In this paper, we investigate the polarization effect of the amplitude phase shift keying(APSK) constellations. We find that the polarization effect of the APSK constellations is affected by the bit mapping and the bit loading. Traditionally, the Gray mapping is usually adopted in APSK constellations. Based on the given Gray mapping, we firstly propose the bit interleaved coded modulation polar-APSK(BICM-PA) scheme, which neglects the correlations between the bit levels by using the bit interleaver. In the BICM-PA scheme, a new metric called cumulative Bhattacharyya parameter is introduced to optimize the bit loading of the APSK constellations under the Gray mapping. Second, the multilevel coded modulation polar-APSK(MLCM-PA) is proposed to further improve the performance. A twostage optimization approach is adopted to select the bit mapping and the bit loading in the MLCM-PA scheme. The semi-set partitioning mapping is introduced to achieve a better system performance in the MLCM-PA scheme. Simulation results verify the effectiveness of all the proposed metrics. In addition to this, BICM-PA and MLCM-PA both outperform the coded modulation Turbo-APSK scheme up to 1 dB. In this paper, we investigate the polarization effect of the amplitude phase shift keying (APSK) constellations. We find that the polarization effect of the APSK constellations is affected by the bit mapping and the bit loading. Traditionally, the Gray mapping is usually adopted in APSK constellations. Based on the given Gray mapping, we referred propose the bit interleaved coded modulation polar-APSK (BICM-PA) scheme, which neglects the correlations between the bit levels by using the bit interleaver. In the BICM-PA scheme, a new metric called cumulative Bhattacharyya parameter is introduced to optimize the bit loading of the APSK constellations under the Gray mapping. Second, the multilevel coded modulation polar-APSK (MLCM-PA) is proposed to further improve the performance. A twostage optimization approach to select the bit mapping and the bit loading in the MLCM-PA scheme. The semi-set partitioning mapping is introduced to achieve a better system performance in the MLCM-PA s cheme. Simulation results verify the effectiveness of all the proposed metrics. In addition to this, BICM-PA and MLCM-PA both outperform the coded modulation Turbo-APSK scheme up to 1 dB.