At continental mid-latitude, soil moisture (SM) is a key component of the climate systems and land surface initialization is crucial for subseasonal-to-seasonal (S2S) predictions. We introduce a new stochastic global land reanalysis system called the Norwegian Climate Prediction Model Land (NorCPM-Land), which will be used to initialize the land component of the Norwegian Climate Prediction Model (NorCPM). We assimilate the blended SM from the European Space Agency’s Climate Change Initiative (ESA CCI) into a 30-member offline simulation of the land surface Community Land Model (CLM). Fluxes are provided by 30-member historical simulations of the full coupled NorCPM. The Ensemble Kalman Filter (EnKF) updates daily the soil column from the SM data using the cumulative density function matching method.The NorCPM-Land is currently produced for 40 years from 1980 to 2019. Assimilation significantly improves the land surface state variability and reduces error by 10.5% when validated using independent SM observations and by reanalysis estimates from ERA5-Land. It also yields an improvement of land surface energy, runoff and net primary production. We demonstrate that adjusting the underlying soil moisture considerably enhances the ability to simulate land surface state dynamics.