Refine Statement of Need for UNSAFE framework

Revised the statement of need for the UNSAFE framework, enhancing clarity and precision in describing flood-risk estimation tools and their limitations. Updated references and improved the overall structure of the section.

Author: atrisovic

joss_submission/paper.md

@@ -35,13 +35,13 @@ bibliography: paper.bib
 
 ---
 # Statement of Need
-Flooding is a frequent, widespread, and damaging natural hazard in the United States [@Kousky2020]. Researchers and practitioners increasingly rely on flood-risk estimates to analyze dynamics and inform decisions [@Merz2010; @Trigg2016; @Bates2023; @Mulder2023]. There is increasing demand for flood-risk estimates at the scale of individual assets [@Condon2023]. One driver of this demand is that flood-risk estimates at coarser scales are susceptible to aggregation biases [@Pollack2022; @Condon2023]. Sound flood-risk estimates hinge on careful representation of uncertainties surrounding key inputs driving hazards, exposures, and vulnerabilities at relevant scales [@Bates2023; @Sieg2023; @Saint-Geours2015; @Tate2015; @Rozer2019; @Hosseini-Shakib2024].
+Flooding is among the most frequent and damaging natural hazards in the United States [@Kousky2020]. Researchers and practitioners increasingly use flood-risk estimates to analyze dynamics and inform decisions [@Merz2010; @Trigg2016; @Bates2023; @Mulder2023]. There is increasing demand for flood-risk estimates at the scale of individual assets [@Condon2023]. One driver of this demand is that flood-risk estimates at coarser scales are susceptible to aggregation biases [@Pollack2022; @Condon2023]. Robust flood-risk estimates require explicit representation of uncertainties surrounding key inputs driving hazards, exposures, and vulnerabilities at relevant scales [@Bates2023; @Sieg2023; @Saint-Geours2015; @Tate2015; @Rozer2019; @Hosseini-Shakib2024].
 
-Many property-level flood-risk estimation frameworks are silent on the effects of uncertainties  surrounding key inputs [@Schneider2006; @Merz2010; @Saint-Geours2015; @Tate2015; @Sieg2023]. The predominant approach  to estimating economic flood damages in a U.S. setting relies on either the U.S. Army Corps of Engineers (USACE) or Federal Emergency Management Agency (FEMA) depth-damage functions (DDFs) [@Scawthorn2006; @Merz2010; @Tate2015]. In this DDF approach, flood-risk estimates depend on assumptions about several exposure and vulnerability characteristics. Risk estimates are sensitive to the spatial precision of linking a structure to a certain flood depth, the first-floor elevation of a structure, the structure's foundation type, the number of stories of the structure, the main function of the structure (i.e. residential or commercial), and the value of the structure [@Merz2010; @Tate2015; @Wing2020; @Pollack2022; @Xia2024]. Risk estimates are also sensitive to the expected damage for a given depth and the shape of the depth-damage relationship [@Merz2010; @Tate2015; @Rozer2019; @Wing2020]. All of these characteristics are subject to uncertainties which propagate to the resulting risk estimates [@Tate2015; @Saint-Geours2015; @Pollack2022; @Hosseini-Shakib2024].
+Many property-level flood-risk estimation frameworks are silent on the effects of uncertainties surrounding key inputs [@Schneider2006; @Merz2010; @Saint-Geours2015; @Tate2015; @Sieg2023]. The predominant approach  to estimating economic flood damages in a U.S. setting relies on either the U.S. Army Corps of Engineers (USACE) or Federal Emergency Management Agency (FEMA) depth-damage functions (DDFs) [@Scawthorn2006; @Merz2010; @Tate2015]. In this DDF approach, flood-risk estimates depend on assumptions about several exposure and vulnerability characteristics. Risk estimates are sensitive to the spatial precision of linking a structure to a certain flood depth, the first-floor elevation of a structure, the structure's foundation type, the number of stories of the structure, the main function of the structure (i.e. residential or commercial), and the value of the structure [@Merz2010; @Tate2015; @Wing2020; @Pollack2022; @Xia2024]. Risk estimates are also sensitive to the expected damage for a given depth and the shape of the depth-damage relationship [@Merz2010; @Tate2015; @Rozer2019; @Wing2020]. All of these characteristics are subject to uncertainties which propagate to the resulting risk estimates [@Tate2015; @Saint-Geours2015; @Pollack2022; @Hosseini-Shakib2024].
 
-Here we implemented the Uncertain Structure and Fragility Ensemble (UNSAFE) framework to provide the U.S. flood-risk assessment community with a free and open-source tool for estimating property-level flood risks under uncertainty. UNSAFE represents exposure and vulnerability inputs under uncertainty using entirely free data. This improves on the closest tool we were able to find,[“go-consequences”](https://github.com/USACE/go-consequences) [@USACE-go2024] from the USACE. We could not find documentation, example usage, or an official release for this tool. From our inspection of the GitHub repository, it appears that analysts can use go-consequences to produce stochastic representations of a subset of exposure and vulnerability characteristics in the DDF paradigm. However, this functionality is not available for key drivers like structure value or the functional form of the DDF for a given structure.  
+Here we implemented the Uncertain Structure and Fragility Ensemble (UNSAFE) framework to provide the U.S. flood-risk assessment community with a free and open-source tool for estimating property-level flood risks under uncertainty. UNSAFE represents exposure and vulnerability inputs under uncertainty using entirely free data. This extends the functionality of the most comparable publicly available tool, [“go-consequences”](https://github.com/USACE/go-consequences) [@USACE-go2024], developed by the USACE, for which limited documentation and usage examples are available. Examination of the repository suggests that go-consequences can produce stochastic representations of selected exposure and vulnerability characteristics in the DDF paradigm. However, this functionality is not available for key drivers like structure value or the functional form of the DDF for a given structure.  
 
-A few other tools are worth mentioning to contextualize the gap `UNSAFE` fills. The most prominent is [Hazus](https://www.fema.gov/flood-maps/products-tools/hazus) [@Scawthorn2006; @Schneider2006; @Tate2015]. Hazus is freely available as a GIS-based desktop application running Windows but cannot be easily modified to accommodate uncertainty in exposure and vulnerability. FEMA also developed the [“Flood Assessment Structure Tool”](https://github.com/nhrap-hazus/FAST?tab=readme-ov-file) [@FEMA2021] to facilitate more efficient deterministic Hazus analyses in Python. This tool appears deprecated. The USACE maintains two published tools for deterministic analyses,([HEC-FIA](https://www.hec.usace.army.mil/confluence/fiadocs/fiaum/latest) [@USACE2021] and [HEC-FDA](https://www.hec.usace.army.mil/software/hec-fda/documentation/CPD-72_V1.4.1.pdf) [@USACE2024]). Lastly, there is the [Delft-FIAT (Fast Impact Assessment Tool)](https://deltares.github.io/Delft-FIAT/stable/) [@Deltares2024], developed and maintained by Deltares. It currently does not accommodate flood risk estimation with uncertainty in exposure and vulnerability inputs, but is open-source and well-documented. Sophisticated users could likely modify the tool to account for uncertainty in inputs; `UNSAFE` streamlines this workflow.
+Several existing tools help contextualize the methodological gap that UNSAFE addresses. The most prominent is [Hazus](https://www.fema.gov/flood-maps/products-tools/hazus) [@Scawthorn2006; @Schneider2006; @Tate2015], a freely available GIS-based desktop application for Windows that supports deterministic flood-damage assessments but cannot be readily modified to incorporate uncertainty in exposure and vulnerability. FEMA also developed the [“Flood Assessment Structure Tool”](https://github.com/nhrap-hazus/FAST?tab=readme-ov-file) [@FEMA2021] to facilitate more efficient deterministic Hazus analyses in Python, although this tool appears to be deprecated. The USACE maintains two published tools for deterministic analyses,([HEC-FIA](https://www.hec.usace.army.mil/confluence/fiadocs/fiaum/latest) [@USACE2021] and [HEC-FDA](https://www.hec.usace.army.mil/software/hec-fda/documentation/CPD-72_V1.4.1.pdf) [@USACE2024]). Lastly, there is the [Delft-FIAT (Fast Impact Assessment Tool)](https://deltares.github.io/Delft-FIAT/stable/) [@Deltares2024], developed and maintained by Deltares. It currently does not accommodate flood risk estimation with uncertainty in exposure and vulnerability inputs, but is open-source and well-documented. Advanced users could, in principle, extend its functionality to represent such uncertainty. `UNSAFE` provides a direct and streamlined framework for doing so.
 
 # Summary
 `UNSAFE` adopts and expands on a property-level risk assessment framework common in academic research and practice (e.g., [Federal Emergency Management Agency (FEMA) loss avoidance studies](https://www.fema.gov/grants/mitigation/loss-avoidance-studies) [@FEMA2024], [United States Army Corps of Engineers (USACE) feasibility studies](https://www.nad.usace.army.mil/Portals/40/docs/NACCS/10A_PhysicalDepthDmgFxSummary_26Jan2015.pdf) [@USACE2015]). UNSAFE allows users to add parametric uncertainty to the widely used [National Structure Inventory dataset](https://www.hec.usace.army.mil/confluence/nsi/technicalreferences/2019/technical-documentation) [@USACE-nsi2024] (i.e. uncertainty in exposure), and facilitates the use of multiple, potentially conflicting, expert-based DDFs (i.e. deep uncertainty in vulnerability). A tutorial that demonstrates the functionality of `UNSAFE` is available in [this Jupyter notebook](https://github.com/abpoll/unsafe/blob/main/examples/phil_frd_partial/notebooks/partial_data_example.ipynb). The corresponding [GitHub repository](https://github.com/abpoll/unsafe) includes detailed technical documentation on the data sources, distributional assumptions of key parameters, and peer-reviewed methods used in `UNSAFE`.
@@ -56,4 +56,4 @@ This research was supported by the U.S. Department of Energy, Office of Science,
 
 Software License: `UNSAFE` is distributed under the BSD-2-Clause license. The authors do not assume responsibility for any (mis)use of the provided code.
 
-# References
+# References