How to make inventory initial file of FATES from forest inventory data

[suger-WU]

Dear model developers,

I am currently working on preparing a forest inventory file for inventory initialization in FATES and have a few questions regarding the process. My research site has forest inventory data where large, medium, and small trees were sampled differently:

(1) Large trees: All trees with a DBH greater than 40 cm were sampled across the entire treatment area.
(2) Medium trees: Trees and lianas with a DBH between 20–39.9 cm were sampled within 20 meters of six transects.
(3) Small trees: Trees with a DBH between 10–19.9 cm were sampled within 4 meters of the same six transects.

Since the forest inventory data was collected using transect sampling rather than quadrat/plot sampling, I manually defined plots of 50m × 50m, treating each defined plot as a patch. Specifically:
(1) all_plot_area = 50m × 50m (for large trees)
(2) sub_plot_area = 50m × 20m (for medium trees)
(3) small_plot_area = 50m × 4m (for small trees)

In generating the CSS file, I calculated the stem density (n) for the forest inventory data as follows:
(1)For large trees: n = 1 / all_plot_area
(2)For medium trees: n = 1 / sub_plot_area
(3)For small trees: n = 1 / small_plot_area

I wanna know if my above understanding and setup is correct? I have some concerns: (1) Given that the number of patches for medium and small trees is less than that for large trees, could this affect the overall representation of forest structure and composition. (2) Does the plot size I manually set (50m*50m) affect the most important result?

I would greatly appreciate your insights or suggestions on this matter.

Best,
Xiuyi

[mpaiao]

@suger-WU You are on the right track, the stem number density calculation must account for the sampling effort, so your numbers for the calculation of n are correct.

However, make sure each patch has all trees living in the same patch unit. I don't know your exact configuration, but let's say you have 10 plots with the sampling effort you described. In this case you want to create 10 patches, each of them with relative area of 0.1. Using plot 1 as an example, every tree in plot 1 should be part of the same patch 1, but medium- and small-sized trees will have a higher stem number density.

The main impact of the original plot size is how much local heterogeneity you will be able to capture. If you have the coordinates of individual trees, you can further divide your plots in smaller units. For tropical forests, I typically try to get something like 25x25m or so, but I adjust the exact size depending on the plot geometry. The idea behind this is to be consistent with the original ED assumption that patches are collections of gaps that have roughly the area of a large tree crown. That said, splitting a plot into smaller units is easier when you don't have multiple sampling efforts like in your case.

[suger-WU]

Hi Marcos,

Thanks for your reply!

My entire treatment area is rectangle area of 500m800m, and I manually defined plots of 50m × 50m. So I have 160 plots in total ((500800)/(50*50)). I have created corresponding patches with unique ID, with a relative area is 0.00625(=1/160).

I have individual tree coordinates from field sampling. In my current CSS file, not all patches contain large, medium, and small trees. Only patches near sample transects include all three size classes, as I defined patches based on the coordinates of medium and small trees. However, all patches contain large trees. Now, I realize this approach might introduce issues—some patches may contain only large trees without any medium or small trees.

To address this, I am considering whether I should copy the medium and small trees to all patches to ensure a more representative mix of tree sizes. Would this be a valid approach?

Additionally, I am thinking about reducing the plot size to 25m × 25m to better capture local heterogeneity.

I would greatly appreciate your insights.

[mpaiao]

I agree with you that the current approach will likely bias the forest structure. Copying the data from medium and small trees may be an option, but this has to be done carefully to avoid mixing understory with canopy of very different forests.

I believe you are using data from Tanguro, correct? For reference, I used Tanguro forest inventory to drive ED2 a long time ago. I think I ended up using only the transect areas where all trees with DBH ≥ 20cm were measured (but scaled the stem density for trees with DBH ≥ 5cm, similarly to what you described). That solution was not ideal either because I discarded data, but I thought it was safer because the sub-plots were still near the area I included.

As for the size, I would say, try to reduce the size further only if the plot/transect geometry allows for it to be done easily, otherwise, 50x50m should still be fine.