Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data
This study presents the result of topsoil (010 cm) soil organic carbon (SOC) mapping in two areas of Moscow Region (2007 status): 1096 km2 Podolsky District, and 1101 km2 Serpukhovsky District. Based on 2007 legacy soil sampling data (n= 282) within these areas, we have created astatistical model...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Peoples’ Friendship University of Russia (RUDN University)
2024-12-01
|
| Series: | RUDN Journal of Agronomy and Animal Industries |
| Subjects: | |
| Online Access: | https://agrojournal.rudn.ru/agronomy/article/viewFile/20127/16458 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849337914224279552 |
|---|---|
| author | Yury A. Dvornikov Lukyan A. Mirniy Ekaterina S. Mukvich Kristina V. Ivashchenko |
| author_facet | Yury A. Dvornikov Lukyan A. Mirniy Ekaterina S. Mukvich Kristina V. Ivashchenko |
| author_sort | Yury A. Dvornikov |
| collection | DOAJ |
| description | This study presents the result of topsoil (010 cm) soil organic carbon (SOC) mapping in two areas of Moscow Region (2007 status): 1096 km2 Podolsky District, and 1101 km2 Serpukhovsky District. Based on 2007 legacy soil sampling data (n= 282) within these areas, we have created astatistical model between the target variable (SOC stocks, kg/m2) and numerous covariates (legacy maps and remote sensing data). GBM model has explained 56% of soil organic carbon stocks variability. Differences in stocks within different land use types were shown quantitatively. At the same time, the spectral reflectance in the near infrared band (B5) of Landsat‑5 TM made the greatest contribution in explaining the differences within individual types (among fallow lands and urbanized areas), and the spectral index NDVI has explained the spatial variability of soil organic carbon among forest ecosystems. The root mean square error of cross-validation (RMSEcv = 0.67 kg/m2) was chosen to describe the uncertainty of soil organic carbon stock prediction. According to the model, the total soil organic carbon stocks in the upper 10 cm soil layer of the Podolsky District were 2.65 0.72 Tg, for the Serpukhovsky District 2.77 0.73 Tg. |
| format | Article |
| id | doaj-art-80279348aa6f40b4ade24f7b729d5c9f |
| institution | Kabale University |
| issn | 2312-797X 2312-7988 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Peoples’ Friendship University of Russia (RUDN University) |
| record_format | Article |
| series | RUDN Journal of Agronomy and Animal Industries |
| spelling | doaj-art-80279348aa6f40b4ade24f7b729d5c9f2025-08-20T03:44:33ZengPeoples’ Friendship University of Russia (RUDN University)RUDN Journal of Agronomy and Animal Industries2312-797X2312-79882024-12-0119460261710.22363/2312-797X-2024-19-4-602-61717104Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy dataYury A. Dvornikov0https://orcid.org/0000-0003-3491-4487Lukyan A. Mirniy1Ekaterina S. Mukvich2https://orcid.org/0009-0004-5378-8775Kristina V. Ivashchenko3https://orcid.org/0000-0001-8397-158XRUDN UniversityInstitute of Physicochemical and Biological Problems of Soil Science RASInstitute of Physicochemical and Biological Problems of Soil Science RASInstitute of Physicochemical and Biological Problems of Soil Science RASThis study presents the result of topsoil (010 cm) soil organic carbon (SOC) mapping in two areas of Moscow Region (2007 status): 1096 km2 Podolsky District, and 1101 km2 Serpukhovsky District. Based on 2007 legacy soil sampling data (n= 282) within these areas, we have created astatistical model between the target variable (SOC stocks, kg/m2) and numerous covariates (legacy maps and remote sensing data). GBM model has explained 56% of soil organic carbon stocks variability. Differences in stocks within different land use types were shown quantitatively. At the same time, the spectral reflectance in the near infrared band (B5) of Landsat‑5 TM made the greatest contribution in explaining the differences within individual types (among fallow lands and urbanized areas), and the spectral index NDVI has explained the spatial variability of soil organic carbon among forest ecosystems. The root mean square error of cross-validation (RMSEcv = 0.67 kg/m2) was chosen to describe the uncertainty of soil organic carbon stock prediction. According to the model, the total soil organic carbon stocks in the upper 10 cm soil layer of the Podolsky District were 2.65 0.72 Tg, for the Serpukhovsky District 2.77 0.73 Tg.https://agrojournal.rudn.ru/agronomy/article/viewFile/20127/16458landsatstochastic gradient boostingreliefsoil organic carbonparameterizationspectral transformationmoscow region |
| spellingShingle | Yury A. Dvornikov Lukyan A. Mirniy Ekaterina S. Mukvich Kristina V. Ivashchenko Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data RUDN Journal of Agronomy and Animal Industries landsat stochastic gradient boosting relief soil organic carbon parameterization spectral transformation moscow region |
| title | Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data |
| title_full | Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data |
| title_fullStr | Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data |
| title_full_unstemmed | Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data |
| title_short | Mapping soil organic carbon stocks of different land use types in the Southern Moscow region by applying machine learning to legacy data |
| title_sort | mapping soil organic carbon stocks of different land use types in the southern moscow region by applying machine learning to legacy data |
| topic | landsat stochastic gradient boosting relief soil organic carbon parameterization spectral transformation moscow region |
| url | https://agrojournal.rudn.ru/agronomy/article/viewFile/20127/16458 |
| work_keys_str_mv | AT yuryadvornikov mappingsoilorganiccarbonstocksofdifferentlandusetypesinthesouthernmoscowregionbyapplyingmachinelearningtolegacydata AT lukyanamirniy mappingsoilorganiccarbonstocksofdifferentlandusetypesinthesouthernmoscowregionbyapplyingmachinelearningtolegacydata AT ekaterinasmukvich mappingsoilorganiccarbonstocksofdifferentlandusetypesinthesouthernmoscowregionbyapplyingmachinelearningtolegacydata AT kristinavivashchenko mappingsoilorganiccarbonstocksofdifferentlandusetypesinthesouthernmoscowregionbyapplyingmachinelearningtolegacydata |