How rvalavi/blockcv works
blockCV is an R package for creating spatially or environmentally separated cross-validation folds and evaluating their suitability for spatial prediction. The main workflow is to provide spatial observations and optional raster or prediction-domain data, choose a fold strategy, then inspect fold balance, spatial separation, and environmental extrapolation; folds can also be visualised. Separate capabilities estimate spatial autocorrelation and explore block sizes. The README documents the feature relationships; the sampled source shows the registered C++ routines and their matrix helper, but not the R implementations or their internal calls.
Main entry points: cv_spatial.R Spatial blocks · cv_cluster.R Clustering folds · cv_group.R Group folds · cv_buffer.R Buffer folds
Components 19
Fold construction
- Spatial blocks
R/cv_spatial.R - Clustering folds
R/cv_cluster.R - Group folds
R/cv_group.R - Buffer folds
R/cv_buffer.R - LOO NNDM
R/cv_nndm.R - K-fold NNDM
R/cv_knndm.R - Fold results
Fold diagnostics
- Fold summary
R/cv_summary.R - Distance check
R/cv_distance.R - Similarity check
R/cv_similarity.R
Spatial scale
- Autocorrelation
R/cv_spatial_autocor.R - Block-size explorer
R/cv_block_size.R
Visualization
- Fold plots
R/cv_plot.R
Numerical routines
- Rcpp dispatch
src/RcppExports.cpp - NNDM routine
src/nndm.cpp - Similarity routine
src/similarity.cpp - Matrix helper
src/Lightweight_matrix.h
Other components
- Analyst
- Spatial inputs
Connections 28
- Analyst → Spatial blocks: chooses strategy
README.md - Analyst → Clustering folds: chooses strategy
README.md - Analyst → Group folds: chooses strategy
README.md - Analyst → Buffer folds: chooses strategy
README.md - Analyst → LOO NNDM: chooses strategy
README.md - Analyst → K-fold NNDM: chooses strategy
README.md - Spatial inputs → Spatial blocks: supplies observations
README.md - Spatial inputs → Clustering folds: supplies observations
README.md - Spatial inputs → Group folds: supplies observations
README.md - Spatial inputs → Buffer folds: supplies observations
README.md - Spatial inputs → LOO NNDM: supplies prediction domain
README.md - Spatial inputs → K-fold NNDM: supplies prediction domain
README.md - Spatial blocks → Fold results: creates folds
README.md - Clustering folds → Fold results: creates folds
README.md - Group folds → Fold results: creates folds
README.md - Buffer folds → Fold results: creates folds
README.md - LOO NNDM → Fold results: creates folds
README.md - K-fold NNDM → Fold results: creates folds
README.md - Fold results → Fold summary: summarised by
README.md - Fold results → Distance check: checked by
README.md - Fold results → Similarity check: checked by
README.md - Fold results → Fold plots: visualised by
README.md - Analyst → Autocorrelation: explores scale
README.md - Analyst → Block-size explorer: explores scale
README.md - Rcpp dispatch → NNDM routine: dispatches to
src/RcppExports.cpp - Rcpp dispatch → Similarity routine: dispatches to
src/RcppExports.cpp - NNDM routine → Matrix helper: uses
src/nndm.cpp - Similarity routine → Matrix helper: uses
src/similarity.cpp