Saffron yield estimation by adaptive neural-fuzzy inference system and particle swarm optimization (ANFIS-SCM-PSO) hybrid model

• Published in: Archiv für Acker- und Pflanzenbau und Bodenkunde Vol. 69; no. 3; pp. 461 - 475
• Main Authors: Nazari, Hosnie, Mohammadkhani, Nayer, Servati, Moslem
• Format: Journal Article
• Language: English
• Published: Abingdon Taylor & Francis 23.02.2023; Taylor & Francis Ltd
• Subjects: agronomy; algorithms; Artificial neural networks; Calcium carbonate; Cation exchange; cation exchange capacity; Cation exchanging; climate; Combination model; Electrical conductivity; Electrical resistivity; Estimation; Food and Agriculture Organization; Fuzzy logic; Genetic analysis; Gypsum; Inference; Iran; lakes; land characteristics; Meteorological data; Optimization; Optimization models; Organic carbon; Particle swarm optimization; Saffron; Siminehrood catchment; Soil analysis; Soil profiles; Soil properties; sustainable agriculture; Urmia Lake; watersheds; yield mapping; Iran
• ISSN: 0365-0340; 1476-3567; 1476-3567
• DOI: 10.1080/03650340.2021.2004588

The aim of the present research is to estimate the saffron yield by land characteristics through a combination of adaptive neural-fuzzy inference system and particle swarm optimization in Siminehrood catchment, south of Urmia Lake, Iran. To achieve this target, 150 representative soil profiles were descripted in saffron fields. Then each genetic horizon was sampled for soil analysis. Climate rating was calculated from meteorological data by Food and Agriculture Organization framework. Saffron observed yield obtained from saffron field data. The results showed that coarse fragment, gypsum, electrical conductivity, organic carbon, cation exchange capacity, pH, calcium carbonate equivalent and climate rating have the highest correlation with saffron yield. The range of saffron estimated yield was between 1937-4124 and 1843-4025 kg per hectare for combination model and fuzzy model, respectively. Combination model doses a more accurate estimation compared with the observed yield values (2020-4200), statistical validation indicator results confirm this too. High agreement obtained from combination model between estimated saffron yield map with observed yield map. Finally, a combination of adaptive neural-fuzzy inference system and particle swarm optimization model can be employed as a powerful, low time-consuming and accurate method for estimating saffron yield in pre-decision for saffron cultivation.