Evapotranspiration (ET) plays a pivotal role in terrestrial water, energy, and carbon cycles, serving as a critical linkage between the Earth's surface and the atmosphere. However, uncertainties in global ET datasets persist due to the empirical parameterization of resistances in conventional models. To address this limitation, we improved the Remote Sensed Non-Parametric (RSNP) model based on Hamilton's principle, which provides a diagnostic estimation of global ET without requiring resistance parameterization. The RSNP model integrates remote sensing and reanalysis data to generate a global monthly ET dataset at 0.1° spatial resolution from 2001 to 2019. The RSNP ETs were validated globally with eighty-eight FLUXNET2015 sites, thirty-eight basins, and another five global ET datasets. The results showed that: (a) compared with ground observations at the in situ sites, RSNP showed a Root Mean Square Error (RMSE) value of 23.19 mm/month, a bias value of -3.81 mm/month and an R² value of 0.65, and represents relatively great capabilities in vegetated landcovers; (b) compared with water-balance-based ET at the basin scale, the RSNP model displayed a great correlation, with an RMSE value of 113.04 mm/yr, RE value of 16 %, and R² value of 0.89; (c) RSNP provides continuous and gap-free global ETs, which were comparable to other global ET datasets and effectively capture spatial details of land surface ET. This study advances global ET estimation by eliminating the need for resistance parameterization, and the RSNP ET dataset directly supports improved water resource management and climate modelling efforts. The dataset presented in this article has been published in the National Tibetan Plateau Data Center at https://doi.org/10.11888/Terre.tpdc.301343 (Pan, Liu and Yuan, 2024).