Simultaneous identification of groundwater contaminant source and simulation model parameters based on an ensemble Kalman filter – Adaptive step length ant colony optimization algorithm

•Construct an EnKF-ASACO algorithm to identify the pollution source characteristics and simulation model parameters simultaneously.•Adopt the EnKF algorithm to provide an initial point for the ASACO algorithm.•Improve the previous ACO algorithm by the adaptive step length search strategy.•Embed the Kriging surrogate model into the EnKF and ASACO algorithms. The simulation–optimization (SO) method is the most widely used method for groundwater contaminant source identification. However, the heuristic optimization algorithms in the SO method often rely on the selection of the initial point (the initial estimate of each variable to be solved). If the initial point is far from the actual value, the convergence speed of the algorithm will be slow, and it will easily converge prematurely. It is challenging to obtain the accurate identification results of the contaminant source. Therefore, providing a good initial point for the optimization algorithm is critical. The ensemble Kalman filter (EnKF) algorithm has the advantages of a simple calculation process and fast running speed. However, when the inverse problem has a strong nonlinearity, its identification accuracy still needs improvement. This study combined the EnKF and adaptive step length ant colony optimization (ASACO) algorithms to construct an EnKF-ASACO algorithm. This was used to identify pollution source characteristics and simulation model parameters simultaneously. First, the EnKF algorithm was adopted to provide a good initial point for the optimization algorithm. On this basis, the adaptive step length search strategy was used to improve the ant colony optimization (ACO) algorithm, and the ASACO algorithm was obtained. As a result, the step length changed adaptively to avoid falling into the local optimum. At the same time, to reduce the substantial computational load caused by repeatedly calling the numerical simulation model, the Kriging method was used to establish a surrogate model of the simulation model. Finally, the identification results of the EnKF-ASACO algorithm were compared and analyzed with those of the EnKF, previous ACO, and ASACO algorithms to verify the effectiveness of the proposed new algorithm. The results show that compared with the three single algorithms, the EnKF-ASACO algorithm constructed in this study can speed up the convergence speed and significantly improve the search efficiency and identification accuracy. Additionally, the EnKF-ASACO algorithm was also applied in