Assessment of Relationships between Rose Yield and Soil and Topography Properties Using Multivariate Regression, Artificial Neural Network and Structure Equation Modeling

Due to the relationship between crop yield, and soil characteristics and land topography, knowledge and awareness of these characteristics is necessary to achieve sustainable development in agriculture. This study was performed to evaluate and determine the relationships between Rose yield (Rosa Damasceneea Mill) with soil properties and land topography by using Multivariate Linear Regression models (MLR), Artificial Neural Network (ANN) and Structural Equation Modeling (SEM) in Bardsir City, Kerman Province. For this purpose, soil sampling and crop yield were performed in the form of a regular grid pattern. Besides, some topographic features of the land were calculated using digital elevation model (DEM) of the region, and to implement the conceptual models, three theoretical models were designed and tested. The results showed that MLR and ANN models were able to justify 68 and 87 % of the yield variability, respectively, which indicates the higher accuracy of ANN model than MLR in yield estimation. The results of SEM illustrated that Rose yield is mainly controlled by soil chemical properties, topographic features, and soil physical properties, respectively. Different scenarios for SEM showed that simpler models with fewer hidden structures could have a better fitting. Therefore, the first conceptual model of this method with the values of root mean square error, goodness of fit index and comparative fit index of 0.033, 0.88 and 0.94, respectively, was selected as the best model. The overall results showed that the ANN model was more efficient than MLR in yield prediction due to consideration of the nonlinear relationship between crop yield and the factors affecting it. In addition to the ability of the ANN model to estimate crop yield, the SEM also showed that the latter method can provide more explanations about the relationships and simultaneous interactions between variables. In general, the application of SEM method, relying on the capabilities of this method, can improve the yield of various crops.