Improvement of crop growth simulations under different drip irrigation modes by jointly assimilating UAV multimodal data into crop models

Abstract

Accurate crop monitoring is essential for agricultural planning and food security. This study developed a coupling framework of unmanned aerial vehicle (UAV) multimodal data and crop models based on a sequential data assimilation method, offering technical support for crop growth simulation and precision management under drip irrigation modes in the Hexi Corridor of Northwest China. Multispectral and thermal infrared image data of spring maize at different growth stages were acquired via UAVs. The UAV-derived leaf area index (LAI) and soil moisture (SM) were assimilated into the WOFOST model using the ensemble Kalman filter (EnKF). Three assimilation schemes including (a) LAI, (b) SM, and (c) LAI+SM were compared to explore the effects of different mulching treatments (mulched vs. non-mulched) and irrigation gradients on assimilation performance under drip irrigation modes. Our results showed that the fusion of UAV-based multispectral and thermal infrared multimodal data enabled accurate retrieval of LAI and SM, with a maximum R² of 0.85. The three assimilation schemes exhibited significant differences, and the joint assimilation of LAI and SM outperformed the others. This may be since LAI and SM, as key indicators of crop growth and development, undergo dynamic changes throughout the growth period, and their joint assimilation fully captures the temporal variability of crops and soil. In addition, the proposed framework demonstrated marked variations in simulation accuracy across different drip irrigation modes. Overall, the performance for shallow buried drip irrigation (SBDI) was superior to that for surface drip irrigation (SDI) and film-mulched drip irrigation (FDI). This may be attributed to the direct influence on soil evaporation and evapotranspiration under the latter two modes, which in turn modifies crop growth and development processes and ultimately affects the model's simulation accuracy.

Plant phenotyping relevance

UAVマルチスペクトル・熱赤外データからLAIを推定し、作物モデルへ同化する手法の開発と性能比較が研究の中心であり、植物形質取得の技術的評価を含む。

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.