Potential of the hazelnut tree search–ELM hybrid approach in estimating yield and water productivity

Abstract

Wheat plays a vital role in the food security of society, and early estimation of its yield will be a great help to macro-decisions. For this purpose, wheat yield and water productivity (WP) by considering soil data, irrigation, fertilizer, climate, and crop characteristics and using a novel hybrid approach called hazelnut tree search algorithm (HTS) and extreme machine learning method (ELM) was examined under the drip (tape) irrigation. A dataset including 125 wheat yield data, irrigation and meteorological data of Mahabad plain located southeast of Lake Urmia, Iran, was used as input parameters for crop year 2020–2021. Eighty percentage of the data were used for training, and the remaining 20% for model testing. Nine different input scenarios were presented to estimate yield and WP. The efficiency of the proposed model was calculated with the statistical indices coefficient of determination (R2), root-mean-square error (RMSE), normalized root-mean-square error, and efficiency criterion. Sensitivity analysis result showed that the parameters of irrigation, rainfall, soil moisture, and crop variety provide better results for modeling. There was good agreement between the practical values (field management data) and the estimated values with the HTS–ELM model. The results also showed that the HTS–ELM method is very efficient in selecting the best input combination with R2 = 0.985 and RMSE = 0.005. In general, intelligent hybrid methods can enable optimal and economical use of water and soil resources.