Dynamic constrained multiobjective optimization problems (DCMOPs) pose significant challenges due to shifting objectives and fragmented feasible regions. To address these, we propose DCEMOA, a lightweight evolutionary algorithm that decomposes the search space into fixed subspaces and steers evolution through both historical information and adaptive learning. Each subspace maintains a representative solution, guided by two dynamic directives: global learning targets from neighbor subspaces and local velocity estimated by temporal changes. Environmental changes are handled by a simplified multiKalman filter predictor that enables fast and robust population reinitialization. Additionally, a score-based feedback mechanism regulates inter-subspace influence by evaluating feasibility and improvement. Experiments on standard DCMOP benchmarks demonstrate that DCEMOA achieves competitive performance, balancing adaptability, constraint handling, and computational cost. The algorithm maintains strong convergence and diversity across varying degrees of environmental dynamics, often outperforming existing baselines.