Wildfire prediction gained much attention in recent years, but, it remains a computational challenge, since it requires the collection of several environmental parameters in real time, from diverse environmental data sources. Moreover, since the globalization of environmental Web Services accessing the environmental data sources, services can be functionally similar, but with different qualities of services. Therefore, environmental data sources compete to provide these functionally similar services with different levels of qualities related to the services (QoS) and also to the data sources and their related data (QoDS). Furthermore, automating and dynamically composing services, while taking prior domain knowledge into account remains an ongoing challenge, since the service composition process is data-driven in the context of wildfire prediction. Subsequently, we present an approach to fulfill the requirement of ranking and selecting the optimal services based on the data received from previous services calls and their related quality dimensions. Our contribution is an automatic, knowledge-driven approach that relies on the ELECTRE III MCDM (Multi-Criteria Decision Making) method and on quality-aware service selection, to automatically and dynamically compose services for the purpose of predicting wildfire catastrophe and triggering alerts.