Disentangling spectral–terrain controls of soil organic carbon in tropical highland agroecosystems through comparative bagging and boosting ensemble learning

Accurate spatial identification of soil organic carbon (SOC) in tropical highland agroecosystems is essential for sustainable soil management but remains challenging due to complex terrain, intensive land use, and scale limitations of conventional methods. We develop a spatially explicit comparative evaluation of ensemble learning algorithms to map SOC and disentangle spectral–terrain controls in Batu City, East Java, a representative tropical highland agroecosystem. Sentinel-2A spectral indices and DEM-derived terrain derivatives were integrated with Random Forest, ExtraTrees, Gradient Boosting, XGBoost, LightGBM, CatBoost, and a Generalised Additive Model (GAM) as the statistical baseline. Tree-based ensembles showed stable performance under five-fold random cross-validation (RMSE = 0.55–0.65 g kg−1; MAE = 0.39–0.47 g kg−1; R2 up to 0.70) despite pronounced terrain-driven heterogeneity. Terrain attributes were consistently the highest-ranked individual predictors across all ensemble models. However, spectral-only models outperformed terrain-only models (R2 = 0.626 vs 0.456), while the combined model outperformed both (R2 = 0.679), indicating complementary rather than strictly hierarchical contributions of terrain and spectral predictors. The framework provides a methodologically reproducible basis for operational SOC monitoring, pending independent spatial validation in comparable tropical highland landscapes, enabling landscape-position–specific soil management, erosion-aware restoration planning, and climate-smart land-use strategies in rapidly transforming tropical highland regions.

Authors

Institutions

Publication Details

Journal
All Earth
Published
2026-11-16
DOI
https://doi.org/10.1080/27669645.2026.2762633
Primary Topic
Soil Geostatistics and Mapping
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Disentangling spectral–terrain controls of soil organic carbon in tropical highland agroecosystems through comparative bagging and boosting ensemble learning

Fuad Hartono, Kashif Ali Solangi, Erwin Ismu Wisnubroto, Aditya Nugraha Putra et al.
2 links
All Earth
Soil Geostatistics and Mapping
article

Disentangling spectral–terrain controls of soil organic carbon in tropical highland agroecosystems through comparative bagging and boosting ensemble learning

Fuad Hartono, Kashif Ali Solangi, Erwin Ismu Wisnubroto, Aditya Nugraha Putra, Subhan Arif Budiman, Putri Rizka Lestari, Anita Dwy Fitria, Hana Hanifa, Novandi Rizky Prasetya, Michelle Talisia Sugiarto, Dimas Sultan Syawala, Mochtar Lutfi Rayes, Imaniar Ilmi Pariasa, Tias Fikri Fadhil, Sudarto
article en

Abstract

Accurate spatial identification of soil organic carbon (SOC) in tropical highland agroecosystems is essential for sustainable soil management but remains challenging due to complex terrain, intensive land use, and scale limitations of conventional methods. We develop a spatially explicit comparative evaluation of ensemble learning algorithms to map SOC and disentangle spectral–terrain controls in Batu City, East Java, a representative tropical highland agroecosystem. Sentinel-2A spectral indices and DEM-derived terrain derivatives were integrated with Random Forest, ExtraTrees, Gradient Boosting, XGBoost, LightGBM, CatBoost, and a Generalised Additive Model (GAM) as the statistical baseline. Tree-based ensembles showed stable performance under five-fold random cross-validation (RMSE = 0.55–0.65 g kg−1; MAE = 0.39–0.47 g kg−1; R2 up to 0.70) despite pronounced terrain-driven heterogeneity. Terrain attributes were consistently the highest-ranked individual predictors across all ensemble models. However, spectral-only models outperformed terrain-only models (R2 = 0.626 vs 0.456), while the combined model outperformed both (R2 = 0.679), indicating complementary rather than strictly hierarchical contributions of terrain and spectral predictors. The framework provides a methodologically reproducible basis for operational SOC monitoring, pending independent spatial validation in comparable tropical highland landscapes, enabling landscape-position–specific soil management, erosion-aware restoration planning, and climate-smart land-use strategies in rapidly transforming tropical highland regions.

All EarthVol. 38(1)
Slovak University of Technology in Bratislava (SK), University of Brawijaya (ID), Hainan University (CN), Jenderal Soedirman University (ID), Universitas Jember (ID), Universitas Tribhuwana Tunggadewi (ID), Siliwangi University (ID)
Life on land, Climate action, Zero hunger
Soil Geostatistics and Mapping
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