Fits tree growth or mortality models using neighborhood metrics based on differing neighborhood sizes, calculating mean square error of each model, and thereby providing useful data on the appropriate neighborhood size for analysis. It is strongly recommended that this sensitivity analysis or a similar one is used to select a neighborhood size for modeling because the appropriate neighborhood size varies between study systems and ecological processes.
select_nbhd_size( radii, map_data, model_process, tree_data, abiotic_data = NULL, focal_sps, dens_type, edge_handling = F, max_x, max_y, rare_sps = 100 )
| radii | Vector of the neighborhood radii to try. |
|---|---|
| map_data | Data frame containing tree coordinates. |
| model_process | Character determining type of process to be modeled
(i.e. |
| tree_data | Tree measurement data frame containing size or mortality
data for trees in |
| abiotic_data | Data frame containing abiotic data for each of the stands
in |
| focal_sps | Vector of species names (i.e. character strings) for which a neighborhood size should be selected. |
| dens_type | Method for calculating tree densities in each neighborhood.
Must take one of the values: |
| edge_handling | Boolean indicating whether trees whose neighborhood overlaps the stand boundary should have their neighborhood metrics adjusted to account for the unsampled portion of their neighborhood. If FALSE, these trees will be excluded from the analysis as focals. |
| max_x | Maximum expected x coordinate (i.e. should be 100 if the stands are 100 x 100 m). |
| max_y | Maximum expected y coordinate. |
| rare_sps | Minimum number of interactions a competitor species must
appear in to remain separate from the "RARE" category in the models
(see |
A list containing a number of named elements (exact number will be
the number of focal_sps plus one:
mse_vals is a data frame containing the mean square error of
the fitted model for each focal_sps using each neighborhood size
listed in radii.
each additional element of the list will be a plot of model mean square
error vs. neighborhood size for one of the focal_sps.
This function uses a data frame of tree coordinates (map_data) to
calculate a series of metrics to describe the neighborhoods around each tree.
It does this by applying the neighborhoods and then the
neighborhood_summary functions. It is therefore necessary to provide,
under the argument dens_type, instruction for the method that
neighborhood_summary should use to calculate densities. It is also
possible to conduct edge handling by setting edge_handling = T. This
process is repeated for each of the neighborhood sizes specified by the
argument radii. If edge_handling = F, trees whose neighborhood
overlaps the stand boundary will be excluded from modeling. Boundary overlap
is determined using the largest neighborhood size in radii to ensure
the number of focal trees is constant across neighborhood sizes.
The neighborhood metrics are then combined with the growth rates or mortality
status of their focal tree. If model_process = "growth" the
size_corr_growth output by growth_summary will be used as
the dependent variable (any trees with illogical negative annual growth rates
are excluded as focal trees). If the user also provides abiotic data for the
stands using the argument abiotic_data, these are also joined to the
neighborhood metric and growth/mortality data. The result is a design matrix
that is given to the growth_model or mortality_model function,
depending on the value of model_process. This entire process is
repeated for each neighborhood size (radii) and focal_sps,
with the mean square error of each resulting model being recorded. The
function outputs a list containing a data frame of all mean square error
values and a plot of mean square error vs. neighborhood size for each
focal_sps.
# See vignette "Selecting neighborhood size"