Web applications are complex systems that are in the core of many businesses. However, their development is, contrary to other domains, not characterized by rigorous software engineering methods. The consequence is that many web applications are poorly structured and are not adequately documented, which leads to difficult maintenance. One way for dealing with web application complexity is their modeling on a higher level of abstraction. This also provides the possibility to cope with continuous evolution of web applications. In this paper we present phpModeler, a tool for reverse engineering of legacy PHP web applications. It generates static UML diagrams showing resources that each web page is using, web page's functions and dependencies it has on other web pages. Once the models have been created, phpModeler can analyze them and generate dependency models for each entity in every web page model. phpModeler can also be used to highlight the difference between page models - a feature that, when combined with an SVN repository shows the way how a web page has evolved over time. Tool usability has been tested on a case study application - iForestFire.