ABSTRACT

Saffron (Crocus sativus L.) is one of the most important crops produced globally, and in only a limited number of countries. Determining the best conditions for cultivating this crop is important. Prediction of saffron yield according to soil characteristics can help to evaluate the land’s ability to cultivate this valuable plant. To achieve this objective, 100 soil samples were taken. Physicochemical properties, such as soil texture, nutrients, soil acidity, electrical conductivity, organic matter and lime were measured. After harvesting saffron, fresh weight of the saffron flower was measured in kg ha⁻¹. Using artificial neural networks and creating different models with different data sets of soil properties as the input and saffron yield as the output, the ability of this network was evaluated in prediction of the saffron yield. Available phosphorus and organic matter based on the results and the Pearson coefficient are the most effective factors on saffron yield. Evaluation of the model results indicated that the coefficient varied was obtained from 0.45 to 0.89. The best model for saffron yield estimation was obtained when phosphorus, organic matter, potassium and electrical conductivity were used as the input, so that values of R² and root mean square error (RMSE) were obtained at 0.891 and 0.89 kg ha⁻¹, respectively.

摘要

藏红花 ( Crocus sativus L.) 是全球最重要的作物之一，仅在少数国家生产。确定种植这种作物的最佳条件很重要。根据土壤特征预测藏红花产量有助于评估土地种植这种有价值植物的能力。为实现这一目标，采集了 100 个土壤样品。测量了土壤质地、养分、土壤酸度、电导率、有机质和石灰等物理化学特性。收获藏红花后，以kg ha ^-1为单位测量藏红花的鲜重. 使用人工神经网络并以不同的土壤特性数据集作为输入，以藏红花产量作为输出创建不同模型，评估了该网络在藏红花产量预测中的能力。基于结果的有效磷和有机质以及 Pearson 系数是影响藏红花产量的最有效因素。模型结果的评估表明，系数从 0.45 变化到 0.89。当以磷、有机质、钾和电导率作为输入时，获得了藏红花产量估算的最佳模型，因此在 0.891 和 0.89 kg ha ^{-1 时}获得了 R ²和均方根误差 (RMSE) 值，分别。