Application of a hybrid neural-fuzzy inference system for mapping crop suitability areas and predicting rice yields. (April 2019)

• Record Type: Journal Article
• Authors: Dang, Kinh Bac; Burkhard, Benjamin; Windhorst, Wilhelm; Müller, Felix
• Is Part Of: Environmental modelling & software. Volume 114(2019)
• Page Start: 166
• Page End: 180
• Publication Date: 2019-04
• Subjects: Agriculture -- Crop -- HyFIS -- Neural network -- Regional scale -- Plot scale; HyFIS Hybrid neural-Fuzzy Inference System -- SVM Support Vector Machine -- GLM General Logistic Regression -- GIS Geographical Information System -- LULC Land Use and Land Cover -- GSO General Statistics Office of Vietnam -- fractp ratio of evapotranspiration to precipitation -- TWI Topographic Wetness Index -- IRRI International Rice Research Institute; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2019.01.015
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Environmental stressors and population growth have significantly affected terraced rice ecosystems, such as in the Sapa district in northern Vietnam. The question arises how natural and socio-economic components determine the amount of rice yields. This study combines a hybrid neural-fuzzy inference system (HyFIS) with GIS-based methods to generate two models that can map suitability areas for rice cultivation at a regional scale and predict actual rice yields at a plot scale. Semi-structured interviews, the "Integrated Valuation of Ecosystem Services and Tradeoffs" tool and different statistical models were used to investigate the impacts of eight environmental variables and three socio-economic variables on rice production. Subsequently, two HyFIS models were trained with an accuracy higher than 88%. Because the predictive power values of the two proposed HyFIS models were higher than those of benchmark models, they are considered as useful tools to assess and optimize land use and related rice productivity. Highlights: Hybrid models integrating neural network, fuzzy logic and geographical information system approaches were developed to find areas suitable for rice cultivation (a so-called S-HyFIS model) and to predict rice yields (a so-called I-HyFIS model). Impacts of environmental characteristics and additional human inputs on rice production were quantified. The S-HyFIS model can extrapolate results from plot to regional scales in order to generate a map representing crop suitability areas for rice cultivation. The potential of conversion from different land uses/covers to arable lands and vice versa were analyzed in the Sapa district, Lao Cai province, Vietnam.