Multitemporal multispectral imagery for rice yield and phenology prediction

Timeseries data captured by unoccupied aircraft systems (UASs) are increasingly used for agricultural applications requiring accurate prediction of plant phenotypes from remotely-sensed imagery. This benchmark dataset for rice supports the development of improved analytical approaches for phenotype prediction from multispectral timeseries of drone imagery. The dataset includes five experiments conducted at the USDA-ARS Dale Bumpers National Rice Research Center in Stuttgart, AR in 2021 and 2022: two nitrogen rate studies, a private hybrid study, an inbred study, and a genetic diversity study. A randomized block design was established in both years, with 252 total plots in 2021 and 180 plots in 2022. Plots were imaged at 12 timepoints throughout the season in both years. The dataset includes images for each plot as well as extracted features (49 features including vegetation indices, texture properties, and thermal features), and per-plot yield and phenology data., A team of licensed and certified drone pilots collected UAS imagery during the 2021 and 2022 growing seasons. Multispectral and thermal imagery data were acquired using the DJI Matrice 210 V2 quadcopter (DJI, Shenzhen, Nanshan District, China) equipped with an Altum sensor (MicaSense, Seattle, Washington). The Altum sensor captured six bands across the electromagnetic spectrum, including blue (459-491 nm), green (547-573 nm), red (661-675 nm), red edge (711-723 nm), near-infrared (814-870 nm), and long-wave infrared (8,000-14,000 nm). Data were collected at approximately weekly intervals, weather permitting, within two hours of solar noon. Prior to each flight, images of a calibrated reflectance panel were collected using the Altum sensor and the downwelling light sensor (DLS) to generate radiometrically calibrated reflectance maps. Flights were conducted at a height of 120 m above ground level, providing a spatial resolution of 5 cm/px for the spectral bands and 81 cm/px for the therma..., , # Multitemporal multispectral imagery for rice yield and phenology prediction

[https://doi.org/10.5061/dryad.v41ns1s4z](https://doi.org/10.5061/dryad.v41ns1s4z)

The dataset includes five experiments conducted at the USDA-ARS Dale Bumpers National Rice Research Center in Stuttgart, AR in 2021 and 2022: two nitrogen rate studies, a private hybrid study, an inbred study, and a genetic diversity study. A randomized block design was established in both years, with 252 total plots in 2021 and 180 plots in 2022. Plots were imaged at 12 timepoints throughout the season in both years.

## Description of the data and file structure

The dataset includes images for each plot as well as extracted features (49 features including vegetation indices, texture properties, and thermal features), and per-plot yield and phenology data. Images are named by plot within separate directories for each flyover date. Test and train sets are based on a 'leave-one-replicate-out' approach, where one replicate plot...