Multispectral imaging of winter wheat leaves for leaf nitrogen prediction: data and analysis code

This repository contains the datasets and Python analysis code accompanying the manuscript entitled “Beyond random cross-validation: Transferability of multispectral imaging models for winter wheat leaf nitrogen across cultivars and growing environments”. The dataset includes leaf-level multispectral measurements, plot- and sample-level aggregated data, and measured total leaf nitrogen content (TNC) from two experimental settings: Field experiment: Ten winter wheat cultivars grown under two nitrogen fertilisation treatments in a split-plot design with four blocks during the 2022/2023 growing season in Rugvica, Croatia. Growth chamber experiment: Two winter wheat cultivars grown under three nitrogen treatments, with multispectral measurements obtained from fresh and dried leaves. Multispectral measurements were acquired using the CropReporter™ imaging system (PhenoVation B.V.), while total leaf nitrogen content was determined using the Kjeldahl method. The repository also provides Python scripts for data preparation, statistical analyses, correlation analyses, machine-learning model development and validation, cross-environment transferability assessment, stability selection, and sensitivity analyses. The accompanying README file describes the experimental design, dataset structure, measured variables, vegetation indices, analytical procedures, and instructions for reproducing the reported results. Software: Python (version 3.14 used for the reported analyses). Data licence: Creative Commons Attribution 4.0 International (CC BY 4.0). Code licence: MIT License.