Spatiotemporal Evaluation and Uncertainty Assessment of Multi-Source Evapotranspiration Products Across Hong Kong: Implications for Mai Po Wetland Water-Balance Monitoring

Accurate evapotranspiration (ET) estimates are essential for hydrological assessment, water-resource management, and wetland conservation, yet global ET products often disagree in heterogeneous coastal environments. This study compared ERA5-Land, GLDAS Noah, TerraClimate, MERRA-2, MOD16A2GF, and PML_V2 across Hong Kong during 2001–2020 and examined their applicability to the Mai Po Marshes. The products were standardized to annual water-depth equivalents and evaluated using descriptive statistics, the Friedman repeated-measures test, Holm-adjusted Wilcoxon signed-rank tests, lag-1 autocorrelation analysis, and standard and Hamed–Rao modified Mann–Kendall tests. Spatial uncertainty was quantified using the inter-product coefficient of variation and summarized by land-cover class. Mean annual ET ranged from 530.09 mm yr−1 for GLDAS to 1182.80 mm yr−1 for ERA5-Land. Product differences were significant (χ2(5) = 94.74, p < 0.001; Kendall’s W = 0.947), and all pairwise contrasts remained significant after Holm correction. GLDAS was the only product with a significant increasing trend (Sen’s slope = 4.88 mm yr−2, p < 0.001), while the remaining products showed no significant trends. Within Mai Po, ET estimates differed among products but showed negligible variation among aggregated wetland habitat classes. The findings support uncertainty-aware, multi-product interpretation for regional ET monitoring and cautious application to compact coastal wetlands.