Type: Package
Title: Fitting and Assessing Neighborhood Models of the Effect of Interspecific Competition on the Growth of Trees
Version: 0.2.0
Maintainer: Albert Y. Kim <albert.ys.kim@gmail.com>
Description: Code for fitting and assessing models for the growth of trees. In particular for the Bayesian neighborhood competition linear regression model of Allen (2020): methods for model fitting and generating fitted/predicted values, evaluating the effect of competitor species identity using permutation tests, and evaluating model performance using spatial cross-validation.
License: MIT + file LICENSE
Encoding: UTF-8
LazyData: true
RoxygenNote: 7.1.1
URL: https://github.com/rudeboybert/forestecology
BugReports: https://github.com/rudeboybert/forestecology/issues
Depends: R (≥ 3.2.4)
Suggests: testthat (≥ 3.0.0), knitr, rmarkdown, covr
Imports: magrittr, rlang, glue, ggplot2, stringr, dplyr, tidyr, purrr, ggridges, mvnfast, sf, sfheaders, snakecase, tibble, yardstick, blockCV, forcats, patchwork
Config/testthat/edition: 3
VignetteBuilder: knitr
NeedsCompilation: no
Packaged: 2021-10-01 20:47:00 UTC; akim04
Author: Albert Y. Kim ORCID iD [aut, cre], David Allen ORCID iD [aut], Simon Couch ORCID iD [aut]
Repository: CRAN
Date/Publication: 2021-10-02 13:30:05 UTC

Pipe operator

Description

See magrittr::%>% for details.

Usage

lhs %>% rhs

Identify trees in the buffer region

Description

Identify trees in the buffer region

Usage

add_buffer_variable(growth_df, direction = "in", size, region)

Arguments

growth_df

sf data frame

direction

"in" for buffers that are contained within region (default), "out" for buffers that contain region.

size

Distance to determine which neighboring trees to a focal tree are competitors. The units are assumed to be the same as the geometry variables in growth_df.

region

An sf polygon object of region to be buffered

Value

The same growth_df data frame but with a new boolean variable buffer indicating if a tree is in the study region buffer area. This uses compute_buffer_region() to define the boundary of the buffer region.

See Also

Other spatial functions: compute_buffer_region(), focal_vs_comp_distance()

Examples

library(tibble)
library(sfheaders)
library(ggplot2)

# Example square region to be buffered
region <- tibble(
  x = c(0, 0, 1, 1),
  y = c(0, 1, 1, 0)
) %>%
  sf_polygon()

# Example points
study_points <- tibble(
  x = runif(50),
  y = runif(50)
) %>%
  sf_point()

# Size of buffer
size <- 0.05

# Identify whether points are within size of boundary
study_points <- study_points %>%
  add_buffer_variable(direction = "in", size = size, region = region)

# Plot study points coded by whether they are within size of boundary
p <- ggplot() +
  geom_sf(data = region, fill = "transparent") +
  geom_sf(data = study_points, aes(col = buffer))
p

# Additionally, show buffer boundary in red
buffer_boundary <- region %>%
  compute_buffer_region(direction = "in", size = size)
p +
  geom_sf(data = buffer_boundary, col = "red", fill = "transparent")

Plot Bayesian model parameters

Description

Plot Bayesian model parameters

Usage

## S3 method for class 'comp_bayes_lm'
autoplot(object, type = "intercepts", sp_to_plot = NULL, ...)

Arguments

object

Output of comp_bayes_lm

type

A single character string for plot type with possible values "intercepts", "dbh_slopes", or "competition".

sp_to_plot

Vector of subset of species to plot

...

Currently ignored—only included for consistency with generic.

Value

focal_vs_comp with new column of predicted growth_hat

Examples

library(ggplot2)
library(ggridges)

# Load in posterior parameter example
data(comp_bayes_lm_ex)

# Plot beta_0, growth intercepts
autoplot(comp_bayes_lm_ex, type = "intercepts")

# Plot beta_dbh, growth-dbh slope
autoplot(comp_bayes_lm_ex, type = "dbh_slopes")

# Plot lambdas, competition coefficients
autoplot(comp_bayes_lm_ex, type = "competition")

Example cross validation grid

Description

This is an example cross validation grid. This is needed to create the focal versus comp data frame and run cross-validated models

Usage

blocks_ex

Format

A sf polygons

foldID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

geometry

Point location of the individual

See Also

Other example data objects: census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(ggplot2)
library(sf)
library(dplyr)

comp_dist <- 1

ggplot(blocks_ex) +
  geom_sf() +
  geom_sf(data = growth_spatial_ex)

focal_vs_comp_ex <- growth_spatial_ex %>%
  mutate(basal_area = 0.0001 * pi * (dbh1 / 2)^2) %>%
  create_focal_vs_comp(comp_dist, blocks = blocks_ex, id = "ID", comp_x_var = "basal_area")

Example input census data for package use

Description

This is example forest census data to be analyzed with this package.

Usage

census_1_ex

Format

A tibble

ID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

sp

Species of the individual

gx

x-coordinate meters from reference point

gy

y-coordinate meters from reference point

date

Date the stem was measured

codes

Code for additional information on the stem: M means the main stem of the individual tree and R means the stem was lost, but the tag was moved to another stem greater than DBH cutoff, this stands for resprout.

dbh

Diameter at breast height of the tree in cm

See Also

Other example data objects: blocks_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(dplyr)
library(stringr)

growth_ex <-
  compute_growth(
    census_1 = census_1_ex %>%
      mutate(sp = to_any_case(sp) %>% factor()),
    census_2 = census_2_ex %>%
      filter(!str_detect(codes, "R")) %>%
      mutate(sp = to_any_case(sp) %>% factor()),
    id = "ID"
  )

Michigan Big Woods research plot data

Description

The Big Woods data come from three censuses of a 23 ha forest research plots. All free-standing vegetation greater than 1 cm diameter at 1.3 m height (diameter at breast height; DBH) were tagged, identified, spatially mapped and had their DBH measured. The original census took place in 2003 and covered only 12 ha. A second census took place from 2008-2010 and expanded the plot to its current 23 ha. In the first and second censuses trees larger than 3.1 cm DBH were included. Finally a third census took place in 2014. In this census trees larger than 1 cm DBH were included. In the second and third censuses the original trees were found, recorded for survival, remeasured, and new individuals were tagged. This data frame has data from the second census (2008-2010).

Usage

census_2008_bw

Format

A data frame with 27193 rows and 8 variables:

treeID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

stemID

Stem number for a multi-stemmed individual. For all trees this starts at 1 and continues up from there. To uniquely identify a stem across the plot this value must be combined with treeID.

dbh

Diameter at breast

sp

Code for the species. See species_bw for scientific name.

gx

x-coordinate meters from reference point

gy

y-coordinate meters from reference point

date

Approximate date the stem was measured

codes

Code for additional information on the stem: M means the main stem of the individual tree; AL means the stem is alive but leaning or completely fallen over; B means the stem is broken and over half the canopy is assumed to be missing; and R means the stem was lost, but the tag was moved to another stem greater than DBH cutoff, this stands for resprout.

Details

This plot is part of the Smithsonian Institution's Forest Global Earth Observatory (ForestGEO) global network of forest research sites. For complete details on this dataset see its Deep Blue Data repository page.

See Also

Other Big Woods data: census_2014_bw, species_bw, study_region_bw

Other example data objects: blocks_ex, census_1_ex, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(ggplot2)
library(sf)

# Convert all 2008 plot stems to sf object
census_2008_bw_sf <- census_2008_bw %>%
  st_as_sf(coords = c("gx", "gy"))

# Plot stems with plot boundary
ggplot() +
  geom_sf(data = census_2008_bw_sf, size = 0.25)

Michigan Big Woods research plot data

Description

The Big Woods data come from three censuses of a 23 ha forest research plots. All free-standing vegetation greater than 1 cm diameter at 1.3 m height (diameter at breast height; DBH) were tagged, identified, spatially mapped and had their DBH measured. The original census took place in 2003 and covered only 12 ha. A second census took place from 2008-2010 and expanded the plot to its current 23 ha. In the first and second censuses trees larger than 3.1 cm DBH were included. Finally a third census took place in 2014. In this census trees larger than 1 cm DBH were included. In the second and third censuses the original trees were found, recorded for survival, remeasured, and new individuals were tagged. This data frame has data from the third census (2014).

Usage

census_2014_bw

Format

A data frame with 48371 rows and 8 variables:

treeID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

stemID

Stem number for a multi-stemmed individual. For all trees this starts at 1 and continues up from there. To uniquely identify a stem across the plot this value must be combined with treeID.

sp

Code for the species. See species_bw for scientific name.

dbh

Diameter at breast

gx

x-coordinate meters from reference point

gy

y-coordinate meters from reference point

date

Approximate date the stem was measured

codes

Code for additional information on the stem: M means the main stem of the individual tree; AL means the stem is alive but leaning or completely fallen over; B means the stem is broken and over half the canopy is assumed to be missing; and R means the stem was lost, but the tag was moved to another stem greater than DBH cutoff, this stands for resprout.

Details

This plot is part of the Smithsonian Institution's Forest Global Earth Observatory (ForestGEO) global network of forest research sites. For complete details on this dataset see its Deep Blue Data repository page.

See Also

Other Big Woods data: census_2008_bw, species_bw, study_region_bw

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(ggplot2)
library(sf)
library(dplyr)

# Convert all 2008 plot stems to sf object
census_2008_bw_sf <- census_2008_bw %>%
  st_as_sf(coords = c("gx", "gy"))

# Plot stems with plot boundary
ggplot() +
  geom_sf(data = census_2008_bw_sf, size = 0.25)

# Species-specific mortality between 2008 and 2014 censuses
census_2008_bw %>%
  left_join(census_2014_bw, by = c("treeID", "stemID"), suffix = c("_2008", "_2014")) %>%
  mutate(mortality = ifelse(is.na(dbh_2014), 1, 0)) %>%
  group_by(sp_2008) %>%
  summarize(mortality = mean(mortality), n = n()) %>%
  arrange(desc(n))

Example input census data for package use

Description

This is an example second census to be analyzed with the package.

Usage

census_2_ex

Format

A tibble

ID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

sp

Species of the individual

gx

x-coordinate meters from reference point

gy

y-coordinate meters from reference point

date

Date the stem was measured

codes

Code for additional information on the stem: M means the main stem of the individual tree and R means the stem was lost, but the tag was moved to another stem greater than DBH cutoff, this stands for resprout.

dbh

Diameter at breast height of the tree in cm

See Also

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(dplyr)
library(stringr)

growth_ex <-
  compute_growth(
    census_1 = census_1_ex %>%
      mutate(sp = to_any_case(sp) %>% factor()),
    census_2 = census_2_ex %>%
      filter(!str_detect(codes, "R")) %>%
      mutate(sp = to_any_case(sp) %>% factor()),
    id = "ID"
  )

Fit Bayesian competition model

Description

Fit a Bayesian linear regression model with interactions terms where

y = X \beta + \epsilon

\mu mean hyperparameter vector for \beta of length p + 1
V covariance hyperparameter matrix for \beta of dimension (p + 1) x (p + 1)
a shape hyperparameter for \sigma^2 > 0
b scale hyperparameter for \sigma^2 > 0

Usage

comp_bayes_lm(focal_vs_comp, prior_param = NULL, run_shuffle = FALSE)

Arguments

focal_vs_comp

data frame from create_focal_vs_comp()

prior_param

A list of ⁠{a_0, b_0, mu_0, V_0}⁠ prior hyperparameters. Defaults to a_0 = 250, b_0 = 250, mu_0 a vector of zeros of length p + 1, V_0 an identity matrix of dimension (p + 1) x (p + 1)

run_shuffle

boolean as to whether to run permutation test shuffle of competitor tree species within a particular focal_ID

Value

A list of ⁠{a_star, b_star, mu_star, V_star}⁠ posterior hyperparameters

Source

Closed-form solutions of Bayesian linear regression doi: 10.1371/journal.pone.0229930.s004

See Also

Other modeling functions: create_bayes_lm_data(), predict.comp_bayes_lm(), run_cv()

Examples

library(dplyr)

# Load in focal versus comp
data(focal_vs_comp_ex)

comp_bayes_lm_ex <- focal_vs_comp_ex %>%
  comp_bayes_lm(prior_param = NULL, run_shuffle = FALSE)

Example bayesian competition model fit

Description

This object contains an example fitted Bayesian competition model outputted by comp_bayes_lm().

Usage

comp_bayes_lm_ex

Format

A list subclass containing the following elements:

prior_params

Prior parameters supplied to comp_bayes_lm()

post_params

Posterior parameters outputted by comp_bayes_lm()

terms

The formula object used in model fitting

See Also

comp_bayes_lm()

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(dplyr)
library(yardstick)

# Compare model predictions to observation
predictions <- focal_vs_comp_ex %>%
  mutate(growth_hat = predict(comp_bayes_lm_ex, focal_vs_comp_ex))

predictions %>%
  rmse(truth = growth, estimate = growth_hat) %>%
  pull(.estimate)

# Plot posterior parameters
comp_bayes_lm_ex %>%
  autoplot()

Compute buffer to a region.

Description

Compute buffer to a region.

Usage

compute_buffer_region(region, direction = "in", size)

Arguments

region

An sf polygon object of region to be buffered

direction

"in" for buffers that are contained within region (default), "out" for buffers that contain region.

size

Distance to determine which neighboring trees to a focal tree are competitors. The units are assumed to be the same as the geometry variables in growth_df.

Value

An sf polygon object of buffer

See Also

Other spatial functions: add_buffer_variable(), focal_vs_comp_distance()

Examples

library(tibble)
library(sfheaders)
library(ggplot2)

# Example square region to be buffered (as sf object)
region <- tibble(
  x = c(0, 0, 1, 1),
  y = c(0, 1, 1, 0)
) %>%
  sf_polygon()

# Size of buffer
size <- 0.05

# Compute "inwards" buffer
inwards_buffer_region <- region %>%
  compute_buffer_region(direction = "in", size = size)

# Compute "outwards" buffer
outwards_buffer_region <- region %>%
  compute_buffer_region(direction = "out", size = size)

# Plot all three regions:
ggplot() +
  geom_sf(data = outwards_buffer_region, col = "blue", fill = "transparent") +
  geom_sf(data = region, fill = "transparent") +
  geom_sf(data = inwards_buffer_region, col = "orange", fill = "transparent") +
  labs(title = "Regions: original (black), inwards buffer (orange), and outwards buffer (blue)")

Compute growth of trees

Description

Based on two tree censuses, compute the average annual growth in dbh for all trees.

Usage

compute_growth(census_1, census_2, id)

Arguments

census_1

A data frame of the first census.

census_2

A data frame of the second (later) census

id

Name of variable that uniquely identifies each tree common to census_1 and census_2 allowing you to join/merge both data frames.

Value

An sf data frame with column growth giving the average annual growth in dbh.

See Also

Other data processing functions: create_bayes_lm_data(), create_focal_vs_comp()

Examples

library(dplyr)
library(stringr)

growth_ex <-
  compute_growth(
    census_1 = census_1_ex %>%
      mutate(sp = to_any_case(sp) %>% factor()),
    census_2 = census_2_ex %>%
      filter(!str_detect(codes, "R")) %>%
      mutate(sp = to_any_case(sp) %>% factor()),
    id = "ID"
  )

Create input data frame for Bayesian regression

Description

This function "widens" focal-competitor data frames for use inside of package modeling functions, where each comp_sp inside of the comp list-column receives its own column with its associated total basal area.

This function is used internally by comp_bayes_lm() and predict.comp_bayes_lm() exported as a convenience for applications extending this package's functionality.

Usage

create_bayes_lm_data(focal_vs_comp, run_shuffle = FALSE)

Arguments

focal_vs_comp

data frame from create_focal_vs_comp()

run_shuffle

boolean as to whether to run permutation test shuffle of competitor tree species within a particular focal_ID

Value

Data frame for internal package use.

See Also

Other modeling functions: comp_bayes_lm(), predict.comp_bayes_lm(), run_cv()

Other data processing functions: compute_growth(), create_focal_vs_comp()

Examples

create_bayes_lm_data(focal_vs_comp_ex)

Create focal versus competitor trees data frame

Description

"Focal versus competitor trees" data frames are the main data frame used for analysis. "Focal trees" are all trees that satisfy the following criteria

  1. Were alive at both censuses

  2. Were not part of the study region's buffer as computed by add_buffer_variable()

  3. Were not a resprout at the second census. Such trees should be coded as "R" in the codes2 variable (OK if a resprout at first census)

For each focal tree, "competitor trees" are all trees that (1) were alive at the first census and (2) within comp_dist distance of the focal tree.

Usage

create_focal_vs_comp(growth_df, comp_dist, blocks, id, comp_x_var)

Arguments

growth_df

A compute_growth() output converted to sf object

comp_dist

Distance to determine which neighboring trees to a focal tree are competitors.

blocks

An sf object of a blockCV block output

id

A character string of the variable name in growth_df uniquely identifying each tree

comp_x_var

A character string indicating which numerical variable to use as competitor explanatory variable

Value

focal_vs_comp data frame of all focal trees and for each focal tree all possible competitor trees. In particular, for each competitor tree we return comp_x_var. Potential examples of comp_x_var include basal area or estimate above ground biomass.

Note

In order to speed computation, in particular of distances between all focal/competitor tree pairs, we use the cross-validation blockCV object to divide the study region into smaller subsets.

See Also

Other data processing functions: compute_growth(), create_bayes_lm_data()

Examples

library(ggplot2)
library(dplyr)
library(stringr)
library(sf)
library(sfheaders)
library(tibble)

# Create fold information sf object.
SpatialBlock_ex <-
  tibble(
    # Study region boundary
    x = c(0, 0, 5, 5),
    y = c(0, 5, 5, 0)
  ) %>%
  # Convert to sf object
  sf_polygon() %>%
  mutate(folds = "1")

# Plot example data. Observe for comp_dist = 1.5, there are 6 focal vs comp pairs:
# 1. focal 1 vs comp 2
# 2. focal 2 vs comp 1
# 3. focal 2 vs comp 3
# 4. focal 3 vs comp 2
# 5. focal 4 vs comp 5
# 6. focal 5 vs comp 4
ggplot() +
  geom_sf(data = SpatialBlock_ex, fill = "transparent") +
  geom_sf_label(data = growth_toy, aes(label = ID))

# Return corresponding data frame
growth_toy %>%
  mutate(basal_area = 0.0001 * pi * (dbh1 / 2)^2) %>%
  create_focal_vs_comp(comp_dist = 1.5, blocks = SpatialBlock_ex, id = "ID",
                       comp_x_var = "basal_area")

# Load in growth_df with spatial data
# See ?growth_ex for attaching spatial data to growth_df
data(growth_spatial_ex)
# Load in blocks
data(blocks_ex)

focal_vs_comp_ex <- growth_spatial_ex %>%
  mutate(basal_area = 0.0001 * pi * (dbh1 / 2)^2) %>%
  create_focal_vs_comp(comp_dist = 1, blocks = blocks_ex, id = "ID",
                       comp_x_var = "basal_area")

Return all pairwise distances between two data frames of trees

Description

Return all pairwise distances between two data frames of trees

Usage

focal_vs_comp_distance(focal_trees, comp_trees)

Arguments

focal_trees

An sf polygon object of the focal trees of interest

comp_trees

An sf polygon object of the competitor trees

Value

A data frame with three columns: focal_ID of focal tree, comp_dist of competitor tree, and dist of distance between them.

See Also

Other spatial functions: add_buffer_variable(), compute_buffer_region()

Examples

library(tibble)
library(ggplot2)
library(sf)

# Create toy example focal and competitor trees
focal_trees <- tibble(
  focal_ID = c(1, 2, 3),
  x = c(0.3, 0.6, 0.7),
  y = c(0.1, 0.5, 0.7)
) %>%
  st_as_sf(coords = c("x", "y"))

comp_trees <- tibble(
  comp_ID = c(4, 5, 6, 7),
  x = c(0, 0.2, 0.4, 0.6),
  y = c(0.6, 0.7, 1, 0.2)
) %>%
  st_as_sf(coords = c("x", "y"))

# Plot both sets of trees
ggplot() +
  geom_sf_label(data = focal_trees, aes(label = focal_ID), col = "black") +
  geom_sf_label(data = comp_trees, aes(label = comp_ID), col = "orange") +
  labs(title = "Focal trees in black, competitor trees in orange")

# Compute corresponding distances between the 3 focal trees and 4 competitor trees
focal_vs_comp_distance(focal_trees, comp_trees)

Example focal versus comp data frame

Description

This is an example focal versus comp data frame. The rows are focal trees which are repeated for all competitor trees within a specified distance from them. In this case that distance is 1. This is the focal versus comp for growth_spatial_ex.

Usage

focal_vs_comp_ex

Format

A tibble::tbl_df:

focal_ID

Tree identification number for the focal tree

focal_sp

Species of the focal tree

dbh

Diameter at breast height of the focal tree at the first census

foldID

The CV-fold that the focal tree is in

geometry

The point location of the focal tree

growth

The average annual growth of the focal tree between censuses

comp

A list-column: characteristics of the relevant competitor trees

The comp list-column contains tibble::tbl_dfs with columns:

comp_ID

Tree identification number for the competitor tree

dist

The distance between the focal and comp tree, this will be less than the max distance specified.

comp_sp

Species of the comp tree

comp_x_var

Numerical variable associated with comp tree

See Also

create_focal_vs_comp()

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

comp_bayes_lm_ex <- focal_vs_comp_ex %>%
  comp_bayes_lm(prior_param = NULL, run_shuffle = FALSE)

forestecology package

Description

Methods and data for forest ecology model selection and assessment

Details

See the README on GitHub

Example growth data frame for small example

Description

This is an example growth data frame formed from two census data frames. In this case it is made by combining census_1_ex and census_2_ex. The individuals alive in both censuses were linked by their tree ID.

Usage

growth_ex

Format

A sf spatial tibble

ID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

sp

Species of the individual

codes1

Code for additional information on the stem during the first census: M means the main stem of the individual tree and R means the stem was lost, but the tag was moved to another stem greater than DBH cutoff, this stands for resprout.

dbh1

Diameter at breast height of the tree in cm at the first census

dbh2

Diameter at breast height of the tree in cm at the second census

growth

Average annual growth between the two censuses in cm per year

codes2

Codes at the second census

geometry

Point location of the individual

See Also

compute_growth()

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(ggplot2)
library(dplyr)
library(sf)
library(sfheaders)
library(blockCV)

growth_ex %>%
  ggplot() +
  geom_sf()

growth_ex %>%
  group_by(sp) %>%
  summarize(mean(growth))

# Add buffer
growth_spatial_ex <- growth_ex %>%
  add_buffer_variable(direction = "in", size = 1, region = study_region_ex)

# Add cross-validation folds
fold1 <- rbind(c(0, 0), c(5, 0), c(5, 5), c(0, 5), c(0, 0))
fold2 <- rbind(c(5, 0), c(10, 0), c(10, 5), c(5, 5), c(5, 0))
blocks <- bind_rows(
  sf_polygon(fold1),
  sf_polygon(fold2)
) %>%
  mutate(foldID = c(1, 2))

SpatialBlock_ex <- spatialBlock(
  speciesData = growth_ex,
  verbose = FALSE,
  k = 2,
  selection = "systematic",
  blocks = blocks
)

# Add foldID to data
growth_spatial_ex <- growth_spatial_ex %>%
  mutate(foldID = SpatialBlock_ex$foldID %>% as.factor())

Example growth data frame with spatial data for small example

Description

This is an example growth data frame formed from two census data frames which has been updated with spatial data. It starts from growth_ex.

Usage

growth_spatial_ex

Format

A sf spatial tibble

ID

Tree identification number. This identifies an individual tree and can be used to connect trees between the two censuses.

sp

Species of the individual

codes1

Code for additional information on the stem during the first census: M means the main stem of the individual tree and R means the stem was lost, but the tag was moved to another stem greater than DBH cutoff, this stands for resprout.

dbh1

Diameter at breast height of the tree in cm at the first census

dbh2

Diameter at breast height of the tree in cm at the second census

growth

Average annual growth between the two censuses in cm per year

codes2

Codes at the second census

geometry

Point location of the individual

buffer

A boolean variable for whether the individual is in the buffer region or not

foldID

Which cross-validation fold the individual is in

See Also

compute_growth()

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_toy, species_bw, study_region_bw, study_region_ex

Examples

library(ggplot2)
library(dplyr)
library(sf)

comp_dist <- 1

ggplot() +
  geom_sf(data = growth_spatial_ex, aes(col = buffer), size = 2)

ggplot() +
  geom_sf(data = growth_spatial_ex, aes(col = foldID), size = 2)

# Create the focal versus comp data frame
focal_vs_comp_ex <- growth_spatial_ex %>%
  mutate(basal_area = 0.0001 * pi * (dbh1 / 2)^2) %>%
  create_focal_vs_comp(comp_dist, blocks = blocks_ex, id = "ID", comp_x_var = "basal_area")

Example input data for create_focal_vs_comp()

Description

An example sf of type generated by compute_growth()

Usage

growth_toy

Format

A sf spatial features polygon

See Also

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, species_bw, study_region_bw, study_region_ex

Make predictions based on fitted Bayesian model

Description

Applies fitted model from comp_bayes_lm() and returns posterior predicted values.

Usage

## S3 method for class 'comp_bayes_lm'
predict(object, newdata, ...)

Arguments

object

Output of comp_bayes_lm(): A list of ⁠{a_star, b_star, mu_star, V_star}⁠ posterior hyperparameters

newdata

A data frame of type focal_vs_comp in which to look for variables with which to predict.

...

Currently ignored—only included for consistency with generic.

Value

A vector of predictions with length equal to the input data.

Source

Closed-form solutions of Bayesian linear regression doi: 10.1371/journal.pone.0229930.s004

See Also

Other modeling functions: comp_bayes_lm(), create_bayes_lm_data(), run_cv()

Examples

library(dplyr)
library(sf)
library(ggplot2)

# Load in posterior parameter example
# and growth data to compare to
data(comp_bayes_lm_ex, growth_ex)

predictions <- focal_vs_comp_ex %>%
  mutate(growth_hat = predict(comp_bayes_lm_ex, focal_vs_comp_ex))

predictions %>%
  ggplot(aes(growth, growth_hat)) +
  geom_point() +
  geom_abline(slope = 1, intercept = 0)

Objects exported from other packages

Description

These objects are imported from other packages. Follow the links below to see their documentation.

ggplot2

autoplot

sfheaders

sf_polygon

snakecase

to_any_case

yardstick

rmse

Run the bayesian model with spatial cross validation

Description

This function carries out the bayesian modeling process with spatial cross-validation as described in Allen and Kim (2020). Given a focal-competitor data frame, it appends a column with predicted growth values.

Usage

run_cv(
  focal_vs_comp,
  comp_dist,
  blocks,
  prior_param = NULL,
  run_shuffle = FALSE
)

Arguments

focal_vs_comp

data frame from create_focal_vs_comp()

comp_dist

Distance to determine which neighboring trees to a focal tree are competitors.

blocks

An sf object of a blockCV block output

prior_param

A list of ⁠{a_0, b_0, mu_0, V_0}⁠ prior hyperparameters. Defaults to a_0 = 250, b_0 = 250, mu_0 a vector of zeros of length p + 1, V_0 an identity matrix of dimension (p + 1) x (p + 1)

run_shuffle

boolean as to whether to run permutation test shuffle of competitor tree species within a particular focal_ID

Value

focal_vs_comp with new column of predicted growth_hat

See Also

Other modeling functions: comp_bayes_lm(), create_bayes_lm_data(), predict.comp_bayes_lm()

Examples


run_cv(
  focal_vs_comp_ex,
  comp_dist = 1,
  blocks = blocks_ex
)

Phylogenic groupings and trait based clustering of various tree species

Description

A date frame mapping the species codes to their common names, scientific names, and families. This also includes a trait-based clustering of the species.

Usage

species_bw

Format

A data frame with 46 rows and 6 variables:

sp

The code for the species. Link to census_2008_bw and census_2014_bw with sp variable.

genus

Genus

species

Species epithet

latin

Scientific name

family

Family

trait_group

Clustering of species based on three traits rather than their evolutionary relationships. The traits are specific leaf area, maximum height, and wood density

Source

For more information on trait clustering see Allen and Kim 2020 "A permutation test and spatial cross-validation approach to assess models of interspecific competition between trees." doi: 10.1371/journal.pone.0229930PLOS One 15: e0229930.

See Also

Other Big Woods data: census_2008_bw, census_2014_bw, study_region_bw

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, study_region_bw, study_region_ex

Examples

library(dplyr)

# Original 2008 census data
census_2008_bw

# 2008 census data with additional species information
census_2008_bw %>%
  left_join(species_bw, by = "sp")

Bigwoods forest study region boundary

Description

Boundary region for Bigwoods defined in terms of (x,y) vertices of a polygon.

Usage

study_region_bw

Format

A sf spatial features polygon

See Also

Other Big Woods data: census_2008_bw, census_2014_bw, species_bw

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_ex

Examples

library(ggplot2)
library(sf)

# Convert all 2008 plot stems to sf object
census_2008_bw_sf <- census_2008_bw %>%
  st_as_sf(coords = c("gx", "gy"))

# Plot stems with plot boundary
ggplot() +
  geom_sf(data = census_2008_bw_sf, size = 0.25) +
  geom_sf(data = study_region_bw, color = "red", fill = "transparent")

Study region for example data

Description

Boundary region for small example data set defined in terms of (x,y) vertices of a polygon.

Usage

study_region_ex

Format

A sf spatial features polygon

See Also

Other example data objects: blocks_ex, census_1_ex, census_2008_bw, census_2014_bw, census_2_ex, comp_bayes_lm_ex, focal_vs_comp_ex, growth_ex, growth_spatial_ex, growth_toy, species_bw, study_region_bw

Examples

library(ggplot2)
library(sf)

# Convert stems to sf object
census_1_ex_sf <- census_1_ex %>%
  st_as_sf(coords = c("gx", "gy"))

# Plot stems with plot boundary
ggplot() +
  geom_sf(data = study_region_ex) +
  geom_sf(data = study_region_bw, color = "red", fill = "transparent")