Fusion of EnMAP/PRISMA hyperspectral and Sentinel-2 multispectral imagery for estimating crop traits

Accurate crop parameter estimation is crucial for precision agriculture and food security. Hyperspectral (HS) and multispectral (MS) satellite imagery provide complementary spectral, spatial and temporal information, and their fusion offers improved agricultural monitoring quality. This study presents the first comprehensive assessment of spatially enhanced crop parameter estimation using Environmental Mapping and Analysis Program (EnMAP) and Precursore Iperspettrale della Missione Applicativa (PRISMA) HS data (30 m) fused with Sentinel-2 (S2) MS data (10 m). Ten image fusion techniques, including Gram-Schmidt adaptive (GSA), smoothing filter-based intensity modulation (SFIM), generalized Laplacian pyramid (GLP) with modulation transfer function (MTF) matched filter (MTFGLP), full-scale regression-based MTFGLP (MTFGLPFS), MTFGLP with high-pass modulation (MTFGLPHPM), HS superresolution (HySure), maximum a posteriori (MAP) with stochastic mixing model (SMM) (MAPSMM), fast fusion based on Sylvester equation (FUSE), coupled nonnegative matrix factorization (CNMF) and coupled sparse tensor factorization (CSTF), were evaluated for their ability to improve spatial resolution while preserving spectral fidelity. The fused images were assessed using standard quality metrics and then used to estimate 18 wheat biophysical and biochemical parameters across vegetative and reproductive stages using partial least squares regression (PLSR). Results indicate that fusion significantly improved parameter estimation, with HySure delivering the highest accuracy, returning mean R2 values of 0.629, 0.620, 0.406 and 0.450, in EnMAP-S2 fusion in the vegetative phase, PRISMA-S2 fusion in the vegetative phase, EnMAP-S2 fusion in the reproductive phase and PRISMA-S2 fusion in the reproductive phase, respectively. EnMAP-S2 fusion generally outperformed PRISMA-S2. Findings demonstrate that fused satellite imagery can serve as a cost-efficient alternative when in-situ data collection is limited, highlighting the potential of image fusion techniques to advance operational crop monitoring and support decision-making in precision agriculture.