UAV applications in agricultural statistics
Outline
The “UAV applications in agricultural statistics” section (Chapters 30–33) explores the emerging role of Unmanned Aerial Vehicles (drones) in filling the data gap between ground surveys and satellite remote sensing. While satellites provide global coverage, they often lack the spatial resolution or temporal flexibility required for smallholder farming, complex cropping patterns, or persistent cloud cover. This section details how drones can be used operationally to delineate field boundaries, monitor crop health, and validate census data.
Chapter 27, “Field Parcel Identification Using UAV Imagery Based on the DCP-MTL Model” introduces the DCP-MTL (Drone-based Cultivation Parcel extraction Multi-Task Learning) model. This architecture is designed to automatically map agricultural fields from very high-resolution drone imagery. Traditional segmentation models often fail to separate adjacent fields (they merge into one large blob) or miss the precise boundaries. The authors propose a multi-task learning approach that forces the neural network to learn two things simultaneously: (a) region extraction for identifying the interior pixels of a field; (b) boundary detection: explicitly identifying the edges/hedges between fields. This allows for the automated creation of digital field maps (vector polygons) which serve as the basis for area sampling frames or for verifying farmer declarations in subsidy systems.
Chapter 28, “Application of UAV in Wheat Growth Monitoring” focus on use of UAV for precision agriculture. This chapter details the use of UAVs equipped with multispectral sensors to monitor the specific phenological stages of wheat. It focuses on extracting quantitative traits such as plant height, leaf area index, and biomass. By correlating these drone-measured parameters with ground samples, statisticians can build robust local yield models to calibrate the coarser, regional estimates derived from satellites.
Chapter 29, “Automated Monitoring of Crop Growth Index using UAV” focuses on the computational workflow to generate crop growth indices. It describes how to do flight planning for consistent temporal sampling. The text describes the generation of orthomosaics combining drone photos into a single map. Vegetation indices (e.g., NDVI, GNDVI) track the greenness curve ig a crop throughout the season. This provides a dataset that is far more detailed than satellite pixels, and can be used to cross-validate satellite-based crop growth profiles.
Chapter 30, “Integrating UAV Imagery into Agricultural Statistics in the Cook Islands” provides an example of agricultural census validation in small island developing states using UAVs. Pacific islands face two major hurdles for using remote sensing for agricultural statistics. Persistent cloud cover makes optical satellite monitoring nearly impossible. Agricultural production is fragmented, with small irregular plots hidden under tree canopies (agroforestry) or in rough terrain. The chapter shows the use of UAVs to conduct a post-enumeration survey following the 2021 Agricultural Census.Drones were used to fly over selected sample enumeration areas to capture high-resolution imagery. The drone measurements of plot sizes were compared against farmer-reported areas from the census interviews. This allowed the statistical office to calculate a correction factor for respondent bias.