An ecological niche model considers species occurrence points and environmental data for generating models that represent the potential geographical distribution of biological species in specific areas. Theoretically, it is thougth to be similar to those areas where the species in fact occur. For an ecological niche model to be generated, a detailed process, well known by researchers and already formally described in former publications, has to be followed. A Service-Oriented Architecture, SOA, may be seen as a paradigm to organize distributed competences, which may be controlled by different domains. In a distributed computational scenario, which is clearly the case of ecological niche modelling (e.g. distributed data collection available on the internet), a requirement of one party may be compatible with a competence furnished by another, even if both parties are not directly linked. If these competences are organized as services (such as web-services) with well-described interfaces, a match can be found and one can be used by the other. In order to implement a software package for ecological niche modelling based on SOA, a specific component to command the request and response of other (potentially distributed) services is required. Otherwise, the user will have to deal with the process management difficulty to generate a model. In other monolithic solutions, such as DesktopGARP, MaxEnt and openModeller, the process management is already implemented, so a solution with an architectural evolution, represented by a SOA-based system, should improve, instead of reducing, the usability of the software package. Despite seeming to be just a simple system implementation instead of a relevant research task, a SOA-based architecture proposed to develop an ecological niche modelling software package has highly complex issues to be solved in order to meet problem requirements. The management process described in this paper requires dealing with many different technologies and interfaces, so very different aspects are discussed and solutions are suggested for each of the problems found. For instance, this paper describes interfaces for cluster integration so as to implement parallel algorithms as services, which nowadays is a demand and also a complex but very interesting problem. The management component still controls other aspects of ecological niche modelling to provide information to the user about the modelling evolution and, when necessary, to interact with him/her. The aim of the complete proposal is to allow smooth introduction of new aspects, algorithms and steps into ecological niche modelling to improve the modelling technique itself.