Skip to contents

Compute fuzzy similarity of differences for transition validation

Source: R/validation.R
calc_transition_similarity.Rd

Implements the "similarity of differences" approach from Dinamica EGO (Soares-Filho et al. 2009), after the fuzzy set comparison of Hagen (2003). Compares the spatial pattern of a specific transition in observed vs simulated maps, allowing for spatial tolerance.

Usage

calc_transition_similarity(
  initial_map,
  observed_map,
  simulated_map,
  from_class,
  to_class,
  window_size = 11L,
  use_exp_decay = TRUE,
  decay_divisor = 2
)

Arguments

initial_map

SpatRaster, initial LULC state

observed_map

SpatRaster, observed final LULC state

simulated_map

SpatRaster, simulated final LULC state

from_class

Integer, initial class of transition

to_class

Integer, final class of transition

window_size

Integer, size of moving window (must be odd). Default 11.

use_exp_decay

Logical, use exponential decay? Default TRUE.

decay_divisor

Numeric, attenuation factor. Default 2.

Value

List with:

  • observed_change: Logical SpatRaster, TRUE where the transition was observed

  • simulated_change: Logical SpatRaster, TRUE where the transition was simulated

  • similarity: Minimum of the two directional similarities; NA if the transition happened in neither map

  • similarity_observed_to_simulated: Mean membership of observed changes in the simulated change

  • similarity_simulated_to_observed: Mean membership of simulated changes in the observed change

  • similarity_map: SpatRaster of the observed-to-simulated membership at each observed change (NA elsewhere)

  • n_observed: Number of cells with observed transition

  • n_simulated: Number of cells with simulated transition

Details

Both maps are reduced to where the transition happened. For each cell that changed in one map, its membership in the other map's change is the distance decay to the nearest cell that changed there, within the window: 1 for a change in the same cell, exp(-d / decay_divisor) for one d cells away (or 1 anywhere in the window without decay), 0 if there is none. The directional similarity is the mean of these memberships over the changed cells of the first map only; cells that changed in neither map do not enter it. The overall similarity is the minimum of the two directions, so a simulation cannot score well by placing much more (or much less) change than was observed.

References

Hagen, A. (2003). Fuzzy set approach to assessing similarity of categorical maps. International Journal of Geographical Information Science, 17(3), 235-249. https://doi.org/10.1080/13658810210157822

Soares-Filho, B. S., Rodrigues, H. O., & Costa, W. L. (2009). Modeling Environmental Dynamics with Dinamica EGO. Centro de Sensoriamento Remoto, UFMG.