Firstly, I want to express my gratitude for giving me the opportunity to review this excellent piece of work. I must admit that as I am not a risk modeler, I may not be able to provide detailed commentary on the models integrated into this paper. Overall, I found the paper to be well-structured and thoughtfully presented. However, I do have a few suggestions for modifications that I believe could enhance clarity and readability. These suggestions are outlined below.

In the introduction but more importantly in the title, the word "humanitarian" in this context might be misleading. While floods can indeed have significant humanitarian impacts, the term "humanitarian" typically refers to actions or interventions aimed at alleviating human suffering, particularly in emergency situations. I will highly recommend revising the introduction to better convey the challenges faced by communities affected by floods.

The second paragraph of the introduction refers to the Sendai Framework for Disaster Risk Reduction. The correct footnote should cite UNDRR instead of UNISDR.

It would also be beneficial to define displacement in the introduction, particularly in the context of flooding, such as during monsoon seasons. This is important because individuals may experience multiple waves of flooding, resulting in repeated displacement.

I would suggest merging sections 2) "Data" and 3) "Flood Model" into one section. The rationale behind this suggestion is that the data section currently contains a significant amount of information on various flood modeling techniques. I recommend using the data section as an introductory section instead. Later in the document, this data could then be utilized to present empirical evidence from past events from multiple sources.

Section 3 flood model: I would suggest revising the 1^st sentence:

To compute a flood inundation footprint from gridded, geo-located river discharge data, said data is related to the historical 85 discharge time series via an extreme value analysis, and the corresponding return period is used to look up flood depths in flood hazard maps.”

In Section 4, Implementation, I suggest refraining from using the term 'natural,' especially when discussing exposure. You can explore the 'no natural disasters' campaign, which emphasizes that while some hazards are inherent in nature and unavoidable, the resulting disasters are often influenced by human actions and decisions. - https://www.nonaturaldisasters.com/

Regarding section 5.3 on historical times series. Have you reach out to the Pakistan government to explore further historical trends? I think data are available only since 2017 https://pdma.punjab.gov.pk/

I highly recommend adding a section on Vulnerability to explain how it is defined for flood displacement as an impact function. We can find some information in section 6.2 where you mention that vulnerability is determined by calibrating impact functions to impact data from past events. It would be beneficial to expand on this further. How many events were assessed? What were the dates and magnitudes? Additionally, the choice of the step function for simplicity may require additional information.

Later, you mention in your calibration process that you decided to use 0.5m as a threshold for Pakistan. Having this information centralized in one section would make it easier to understand. It appears that you calibrated to 0.5m for Pakistan. Do you have any measurements that can corroborate this hypothesis? Additionally, it could be interesting to explain in the conclusion how, as you aim to develop a global impact forecasting system, parameters may be adjusted depending on the context.

At around line 390, I would recommend changing the wording from "historical time series of displacement" to "historical trends". The reason for this change is that each displacement has a duration that can vary from hours to days, months, or years. The term "time series" typically refers to situations where we have multiple snapshots of information over time. Regarding the flooding events of 2022, there are still more than 1.1 million people living in displacement situations in Sindh province. They were not able to return home due to many obstacles.

In the conclusion I personally welcome the suggestion of a range of people at risk of displacement between the worst case and best-case scenarios.

general comments

The paper presents a valuable contribution to the field of fluvial flood modeling and forecasting, addressing a critical need for efficient and accurate methods to assess flood impacts globally. The use of extreme value analysis coupled with openly available data within a catastrophe modeling framework offers a promising approach to rapidly compute flood inundation footprints and estimate associated impacts. The application of the model in Pakistan exemplifies its utility in assessing flood depths, extents, and population displacement. The findings regarding the incorporation of estimated flood protection standards and the calibration of vulnerability models provide important insights for future research and disaster preparedness efforts.

I would be excited to see how this model could translate into better impact-based early warning systems, potentially revolutionizing the way we prepare for and respond to fluvial flood events. Additionally, while the methodology is tested in Pakistan, it would be beneficial to explore how it could be extended to other countries and contexts, considering varying environmental and socioeconomic factors.

The paper addresses the main sources of uncertainty, including uncertainty in displacement data, river discharge, and flood footprint, in a convincing manner. However, it is acknowledged that these factors remain significant limiting factors that could influence the accuracy and reliability of the model's predictions. Further research and improvements in handling uncertainty will be crucial for enhancing the robustness and applicability of the model on a global scale.

specific comments

Line 228: maximum displaced population is reported a month after the end of the peak season. Is this due to a delay in reporting or because of multiple waves in displacements. Typically, we see an initial wave for people directly impacted and a ‘late’ displacement wave due to other socioeconomic impacts such as loss of livelihoods, services, market disruptions etc. The latter is not expected to be captured by the model. 

Line 369: It is indeed surprising that such a large variation in the impact functions result in a relatively small difference in displacement. I am wondering if this could be due to some sort of overfitting of the data as we have limited displacement data and a pretty wide parameter space that is considered?