Mapping the geogenic radon potential for Germany by machine learning

Petermann, Eric; Meyer, Hanna; Nussbaum, Madlene; Bossew, Peter (2021). Mapping the geogenic radon potential for Germany by machine learning Science of The Total Environment, 754, p. 142291. Elsevier 10.1016/j.scitotenv.2020.142291

1-s2.0-S0048969720358204-main.pdf - Published Version
Available under License Creative Commons: Attribution-Noncommercial-No Derivative Works (CC-BY-NC-ND).

Download (7MB) | Preview

The radioactive gas radon (Rn) is considered as an indoor air pollutant due to its detrimental effects on humanhealth. In fact, exposure to Rn belongs to the most important causes for lung cancer after tobacco smoking. Thedominant source of indoor Rn is the ground beneath the house. The geogenic Rn potential (GRP) - a functionof soil gas Rn concentration and soil gas permeability - quantifies what“earth delivers in terms of Rn”and rep-resents a hazard indicator for elevated indoor Rn concentration. In this study, we aim at developing an improvedspatial continuous GRP map based on 4448field measurements of GRP distributed across Germany. Wefittedthree different machine learning algorithms, multivariate adaptive regression splines, random forest and supportvector machines utilizing 36 candidate predictors. Predictor selection, hyperparameter tuning and performanceassessment were conducted using a spatial cross-validation where the data was iteratively left out by spatialblocks of 40 km*40 km. Thisprocedure counteracts the effectofspatial auto-correlation in predictorand responsedata and minimizes dependence of training and test data. The spatial cross-validated performance statistics re-vealed that random forest provided the most accurate predictions. The predictors selected as informative reflectgeology, climate (temperature,precipitation and soil moisture), soil hydraulic, soilphysical (field capacity, coarsefraction) and soil chemical properties (potassium and nitrogen concentration). Model interpretation techniquessuch as predictor importance as well as partial and spatial dependence plots confirmed the hypothesized domi-nant effect of geology on GRP, but also revealed significant contributions of the other predictors. Partial and spa-tial dependence plots gave further valuable insight into the quantitative predictor-response relationship and itsspatial distribution. A comparison with a previous version of the German GRP map using 1359 independent testdata indicates a significantly better performance of the random forest based map.

Item Type:

Journal Article (Original Article)


School of Agricultural, Forest and Food Sciences HAFL
School of Agricultural, Forest and Food Sciences HAFL > Resource-efficient agricultural production systems
School of Agricultural, Forest and Food Sciences HAFL > Agriculture
School of Agricultural, Forest and Food Sciences HAFL > Agriculture > Soils and Geoinformation


Petermann, Eric;
Meyer, Hanna;
Nussbaum, Madlene0000-0002-6808-8956 and
Bossew, Peter


G Geography. Anthropology. Recreation > G Geography (General)
Q Science > QD Chemistry
S Agriculture > SB Plant culture






Madlene Nussbaum

Date Deposited:

17 Nov 2020 14:23

Last Modified:

06 Jan 2023 12:12

Publisher DOI:





Actions (login required)

View Item View Item
Provide Feedback