Scholarlink

Purpose <p>Accurate crop yield prediction across spatial scales remains a fundamental challenge in precision agriculture, particularly when integrating field measurements with satellite observations. Current approaches often fail to systematically bridge the gap between ground-based measurements and satellite-derived predictions, leading to reduced accuracy and increased uncertainty.</p> Objective <p>To develop and evaluate a novel upscaling framework that systematically integrates ground-based, drone, and satellite data to enhance corn yield prediction accuracy and reduce prediction uncertainty across multiple spatial scales.</p> Methods <p>The study was conducted at two farms in Ontario, Canada, during the 2019–2020 growing seasons. Data collection included Sentinel-2 satellite imagery and high-resolution drone observations. Two prediction strategies were developed and compared: (1) a conventional direct approach using satellite data alone, and (2) an innovative upscaling approach that bridges multiple spatial scales through systematic integration of multi-platform observations.</p> Results <p>The upscaling strategy demonstrated superior performance compared to the direct approach, improving correlation coefficients from 0.82 to 0.88 at Farm A and from 0.81 to 0.85 at Farm B. The method substantially reduced prediction uncertainty, with high uncertainty areas (&gt;1%) decreasing from 43.3% to 20.4% at Farm A and from 74.5% to 9.2% at Farm B. Analysis revealed scale-dependent relationships between spectral variables and yield, with vegetation indices (CCCI, CIgreen, and GNDVI) showing varying importance across spatial resolutions.</p> Results <p>Systematic integration of multi-scale observations significantly improves yield prediction accuracy while reducing uncertainty. The developed framework provides insights into uncertainty propagation across scales and offers practical implications for precision agriculture and sustainable crop management, addressing a critical gap in current precision agriculture applications.</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

Reducing corn yield prediction uncertainty through multi-scale integration of ground, drone, and satellite data

  • 期刊论文

  • 发表日期：

    2025年9月17日

查看全文
  • Solmaz Fathololoumi,
  • Hiteshkumar B. Vasava,
  • Mohammad Karimi Firozjaei,
  • Prasad Daggupati,
  • John Sulik,
  • Asim Biswas

摘要

Purpose

Accurate crop yield prediction across spatial scales remains a fundamental challenge in precision agriculture, particularly when integrating field measurements with satellite observations. Current approaches often fail to systematically bridge the gap between ground-based measurements and satellite-derived predictions, leading to reduced accuracy and increased uncertainty.

Objective

To develop and evaluate a novel upscaling framework that systematically integrates ground-based, drone, and satellite data to enhance corn yield prediction accuracy and reduce prediction uncertainty across multiple spatial scales.

Methods

The study was conducted at two farms in Ontario, Canada, during the 2019–2020 growing seasons. Data collection included Sentinel-2 satellite imagery and high-resolution drone observations. Two prediction strategies were developed and compared: (1) a conventional direct approach using satellite data alone, and (2) an innovative upscaling approach that bridges multiple spatial scales through systematic integration of multi-platform observations.

Results

The upscaling strategy demonstrated superior performance compared to the direct approach, improving correlation coefficients from 0.82 to 0.88 at Farm A and from 0.81 to 0.85 at Farm B. The method substantially reduced prediction uncertainty, with high uncertainty areas (>1%) decreasing from 43.3% to 20.4% at Farm A and from 74.5% to 9.2% at Farm B. Analysis revealed scale-dependent relationships between spectral variables and yield, with vegetation indices (CCCI, CIgreen, and GNDVI) showing varying importance across spatial resolutions.

Results

Systematic integration of multi-scale observations significantly improves yield prediction accuracy while reducing uncertainty. The developed framework provides insights into uncertainty propagation across scales and offers practical implications for precision agriculture and sustainable crop management, addressing a critical gap in current precision agriculture applications.