The paper presents two approaches-RSM (response surface meta-model) and the niche pool technique with adaptive radii applied to the covariance matrix adaptation evolutionary strategy (CMA-ES) to locate all global and local optima on a real-valued multi-model landscape. RSM is used to estimate the covariance matrix for CMA-ES, while it is also very helpful for determining the location of optima. The niche pool is used to store the potential up-to-date global or local optima found. Each individual in the niche pool is a 5-tuple elements that contains the information of a potential global or local optimal solution. The activity intensity parameter and niche radius play an important role in the newly proposed CMA-ES-MA (Covariance Matrix Adaptation Evolutionary Strategy with Meta-model and Adaptive niche pool) algorithm. The former is used to determine which niche should be explored when CMA-ES falls into stagnation. The niche radius is the function of covariance matrix and step length, and it is the estimation of the landscape shape corresponding to the location of the niche. Two kinds of experiments are performed: the first one is to verify the performance of locating all the optima; the second one is to compare our approach with CMA-ES based on the real-Parameter Black-Box Optimization Benchmarking software. The experiment results show that this approach can effectively find and are reserve the local and global optima, and the performance of our approach is better than the traditional CMA-ES