Where will urban expansion occur in El Salvador? A temporally validated gradient-boosting model with a spatial noise floor

We develop and validate a model that predicts where urban expansion will occur across the whole of El Salvador at one-hectare resolution. The problem is framed as supervised classification over cells that were not built up in the baseline year: the model is trained on the 1990–2010 transition derived from the Global Human Settlement Layer and validated on the observed 2010–2020 transition, which it never saw during fitting. Four specifications are compared on identical data: a null model based only on distance to existing built-up land, a logistic regression, a random forest and a gradient-boosting classifier. Because only 0.66% of eligible cells urbanized in the validation decade, performance is measured with the Figure of Merit, the area under the precision–recall curve read against the base rate, and precision at the top of the ranking; overall accuracy is not used. Gradient boosting reaches a Figure of Merit of 0.267 and a precision–recall area of 0.381, 58 times the base rate, against 0.056 and 0.067 for the null model. A bootstrap over 77 spatial blocks of 20 km gives a 95% interval of 0.355–0.412 for the precision–recall area and a minimum detectable improvement of 0.029. Judged against this noise floor, adding statutory, hazard and informal-settlement layers produces no demonstrable gain, and an improvement accepted earlier in the project under a relative-percentage criterion is withdrawn. Predictive capacity is concentrated in spatial inertia. A cellular automaton then allocates 14,893 ha of expected expansion to 2035, yielding a ranking of 11,684 candidate polygons with calibrated probabilities and a suitability index.Non-peer-reviewed preprint.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-18
DOI
https://doi.org/10.5281/zenodo.24345448
Primary Topic
Land Use and Ecosystem Services
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Where will urban expansion occur in El Salvador? A temporally validated gradient-boosting model with a spatial noise floor

Julio César Serrano Sandoval
2 links
Zenodo (CERN European Organization for Nuclear Research)
Land Use and Ecosystem Services
preprint

Where will urban expansion occur in El Salvador? A temporally validated gradient-boosting model with a spatial noise floor

Julio César Serrano Sandoval
preprint en

Abstract

We develop and validate a model that predicts where urban expansion will occur across the whole of El Salvador at one-hectare resolution. The problem is framed as supervised classification over cells that were not built up in the baseline year: the model is trained on the 1990–2010 transition derived from the Global Human Settlement Layer and validated on the observed 2010–2020 transition, which it never saw during fitting. Four specifications are compared on identical data: a null model based only on distance to existing built-up land, a logistic regression, a random forest and a gradient-boosting classifier. Because only 0.66% of eligible cells urbanized in the validation decade, performance is measured with the Figure of Merit, the area under the precision–recall curve read against the base rate, and precision at the top of the ranking; overall accuracy is not used. Gradient boosting reaches a Figure of Merit of 0.267 and a precision–recall area of 0.381, 58 times the base rate, against 0.056 and 0.067 for the null model. A bootstrap over 77 spatial blocks of 20 km gives a 95% interval of 0.355–0.412 for the precision–recall area and a minimum detectable improvement of 0.029. Judged against this noise floor, adding statutory, hazard and informal-settlement layers produces no demonstrable gain, and an improvement accepted earlier in the project under a relative-percentage criterion is withdrawn. Predictive capacity is concentrated in spatial inertia. A cellular automaton then allocates 14,893 ha of expected expansion to 2035, yielding a ranking of 11,684 candidate polygons with calibrated probabilities and a suitability index.Non-peer-reviewed preprint.

Zenodo (CERN European Organization for Nuclear Research)
Sustainable cities and communities
Land Use and Ecosystem Services
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon