Identifying the determinant characteristics influencing soil compactibility indices using neural networks and path analysis Running title: Investigation of Soil Compaction by ANN and PA

Soil compactibility can be quantified using different indices such as maximum dry bulk density (BDmax) and critical water content (θcritical) in a compaction test. The objective of this study was to determine soil properties influencing soil compactibility by evaluate pedotranfer functions (PTFs) with respect to their accuracy and usefulness for the prediction of BDmax and θcritical using linear regression and ANN methods. 100 soil samples were collected from arable and virgin lands in southeast Iran. Primary particle size distribution, CaSO4, CaCO3, organic matter (OM) contents and natural bulk density were used as predictors. Two PTFs were developed using linear multiple stepwise regression: a PTF that estimates BDmax using clay and sand contents and natural bulk density as predictors (R2 = 0.45), the other one for the estimation of θcritical using clay and CaSO4 contents as predictors (R2 = 0.51). Furthermore, an attempt was made to construct PTFs for the prediction of the BDmax and θcritical using ANNs. High prediction efficiencies were achieved using the ANN models. Generally, when all of the easily-available soil properties were included as predictors, much more accurate estimates were obtained by the ANN models for the θcritical and BDmax as compared with the linear regression method. Sensitivity analysis showed that the most important variable in BDmax prediction using ANNs is the BDnatural followed by sand and clay, CaCO3 and CaSO4 contents. The θcritical had the highest sensitivity to clay content and the lowest sensitivity to OM content in the studied soils.