# Importance of spatial predictor variable selection in machine learning applications -- Moving from data reproduction to spatial prediction

arXiv: https://arxiv.org/abs/1908.07805 · HTML: https://mlsanity.com/paper/1908.07805/

## Key facts

- Cited by 24 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 21 Aug 2019.

## Built on by (most cited first, 24 of 24)

- [Better, Not Just More: Data-Centric Machine Learning for Earth Observation](https://mlsanity.com/paper/2312.05327/) (2023)
- [Evaluation Challenges for Geospatial ML](https://mlsanity.com/paper/2303.18087/) (2023)
- [Contrasting local and global modeling with machine learning and satellite data: A case study estimating tree canopy height in African savannas](https://mlsanity.com/paper/2411.14354/) (2024)
- [CITYREP: A Unified Benchmark for Urban Representations Across Cities, Tasks, and Modalities](https://mlsanity.com/paper/2605.26036/) (2026)
- [Continental-scale land cover mapping at 10 m resolution over Europe (ELC10)](https://mlsanity.com/paper/2104.10922/) (2021)
- [Earthquake Nowcasting with Deep Learning](https://mlsanity.com/paper/2201.01869/) (2021)
- [DeepTopoNet: A Framework for Subglacial Topography Estimation on the Greenland Ice Sheets](https://mlsanity.com/paper/2505.23980/) (2025)
- [Using Machine Learning to generate an open-access cropland map from satellite images time series in the Indian Himalayan Region](https://mlsanity.com/paper/2203.14673/) (2022)
- [A path in regression Random Forest looking for spatial dependence: a taxonomy and a systematic review](https://mlsanity.com/paper/2303.04693/) (2023)
- [Challenges in data-based geospatial modeling for environmental research and practice](https://mlsanity.com/paper/2311.11057/) (2023)
- [The CAST package for training and assessment of spatial prediction models in R](https://mlsanity.com/paper/2404.06978/) (2024)
- [Improving Greenland Bed Topography Mapping with Uncertainty-Aware Graph Learning on Sparse Radar Data](https://mlsanity.com/paper/2509.08571/) (2025)
- [Autocart -- spatially-aware regression trees for ecological and spatial modeling](https://mlsanity.com/paper/2101.08258/) (2021)
- [A Machine Learning Approach for Material Type Logging and Chemical Assaying from Autonomous Measure-While-Drilling (MWD) Data](https://mlsanity.com/paper/2202.02959/) (2022)
- [Variable importance measure for spatial machine learning models with application to air pollution exposure prediction](https://mlsanity.com/paper/2406.01982/) (2024)
- [Enhancing Poverty Targeting with Spatial Machine Learning: An application to Indonesia](https://mlsanity.com/paper/2503.04300/) (2025)
- [Mapping Drivers of Greenness: Spatial Variable Selection for MODIS Vegetation Indices](https://mlsanity.com/paper/2602.07681/) (2026)
- [Decoupling Distance and Networks: Hybrid Graph Attention-Geostatistical Methods for Spatio-temporal Risk Mapping](https://mlsanity.com/paper/2603.08393/) (2026)
- [Evaluating Supervised Machine Learning Models: Principles, Pitfalls, and Metric Selection](https://mlsanity.com/paper/2604.13882/) (2026)
- [STeMP: Spatio-Temporal Modelling Protocol](https://mlsanity.com/paper/2607.20592/) (2026)
- [The Cost of a Physics Prior Is Bounded by the Ablation Gap](https://mlsanity.com/paper/2608.21059/) (2026)
- [Forward-Facing Near-Infrared Adds Little to Colour for Farm-Machinery Traversability: A Site-Disjoint Evaluation of Sensor-Dependent Spatial Leakage](https://mlsanity.com/paper/2609.13265/) (2026)
- [GeoDose-CP: Graph-Local Conformal Inference for Continuous-Treatment Earth Observation](https://mlsanity.com/paper/2609.28895/) (2026)
- [Autoregressive Drillhole Modelling Under Distribution Shift](https://mlsanity.com/paper/2610.01204/) (2026)

## Abstract (arXiv)

Machine learning algorithms find frequent application in spatial prediction of biotic and abiotic environmental variables. However, the characteristics of spatial data, especially spatial autocorrelation, are widely ignored. We hypothesize that this is problematic and results in models that can reproduce training data but are unable to make spatial predictions beyond the locations of the training samples. We assume that not only spatial validation strategies but also spatial variable selection is essential for reliable spatial predictions. We introduce two case studies that use remote sensing to predict land cover and the leaf area index for the "Marburg Open Forest", an open research and education site of Marburg University, Germany. We use the machine learning algorithm Random Forests to train models using non-spatial and spatial cross-validation strategies to understand how spatial variable selection affects the predictions. Our findings confirm that spatial cross-validation is essential in preventing overoptimistic model performance. We further show that highly autocorrelated predictors (such as geolocation variables, e.g. latitude, longitude) can lead to considerable overfitting and result in models that can reproduce the training data but fail in making spatial predictions. The problem becomes apparent in the visual assessment of the spatial predictions that show clear artefacts that can be traced back to a misinterpretation of the spatially autocorrelated predictors by the algorithm. Spatial variable selection could automatically detect and remove such variables that lead to overfitting, resulting in reliable spatial prediction patterns and improved statistical spatial model performance. We conclude that in addition to spatial validation, a spatial variable selection must be considered in spatial predictions of ecological data to produce reliable predictions.