Parameter estimation (PE) with a global coupled data assimilation (CDA) system can improve the runs, but the improvement remains in a limited range. We have to come back to simple models to sort out the sources of noises. Incomplete observations and the chaotic nature of the atmosphere have much stronger influences on the PE through the state estimation (SE) process. Here, we propose the guidelines of how to enhance the signal-to-noise ratio under partial SE status.