create_training_data_with_buffer: Create training data from a negative buffer region
In tabrasel/WetlandTools: General utility functions for TerrainWorks

create_training_data_with_buffer

R Documentation

Create training data from a negative buffer region

Description

Given an input dataset and set of positive points, generate randomly sampled negative points and extract predictor values for positive and negative points. This sampling method is designed for a data set of positive points that is not considered a full inventory, such as the DOGAMI landslide inventory. First, we build a presumed negative region, with enough proximity to recorded landslide initiation points that we assume if a landslide had occurred there, it would have been recorded.

Usage

create_training_data_with_buffer(
  positive_points,
  predictors_raster,
  pos_buffer = 15,
  neg_buffer = 50,
  analysis_region_mask = NULL,
  negative_proportion = 1,
  extraction_method = "centroid",
  extraction_layer = NULL,
  rseed = NULL
)

Arguments

`positive_points`	SpatVector with locations of all points with positive class; that is landslide initiation points
`predictors_raster`	SpatRaster with a layer for each predictor variable
`pos_buffer`	The radius in meters around a positive point that is considered within the landslide initiation zone; negative sample points are excluded from within this buffer
`neg_buffer`	The radius around a positive point within which we are confident that no landslides were observed; negative (nonlandslide) points are sampled from the area between the pos_buffer and neg_buffer radius.
`analysis_region_mask`	SpatRaster with data values everywhere that points can be sampled from. All locations that should be excluded from sampling should be NA. If NULL, all cells which are non-NA for all layers of the predictors_raster will be used.
`negative_proportion`	Proportion of negative points to be generated compared to number of positive points
`extraction_method`	Method to use for extracting values from each point: "all", "centroid", "max", or "min". Ignored if extractionPoints is not polygon
`extraction_layer`	Layer to use for extracting value. Ignored if extraction_method = "centroid". Ignored if extraction_method is "all" or "centroid" or if extractionPoints is not polygon.
`rseed`	Optional integer to seed the random sampling. This allows exact "random" results to be reproduced multiple times. If no number is given, a random number will be chosen as a seed.