Extended Data Table 1 XGBoost model performance measured using a spatial leave-one-out approach, as described in the Methods section

From: A globally consistent negative effect of edge on aboveground forest biomass

  1. Root Mean Squared Error (RMSE), R-squared (R2), and Mean Absolute Error (MAE) metrics were calculated using various spatial buffers on the test dataset. The test dataset consisted of a randomly selected 20% subset of the full, original data; however, each test point was evaluated individually using a spatially buffered leave-one-out cross-validation approach. Specifically, for each test point, a model was trained after removing all nearby data points within a specified spatial buffer. The mean number of the nearest grid cells from each test grid cell that were not used to train the model (that is excluded by the spatial buffer) is shown in ‘Number of grid cells excluded from the training dataset’.
Back to article page