Search

Evaluating soil quality in the alluvial plains of the semi-arid tropics is vital to understand the impact of existing land management practices, thus facilitating the adoption of more sustainable agricultural methods. We analysed soil profiles using 146 horizon-specific samples from 24 pedons, representing seven soil series from the sub-tropical to semi-arid River Cauvery Delta basin, Tamil Nadu, India. Using principal component analysis (PCA), we identified minimum datasets (MDS) from 30 different soil properties. The plough layer's MDS were clay content, exchangeable sodium percentage (ESP), very coarse sand content, soil pH and soil organic carbon (SOC). For the subsurface layer, sand content, ESP, SOC, diethylene-triamine-pentaacetic acid extractable zinc (DTPA-Zn), and very coarse sand content were identified as the MDS. Soil quality index (SQI) values were 0.47-0.72 for the plough layer and 0.46-0.56 for the subsurface soil, suggesting limited soil function support in subsurface soils. Surface soil-site suitability criteria displayed a moderate correlation with SQI values derived via PCA, with additive (R2 = .545) and weighted additive indices (R2 = .659). Both simple additive index and weighted derived SQIs demonstrated a moderate association with rice yield (R2 = .50 and .55, respectively). This research indicates the potential to identify soil quality indicators in alluvial plains of semi-arid tropics. [ABSTRACT FROM AUTHOR]

In this paper, a multilayer perceptron neural network has been used to represent temperature measurement during rotary drilling of five types of rock samples. To forecast the temperature at various thermocouple depths, the experimentally collected data was standardized. Indicators of model performance was also obtained in order to assess the correctness of the model. One hidden layer and one output layer were employed with MLPANN, which has ten input parameters (bit diameter (DD), Spindle Speed (SS), Penetration Rate (PR), thrust, and torque) and rock properties. Levenberg Marquardt learning algorithm with transfer function of logsig is the most optimal neuron number of 10-16-1 was successfully forecasting the temperature with a correlation of 0.9936 and 0.9941 for training and testing algorithm during drilling after analysis based on the trial-anderror approach to identify the optimum algorithm. Ten input parameters, a logsig sigmoid transfer function, and the trainlm algorithm in this study provide good prediction ability with tolerable accuracy. [ABSTRACT FROM AUTHOR]