This document describes a protocol for the verification and validation of building fire evacuation models. This document mostly addresses evacuation model components as they are in microscopic (agent-based) models. Nevertheless, it can be adopted (entirely or partially) for macroscopic models if the model is able to represent the components under consideration.

The area of application of the evacuation models discussed in this document includes performance-based design of buildings and the review of the effectiveness of evacuation planning and procedures. The evacuation process is represented with evacuation models in which people's movement and their interaction with the environment make use of human behaviour in fire theories and empirical observations^[5]. The simulation of evacuation is represented using mathematical models and/or agenttoagent and agent-to-environment rules.

The area of application of this document relates to buildings. This document is not intended to cover aspects of transportation systems in motion (e.g. trains, ships) since specific ad-hoc additional tests may be required for addressing the simulation of human behaviour during evacuation in these types of systems^[6].

This document includes a list of components for verification and validation testing as well as a methodology for the analysis and assessment of accuracy associated with evacuation models. The procedure for the analysis of acceptance criteria is also included.

A comprehensive list of components for testing is presented in this document, since the scope of the testing has not been artificially restricted to a set of straightforward applications. Nevertheless, the application of evacuation models as a design tool can be affected by the numbers of variables affecting human behaviour under consideration. A high number of influences can hamper the acceptance of the results obtained given the level of complexity associated with the results. Simpler calculation methods, such as macroscopic models, capacity analyses or flow calculations, are affected to a lower extent by the need to aim at high fidelity modelling. In contrast, more sophisticated calculation methods (i.e. agent-based models) rely more on the ability to demonstrate that the simulation is able to represent different emergent behaviours. For this reason, the components for testing are divided into different categories, enabling the evacuation model tester to test an evacuation model both in relation to the degree of sophistication embedded in the model as well as the specific scope of the model application.

In Annex A, a reporting template is provided to provide guidance to users regarding a format for presenting test results and exemplary application of verification and validation tests are presented in Annex B.