape measure drifting, which is especially problematic for square/rectangular plots where the corners have to be at 90 degrees. After a plot gets to be more than about 25 m in radius, I would advocate for square plots instead, as it actually becomes easier to set them up as squares at this size.

The problem is that small plots can lead to un-representative estimates of the biomass and species diversity of the landscape they are sampling.

Measuring competition effects: There’s been some work in our group on miombo woodland plots in Tanzania which suggests that tree-tree competition is an important determinant of tree growth, and that this competition effect peaks at about ~10-15 m radius from the focal tree, on average. In the case of a small plot (e.g. 20 m radius circle), you will only really be able to understand competition effects for a few individuals right at the centre of the plot. I’m particularly interested in spatial distribution of trees, and it’s something I would like SEOSAW to do more of. But we can’t do that with small plots.

Edge effects and the representative-ness of biomass estimates: The woody biomass per area estimates from small plots are much more influenced by the presence or absence of large trees than larger plots. To gather a representative estimate of the biomass per area of a landscape, it can actually require greater effort per unit area sampled to set up many small plots, than fewer big plots, as you need to sample less area overall from big plots. Additionally, edge effects are magnified when measuring smaller plots, as the perimeter:area ratio doesn’t scale geometrically, thus subjective decisions about whether a tree/stem is inside or outside a plot are more influential on plot level estimates of abundance, diversity, and biomass. Finally, from personal experience there is more chance that small plots will be opportunistically sited to include or exclude certain features of the woodland, like a thicket area, or large impressive trees, introducing bias when estimating landscape processes. So, replication is not such an issue as compared to sampling a few representative plots.

Matching with satellite data: In our lab group there are people working on estimating biomass from L-band radar backscatter relationships. Small plots are much worse for matching with these remotely sensed data. A previous PhD student found that the R^2 of the biomass estimation uncertainty peaked at about 1 ha (100x100 m), and was almost zero for plots lt0.2 ha (lt~25m radius circle). You just can’t fit as many data points in a smaller plot. These papers highlight similar issues. https://doi.org/10.3390/rs5031001 , https://doi.org/10.3390/rs10101586 .