Evaluating UAV Pansharpening for Soil Water Content Estimation Using Spectral Indices in Generative Sweet Maize

Soil water content monitoring during the generative phase of sweet maize is critical for irrigation decisions in precision agriculture, yet conventional destructive sampling is too slow and spatially limited for field-scale management. This study evaluated whether pansharpening improves or degrades the accuracy of eight spectral indices for estimating SWC through an index-by-index comparison. The study was conducted on a 720 m² field of fertilizer-dose treatment plots in Ciparanje, Jatinangor, Indonesia, with 15 purposively selected sampling points. Multispectral imagery was acquired with a DJI Matrice 350 RTK UAV and MicaSense Altum-PT sensor at 30 m altitude and 80% overlap, yielding 1.24 cm/pixel ground sample distance before and 0.62 cm/pixel after pansharpening. Eight indices were correlated with gravimetric SWC via Spearman correlation and linear regression, cross-checked with classical assumption tests. Pansharpening improved spatial resolution with generally preserved spectral quality, though the NIR and RedEdge bands showed the weakest fidelity. The effect on SWC estimation was index-dependent: NDVI was essentially preserved, whereas SAVI and MSAVI lost 25.0% and 22.7% of their explanatory power; red-edge and water-based indices remained unchanged, and TVDI was non-significant. Only the red-edge indices CIre and NDRE satisfied all classical regression assumptions in both products; the pansharpened NDRE model (R² = 0.4177, RMSE = 1.5021, p = 0.009) was therefore used to generate a field-capacity-based SWC map of the 720 m² field, providing zone-specific irrigation guidance. Pansharpening thus benefits SWC estimation depending on index formulation, supporting an index-specific processing rule for irrigation scheduling in regenerative sweet maize.