Scholarlink

Purpose <p>Spatial variability within fields and unpredictable rainfall patterns make nitrogen (N) management challenging, with up to 65% of applied N being lost to the environment. Post-emergence sidedress applications of N fertilizer can improve plant uptake and reduce N losses, making it critical to efficiently identify corn (<i>Zea mays</i> L.) N status at early growth stages. We hypothesized that indicators of plant structure (plant height and canopy cover fraction), canopy greenness (vegetation indices), and their integration with soil and topography-related properties would improve the prediction of early-season corn N status. The objectives of this study were to: (1) evaluate plant height, canopy cover fraction (CCF), and vegetation indices (VI) as indicators of biomass, N concentration, and N uptake at early growth stages (~ V4); (2) assess whether linear models integrating UAV-derived CCF with VI improve N uptake prediction; and (3) determine whether incorporating soil and topographic parameters from publicly available datasets into machine learning (ML) models improves performance over linear regressions.</p> Methods <p>Two large-scale field trials were conducted in Indiana during the 2019 growing season. Multispectral UAV (MicaSense Altum, 0.03 m resolution) and satellite imagery (Planet, 3 m resolution) were acquired and processed to extract CCF and calculate VI. Biomass samples were collected to determine N uptake. Linear regressions and three ML models were evaluated.</p> Results <p>Plant structural metrics, CCF and plant height, were the most reliable predictors of biomass and N uptake (R² up to 0.95). Integrating CCF with NIR-based VI improved or maintained model performance. Adding soil and topographic metrics provided limited improvement.</p> Conclusion <p>Linear regression models performed comparably to ML approaches, emphasizing the utility of simpler models for supporting more efficient in-season fertilizer applications. Performance differences across sites reflected variability in crop development and underscore challenges in model generalization.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integration of satellite, UAV, soil, and topographic data for assessing corn nitrogen uptake at early vegetative growth stages

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年11月16日

查看全文
  • Ana Morales-Ona,
  • James Camberato,
  • Robert Nielsen,
  • Siddhartho Paul,
  • Daniel Quinn

摘要

Purpose

Spatial variability within fields and unpredictable rainfall patterns make nitrogen (N) management challenging, with up to 65% of applied N being lost to the environment. Post-emergence sidedress applications of N fertilizer can improve plant uptake and reduce N losses, making it critical to efficiently identify corn (Zea mays L.) N status at early growth stages. We hypothesized that indicators of plant structure (plant height and canopy cover fraction), canopy greenness (vegetation indices), and their integration with soil and topography-related properties would improve the prediction of early-season corn N status. The objectives of this study were to: (1) evaluate plant height, canopy cover fraction (CCF), and vegetation indices (VI) as indicators of biomass, N concentration, and N uptake at early growth stages (~ V4); (2) assess whether linear models integrating UAV-derived CCF with VI improve N uptake prediction; and (3) determine whether incorporating soil and topographic parameters from publicly available datasets into machine learning (ML) models improves performance over linear regressions.

Methods

Two large-scale field trials were conducted in Indiana during the 2019 growing season. Multispectral UAV (MicaSense Altum, 0.03 m resolution) and satellite imagery (Planet, 3 m resolution) were acquired and processed to extract CCF and calculate VI. Biomass samples were collected to determine N uptake. Linear regressions and three ML models were evaluated.

Results

Plant structural metrics, CCF and plant height, were the most reliable predictors of biomass and N uptake (R² up to 0.95). Integrating CCF with NIR-based VI improved or maintained model performance. Adding soil and topographic metrics provided limited improvement.

Conclusion

Linear regression models performed comparably to ML approaches, emphasizing the utility of simpler models for supporting more efficient in-season fertilizer applications. Performance differences across sites reflected variability in crop development and underscore challenges in model generalization.