This paper develops an integrated socio-hydrological modeling framework that couples a hydrodynamic model, an agent-based model, and a transportation model to examine household evacuation outcomes under various shelter location plans and human behavior scenarios. The results demonstrate the unique functionality of the model to support flood risk assessment and to advance the understandings of evacuation performances. The manuscript is well organized and written. The logic flow is easy to follow. Tables and Figures are clear and well presented. I think this is a high-quality manuscript, which will contribute to the flood managment practice. I have only few minor concerns, as follows:

• Lines 184-187: May agents also consider the shelter with least travelling time? please check the assumption.
• Lines 209-212: Will family agents consider at system level? Why will agents want to contribute to system efficiency? Mode 2 should be re-interpreted based on rational an assumption.
• In the results section, I think it is better to discuss the specific policy implications and recommendations following each result, from the perspectives of both emergency responders and family agents. In this way, readers can easily link the new findings to management practice.

Summary: the authors evaluate tradeoffs between behavioral heterogeneity in departure time, which route they choose and where evacuation centers are location on traffic congestion and time to evacuate the affected population using a coupled model of flood inundation and traffic routing.

Comments: The paper is very interesting and well organized. Following are few of my comments that the authors may want to consider before resubmitting their manuscript:

 - at places authors may want to check for languange

- there is reference to sociohydrology - perhaps the authors can spend some space on why it is sociohydrology (e.g. because of bidirectional feedbacks between agents decisions and travel times). Also more references, placing this study in the landscape of other sociohydrological studies would be helpful.

- It remains a semi-empirical study. The authors may want to discuss what next steps should be taken to make it more realistic in terms of mapping household behaviour. For example using household surveys on psychological factors that may influence such behavior. Can then behavior of others influences the psychology of those who have not yet started to evacuate (e.g. "others are evacuating with urgency so I better hurry"). This may be a more concious feedback than the traveltime congestion feedback due to heterogeniety in time of departure)

- Table 1 is secondary not primary data. Primary data is self measured, e.g. trhough field campaigns

- How are the travel time results affected when shelter locations are designed to be located close to denser parts of the population than when they are randomly assigned in space? Here, perhaps simulations with more number of shelters and where they are designed to be located are needed to conclude that marginal gains reduce as number of shelters are increased.

- Are system wide shortest routes calculated for each of households that have yet to decide to evacuate at each time step of the simulation? This is not clear and perhaps affects the interpretation of the results regarding the superiority of centrally planned routes. What if households are just given live updates on congestion and then let them decide on their own vs a route that is centrally planned before the flood hits. Centrally planned routes may still be better if they are repeatedly calculated at each time step of simulation where central planners also have information on congestion on various routes and it would be interesting to see how this fares compared to agents deciding on their own route but with live information on congestions. Perhaps the authors may want to provide result on this so the two cases can be fairly evaluated (self organization for evaucation vs centrally planned one - which one is better?)