Multi-Scale Performance Assessment

Evaluates model accuracy at increasing spatial aggregation scales. Predictions and observations are averaged within grid cells of increasing size, and performance is computed at each scale. Reveals scale-dependent error patterns.

Usage

borg_multiscale(
  data,
  predictions,
  target,
  coords,
  scales = NULL,
  metric = c("rmse", "mae", "rsq")
)

Arguments

data

Data frame with coordinates and target variable.

predictions

Numeric vector of predicted values (same length as nrow(data)).

target

Character. Target variable name.

coords

Character vector of length 2. Coordinate columns.

scales

Numeric vector. Grid cell sizes to evaluate. If NULL, auto-generated from 5 percent to 50 percent of extent in 8 steps.

metric

Character. "rmse" (default), "mae", or "rsq".

Value

A data frame with class "borg_multiscale" containing:

scale

Grid cell size

metric_value

Performance at this scale

n_cells

Number of occupied grid cells

Has an autoplot() method.

References

Riemann, R., Wilson, B. T., Lister, A., & Parks, S. (2010). An effective assessment protocol for continuous geospatial datasets of forest characteristics using USFS Forest Inventory and Analysis (FIA) data. Remote Sensing of Environment, 114(10), 2337-2352. doi:10.1016/j.rse.2010.05.010

Examples

set.seed(42)
d <- data.frame(x = runif(200, 0, 100), y = runif(200, 0, 100),
                 z = rnorm(200))
preds <- d$z + rnorm(200, sd = 0.5)
ms <- borg_multiscale(d, preds, target = "z", coords = c("x", "y"))
ms