Monitoring Double-Cropped Extent with Remote Sensing in Areas with High Crop Diversity.

Area of Science:

• Agricultural remote sensing
• Machine learning applications in agriculture
• Environmental resource management

Background:

• Multi-cropping systems significantly impact food security and resource use, but accurate data is lacking, especially in diverse agricultural regions.
• Existing methods for assessing cropping intensity often fail in areas with high crop diversity, leading to overestimations of double-cropping.
• Reliable field-scale data on cropping practices is crucial for informed land and water management decisions.

Purpose of the Study:

• To develop and apply a scalable, low-cost method for identifying double-cropping at the field scale using satellite imagery.
• To evaluate the effectiveness of machine learning models compared to traditional methods in diverse agricultural landscapes.
• To provide accurate data for monitoring cropping intensity and informing policy related to food production and resource use.

Main Methods:

• Utilized Landsat satellite imagery and developed a process combining machine learning with expert labeling for cropping intensity assessment.
• Evaluated multiple machine learning models, including time-series image classification and deep learning, trained on expert-identified data.
• Applied the developed method to a diverse agricultural region in Washington State, USA, to measure double-cropping extent.

Main Results:

• Traditional rule-based methods using vegetation indices performed poorly in diverse crop regions, overestimating double-cropping.
• Machine learning models, particularly a deep learning approach, accurately captured nuances and achieved high accuracy (96-99% overall, 83-93% producer accuracy for double-cropped class).
• The expert labeling process proved effective and scalable for remote sensing applications, validating model outputs.

Conclusions:

• The developed machine learning approach offers a robust and accurate solution for mapping double-cropping in complex agricultural environments.
• This method provides valuable, high-resolution data for long-term monitoring of cropping intensity and supports evidence-based policy decisions.
• The low-cost, scalable nature of the approach makes it suitable for broad application in agricultural research and management.