Quantification of Loss of Access to Critical Services during Floods in Greater Jakarta: Integrating Social, Geospatial, and Network Perspectives
Round 1
Reviewer 1 Report
The paper introduces an intriguing approach to flood modeling, utilizing network analysis in conjunction with GIS data. The focal point of the study revolves around the 2020 floods in Indonesia, specifically delving into the complexities of Jakarta's infrastructure vulnerabilities and their interplay with landscape and socio-economic factors. To accomplish this, the study extensively simulates potential flood scenarios within their developed model. This content offers a fresh and captivating perspective, catering to specialists in risk analysis and environmental modelers who rely on remote sensing data. Moreover, the implications for policymakers are significant, as it furnishes a crucial backdrop for comprehending how floods may impact the critical services of urban infrastructure. While the manuscript holds promise, there remains room for improvement prior to publication, and I have outlined my comments below.
Abstract
Line 18: "The entire framework can be applied to other cities and urban areas." - Clarify whether this refers to cities in Indonesia or globally. Further elaboration on data requirements can be provided in the Discussion section.
Introduction
Consider adding a brief paragraph explaining the use of remote sensing data in modeling for the benefit of readers.
Line 50: Provide examples of facilities and installed devices to illustrate your point.
Background
Lines 97-98: Add a reference to support the statement.
Lines 210-212: Evaluate the reasonableness of this assumption and provide supporting context.
Data
This section is well-structured and informative.
Methods
Lines 301-308: Enhance the manuscript with a rigorous mathematical formulation of the network, including dimensionality and potential constraints. Elaborate on these aspects for better clarity.
Line 352: Explain how depth levels are calculated for the readers' understanding.
Results
Line 415: Consider using commas to separate large numbers throughout the text for readability.
Discussion
Include a paragraph discussing the usage of remote sensing data and the framework's applicability to other cities, as mentioned in the Abstract.
Conclusions
Lines 508-513: Work through this paragraph to provide a clearer summary of the main points and the key takeaways of your research. Explain the significance and implications of your findings in a concise manner.
Author Response
Thank you for your valuable comments. We have addressed all of them. Please see our responses under each of your question below. Please see attached an improved manuscript.
Abstract
Line 18: "The entire framework can be applied to other cities and urban areas." - Clarify whether this refers to cities in Indonesia or globally. Further elaboration on data requirements can be provided in the Discussion section.
--- added word "globally" in the Abstract.
--- this text was added in the Discussion (lines 494-499): "The framework can be applied to other cities with minimal adaptation. The physical infrastructure data sources will remain the same. What will require some changes is the social dimension, which includes information on population density and poverty. In some cases these data may not be available, but in other cases it is possible that even more detailed data can be found. Overall, the better the quality of the data, the more accurate the results."
Introduction
Consider adding a brief paragraph explaining the use of remote sensing data in modeling for the benefit of readers.
-- the following text was added (lines 65-68): "Satellite and airborne remote sensing systems can provide much of the information needed to delineate flood-prone areas, assess damage, and feed models that can forecast the vulnerability of inland and coastal areas to flooding \cite{klemas2015}".
Line 50: Provide examples of facilities and installed devices to illustrate your point.
-- The paragraph (lines 50-55) changed to "The growth of cloud computing, big data analytics and AI infrastructure, as well as IoT, coupled with the massive expansion of transportation infrastructure, is making critical components of cities more interdependent and more vulnerable to attack \cite{telo2023}. Over-reliance on various sensors and centralized data collection and decision-making mechanisms can lead to serious problems in the event of a minor power outage or cyberattack, so the more complex a city is, the more vulnerable it is \cite{vale2005resilient}."
Background
Lines 97-98: Add a reference to support the statement.
--addressed, references added, text changed (lines 101-103): The risk of catastrophes is related to a power-law distribution, where the probability of an event typically varies as a power of its size \cite{malamud2006}. The power law also governs the outcome of disasters \cite{hanson2008}.
Lines 210-212: Evaluate the reasonableness of this assumption and provide supporting context.
-- the following text added (lines 213-219): "We calculate the proportions of loss of access to services. The floods in our model are static. This provides a simple way to test the framework for calculating network statistics in the course of service loss. A more practical application of this technique will require a more rigorous approach with propagating floods, which will allow a more dynamic assessment of access loss; probability models with input from machine learning investigations will further enrich the calculations."
Data
This section is well-structured and informative.
--Please note that we have adjusted the structure of the paper to mdpi standards; now the section names are different, but the essence is same.
Methods
Lines 301-308: Enhance the manuscript with a rigorous mathematical formulation of the network, including dimensionality and potential constraints. Elaborate on these aspects for better clarity.
-- addressed. A formulation was enriched with more details. Text changed to (lines 308-318):
A mathematical formulation of the network is as follows. Let $G = (V, E)$ be a planar undirected graph of the urban transportation system, where $V$ is a set of vertices representing road junctions (intersections) and $E$ is a set of edges representing parts of the road that link those junctions. The number of $V$ equals to 785,442, the number of $E$ equals to 995,074. Each edge $(u,v)$ in $E$ has a corresponding feature vector $\phi_{uv}$ indicating the elements of critical infrastructure which are adjacent to this edge, in particular, police station (equal to $0$ or $1$), hospital ($0$ or $1$), fire station ($0$ or $1$), shelter ($0$ or $1$), public facility ($0$ or $1$), grocery store ($0$ or $1$), population (in count), poverty severity (as a numerical index), and elevation (in meters) (components listed in Table \ref{tab:systems}). $V_d$ and $E_d$ are the sets of vertices and edges affected by the flood. Let $G' = (V', E')$ be a copy of the initial network with all the edges from $E_d$ removed. The number of $E'$ then will be 899,628.
Line 352: Explain how depth levels are calculated for the readers' understanding.
-- These thresholds were the background data of the paper below, which were provided to by its authors. The following text was added (lines 3670373): These intervals were used as background data in the paper \cite{hochrainer2010} and were provided by its authors. The thresholds are based on rainfall-runoff coefficient calculations applied to a 28-year (1980-2008) record of monthly rainfall observations in Jakarta. There is a logarithmic relationship between the frequency of events, measured by return period, and the severity, measured by flood depth. The authors extrapolated the results to unknown return periods, and that allowed us to run ten simulations for each of the intervals. Flood variability is provided by lower and upper caps.
The paper, where the background data come from: Hochrainer-Stigler, S.; Kunreuther, H.; Linnerooth-Bayer, J.; Mechler, R.; Michel-Kerjan, E.; Muir-Wood, R.; Ranger, N.; Vaziri, 701, P.; Young, M. The costs and benefits of reducing risk from natural hazards to residential structures in developing countries. 702 University of Pennsylvania: Philadelphia, PA, USA 2010.
Results
Line 415: Consider using commas to separate large numbers throughout the text for readability.
-- Thank you; addressed everwhere.
Discussion
Include a paragraph discussing the usage of remote sensing data and the framework's applicability to other cities, as mentioned in the Abstract.
-- Text added (lines 499-504): In addition, the proposed framework requires categorized satellite data of actual floods; this research is not possible without these data. The mechanisms for activating disaster monitoring, such as Sentinel Asia and the International Charter on Space and Major Disasters, allow space agencies and specialized private companies to make satellite data on disaster damage more and more accessible, which will also lead to a wider applicability of this framework.
Conclusions
Lines 508-513: Work through this paragraph to provide a clearer summary of the main points and the key takeaways of your research. Explain the significance and implications of your findings in a concise manner.
-- Text added (lines 547-556): We found that the districts Kota Jakarta Barat, Kota Bekasi, Bekasi, Karawang and Tangerang were the most vulnerable to losing access to critical services during the 2020 flood. Kota Jakarta Utara, Kota Tangerang, and Kota Depok were also significantly affected by the flood. The most robust districts were Kota Jakarta Pusat and Bogor. An important contribution of this research is the introduction of the notions of \textit{facility-edge} and \textit{multiedge-facility}, where a \textit{facility-edge} is a segment of the road adjacent to the facility and a \textit{multiedge-facility} is a set of all \textit{facility-edges} located in the vicinity of the building. This allows us to consider ways to reach the facility from different sides, and to distinguish between a \textit{complete loss of access} (if all facility-edges are flooded) and an \textit{impeded access to a facility} (if some but not all facility-edges are flooded).
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This is an interesting and well-written manuscript dealing with the
identification of loss of access to critical infrastructure before and after
the occurrence of floods in Greater Jakarta. The methodology basically relies
on graph theory and flood simulation based on the use of satellite imagery.
The authors carry out a nice discussion on concepts associated with
vulnerability and resilience.
The analysis is conducted at the district level and several graph metrics are
used, such as number and share of roads and services of certain type,
population without access to facilities, betweenness centrality of edges,
average and highest betweenness, meant to quantify and explain the
consequences of flooding. Moreover, the relationship between loss of access,
poverty, and elevation is assessed.
The simulations of flood events are randomly generated, using a probability
model estimated from observations and coping with distinct return periods.
Finally, a vulnerability map based on simulated flood occurrence probability
is generated, considering also random drainage system failure and taking into
account the consequent loss of access.
I have some major concerns which are listed below:
1. The authors ought to explain how the upper and lower thresholds in Table 3 were determined.
2. In the 140 simulations, how were the probabilities converted into flood spots? In other words, how was the map of probabilities sliced into flood patches?
3. Figures 1 and 3: How were the 2020 flood patches mapped? Visually or in a semi-supervised way? Was the Sentinel Asia image subjected to any kind of pre-processing? If so, please describe it. Provide the source of this figure in terms of the road network (OSM, year).
4. In the footnote of Table 9, "gp - population share with access to grocery stores" is missing in the table.
- Minor Comments:
1. Page 3, Line 93: The Methods part...
2. Page 3, Line 93: The Results Section...
3. Page 10, Line 308: ... how the importance of the edges is distributed...
Author Response
Thank you very much for your time reviewing our paper. We have addressed all of your comments. Please see below our response and find attached a new pdf of the manuscript.
1. The authors ought to explain how the upper and lower thresholds in Table 3 were determined.
-- The following text was added to make this clear (lines 367-373): These intervals were used as background data in the paper \cite{hochrainer2010} and were provided by its authors. The thresholds are based on rainfall-runoff coefficient calculations applied to a 28-year (1980-2008) record of monthly rainfall observations in Jakarta. There is a logarithmic relationship between the frequency of events, measured by return period, and the severity, measured by flood depth. The authors extrapolated the results to unknown return periods, and that allowed us to run ten simulations for each of the intervals.
2. In the 140 simulations, how were the probabilities converted into flood spots? In other words, how was the map of probabilities sliced into flood patches?
-- We do not use flood patches, instead, we simulate flooding directly on edges with the specified probabilities (lines 378-385 "probability of flood occurrence on edge...").
3. Figures 1 and 3: How were the 2020 flood patches mapped? Visually or in a semi-supervised way? Was the Sentinel Asia image subjected to any kind of pre-processing? If so, please describe it. Provide the source of this figure in terms of the road network (OSM, year).
-- Sentinel Asia provided already categorized shape files of the flood occurrence, so it is their pre-processing. Slightly chaning text in the Data section for more clarity (lines 240-241): "In this research, we use publicly available categorized and vectorized flood occurrence data provided by \cite{sentinelasia}."
4. In the footnote of Table 9, "gp - population share with access to grocery stores" is missing in the table.
-- The table did not show up completely on the page due to its size; we have decreased the font size, now it is all right. Thank you for spotting this.
- Minor Comments:
1. Page 3, Line 93: The Methods part...
-- thank you; addressed.
2. Page 3, Line 93: The Results Section...
-- addressed.
3. Page 10, Line 308: ... how the importance of the edges is distributed...
-- addressed; thank you!
Author Response File: Author Response.pdf
Reviewer 3 Report
it is an honor to be entrusted with this responsibility, and I am eager to contribute to the rigorous and meticulous peer-review process upheld by your esteemed journal.
comments sheet incuded.
Comments for author File: Comments.pdf
Author Response
Thank you very much for your comments. The methodological flowchart will indeed help to improve clarity and facilitate reproducibility of the framework. We have added a new figure, please see Figure 1 in the attached pdf. A number of minor improvements have also been made while working on revisions.
We provide answers to your other comments below.
• What is the need for adding the simple summary section I think the abstract is sufficient
-- It looks like the journal is stimulating usage of this little section. Let us see if the editor wants it or not. For us, it looks just as good with or without this summary.
• For all figures please add coordinated grid, scale bar, and north direction for more clarity.
-- North arrows have been added everywhere; scale bars have been added where possible. Thank you for this suggestion.
• In the introduction section it will be better to mention the general risk characterization base of hazards, exposure, and vulnerability. Also check the following references.
https://doi.org/10.1007/978-3-030-29635-3_13
-- Our institutions do not have access to this article, so we could not review it.
https://doi.org/10.1016/j.regsus.2023.08.004
-- A reference to this article was added in the discussion session, thank you. The text related to this was added (lines 510-516 in a new pdf): "To compare selected indicators in this paper with some used in the flood risk assessment literature on a more global scale, \cite{guoyi2023} uses data on average annual precipitation, distance to river, elevation, slope, NDVI, population density, GDP, and global land cover. While we take into account elevation, population density, and poverty level (which may correspond to what the authors are trying to show with the GDP index), in the future research other indicators can also be considered when creating urban vulnerability maps."
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
The authors have properly addressed all of my comments. There are no further queries from my side.
Reviewer 3 Report
The author replied to all raised comments and did the required modifications.
the paper can be published in the present form.
