Assessing soil quality is essential for crop management and soil temporal changes. The present study aims to evaluate soil quality in the Ferralitic soils context countrywide. This assessment was done using multivaria...Assessing soil quality is essential for crop management and soil temporal changes. The present study aims to evaluate soil quality in the Ferralitic soils context countrywide. This assessment was done using multivariate soil quality indice (SQI) models, such as additive quality index (AQI), weighted quality indexes (WQI<sub>add</sub> and WQI<sub>com</sub>) and Nemoro quality index (NQI), applied to two approaches of indicator selection: total data set (TDS) and minimum data set (MDS). Physical and chemical soil indicators were extracted from the ORSTOM’s reports resulting from a sampling campaign in different provinces of Gabon. The TDS approach shows soil quality status according to eleven soil indicators extracted from the analysis of 1,059 samples from arable soil layer (0 - 30 cm depth). The results indicated that 87% of all provinces presented a very low soil quality (Q5) whatever the model. Among soil indicators, exchangeable K<sup>+</sup> and Mg<sup>2+</sup>, bulk density and C/N ratio were retained in MDS, using principal component analysis (PCA). In the MDS approach, 50 to 63% of provinces had low soil quality grades with AQI, WQI<sub>add</sub> and NQI, whereas the total was observed with WQI<sub>com</sub>. Only 25% of provinces had medium soil quality grades with AQI and NQI models, while 12.5% (NQI) and 25% (AQI) presented high quality grades. Robust statistical analyses confirmed the accuracy and validation (0.80 r P ≤ 0.016) of AQI, WQI<sub>add</sub> and NQI into the TDS and MDS approaches. The same sensitivity index value (1.53) was obtained with AQI and WQI<sub>add</sub>. However, WQI<sub>add</sub> was chosen as the best SQI model, according to its high linear regression value (R<sup>2</sup> = 0.82) between TDS and MDS. This study has important implications in decision-making on monitoring, evaluation and sustainable management of Gabonese soils in a pedoclimatic context unfavorable to plant growth.展开更多
文摘Assessing soil quality is essential for crop management and soil temporal changes. The present study aims to evaluate soil quality in the Ferralitic soils context countrywide. This assessment was done using multivariate soil quality indice (SQI) models, such as additive quality index (AQI), weighted quality indexes (WQI<sub>add</sub> and WQI<sub>com</sub>) and Nemoro quality index (NQI), applied to two approaches of indicator selection: total data set (TDS) and minimum data set (MDS). Physical and chemical soil indicators were extracted from the ORSTOM’s reports resulting from a sampling campaign in different provinces of Gabon. The TDS approach shows soil quality status according to eleven soil indicators extracted from the analysis of 1,059 samples from arable soil layer (0 - 30 cm depth). The results indicated that 87% of all provinces presented a very low soil quality (Q5) whatever the model. Among soil indicators, exchangeable K<sup>+</sup> and Mg<sup>2+</sup>, bulk density and C/N ratio were retained in MDS, using principal component analysis (PCA). In the MDS approach, 50 to 63% of provinces had low soil quality grades with AQI, WQI<sub>add</sub> and NQI, whereas the total was observed with WQI<sub>com</sub>. Only 25% of provinces had medium soil quality grades with AQI and NQI models, while 12.5% (NQI) and 25% (AQI) presented high quality grades. Robust statistical analyses confirmed the accuracy and validation (0.80 r P ≤ 0.016) of AQI, WQI<sub>add</sub> and NQI into the TDS and MDS approaches. The same sensitivity index value (1.53) was obtained with AQI and WQI<sub>add</sub>. However, WQI<sub>add</sub> was chosen as the best SQI model, according to its high linear regression value (R<sup>2</sup> = 0.82) between TDS and MDS. This study has important implications in decision-making on monitoring, evaluation and sustainable management of Gabonese soils in a pedoclimatic context unfavorable to plant growth.
文摘LeNet-5卷积神经网络在手写数字库上取得了很好地识别效果,但在表情识别中识别率很低.改进了LeNet-5卷积神经网络,使用浅层卷积结构,连续经过1×1和3×3的卷积层,在每一层的卷积后,加上Z-score标准化处理,使用性能更好的Relu激活函数,此函数计算速度快,减少梯度弥散问题;输出层用softmax函数,该层输出表情图像的概率.仿真结果表明,在JAFFE表情数据库上,即使在小样本数据集的情况下,算法识别率达到79.81%,识别单幅人脸表情图像的平均耗时为0.353 s.
