HI intensity mapping is a new and efficient technique for mapping the large-scale-structures in the Universe and its expansion history in three dimensions. Due to the faintness of HI signal, an effective removal of Galactic foregrounds and a careful control of systematics is essential. One way of mitigating systematics is by cross-correlating multiple surveys issued from complementary datasets, which benefits from cancelled systematics, zero-noise-biased cross-spectrum, and complementary mass bias information. The first part of the authorâs work focuses on the cross-correlation between CMB maps and optical galaxy survey, where the thermal SZ cluster residuals in the Planck 2015 NILC CMB map is detected with â 30Ï significance at cluster scale, and overallâ 51Ï significance including large scales. The percentage of thermal SZ emission left over in the NILC CMB map is quantified to be 44 Â± 4%, which, however, is proved to have negligible impact on ISW measurement but can potentially challenge upcoming CMB experiments with higher resolution. In contrast, we provide an alternative CMB map, produced with the 2D-ILC component separation technique, which is shown to be free from thermal SZ contamination. In the second part, the author forecasts the impact of intensity mapping 1/ f noise on cosmological parameter constraints through a Fisher matrix analysis. Without 1/ f noise, constraints of w0 = â1 Â± 0.06, wa = 0 Â± 0.13 and HI bias bHI = 1 Â± 0.02 are obtained from SKA1-MID Band 2+Planck. A representative 1/ f noise model degrades the results by â¼ 50%. To mitigate this, one requires a minimised 1/ f noise spectral slope, a low knee frequency and a large telescope slew speed. A correlation in fre- quency is also preferred. Finally, forecasts have been made on the cross-correlation between HI intensity maps and the galaxy lensing field. Even with two idealised optimal surveys without any noise and a full-sky coverage, the total detection signal-to-noise ratio and parameter constraints are merely comparable to those obtained from Square Kilometre Array (SKA) auto-correlation with the Planck prior. Therefore, the auto-correlation of intensity maps performs better than cross-correlation with galaxy lensing field, in terms of signal detection and cosmological parameter constraints.