Modeling orbital data of soil carbon dioxide efflux from different land uses in Southern Amazon
The dynamics of carbon among atmospheric, soil and biotic stocks are of great importance for ecosystem and climate services. The interdependence of carbon stocks is volatile, since higher atmospheric CO₂ concentrations affect plant development and therefore carbon storage in terrestrial ecosystems....
| Autores: | , , , , , , , , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2025 |
| País: | Brasil |
| Institución: | Universidade Estadual Paulista (UNESP) |
| Repositorio: | Repositório Institucional da UNESP |
| Idioma: | inglés |
| OAI Identifier: | oai:repositorio.unesp.br:11449/300102 |
| Acceso en línea: | http://dx.doi.org/10.1016/j.jsames.2024.105323 https://hdl.handle.net/11449/300102 |
| Access Level: | acceso abierto |
| Palabra clave: | Amazon Carbon dioxide Land use Remote sensing |
| Sumario: | The dynamics of carbon among atmospheric, soil and biotic stocks are of great importance for ecosystem and climate services. The interdependence of carbon stocks is volatile, since higher atmospheric CO₂ concentrations affect plant development and therefore carbon storage in terrestrial ecosystems. In addition, the carbon cycle is related to soil moisture, and sensitivity to moisture differs between ecosystems and climatic regions. In the southern Amazon, agriculture and cattle ranching activities drives anthropogenic actions and for the environmental costs. As a result, those activities impact carbon dynamics and its consequences on the environment. Modeling these dynamics in a spatialized way is possible through remote sensing images, which, together with appropriate modeling tools, allow us to understand the carbon balance at a regional level. The aim of this study is discussing the modeling of the soil carbon dioxide efflux (FCO₂) from different land uses for orbital data predictions using MODIS and PlanetScope imagery. Local data was the reference for the orbital data modeling with partial least squares regression (PLSR). Discussed models are based on soil moisture, temperature, spectral bands and also models with MODIS GPP and CO2Flux were created. Land uses (characterized by high and low productivity soybeans, degraded pasture, productive pasture and native forest) and consisted of different subsets of inputs subsets to design PLSR equations. Results analyzes were based on the statistical metrics of linear regression (R2), mean absolute error (MAE) and root mean square error (RMSE). From those methods, it was observed that the subsets with the lowest error and highest correlation were the subsets related to soybeans. The homogeneity of soybean areas and its spectral characteristics mean greater capacity for predicting FCO₂, since the orbital images and PLSR modeling provide a higher correlation and lower error, both absolute and quadratic. On the other hand, carbon balance modeling in forest areas and pastures is limited and potentially associated with the heterogeneity of that environment. |
|---|