Neuro-deep fuzzy system for estimation of NO ₂ concentration in soil and groundwater on highways from remote sensing images

Remotely sensed data of the contaminated soil and groundwater near the highways around the mountains, deserts, sea side and agricultural fields are collected and analyzed by various traditional and modern techniques to take several purposeful decisions. This article presents a neuro-deep fuzzy system to estimate the concentration of nitrogen dioxide (NO ₂ ) in the soil and nitric acid in the groundwater near the highways due to exhausts of gases from the motor vehicles. The proposed scheme is developed based on algebraic sum of Mexican hat wavelet function that is a universal approximator in the hidden layer’s hybrid with a layer of fuzzy rules. The energy function is generated in an unsupervised manner based upon the diffusion of nitrogen dioxide in the soil surface and nitric acid in the water. A scenario-based case study on various highways has been developed by using various parametric values like air temperature, humidity, pressure and distance of the observations from the environment under consideration as the initial and/or boundary conditions of the diffusion equations developed from the remotely sensed data of the soil and groundwater. Experimental findings indicate that the model quantifies NO ₂ concentrations with a precision of 10 ⁻⁶ g/m ³ , attaining a minimal fitness value of 7.84 × 10 ⁻¹³ for soil and 3.73 × 10 ⁻¹¹ for water. The reliability evaluation across 100 Monte Carlo simulations indicates a Global Mean Square Error (GMSE) of 4.32 × 10 ⁻⁴ and an NSE of 0.979 for soil, whereas for groundwater, the GMSE is 7.93 × 10 ⁻⁴ and the NSE is 0.935. Keeping in view reliability, convergence and computational complexity of the proposed scheme, it can be an effective alternative to analysis of soil and groundwater contamination in the laboratories.