Abstract
In most countries, the site-specific, precise application of N fertilizer in an optimal dose is one of the most challenging tasks for sustainable agriculture. Several recommendation systems have been proposed for N fertilizer as an alternative to the scarce and expensive soil experts. However, none of them exhibited an impressive performance. In this article, we have proposed a highly efficient optimal fuzzy system (OFS) with a novel architecture to recommend crop-specific optimal doses of N fertilizer based on site-specific soil and climatic data. In our proposed OFS, the fuzzy membership functions of each variable were replaced by the respective probability density functions, the rule base was redefined, and finally, conflict resolution was achieved using the probabilistic fuzzy logic controller approach. The output probability density function was optimized with the most popular whale optimization algorithm (WOA). Such an innovative approach to designing an N fertilizer recommendation system has never been well thought out so far. Our designed OFS was empirically validated in terms of four statistical metrics: the co-efficient of determination (R2), the Nash-Sutcliffe efficiency (NSE), the root means squared error (RMSE), and the mean absolute error (MAE) against three varieties of paddy and two varieties of potato cultivated in the Gangetic alluvial plain in West Bengal, India. It was further compared to the other latest N fertilizer recommendation systems designed for different crops worldwide. The study revealed that our system (with R2 ranging from 0.9628 to 0.9880) outperformed all other systems (with R2 ranging from 0.1900 to 0.8400).
Keywords: Fertilizer Recommendation System, Fuzzy Hybrid Systems, Nitrogen Fertilizer, Optimal Fuzzy System, Probability Density Function, Whale Optimization Algorithm.