Data and R code for predictive stability of a pooled length–width equation for leaf area estimation across four citrus types

This dataset contains the measurements and reproducible R code supporting the study “Predictive stability of a pooled length–width equation for leaf area estimation across four citrus types”. The dataset comprises 200 mature, fully expanded leaves from four citrus plant materials: sweet orange ‘Thomson’ (Citrus sinensis), clementine (Citrus clementina), lemon ‘Quatre Saisons’ (Citrus limon), and sour orange (Citrus aurantium), with 50 leaves per citrus type. For each leaf, the dataset provides reference lamina area, maximum lamina length, maximum lamina width, the product of length and width, and the width-to-length ratio. Measurements were obtained from calibrated digital images analysed using ImageJ version 1.54g. The petiole was excluded from all measurements. The accompanying R script reproduces data preparation, descriptive analyses, comparison of linear and allometric models, model diagnostics, a 70/30 internal hold-out validation, repeated stratified 10-fold cross-validation, leave-one-citrus-type-out validation, and the tables and figures supporting the manuscript. The final pooled equation was S = 0.6829 × L × W, operationally rounded to S = 0.683 × L × W. The data were collected in December 2025 at an agricultural experimental facility of the University of Mostaganem, Mostaganem, northwestern Algeria. Tree identifiers were not retained; therefore, the individual leaf is the observational unit, and the validation procedures should be interpreted as internal leaf-level or cross-type validation rather than independent external validation.