Web applications tend to evolve quickly, resulting in errors and failures in test automation scripts that exercise them. Repairing such scripts to work on the updated application is essential for maintaining the quality of the test suite. Updating such scripts manually is a time consuming task, which is often difficult and is prone to errors if not performed carefully. In this paper, we propose a technique to automatically suggest repairs for such web application test scripts. Our technique is based on differential testing and compares the behavior of the test case on two successive versions of the web application: first version in which the test script runs successfully and the second version in which the script results in an error or failure. By analyzing the difference between these two executions, our technique suggests repairs that can be applied to repair the scripts. To evaluate our technique, we implemented it in a tool called WATER and exercised it on real web applications with test cases. Our experiments show that WATER can suggest meaningful repairs for practical test cases, many of which correspond to those made later by developers themselves.