the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 15 Feb 2024
Converging Human Intelligence with AI Systems to Advance Flood Evacuation Decision Making
Abstract. The powers that artificial intelligence (AI) has developed are impressive, with recent success in leveraging human expertise at various stages of model development. AI can attain its full potential only if, as part of its intelligence, it also actively teams with humans to co-create solutions. Combining AI simulation with human intelligence through data convergence can improve decision-making processes and provide a capacity akin to a "teaming intelligence." This research, for the first time, introduces the concepts of Human-AI Convergence (HAC) capabilities for flood evacuation decision-making. The objective of this study was to develop a unique, computationally effective surrogate HAC system for flood evacuation decision-making that integrates the distinctive features of AI with transportation geospatial data, a river hydraulic model, and human data from X (previously Twitter) to visualize flood inundation areas and suggest re-routing. The HAC system is smartly designed to forecast flood stage levels using AI across the US Geological Survey gauging stations and combine the results with Manning's equation results and transportation data, integrated into a web-based Google Earth visualization architecture. The technology has been tested in the Lowcountry of South Carolina, where previous flooding disasters caused considerable damage to the transportation networks and increased traffic on evacuation routes. This state-of-the-art HAC system— a flood evacuation product— stands to advance the frontier of human-AI collaborative research in the context of real-time flood emergency management and response.
- Preprint
(1502 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on nhess-2024-25', Anonymous Referee #1, 02 Jun 2024
Dear authors,
I am not an expert for AI applications in the field of flood risk mitigation;
that's why I will not comment on this; I assume that you developed/applied a scientifically valid approach.
However, even as non-expert in this specific sub-field of natural hazard analysis, I may say that that figures are of poor quality,
maps should not be print screens .. but maps with scale etc.
So, practically all figures have to be revised.
And, this is important - because, even if AI helps you better identify situations and places where, e.g., evacuation is necessary during a flood
event, this information has also to be communicated in an efficient way - and such kind of simplistic representations of your results
would not be very helpful I think.
yours
reviewer H
Citation: https://doi.org/10.5194/nhess-2024-25-RC1 - AC2: 'Reply on RC1', Vidya Samadi, 22 Sep 2024
-
RC2: 'Comment on nhess-2024-25', Anonymous Referee #2, 06 Aug 2024
The research has clearly stated its originality on line 16-17. Line 24: The study has also been tested in the site study which validate the framework. The study aimed to providing more rapid and accurate evacuation modeling using HAC for flood evacuation decisions.
- Line 28-29: Please mention knowledge gaps or limitation in previous evacuation planning (line 28-29). Given these gaps, please clarify ‘real-time flood emergency’ in term of time extent (when is the threshold of time-lag called real-time in flood event, or how previous/current emergency response)
- Line 106-107: Figure 1, there is no indication of three USGS gauging stations. Legend is better to rename following the gauging station name e.g., Turkey Creek (USGS02172035), South Fork Edisto River (USGS02173000), and North Fork Edisto River (USGS02173500).
- Line 124-125: Please clarify line 124-125 for better readability, on the sentence ‘We trained 30 models for each station (overall 180 models [2 models, 3 stations and 30 models each]). It is not clear when you mentioned ‘2 model’ and ‘30 models.’
- Line 130-132: Also please adjust the line 130-132.
- Line 160: Equation should be well-presented (Figure 2) and consistent.
- Line 160: typo Gage or Gauge in Figure 2.
- Line 200: The Rational method (Equation 6) should be Equation 5 instead of 6
- Lines 240-243: The introduction of the HAND model and its purpose is clear. However, the repetition of "HAND model" could be reduced for conciseness.
- Lines 249-253: The process of generating a HAND model using a DEM and flow accumulation map is described accurately. However, the description could benefit from a more concise explanation of each step.
- Lines 257-262: The normalization of the terrain and creation of the nearest drainage chart is explained well but could be split into shorter sentences for better readability.
- Lines 271-279: The step-by-step example of HAND calculation is detailed, but some steps could be simplified for clarity.
- Lines 285-293: The description of the initial data acquisition and integration process is comprehensive. However, the explanation of the amalgamation process and subsequent steps could be more succinct.
- Line 282: It is strongly suggested to change the Figure 3 to more understandable. Indicate number legend.
- Lines 361-364: The introduction to the section is brief but could benefit from a clearer outline of the key findings and their significance.
- Lines 366-377: The description of the model training process is detailed but somewhat cluttered. The inclusion of specific details about the Optuna algorithm and hyperparameters tuning is informative but could be streamlined.
- Lines 379-387: The comparison between LSTM and GRU models is useful, but the reasoning for focusing solely on the LSTM model should be more explicitly justified. Mentioning specific performance metrics in Table 1 helps, but the discussion could be deeper.
- Lines 391-399: The visualization results (Figures 6, 7, and 8) indicate the LSTM model's strengths and weaknesses. The review of its performance is insightful, but the discussion on low gauge height values is somewhat repetitive and could be more concise. Better to provide argument/discussion why Figure 7 is different compare other stations.
- Lines 413-418: The discussion on vanishing or exploding gradient problems is relevant, but it lacks specific examples or evidence from the study. The comparison of performance between gauging stations is useful, but the conclusions drawn should be supported with more detailed analysis.
Another general point:
- The significance/effectiveness a proposed HAC approach is better to compare prior to after implementation of HAC in the case study.
- All figures should be provided in a proper way as mentioned by previous reviewer’s comments.
- Section 2.4 the clarity of the text could be improved by shortening some sentences and avoiding redundancy.
- It is better to provide information on how to address the information on X is a fake or miss information regarding the flood occurrence (Section 3.2).
- In addition to revise the Figure 10, in the section 3.3 should discuss the effectiveness of the routes, or add the cost induced by the floods.
- The conclusion effectively summarizes the study's key findings and highlights the potential and importance of the HAC system. It could benefit from more concise language and a clearer focus on the main points.
Citation: https://doi.org/10.5194/nhess-2024-25-RC2 -
AC1: 'Reply on RC2', Vidya Samadi, 22 Sep 2024
The comment was uploaded in the form of a supplement: https://nhess.copernicus.org/preprints/nhess-2024-25/nhess-2024-25-AC1-supplement.pdf
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 451 | 148 | 43 | 642 | 36 | 30 |
- HTML: 451
- PDF: 148
- XML: 43
- Total: 642
- BibTeX: 36
- EndNote: 30
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1