Gino Licini mapping soil organic carbon stocks in Swiss subalpine Blatt and Ar du Tsan; 25% RMSE improvement with soil and vegetation types

Gino Licini mapping soil organic carbon stocks in Swiss subalpine Blatt and Ar du Tsan; 25% RMSE improvement with soil and vegetation types

New paper: how much does field mapping improve SOC predictions? - EPFL

New paper: how much does field mapping improve SOC predictions?

Import & publish the news

Are you sure you want to import this news into ?

This news will be sent to its subscribers.

Gino’s work is published in Soil Advances.

Freely available maps can only tell you so much about where mountain soil carbon hides. Gino’s new study in Soil Advances tests what field mapping adds on top.

Mapping soil organic carbon (SOC) stocks across mountain landscapes is essential for climate monitoring, but dense field sampling is rarely practical in remote alpine terrain. A new study in Soil Advances, led by Gino Licini, tests how much spatial prediction of SOC stocks actually improves when field-mapped soil and vegetation data are added to freely available environmental covariates.

Working at the same two subalpine sites as Bence’s study (Blatt and Ar du Tsan, roughly 2100-2200 m a.s.l.), the team sampled 49 locations per site and modeled SOC stocks using a regression-kriging approach. Stocks ranged widely, from 16.7 to 622 Mg C

ha, with the highest values concentrated in wet, low-lying areas underlain by organic soils.

Freely available covariates, elevation, curvature, vegetation index, and geology, improved predictions over a naive baseline, cutting prediction error (RMSE) from 137 to 114 Mg C

ha at Ar du Tsan and from 165 to 122 Mg C

ha at Blatt. Geology drove most of this gain, mainly because it separates slope from plain areas rather than through any direct control on carbon itself. Curvature, by contrast, added essentially nothing.

Adding field-mapped soil type covariates produced a further, site-dependent improvement. At Ar du Tsan, including soil and vegetation type together reduced error to 85 Mg C

ha, a 25 percent improvement over external covariates alone. At Blatt, soil type alone gave the best result (97.9 Mg C

ha), and adding vegetation type actually made predictions worse. Across both sites, vegetation type consistently underperformed as a predictor, since vegetation classes did not always align with the organic-versus-mineral soil boundaries that matter most for carbon.

The authors conclude that targeted field mapping of soil type, rather than blanket vegetation surveys, offers the most practical way to capture the small, high-carbon patches that disproportionately shape landscape-scale SOC budgets in subalpine terrain.

Congratulations on this nice study, Gino!

Swiss National Science Foundation (Grant No. 212056 )

Licini, G., Dienes, B., Aeppli, M. (2026). Mapping soil organic carbon stocks in Swiss subalpine soils: External covariates versus field-mapped soil and vegetation types. Soil Advances, 6, 100136. https:

Source: SOIL - Soil biogeochemistry laboratory

All SOIL - Soil biogeochemistry laboratory news

All Architecture, Civil and Environmental Engineering | ENAC news

Mapping soil organic carbon (SOC) stocks across mountain landscapes is essential for climate monitoring, but dense field sampling is rarely practical in remote alpine terrain. A new study in Soil Advances, led by Gino Licini, tests how much spatial prediction of SOC stocks actually improves when field-mapped soil and vegetation data are added to freely available environmental covariates. Working at the same two subalpine sites as Bence’s study (Blatt and Ar du Tsan, roughly 2100-2200 m a.s.l.), the team sampled 49 locations per site and modeled SOC stocks using a regression-kriging approach. Stocks ranged widely, from 16.7 to 622 Mg C

ha, with the highest values concentrated in wet, low-lying areas underlain by organic soils. Freely available covariates, elevation, curvature, vegetation index, and geology, improved predictions over a naive baseline, cutting prediction error (RMSE) from 137 to 114 Mg C

ha at Ar du Tsan and from 165 to 122 Mg C

ha at Blatt. Geology drove most of this gain, mainly because it separates slope from plain areas rather than through any direct control on carbon itself. Curvature, by contrast, added essentially nothing. Adding field-mapped soil type covariates produced a further, site-dependent improvement. At Ar du Tsan, including soil and vegetation type together reduced error to 85 Mg C

ha, a 25 percent improvement over external covariates alone. At Blatt, soil type alone gave the best result (97.9 Mg C

ha), and adding vegetation type actually made predictions worse. Across both sites, vegetation type consistently underperformed as a predictor, since vegetation classes did not always align with the organic-versus-mineral soil boundaries that matter most for carbon. The authors conclude that targeted field mapping of soil type, rather than blanket vegetation surveys, offers the most practical way to capture the small, high-carbon patches that disproportionately shape landscape-scale SOC budgets in subalpine terrain. Congratulations on this nice study, Gino!