02772naa a2200337 a 450000100080000000500110000800800410001902200140006002400540007410000250012824501140015326000090026752018050027665000200208165000170210165300280211865300200214665300130216665300240217965300210220365300100222465300270223470000190226170000190228070000240229970000200232370000210234370000200236470000160238477300340240021751832025-04-28       2024    bl uuuu     u00u1 u    #d  a0341-81627 ahttps://doi.org/10.1016/j.catena.2024.1082192DOI1 aGONÇALVES, D. R. P.  aAccessing and modelling soil organic carbon stocks in Prairies, Savannas, and forests.h[electronic resource]  c2024  aSoils are the third largest carbon pool on Earth and play a crucial role in mitigating climate change. Therefore, understanding and predicting soil carbon sequestration is of major interest to mitigate climate change globally, especially in countries with strong agricultural backgrounds. In this study, we used a new database composed of 5029 samples collected up to 1-meter depth in three biomes that are most representative of agriculture, Pampas (Prairie), Cerrados (Savanna), and Atlantic Forest (Forest), to explore soil organic carbon (SOC) stocks and its environmental drivers. The Cerrado (Savanna) biome was the only one where croplands presented higher SOC stocks than native vegetation (Native vegetation 121.23 Mg/ha and croplands 127.85 Mg/ha or 5 % higher). From the tested models, the Random Forest outperformed the others, achieving an R2 of 0.64 for croplands and 0.56 for native vegetation. The accuracy of the models varied with soil depth, showing better predictions in shallow layers for croplands and deeper layers for native vegetation. Our results highlight the importance of clay content, precipitation, net primary production (NPP), and temperature as key predictors for soil carbon stocks in the studied biomes. The findings emphasize the importance of protecting the surface layers, especially in the Cerrado biome, to enhance SOC stocks and promote sustainable land management practices. Moreover, the results provide valuable insights for the development of nature-based carbon markets and suggest potential strategies for climate change mitigation. Enhancing our understanding of SOC dynamics and adopting precise environmental predictors will contribute to the formulation of targeted soil management strategies and accelerate progress toward achieving climate goals.  aLand use change  aUso da Terra  aAprendizado de máquina  aAtlantic forest  aCerrados  aFloresta Atlântica  aMachine learning  aPampa  aSoil carbon prediction1 aINAGAKI, T. M.1 aBARIONI, L. G.1 aSCALA JUNIOR, N. L.1 aCHERUBIN, M. R.1 aSÁ, J. C. de M.1 aCERRI, C. E. P.1 aANSELMI, A.  tCatenagv. 243, 108219, 2024.