Use of a combination of two clustering and tree algorithms for investigating and predicting alfalfa yield under different conditions of deficit irrigation

Abstract

The largest amount of water consumption in Iran is related to agriculture. Considering that Iran has suffered a drought in recent years, the optimal use of water is necessary, especially in the agricultural sector. For this reason, in this research, the deficit irrigation regarding alfalfa crop was investigated. Alfalfa was cultivated in three years with four replications. Deficit irrigation methods including 40 %, 70 % and 100 % full irrigation (FI) were employed at different stages of alfalfa growth. First, crop per drop (CPD), benefit per drop (BPD) and net benefit per drop (NBPD) values related to alfalfa were clustered. The results of clustering were introduced to the tree algorithm as target data. The tree algorithm predicted the target data according to the factors of 1) the type of cropping year, 2) the time of harvest and 3) the level of deficit irrigation at different stages of alfalfa growth. According to the results of the combination of two algorithms, it was found that The lowest values of CPD, BPD and NBPD are equal to 1.2 (kg/m3), 125 (1000 Rial/m3) and 115 (1000 Rial/m3) respectively in the second harvest and deficit irrigation which was predicted in the two regrowth (RG) and budding (BU) stages. The highest values of CPD, BPD and NBPD are equal to 1.8 (kg/m3), 185 (1000 Rial/m3) and 175 (1000 Rial/m3) respectively in the third harvest and deficit irrigation which was predicted in the branching stage (BR). Thus, to increase the yield, it is better not to perform severe deficit irrigation at RG and BU stages. Also, according to the predictor importance of the software, it was found that the three factors of harvest time, BU and the type of cropping year had the greatest effect on the prediction, respectively.