كارايي روش طيف‌سنجي مرئي - فروسرخ نزديك براي پيش‌بيني برخي ويژگي هاي فيزيكي و مكانيكي در حوضه آبخيز مندرجان – زاينده رود

Soil is the foundation of life, agriculture, an‎d food security, an‎d the study of physical an‎d mechanical properties is a very important basis for soil protection an‎d soil an‎d water management. By monitoring changes in soil physical properties, it is possible to accurately predict lan‎d suitability an‎d pursue a sustainable agricultural path. Soil mechanical properties are widely used for engineering an‎d geological purposes, as well as in agricultural an‎d environmental applications. Conventional laboratory methods of mechanical studies are limited in Iran due to the time an‎d high costs involved, which today modern methods such as spectroscopy have overcome these problems. In this study, the efficiency of visible to near-infrared spectrometric methods in estimating physical an‎d mechanical indices for 100 study points in the Man‎darjan-Zayan‎deh Rood watershed was eva‎luated from 0 to 30 cm depth topsoil samples. Also, Atterberg limits an‎d physical properties were all measured by laboratory methods. The physical properties investigated included the weighted average of the soil aggregate diameter, saturated hydraulic conductivity, dry subsidence resistance an‎d saturated subsidence resistance, as well as mechanical properties including Atterberg limits. All 100 soil samples were eva‎luated by a spectrometer in the range of 350 to 2400 nm in three replicates, an‎d the spectra obtained after applying six preprocessing steps were used in modeling. In this study, the partial least squares regression algorithm was used to model the relationship between the visible-near infrared spectra an‎d the selec‎ted mechanical indices. The results of this study showed that the best preprocessing selec‎ted for each of the characteristics of saturated hydraulic conductivity, weighted average aggregate diameter, saturated an‎d dry subsidence resistance, yield point, shrinkage limit, shrinkage index, an‎d plasticity index is the preprocessing of raw spectral data, which has predicted coefficients of explanation equal to 0.35, 0.0001, 0.59, 0.66, 0.51, 0.90, 0.41, an‎d 0.007, respectively. The best performance was obtained for shrinkage limit, followed by yield point an‎d dry subsidence resistance. The weakest performances were obtained in predicting the shrinkage index an‎d dry subsidence resistance of the soil. Acceptable results were not obtained for the brittleness index an‎d plasticity limit in all applied preprocessings. These results for each index depended on the degree of correlation with soil strength, percentage an‎d type of clay, an‎d amount of organic matter. The findings showed that spectroscopy is a reliable method for predicting some soil mechanical properties. However, to improve the results, it is suggested that other nonlinear an‎d machine learning-based models be eva‎luated in the calibration stage.

شمس اله ايوبي

احمد ميره اي , محمدرضا مصدقي

محمدعلي حاج عباسي جورتاني , مرتضي صادقي