Kerala is the South-western state in the Malabar coast of India that experiences floods. The disaster management in this state is response-centric and the 2018 floods indicated the need for a transition from a response-centric approach to a resilient flood risk management (FRM) approach. According to Vincent Ostrom, multi-level governance (MLG) is required for the effective provision of a public good or service and when considering FRM to be a public good/service, it can be inferred that MLG is required for the effective provision of FRM. Kochi, the largest urban agglomeration in Kerala and the location of interest for this research, has a MLG structure and yet does not have an effective resilient FRM system due to the governance issues within FRM. This research is aimed towards adding an additional layer to the existing researches within the domain linking MLG and FRM by taking a data-driven modelling perspective. In order to achieve this, the existing system is viewed as a complex adaptive system (CAS). The existing flood risk governance arrangements (FRGA) were modelled with the help of MAIA (Modelling Agent Systems based on Institutional Analysis) metamodel and Agent-Based Model (ABM) to derive insights about the characteristics of MLG relevant to the governance issues faced by the resilient FRM in Kochi. The results revealed that the existing power distribution among the State, District and Local authorities must be changed to ensure a better resilient FRM system. Furthermore, the results revealed that the network formation among local communities, non-governmental organizations (NGOs) must be encouraged. The research enabled to propose certain policy recommendations - the existing structure must acknowledge and incorporate the need for monitoring the implementation stage of the policies developed. The existing structure must acknowledge and accommodate the need for a better clarity on the responsibilities of decision-makers. The positive influence of networks is higher than the negative influence of the opportunistic behaviour shown by stakeholders. It is thus recommended to encourage network formation among community members, NGOs and government authorities. This encouragement can be provided through “triggering” events. This research looked into environmental damage which can be referred to as a negative “triggering” event. However, positive “triggering” events must be explored. With respect to the power allocation aspect of decentralization, the research gives three recommendations - when considering the definition of resilience to be the ability of system to absorb shocks and show the least fluctuations, the recommendation is to ensure a power distribution set up where the District authority has the highest power followed by the Local authority and State authority. When considering resilient FRM cycle to involve distinct phases, the recommendation is to adopt “policy switching” which is defined in this research as the approach of considering a combination of policies in terms of power distribution such that the requirement of each phase is addressed. In general, the District and Local authorities must be more involved in the decision-making within resilient FRM cycle.