Performance triggered adaptive model reduction for soil moisture estimation in precision irrigation

TL;DR

The paper tackles the problem of estimating soil moisture across large agricultural fields where high-dimensional agro-hydrological models, based on the discretized Richards equation, are computationally intractable for real-time estimation. It introduces a performance-triggered adaptive model-reduction framework that clusters state trajectories to form reduced-order models via a Petrov-Galerkin projection and uses an error metric $e_L$ with threshold $th_e$ to trigger model re-identification, combined with an adaptive reduced-order EKF. Key contributions include a trajectory-based clustering approach for dynamic model identification, a seamless information transfer between reduced models, and extensive simulation on a 26.4 ha field showing improved computational efficiency and robust soil moisture estimation under realistic disturbances. The approach advances real-time precision irrigation by enabling field-scale soil moisture mapping with manageable computation and adaptive model fidelity.

Abstract

Accurate soil moisture information is crucial for developing precise irrigation control strategies to enhance water use efficiency. Soil moisture estimation based on limited soil moisture sensors is crucial for obtaining comprehensive soil moisture information when dealing with large-scale agricultural fields. The major challenge in soil moisture estimation lies in the high dimensionality of the spatially discretized agro-hydrological models. In this work, we propose a performance-triggered adaptive model reduction approach to address this challenge. The proposed approach employs a trajectory-based unsupervised machine learning technique, and a prediction performance-based triggering scheme is designed to govern model updates adaptively in a way such that the prediction error between the reduced model and the original model over a prediction horizon is maintained below a predetermined threshold. An adaptive extended Kalman filter (EKF) is designed based on the reduced model for soil moisture estimation. The applicability and performance of the proposed approach are evaluated extensively through the application to a simulated large-scale agricultural field.

Figures (16)

• Figure 1: A schematic of an agricultural field agyeman2021soil
• Figure 2: Discretization of the agricultural field where each dot denotes the discretized node and red dots indicate the point sensors
• Figure 3: A schematic of a point sensor
• Figure 4: Proposed performance-triggered adaptive model reduction and state estimation scheme
• Figure 5: A representation of $m^{th}$ model reduction