banana_irrigation_dataset

This dataset supports a multilevel sensing framework for monitoring banana (Musa AAA cv. Williams) responses to controlled irrigation deficit. The experiment was conducted between April and September 2024 at a commercial farm in Varela, Magdalena, Colombia, under four irrigation treatments (T1 = 30%, T2 = 50%, T3 = 75%, T4 = 100% of crop evapotranspiration ETc) applied to 130 plants arranged in a randomised block design.
The dataset integrates three sensing layers collected at weekly resolution across 26 time steps: (1) soil IoT nodes (NPKPHCTH-S, FDR technology) recording relative soil moisture; (2) proximal RGB vision (Raspberry Pi 4 + YOLOv8) providing pseudostem height and functional leaf count; and (3) fortnightly UAV multispectral imagery (DJI Phantom 4 Multispectral, 5 bands) yielding per-plant NDVI and NDRE statistics and canopy area metrics. Meteorological variables (ET₀, VPD, precipitation) were recorded by an iMetos automated weather station.
The file banana_irrigation_dataset_v1.csv contains 3,380 records (130 plants × 26 weeks) with 32 variables representing raw and directly measured values. A README_dataset.txt file describes all variables, units, missing-value conventions, and sensor specifications.