Forward and Inverse Modelling Approaches for Prediction of Light Stimulus from Electrophysiological Response in Plants

TL;DR

This study develops and compares linear and nonlinear dynamical models, especially the Nonlinear Hammerstein-Wiener estimator, to predict light stimulus characteristics from electrical responses in plants, demonstrating effective inverse modeling across multiple species.

Contribution

Introduces a novel inverse modeling approach using NLHW estimators to predict light stimuli from plant electrical responses, validated on multiple plant species.

Findings

NLHW models outperform linear models in data fitting

Accurate detection of light on-off timing and intensity

Validated across diverse plant species

Abstract

In this paper, system identification approach has been adopted to develop a novel dynamical model for describing the relationship between light as an environmental stimulus and the electrical response as the measured output for a bay leaf (Laurus nobilis) plant. More specifically, the target is to predict the characteristics of the input light stimulus (in terms of on-off timing, duration and intensity) from the measured electrical response - leading to an inverse problem. We explored two major classes of system estimators to develop dynamical models - linear and nonlinear - and their several variants for establishing a forward and also an inverse relationship between the light stimulus and plant electrical response. The best class of models are given by the Nonlinear Hammerstein-Wiener (NLHW) estimator showing good data fitting results over other linear and nonlinear estimators in a statistical sense. Consequently, a few set of models using different functional variants of NLHW has been developed and their accuracy in detecting the on-off timing and intensity of the input light stimulus are compared for 19 independent plant datasets (including 2 additional species viz. Zamioculcas zamiifolia and Cucumis sativus) under similar experimental scenario.