Soil structure has a vital role in soil quality and can moderate the environment by motivating soil water retention, hydraulic conductivity, nutrient diffusion, aeration, and temperature. Considering the importance of soil structure, different methods and indices have been proposed to evaluate the aggregate stability for specific objectives and conditions. The present study aimed to evaluate the soil structural stability using high energy moisture characteristic (HEMC) method and its spatial variability, and to determine the most-important factors affecting structural stability of surface soils in Isfahan Province. First, the study area was divided into 20 ´ 20 kilometres network, and 252 soil samples were collected from 0–20 cm depth on the network. The soil structural stability was determined by the HEMC method using sand box apparatus; the soil samples were wetted slowly and quickly. The HEMC data were modelled using the modified van Genuchten model. The values of stability index (SI) were determined for slow-wetted and fast-wetted samples and the stability ratio (SR) was calculated by dividing SI of fast wetting on SI of slow wetting. Moreover, the ratio of volumes of drainable pores (VDPR) and the ratio of matric suctions at inflection point of HEMC (? dR ) were calculated as the relative indices of soil structure stability. Some soil physical and chemical properties including organic matter, gypsum and carbonate contents, texture, pH, EC, SAR and ratio of organic matter to clay (OM/Clay) were also determined. The impacts of different properties on aggregate stability were evaluated using correlation, stepwise regression and spatial variability analyses. In order to evaluate the impact of each variable on soil structural stability, correlation, step-by-step multivariate regression, and neural network (using multilayer perceptron network) analyses and sensitivity analysis were done. In addition, spatial variability analysis was performed by semi-variogram models and kriging interpolation method. Statistical results indicated that structural stability indices and soil properties varied in broad ranges. With an increment in the organic matter, gypsum, carbonate and clay contents, soil pores were less destroyed and the difference between HEMC curves in fast-wetting and slow- wetting treatments were reduced. The Spearman correlation coefficients revealed that SR and VDPR were positively and significantly interrelated, while they were significantly (but negatively) related to ? d R. The stronger relation between SR and VDPR suggested that the structural stability index is mainly governed by the volume of drainable pores rather than by the matric suction at inflection point of HEMC. The SR and VDPR had positive correlations with organic matter, gypsum, OM/Clay and carbonate; because these properties are considered as the agents of formation and stability of soil structure. Contrarily, SAR, known as aggregate-destructive factor, was negatively correlated with SR and VDPR. Step-by-step regression and neural network analyses were used to predict the soil structural stability indices by readily-available soil properties; organic matter, gypsum, OM/Clay, carbonate, EC and SAR were determined as the most-important input variables in both methods. Based on the coefficients of determination, artificial neural network had better efficacy and higher accuracy than multivariate regression in predicting the soil structural stability indices. Mean comparisons of the soil structural stability indices showed that desert land use had the highest value of ? d R (i.e. index of soil structural instability) which was significantly greater than its values in pasture and arable land uses. In addition, means of SR and VDPR were higher in pasture and arable lands. However, no significant difference was observed between the means of soil structural stability indices among different geological deposits. Spatial variability analysis revealed that in northeast part of Isfahan Province, the soil structural stability indices were higher due to gypsum presence in the soils. Also, in southern part with semi-humid climate, soil structure was fairly stable. In south and southeast parts of Isfahan city and in east of Isfahan Province, the soil structure was unstable due to destroying effect of high SAR. Principal component and clustering analyses were also employed to reduce the number of affecting variables. Based on the entered variables, the grouping of cluster analysis was almost similar to that of principal component analysis; the components consisted of three groups of soil physical-climatological properties, soil chemical properties and soil structural stability indices. Keywords: Soil structural stability, High energy moisture characteristic (HEMC), Fast wetting, Slow wetting, Multivariate regression, Artificial neural network