Skip to content

Intercomparison of Machine Learning Algorithms for Remote Sensing-based In-season Crop Mapping

Home / Papers / Intercomparison of Machine Learning Algorithms for Remote Sensing-based In-season Crop Mapping

Intercomparison of Machine Learning Algorithms for Remote Sensing-based In-season Crop Mapping

Authors: August Posch, Jitendra Kumar, Forrest M. Hoffman, Auroop R. Ganguly Date: 2026-06-04 Paper ID: openalex:2606.05731

Summary

This study addresses the critical need for near real-time, in-season crop mapping by comparing ten machine learning algorithms using Harmonized Landsat-Sentinel surface reflectance time series. By incorporating crop rotation history and year-wise cross-validation, the authors identify Support Vector Machines as the most effective model for early-season classification of corn and almonds in California and Iowa. The results quantify significant uncertainty stemming from interannual variability while demonstrating the feasibility of early-season mapping ahead of official post-harvest products.

Key Contributions

  • Introduces a robust intercomparison framework for in-season crop type mapping by evaluating ten machine learning algorithms across interannual variations.
  • Achieves accurate 30m resolution mapping of corn in Iowa and almonds in California by early June using Harmonized Landsat-Sentinel (HLS) imagery.
  • Demonstrates that Support Vector Machines (SVM) outperform other tested machine learning algorithms for in-season classification, reaching mean F1 scores of 0.74 for almonds and 0.59 for corn.

Metadata & Links

url
https://arxiv.org/abs/2606.05731
paper_id
2606.05731
paper_source
openalex
domain
computer-vision
tags
architectures
datasets
concept_slugs
dataset_slugs
skill
TimeSeriesSkill
created_at
2026-06-07T06:32:42Z