Bruno, thanks. Was this study based solely on basal area? And can you explain how the 10% "tolerable uncertainty" was calculated? My Portuguese is non-existent.

Klaus, thanks also. Yes we have established our first pilot plot, which has already helped to refine our size, plus we have 20 1ha plots for stems >=10cm already in place. Is your 2-3ha recommendation based on achieving alpha diversity at a single location? If so, our challenge now is to determine the level of alpha diversity compromise that can be made in order to achieve sufficient sample size for measuring environmental trends and beta diversity. I think we are not alone in this dilemma.

This from GEM regarding plot size for biomass monitoring: "The coefficient of variation of basal area only really increases as sample plot size falls below ~0.4 ha (Phillips et al. 2009). Therefore, a plot size of 1 ha (planimetric) is commonly chosen because it is greater than the scale of typical tree fall events, but sufficiently small to sample individual soil types (although many RAINFOR plots are

Hi Andrew,

We found that 1.5m x 250m works well, though we can even reduce it to 1m in the RAPELD system. A single long thin plot is much easier to implement than many small sublpots. Originally, we had 4m x 250m and that is much too much work for the added precission.

All the Best

Bill

Different forest types would require a different size for properly addresses the natural variation. I found that transects didn't properly catch the variation in the eastern arc forests and switched to half ha plots to reduce the extreme variation in basal area and richness per plot. But I agree with Klaus that for community biodiversity studies, 1 ha or larger is better. Lots of work but if you really want robust estimates and the ability to capture longer term trends I feel you need plots at least 0.5 ha and not transects where spatial heterogeneity within the forest is high. Also if linking to remote sensing, much better with plot data, so much struggle to link transects with remotely sensed data and very poor results.

William, Jaclyn, thanks, and Klaus thanks again. I suspect this is a debate that may rumble.

Like William, I am personally keen on rectangular plots rather than square to better capture local diversity (but perhaps not so thin for the reason stated by Jaclyn). For stems >10cm dbh we have opted for 0.4ha based on the GEM minimum recommendation (also seems just about okay for remote sensing), plus from our existing plots, this size tends to capture most species sampled in 1ha plots placed at the same location. I will try to remember to post some stats on that here later. However, yes, I would agree with Klaus that to really capture the full local community composition we would need something larger. For us this can perhaps come at a later date, but for now we need a decent sample size to investigate environmental influence.

Now, for stems 1-5cm dbh we have ascertained that three 5m x 5m plots at a single location is too small (species accumulation curves not leveling off at all). Joining these plots up into a single long, thin plot may well work, so we will trial that. I am not however too concerned that the sapling plots are of sufficient resolution for remote sensing - they will mostly be under the forest canopy after all, plus they are intended to represent the 0.4ha plots as our unit for satellite reference.

Hi Jaclyn,

That is why we use RAPELD plots, they have the advantage of transects with less internal heterogenity than square plots. They are generally superior to other types of plots for most questions, but not for detailed spatial studies of individual interactions, in which very large (50 ha) square or circular plots would be better.

All the Best

Bill

The attached fractal-based sampling design might also interest you..

Thanks Rachakonda, and Bill again. We may end up with something close to the fractal design (i.e. two-three different plot sizes for different purposes), but adhering too closely to this system would limit our number of independent sample points across our elevation gradient. A modification of the RAPELD method would be more suitable for our purposes.

Thanks again Klaus. Do you have a typical sampling density that you employ for these circular regeneration plots in a given area (or relative to sampling of mature trees)? Presumably that depends on species richness/clumping, stem density, and aims of the study. I can also see the benefit of circular plots for long-term monitoring, as these would simplify difficult decisions regarding whether individual trees lie inside or outside of plots - circular plots would only need a distance measurement from the plot centrum, rather than using a compass to determine plot edge from a permanently marked corner.

Some excellent answers already provided, so I've not much to add, except .. long-term means 3 years, 10 years, 50 years? If long-term is decadal then over that time the questions you'll want to ask will change, so be careful of choosing a sampling method that is optimised too strongly to one question.

Secondly, can you physically mark the edges of the plots with something that will last for, say, 50 years? If you're going to rely on a GPS to relocate the plots a few years hence, then you need big plots (~1ha) to make sure that the errors in positioning will be small (compared to all the other errors). In Sierra Leone we experimented with circular plots but in dense vegetation they were excessively time consuming as we had to keep threading the tape back and forwards; transects are nice, but, there is the risk of turning them into foot-paths or at least damaging the ground flora. In the end we compromised on a few large square plots; replication will be poor, but we're hoping that we can relocate them just using GPS in a few years time, with some degree of confidence.

Thanks Richard. Yes we will certainly be marking plot (and maybe even subplot) corners, although probably not plot edges, instead relying on a compass for that each time. Relocation shouldn't be too much of a problem as our field team is pretty familiar with the area now, but yes we don't want to get too small for this reason, and we will use barrier tape and forestry paint to help make the plots visible.

What size did you choose Andrew Marshall. I am wondering what are the advantages and disadvantages of small (less than 1 Ha), 1 Ha and say 20 Ha plots for biodiversity assessment across the altitudinal gradient. Any pointers to that.

Hi Rishi. We are trialling a few alternative plot sizes and will make a decision soon based on our findings. We will summarise and post here, hopefully next year. However if your research aim is to sample an altitude gradient I would go for 1ha plots or even less. You will understand the gradient better with a large sample size rather than larger plots. A 20ha plot is a big undertaking and you would need several plots to adequately sample the species composition across the whole gradient. I hope that helps.

https://lemonindia.weebly.com/

Hi All, The reason we use long, thin plots in the RAPELD system, despite slightly more difficult localization, is that we can relate the saplings to adults, which may not even occur in small plots. Also, sapplings sometimes occur in groups because of dispersal limitation. If you are interested primarily in local questions (Are there more juveniles near an adult? Does a branch falling affect the density of juveniles?), round plots are OK. However, if you are interested in questions at wider scales (effect of soil, topography, aspect etc.) long thin plots smooth out the very local effects and are more informative, especially if they follow the altitudinal contour lines.

The PSPs I work with use a nested plot strategy that I consider effective (disclaimer - personal opinion). We sample trees >= 20 cm in the full 140 x 140 m plots, which is divided into 20 x 20 m quadrants. Within each 20 x 20m quadrants we have 10 x 10 m, 5 x 5 m, and 2 x 2 m nested plots that we use to sample trees 5 to 20 cm dbh, saplings from 2 to 5 cm dbh, and seedlings 2 cm DBH to 35 cm height, respectively.

There isn't just one answer to this. Any good answer will depend on the overall goals of the study, your constraints (time and costs) and their relation to different scales of variation and needs for statistical independence. If you already know there will be 60 plots then given your time and costs they should capture as many stems (bigger plots are better) and as much critical variation (perhaps more extended plots such as transects or more distributed sub-plots) as possible. But depending on your region of interest you may also need to replicate for any key variables that vary at larger scales and don't simply follow your "climate gradient" (e.g. geology, aspect, soils, slope, distance to roads, disturbance history ...).

If you want to capture geology, aspect, soils, slope etc. the RAPELD system will work well, especially if you plan to include bigger trees or other variables in the future. The explanation in the book that can be downloaded from the PPBio site explains in detail the advantages of long thin (modified Gentry) plots over many small plots, but if you do many small plots they can be placed at equal intervals along the center line of a RAPELD plot and they will still answer many questions.

Wilmer Diaz and I started a monitoring project of mortality and regeneration of saplings 1-5 cm in riparian forests of the Orinoco's floodplain using 5 plots 1x1 m2 in 100m2 plots within each riparian community. Unfortunately we could't continue.

I suggest that you can have 60 plots of 100 m2 (20x50m) distributed in communities of 1- lower (20 plots), 2- middle (20 plots) and 3- higher (20 plots) altitudes each with 5 subplots of 1x1m2. Logistic cost is reduced in that way and you can have 4 replicates in 5 potential different communities in each altitudinal range.