Integrating data assimilation with chlorophyll fluorescence signatures: An innovative framework for real-time monitoring of crop nitrogen dynamics

Abstract

Crop organ-level nitrogen (N) dynamics (accumulation and transport) are strongly associated with final quality and yield. Conventional crop N monitoring methods either have high uncertainty (crop model) or limited capacity to diagnose N status in stems and grains (remote sensing tools). Data assimilation overcomes the shortcomings of crop model and remote sensing tools, but whether it can accurately simulate nitrogen dynamics at the field scale remains unknown. We aimed to develop a novel dual assimilation framework coupling a crop model and unmanned aerial vehicle (UAV) remote sensing, incorporating fluorescence information to enhance the monitoring of crop N dynamics. Firstly, the selection of WOFOST parameters was based on the sensitivity analysis results, and the calibration was conducted through optimization algorithm. Next, machine learning and multi-task neural network (MDNN) were employed to construct the inversion models of four state variables (leaf area index, LAI; leaf dry matter, LDM; leaf N accumulation, LNA; soil moisture content, SMC) based on UAV multispectral data. Meanwhile, a fluorescence operator was constructed using machine learning to capture the complex relationship between fluorescence parameters (actual photochemical efficiency, ΦPSⅡ) and state variables. Finally, the remote sensing inversion results and ΦPSⅡ were incorporated into the dual assimilation framework to update WOFOST. The results showed that MDNN outperformed traditional machine learning in the remote sensing inversion tasks for four state variables. The joint assimilation of LAI, LDM, and LNA improved the simulation accuracy of organ N accumulation. The dual assimilation strategy significantly enhanced the monitoring performance for N accumulation in leaves, stems, and grains (R²: 0.76–0.84, 0.68–0.80, and 0.70–0.75; NRMSE: 15.04–18.74 %, 16.00–25.34 %; 20.40–23.60 %). The treatment of 30 mm irrigation combination with 200 kg ha⁻¹ N fertilizer exhibited the highest N transport (71.22 %) and contribution (60.12 %) to grain. Overall, the dual assimilation framework demonstrated robust performance in monitoring organ-level N dynamics for wheat, providing a promising approach for acquiring spatially variable information about N accumulation and transport.

Plant phenotyping relevance

UAVリモートセンシング、蛍光情報、機械学習、作物モデルを統合した器官レベルの窒素動態推定フレームワークを開発・評価しており、植物状態の取得・推定方法が研究の中心である。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.