spatialcvR: Spatial Cross-Validation for Machine Learning

Overview

spatialcvR provides tools for spatial cross-validation and model evaluation for geospatial machine learning applications. It addresses the fundamental problem that geographic observations are often not independent, which can lead to over-optimistic performance estimates when using standard random cross-validation.

The Problem

In geospatial data analysis, nearby observations tend to be similar (Tobler's First Law of Geography). When using traditional random cross-validation, training and test sets may contain observations that are spatially close, leading to:

• Over-optimistic performance estimates
• Underestimation of generalization error
• Undetected spatial overfitting

The Solution

spatialcvR implements spatial cross-validation methods that explicitly control the separation between training and test observations:

• Spatial Block Cross-Validation: Divide space into rectangular blocks
• Buffered Cross-Validation: Exclude training observations within a buffer radius
• Spatial Clustering Cross-Validation: Group observations spatially
• Random Spatial Split: Baseline comparison method

Installation

# Install from CRAN (when available)
install.packages("spatialcvR")

# Install development version from GitHub
# devtools::install_github("yourusername/spatialcvR")

Quick Start

library(spatialcvR)

# Load sample data
data(sample_spatial_data)

# Create spatial folds
folds <- spatial_folds(
  data = sample_spatial_data,
  x = "longitude",
  y = "latitude",
  k = 5,
  method = "block"
)

# Detect spatial leakage
leakage <- detect_spatial_leakage(
  data = sample_spatial_data,
  folds = folds,
  x = "longitude",
  y = "latitude"
)

# Evaluate model performance
metrics <- spatial_metrics(
  observed = test_values,
  predicted = predictions
)

Key Features

• Multiple spatial CV methods: Block, buffered, clustering approaches
• Spatial leakage detection: Identify risky train/test proximity
• Model evaluation metrics: RMSE, MAE, R², MAPE
• Spatial residual diagnostics: Analyze error distribution in space
• Method comparison: Compare spatial vs. random CV
• Flexible input: Works with sf objects, data frames, or coordinates
• CRS-aware: Proper coordinate system handling

Use Cases

Perfect for researchers and practitioners working with:

• Agriculture and precision farming
• Environmental monitoring
• Climate and meteorology
• Hydrology and water resources
• Forestry and land management
• Urban planning
• Environmental health
• Ecology and biodiversity
• Remote sensing
• Geology and mineral exploration
• Risk assessment
• Any geospatial machine learning application

Documentation

• Introduction vignette
• Spatial cross-validation methods
• Spatial leakage detection
• Model evaluation

Project Status

This package is currently in development (version 0.1.0). The core functionality is being implemented and tested.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request or open an issue for bugs, feature requests, or questions.

License

MIT License - see LICENSE file for details.

Citation

If you use spatialcvR in your research, please cite it:

@software{spatialcvR,
  title = {spatialcvR: Spatial Cross-Validation for Machine Learning},
  author = {Mamadou SOW},
  year = {2026},
  url = {https://github.com/sowsalim01/spatialcvR}
}

Acknowledgments

This package was developed to address the need for robust spatial validation methods in R's geospatial machine learning ecosystem.

Contact

For questions, issues, or suggestions, please open an issue on GitHub.