Cluster points together using the adespatial::constr.hclust algorithm.

Take a dataframe of points containing additional_variable_cols values and edges between points in the dataframe and cluster using the constr.hclust algorithm. This clustering is performed bottom-up based on a combined distance matrix of geographic and additional_variable_cols distances Only points connected by edges are able to cluster together. For additional information see insert citation/link. If points contains over 30,000 observations then interpolation will be used to reduce RAM usage. This process randomly samples 30,000 for clustering and assigns clusters to the remaining points using nearest neighbour interpolation.

Usage

constrained_hclust(
  points,
  edges,
  x_col = "X_standard",
  y_col = "Y_standard",
  additional_variable_cols = c("depth_standard"),
  id_col = "UNIQUE_ID",
  habitat_col = "habitat",
  distance_method = "manhattan",
  distance_alpha = 0.5,
  beta = -1,
  n_points = 204,
  n_clust = (round(nrow(points)/n_points)),
  method = "ward.D2"
)

Arguments

points

data.frame. Contains values for X and Y coordinates, as well as additional_variable_cols.

edges

matrix. Matrix containing edges between points in points.

x_col

character. Name of the column holding X coordinates. Default = "X_standard".

y_col

character. Name of the column holding Y coordinates. Default = "Y_standard".

additional_variable_cols

character vector. Names of additional columns to contribute to the distance matrix. Default = c("depth_standard").

id_col

character. Column holding ID value for the target reef (attached to the site_id values on output). Default = "UNIQUE_ID".

habitat_col

character. Column holding unique habitat values (attached to id_col value and site_id values on output). Default = "habitat".

distance_method

character. Distance matrix creation method. Default = "manhattan" (see dist()).

distance_alpha

float numeric. Weighting applied to additional_variable_cols distance matrix when combining with geographic distances. (1 - alpha) weighting is applied to the geographic distance matrix. Default = 0.5 (symmetric weighting).

beta

float numeric. Beta parameter used by adespatial::constr.hclust. Parameter value is only used if method == "flexible". Default = -1.

n_points

integer numeric. Desired number of points in resulting clusters. Used to calculate n_clust (number of output clusters). Value only used in n_clust specification. Default = 204.

n_clust

integer numeric. Number of clusters in result output. (Point to cut hierarchical clustering tree). Default = (round(nrow(points) / n_points)) (dividing habitat into clusters containing an average of 200 points).

method

character. Clustering method to be applied. See adespatial::constr.hclust() for more details. Default = "ward.D2".

Value

data.frame of points with allocated site_ids based on cluster outputs. site_id values are a combination of the id_col value, habitat_col value and the cluster allocation.