Abstract. Soil bulk density (BD) serves as a fundamental indicator of soil health and quality, exerting a significant influence on critical factors such as plant growth, nutrient availability, and water retention. Due to its limited availability in soil databases, the application of pedotransfer functions (PTFs) has emerged as a potent tool for predicting BD using other easily measurable soil properties, while the impact of these PTFs' performance on soil organic carbon (SOC) stock calculation has been rarely explored. In this study, we proposed an innovative local modeling approach for predicting BD of fine earth (BDfine) across Europe using the recently released BDfine data from the LUCAS Soil (Land Use and Coverage Area Frame Survey Soil) 2018 (0–20 cm) and relevant predictors. Our approach involved a combination of neighbor sample search, forward recursive feature selection (FRFS), and random forest (RF) models (local-RFFRFS). The results showed that local-RFFRFS had a good performance in predicting BDfine (R2 of 0.58, root mean square error (RMSE) of 0.19 g cm−3, relative error (RE) of 16.27 %), surpassing the earlier-published PTFs (R2 of 0.40–0.45, RMSE of 0.22 g cm−3, RE of 19.11 %–21.18 %) and global PTFs using RF models with and without FRFS (R2 of 0.56–0.57, RMSE of 0.19 g cm−3, RE of 16.47 %–16.74 %). Interestingly, we found that the best earlier-published PTF (R2 = 0.84, RMSE = 1.39 kg m−2, RE of 17.57 %) performed close to the local-RFFRFS (R2 = 0.85, RMSE = 1.32 kg m−2, RE of 15.01 %) in SOC stock calculation using BDfine predictions. However, the local-RFFRFS still performed better (ΔR2 > 0.2) for soil samples with low SOC stocks (< 3 kg m−2). Therefore, we suggest that the local-RFFRFS is a promising method for BDfine prediction, while earlier-published PTFs would be more efficient when BDfine is subsequently utilized for calculating SOC stock. Finally, we produced two topsoil BDfine and SOC stock datasets (18 945 and 15 389 soil samples) at 0–20 cm for LUCAS Soil 2018 using the best earlier-published PTF and local-RFFRFS, respectively. This dataset is archived on the Zenodo platform at https://doi.org/10.5281/zenodo.10211884 (S. Chen et al., 2023). The outcomes of this study present a meaningful advancement in enhancing the predictive accuracy of BDfine, and the resultant BDfine and SOC stock datasets for topsoil across the Europe enable more precise soil hydrological and biological modeling.

中文翻译：

---

使用基于机器学习的pedotransfer函数的欧洲表土容重和有机碳库数据库（0–20 cm）

摘要。土壤容重 (BD) 是土壤健康和质量的基本指标，对植物生长、养分利用率和保水性等关键因素产生重大影响。由于其在土壤数据库中的可用性有限，土壤转移函数（PTF）的应用已成为使用其他易于测量的土壤特性来预测 BD 的有效工具，而这些 PTF 的性能对土壤有机碳（SOC）库计算的影响却很少被探索。在这项研究中，我们提出了一种创新的局部建模方法，使用最近发布的 2018 年 LUCAS 土壤（土地利用和覆盖面积框架调查土壤）（0-20 厘米）的 BDfine 数据来预测整个欧洲的细土 BD（BDfine）和相关的预测因子。我们的方法涉及邻近样本搜索、前向递归特征选择 (FRFS) 和随机森林 (RF) 模型 (local-RFFRFS) 的组合。结果表明，local-RFFRFS在预测BDfine方面具有良好的性能（R2为0.58，均方根误差（RMSE）为0.19 g cm−3，相对误差（RE）为16.27%），超越了早期发布的PTF（ R2 为 0.40–0.45，RMSE 为 0.22 g cm−3，RE 为 19.11 %–21.18 %）以及使用带或不带 FRFS 的 RF 模型的全局 PTF（R2 为 0.56–0.57，RMSE 为 0.19 g cm−3，RE 为 16.47 %–16.74 %)。有趣的是，我们发现早期发布的最佳 PTF（R2 = 0.84，RMSE = 1.39 kg m−2，RE 为 17.57 %）的表现接近本地 RFFRFS（R2 = 0.85，RMSE = 1.32 kg m−2，RE） 15.01%）在使用 BDfine 预测的 SOC 库存计算中。然而，对于低 SOC 储量 (< 3 kg m−2) 的土壤样品，局部 RFFRFS 仍然表现更好 (ΔR2 > 0.2)。因此，我们认为 local-RFFRFS 是 BDfine 预测的一种有前途的方法，而当 BDfine 随后用于计算 SOC 库存时，早期发布的 PTF 会更有效。最后，我们分别使用早期发布的最佳 PTF 和本地 RFFRFS 为 LUCAS Soil 2018 生成了 0-20 cm 处的两个表土 BDfine 和 SOC 库数据集（18 945 和 15 389 个土壤样本）。该数据集存档在 Zenodo 平台上：https://doi.org/10.5281/zenodo.10211884（S. Chen 等人，2023）。这项研究的结果在提高 BDfine 的预测准确性方面取得了有意义的进步，由此产生的欧洲表土 BDfine 和 SOC 库数据集可以实现更精确的土壤水文和生物建模。