Flood-Disease ABM: A Multi-Pathway Agent-Based Model for Simulating Flood Disruption and Post-Flood Health Impacts Across the Disaster Lifecycle

Addai, Obeng A.; Spiteri, Raymond J.

doi:10.5194/egusphere-2026-5463

Preprints

https://doi.org/10.5194/egusphere-2026-5463

Preprints

17 Sep 2026

| 17 Sep 2026

Status: this preprint is open for discussion and under review for Natural Hazards and Earth System Sciences (NHESS).

Flood-Disease ABM: A Multi-Pathway Agent-Based Model for Simulating Flood Disruption and Post-Flood Health Impacts Across the Disaster Lifecycle

Obeng A. Addai and Raymond J. Spiteri

Abstract. Flood health burdens extend beyond inundation through displacement, service disruption, contaminated environments, and delayed disease. These processes are often modeled separately or as cause–effect relationships, limiting analysis of feedbacks among human behavior, institutions, and post-flood health. We present the Flood-Disease Agent-Based Model (Flood-Disease ABM), a spatially explicit, hourly modeling framework that integrates flood exposure, human mobility, worldview-informed decision-making, institutional response, economic activity, quality of life, and infectious, vector-borne, and mold-related respiratory disease pathways. Decision-making is differentiated across hierarchist, egalitarian, individualist, and fatalist worldview orientations. A browser-based interactive app enables users to configure scenarios, populations, hazards, service capacity, and policy assumptions and to observe spatial and temporal outcomes. Complementary batch runners support reproducible replications of the same configurations. The framework is implemented for the flood-prone Uvalde, Texas, region using representations of households, businesses, schools, shelters, healthcare facilities, and government. Under the selected assumptions, the compound scenario produces an early flood-related healthcare peak followed by sustained disease-related demand during recovery; mold emerges after floodwater recession; spatially uneven shelter demand exceeds capacity at peak demand; financial stress shifts among households and institutions; and quality-of-life losses are greatest among children and lower-income groups. Results are model-based outcomes and not calibrated forecasts for Uvalde.

Received: 09 Sep 2026 – Discussion started: 17 Sep 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Obeng A. Addai and Raymond J. Spiteri

Status: open (until 29 Oct 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-5463', Anonymous Referee #1, 08 Oct 2026 reply
The manuscript, “Flood-Disease ABM: A Multi-Pathway Agent-Based Model for Simulating Flood Disruption and Post-Flood Health Impacts Across the Disaster Lifecycle,” presents an extension of the Flood-ABM framework previously developed by the authors. In this manuscript, the authors describe the modeling of infectious, vector-borne, and mold-related respiratory disease pathways following a stylized, return-period-based flood event for a case study area spanning part of Uvalde, Texas. Although the ABM presents a novel and interesting integrated modeling approach that captures interactions between post-flood health impacts, I think publication of the manuscript in its current form would be premature.
My main concern is that many model parameter values and parameter weights appear to be based on assumptions rather than empirical estimates, while a formal sensitivity analysis is lacking. The impact of these assumptions on model behavior and results is therefore not sufficiently explored. This limits the interpretability of the model behavior and results and, more importantly, raises concerns about how well the model maps onto real-world processes. The authors also acknowledge this themselves, stating: “Accordingly, neither the reported magnitudes nor the qualitative patterns should be interpreted as empirically established or robust to parameter uncertainty” [line 703]. Therefore, I recommend major revisions, including the following comments:
Major comments:
The manuscript lacks a sensitivity analysis exploring the effects of model assumptions on model behavior and results.

The manuscript would benefit from more in depth discussion of the validity of the integrated model components (floods, mold, infectious disease, economic activity and person-agent behavior) to describe how the model maps onto real-world processes, and where possible, compares to empirical observations or established patterns.

I encourage the authors to include an ODD+D description of the ABM in the supplementary material (Müller et al., 2013). This further could help structure the description of the ABM and its behavior.

A revision of the text to improve clarity and conciseness (also see comments below).

Minor comments:
Abstract
This main finding communicated in the abstract remains somewhat vague: “under the selected assumptions” and “compound scenario”; we do not know what either of these entails [line 11]. Could the authors clarify what behavior the model demonstrates under which scenario characteristics?

Could the authors elaborate on what “spatially uneven shelter demand” entails [line 13]?

Introduction
The authors mention “riverine and urban flood risk” [line 18]. It would be useful to introduce the authors’ definition of risk in this introduction, e.g., risk conceptualized as a product of vulnerability, exposure, and hazard (Koks et al., 2015).

The introduction describes the adverse health impacts of flood events well. However, I think it would strengthen the manuscript if the authors could elaborate on how these health impacts are relevant to the case study area. This would provide a stronger argument for developing and applying the ABM to the context of Uvalde, Texas.

The introduction misses a positioning of the decision framework of the model in the literature. The authors mention “threat appraisal,” hinting at protection motivation theory (Rogers, 1975). I think embedding the behavioral model in the literature would strengthen the manuscript.

The authors state, “We therefore include a contact-transmission pathway so that flood-driven changes in household, institutional, and shelter co-location can interact with the other health pathways.” [line 38]. This sentence appears to be out of place in this introduction paragraph. It would also be fitting to include references to the scientific literature on contact-transmission pathways in the context of flooding events.

I believe that making the research goals listed under Section 1.2 more concise would strengthen the manuscript. Do the authors aim to quantify the effects of certain interactions?

Methods
Could the authors elaborate on what a Python-defined area-of-interest polygon entails? And what are points-of-interest [line 155]?

The authors could be clearer about which mechanisms were already present in Flood-ABM and what their contribution in this paper entails.

The authors indicate that the model is “scenario-based.” Could the authors justify that the scenarios investigated in this paper are plausible? E.g., are these based on historical disease outbreaks in the case study area, or other regions?

The authors state: “The city lies near the Nueces River and its tributary network, and its history of recurrent flood exposure illustrates why flood impacts should be examined beyond the moment of inundation”, and “Severe storms and flooding affected the region in July 2026; the Texas Division of Emergency Management reported that the governor visited Uvalde County to receive updates on flood conditions, survey damage, and meet with state and local response officials (Texas Division of Emergency Management, 2026). This event further underscores the continuing relevance of evacuation, shelter access, healthcare continuity, household displacement, and post-flood environmental exposure” [lines 130-134]. I think the importance of studying long-term flood impacts in the case study area could be better illustrated with concrete examples of historical flood impacts, such as documented effects on displacement, health, infrastructure, shelter demand, or economic activity.

The authors appear to select a study site based on computational constraints [line 135]. Would the selection of a small study site (thereby setting boundary conditions) not affect the modeling of disease dynamics? I think it would strengthen the manuscript if the authors could reflect on this.

Figure 2 (Google Maps screenshot) should be replaced by a figure created using GIS software, which indicates the model extent, scale bar, and north arrow.

The abbreviation AOI is not introduced [line 161].

The authors state that they included buildings themselves to supplement OSM and match census data. How many buildings were drawn by the authors themselves, and on what data did they base the location of these buildings [line 161-165]? In the result section the authors state that the model is initialized with a synthetic population of 5,000 persons (Figure 8). Does this number correspond to the census data as well?

The authors initialize worldview, social capital, and trust at random [line 190]. Could the authors elaborate on the extent these random factors affect model results, and whether they are they randomized in each model run? Additionally, could the authors clarify whether these worldviews are based on the literature, or constructed specifically for this model?

Could the authors explain why there is both a flood-depth adjustment and flood penalty [line 196]?

The authors state “Once local depth falls below the recovery threshold, injury resolves stochastically according to the person’s recovery rate. Prolonged stranding under sufficiently hazardous conditions also creates mortality risk [line 216]”. Could the authors indicate what the recovery threshold represents?

Could the authors describe what “household habitability” represents [line 221]?

The authors state “The implementation defines person-level quality of life as a weighted composite of current activity, health, safety, housing, and financial condition [line 230]”, could the author clarify whether this definition was developed specifically for this model, or whether this definition was derived from literature?

I appreciate the effort that went into developing a user interface for the ABM. However, I believe that moving the detailed description of the GUI and batch runners to the Supplementary Materials would help keep the main manuscript more focused on the model itself and its scientific contributions.

Results
The authors compare a normal day with an average day under the compound scenario. Could the authors elaborate why the average under event conditions is compared with a single day in the baseline conditions?

Other
If applicable under the journal's AI-use and figure-generation policies, could the authors indicate which figures were created or modified using generative AI tools?

References
Koks, E. E., Jongman, B., Husby, T. G., & Botzen, W. J. W. (2015). Combining hazard, exposure and social vulnerability to provide lessons for flood risk management. Environmental Science & Policy, 47, 42–52.
Müller, B., Bohn, F., Dreßler, G., Groeneveld, J., Klassert, C., Martin, R., Schlüter, M., Schulze, J., Weise, H., & Schwarz, N. (2013). Describing human decisions in agent-based models—ODD+D, an extension of the ODD protocol. Environmental Modelling and Software. https://doi.org/10.1016/j.envsoft.2013.06.003
Rogers, R. W. (1975). A Protection Motivation Theory of Fear Appeals and Attitude Change. The Journal of Psychology. https://doi.org/10.1080/00223980.1975.9915803

Reply
Citation: https://doi.org/10.5194/egusphere-2026-5463-RC1

Obeng A. Addai and Raymond J. Spiteri

Viewed

Total article views: 419 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
254	114	51	419	48	49

HTML: 254
PDF: 114
XML: 51
Total: 419
BibTeX: 48
EndNote: 49

Views and downloads (calculated since 17 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	165	55	20	240
Oct 2026	89	59	31	179

Cumulative views and downloads (calculated since 17 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	165	55	20	240
Oct 2026	89	59	31	179

Viewed (geographical distribution)

Total article views: 419 (including HTML, PDF, and XML) Thereof 419 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 27 Oct 2026

Short summary

Floods can harm communities long after the water recedes by disrupting homes, work, health care, and daily life. We developed an agent-based model of Uvalde, Texas, to explore how flooding, injuries, infectious illness, mold, and insect-borne disease interact over time. The results show that flooding causes immediate displacement, while health and financial pressures can continue during recovery. The model can help test preparedness and support decisions.


Total:	0
HTML:	0
PDF:	0
XML:	0