Automated Tsunami Hazard and Exposure Reporting Using Numerical Simulations and WebGIS Visualization ^†

Abstract

The availability of tsunami hazard and exposure information is crucial to support effective emergency response in coastal areas. This study presents an automated framework that integrates tsunami numerical simulation, geospatial exposure analysis, and WebGIS-based visualization to generate standardized hazard and exposure reports for decision support. Using a parallel implementation of the TUNAMI-N2 model, a 6-h tsunami simulation for the Peruvian coast can be completed in ~45 min, with hazard and exposure reports automatically published on a WebGIS platform within ~3 min. Application to the historical 1746 Lima tsunami demonstrates the system’s capability to quantify hazard and exposure for operational decision-making.

1. Introduction

The $M_w$ 8.8 Kamchatka earthquake (29 July 2025) generated a tsunami [1] that triggered evacuations and warnings from Hawaii and Japan to the U.S. West Coast and South America [2]. Like this event, far-field tsunamis provide valuable lead time for evacuations and warnings; by contrast, near-field tsunamis can arrive in a matter of minutes, leaving minimal time for decision-making. This limitation highlights the need for automated systems capable of generating and disseminating tsunami information to support emergency decision-making.

Peru’s tsunami forecast and warning system, operated by the Centro Nacional de Alerta de Tsunamis (CNAT) [3], issues three alert states based on earthquake parameters, such as location, magnitude, and depth (

Table 1. Alert states established by CNAT’s protocol [6].

Source 1	Depth	Informative 2	Watch 3	Warning 4
Near field	≤60 km	5.0 $\le M_w<$ 7.0	7.0 $\le M_w<$ 7.5	7.5 $\le M_w$
	>60 km	5.0 $\le M_w<$ 7.5 5	7.5 $\le M_w$	-
Far field	≤60 km	7.0 $\le M_w<$ 8.0	8.0 $\le M_w<$ 8.5	8.5 $\le M_w$
	>60 km	7.0 $\le M_w<$ 8.5 6	8.5 $\le M_w$	-

¹ Epicenter located offshore or near the coast (trench to 60 km inland). ² Earthquake detected. ³ Potential tsunami. ⁴ Imminent tsunami. ⁵ Enhanced monitoring is implemented when 7.0 $\le M_w<$ 7.5. ⁶ Enhanced monitoring is implemented when 8.0 $\le M_w<$ 8.5.

). Currently, CNAT’s protocol has declared a uniform alert state for the entire Peruvian coast, without differentiating tsunami impacts from one location to another. In contrast, other countries have implemented region-specific warning systems that assign an alert state based on numerical simulations that predict coastal tsunami heights [4,5]. Consequently, upgrading Peru’s tsunami bulletins [3] to this approach would enable more targeted alerts, reducing unnecessary warnings while prioritizing the most threatened coastal communities.

Another key consideration for improving tsunami bulletins is the inclusion of exposure analysis. Historical events evidenced that tsunamis with the highest magnitudes or run-up did not necessarily result in the highest number of fatalities [7], making exposure of people and critical buildings crucial information [8]. Accordingly, integrating hazard with exposure metrics can lead decision-makers to adopt more targeted and effective emergency responses.

In addition, it is well-known that WebGIS systems are useful tools to better disseminate information in Disaster Risk Management [9]. A recent application in Peru for earthquakes demonstrated its advantages to government decision-makers [10]. Therefore, a WebGIS system for tsunamis is needed to enhance emergency response, as implemented in other countries [11,12,13].

In this context, we present an automated system that generates tsunami hazard and exposure reports using numerical simulations and web-based geospatial visualization, with a focus on Peru. The framework links seismic event screening, tsunami numerical simulation, geospatial exposure analysis, and WebGIS visualization into a single workflow that automatically converts earthquake parameters into spatially explicit hazard and exposure information. The resulting outputs are organized into standardized reports at multiple administrative levels, providing decision-oriented information to support tsunami emergency management.

2. Materials and Methods

This study implements an automated workflow (Figure 1) composed of three components: (1) tsunami hazard assessment (Section 2.1), (2) tsunami exposure assessment (Section 2.2), and (3) web-based geospatial visualization (Section 2.3). First, seismic data are automatically extracted and evaluated to compute tsunami generation, propagation, and inundation using numerical simulations, producing key hazard metrics such as arrival times and inundation heights. Second, exposed population and infrastructure are quantified by combining the simulated inundation results with geospatial census and infrastructure databases through spatial intersection. Finally, all hazard and exposure outputs are automatically processed by a backend system and disseminated through a web-based geospatial platform, enabling interactive visualization and the generation of automated reports for decision support.

2.1. Tsunami Hazard Assessment

This section describes an automated workflow for tsunami hazard assessment, in which numerical simulations are executed only for potentially tsunamigenic earthquakes. Seismic parameters, including origin time, epicentral latitude and longitude, focal depth, and magnitude, are continuously retrieved from the Instituto Geofísico del Perú (IGP) [14], and incoming events are automatically screened according to three operational criteria for tsunami generation potential defined in [15]. Specifically, an earthquake is considered potentially tsunamigenic when (1) the epicenter is located offshore or near the coastline; (2) the focal depth is shallower than 60 km; and (3) the magnitude exceeds $M_w$ 7.0.

For earthquakes that satisfy these criteria, source parameterization is performed using the seismic parameters provided by IGP. Additional source parameters (strike, dip, and rake) are assigned based on the geometry of the Peruvian subduction zone and regional earthquake catalogs. Fault rupture dimensions and average slip are estimated from empirical scaling relationships [16] as a function of earthquake magnitude, defining a simplified rectangular fault model.

Bathymetric and topographic data were obtained from the General Bathymetric Chart of the Oceans (GEBCO) and resampled to 15 arc-seconds (~450 m). The elevation dataset covers the South American margin and adjacent Pacific Ocean. For each event, the simulation domain is defined to include the entire Peruvian coast and, when necessary, extended to encompass the full fault geometry for earthquakes occurring outside Peru. The coseismic seafloor deformation is computed with the Okada model [17] and used as the initial sea-surface displacement for tsunami propagation. Tsunami simulation was conducted using the Tohoku University Numerical Analysis Model for Investigation of Near-field Tsunamis No. 2 (TUNAMI-N2) model with a total simulation time of 6 h, a time step of 1 s, and a constant bottom roughness coefficient of 0.025 m^−1/3s. To improve computational efficiency within the automated workflow, a parallelized implementation of TUNAMI-N2 (gWave) [18] is employed, which decomposes the computational domain into subdomains and executes them concurrently across multiple processing cores.

The computational domain covering the Peruvian margin consists of approximately 4000 × 5200 grid cells (~21 million points). Numerical simulations are executed using the parallel gWave implementation on 48 CPU cores (Intel Xeon Gold 6442Y, 2.60 GHz). Under this configuration, a 6 h tsunami propagation simulation requires approximately 45 min, allowing the automated generation of tsunami hazard outputs.

Finally, the tsunami simulation results are processed to obtain spatial distributions of inundation heights and arrival times. Inundation extents are rasterized and exported for use in the exposure module. Arrival time isochrones are generated in the open ocean and exported as shapefiles for web display. To summarize coastal hazard, the shoreline is sampled at 10 km intervals, providing a consistent spatial representation of the coastline and a sampling density comparable to previous large-scale implementations [4]. For each coastal point, the maximum inundation height and earliest arrival time are extracted and stored as point shapefiles. Based on inundation height thresholds following Japan Meteorological Agency (JMA) guidance [5] (

Table 2. Types of tsunami warning/advisory [5].

Alert State	Maximum Tsunami Height (m)	Action Required
Major Tsunami Warning	3 $<h$	Evacuate immediately to high ground or an evacuation building.
Tsunami Warning	1 $<h\le$ 3	Evacuate immediately to high ground or an evacuation building.
Tsunami Advisory	0.2 $<h\le$ 1	Stay away from the coast and leave the water.

), coastal points are assigned alert states, which are then aggregated to the provincial level by selecting the maximum inundation height and minimum arrival time across all coastal points within each province. The resulting provincial indicators are exported as CSV tables for automated web reports and dashboards. This aggregation follows a conservative maximum risk approach, widely adopted in operational tsunami warning systems [4,12] to account for uncertainties in source and model parameters. While this approach inherently leads to some degree of over-warning, it is considered appropriate given these uncertainties, as it reduces the risk of underestimating potentially hazardous conditions.

2.2. Tsunami Exposure Assessment

In this component, the system automates the quantification of potential tsunami impacts by spatially intersecting simulated inundation results with georeferenced census and infrastructure datasets. Exposure is initially assessed at the census block scale and subsequently aggregated to administrative levels relevant for operational decision-making, including district, province, and regional scales. For each tsunami scenario, the system estimates exposed population groups and critical facilities within inundation zones, generating standardized exposure metrics for automated reports and web-based visualization.

Population exposure is quantified using the most recent official dataset available in Peru, the National Population and Housing Census provided by the National Institute of Statistics and Informatics (INEI) [19]. Census block polygons containing population attributes are intersected with the modeled tsunami inundation extent to estimate the number of exposed inhabitants. If a census block intersects the inundation area, its entire population is considered exposed as a conservative assumption consistent with the uncertainty in inundation estimates. To better reflect differential evacuation capacity, the population is disaggregated into three age groups, namely <15 years, 15–65 years, and >65 years, in accordance with the age groups used by INEI. This classification highlights demographic groups that may face greater difficulties during evacuation and emergency response due to limited mobility or health constraints, consistent with commonly used age-based vulnerability groupings in disaster risk assessments [8,20].

In addition to population exposure, the system evaluates the potential impact on selected critical facilities that support evacuation, emergency response, and recovery. Georeferenced datasets from official Peruvian sources are used to identify the location and type of key infrastructure within tsunami inundation zones, including:

• Educational institutions, such as schools and universities, which can serve as community assets and potential vertical shelters [21,22];
• Health facilities, including hospitals and medical centers, which are vital for emergency medical care and disaster response operations;
• Security and emergency services, such as police stations and fire stations, which play a central role in coordinating evacuation, rescue, and response activities.

Overall, the exposure assessment follows the block-scale methodological framework developed for Metropolitan Lima–Callao [23], integrating numerical tsunami inundation simulation with GIS-based socioeconomic and infrastructure layers to estimate the number of exposed people and critical facilities within tsunami inundation zones.

2.3. Web-Based Geospatial Visualization

This component is built upon the seismic monitoring framework implemented by REDACIS–CISMID [24], which continuously monitors seismic information disseminated by IGP. Every few seconds, the platform automatically verifies whether new seismic event updates have been published through the institution’s Twitter REST API. When a new event is detected and satisfies the condition of a tsunamigenic earthquake [15], the system automatically proceeds to execute the subsequent processing stages.

Each processing stage produces hazard and exposure datasets that are automatically inserted into a PostgreSQL–PostGIS spatial database and published through a GeoServer instance, ensuring interoperability with external GIS platforms and services. Once the numerical simulation is completed, the resulting datasets are processed and made available through the WebGIS platform for visualization of the most recently analyzed seismic event. The complete generation of hazard and exposure reports, including database insertion and web visualization, requires approximately 3 min, allowing the automated dissemination of results through the WebGIS platform.

The overall system architecture consists of two main components. The backend is developed using the Python (version 3.12) programming language and the Flask framework and is responsible for managing the previously described processes, orchestrating parallel computations, and inserting newly generated information into the spatial database. The second component is the frontend, which provides interactive visualization and user interaction capabilities, allowing users to explore tsunami hazard layers, exposure results, and alert states through a web-based geospatial interface.

3. Results and Discussion

To illustrate the operation of the automated system, we present an example based on the historical Lima 1746 earthquake ($M_w$ 9.0). Seismic source parameters were taken from [25] and used solely to exemplify the system’s functionality; under operational conditions, the workflow is automatically driven by real-time seismic information from IGP. The resulting hazard and exposure report is shown in Figure 2.

Figure 2 is a screenshot of the automated report produced for the Lima 1746 scenario and exemplifies the three main components delivered to users. On the left, an interactive map allows users to pan and zoom while displaying hazard outputs (Section 2.1), including isochrones of tsunami arrival times in the open ocean and coastal points colored according to their corresponding alert state. The upper-right panel provides a consolidated statistics table summarizing exposure results (Section 2.2). By default, these statistics show results for all of Peru, but the interface supports hierarchical spatial filtering: clicking on a coastal region updates the table with a summary for that region; zooming in reveals provincial boundaries and allows selection at the provincial level; further zooming in displays district boundaries and enables district-level queries. At the district scale, the inundation footprint, cadastral data, and infrastructure are displayed, enabling direct inspection of exposed elements. Finally, the lower-right panel presents a decision-oriented table listing provinces, ordered from north to south, together with their corresponding tsunami arrival time, inundation height, and alert level. This table provides a concise overview that supports prioritization and coordination of emergency response actions across administrative units. In summary, Figure 2 illustrates how this interface consolidates the most relevant information for situational awareness and response planning.

For the Callao province, the simulation yields an approximate tsunami arrival time of 23 min and an inundation height of 16 m. These results are consistent with previous studies of the 1746 event [25], which report arrival times on the order of 23 min and wave heights exceeding 10 m in the Callao area, based on models constrained by historical observations. This agreement indicates that the simulated results are consistent with previously reported tsunami characteristics of that event.

The Lima 1746 demonstration highlights three key strengths of the system: (1) The workflow (Figure 1) automatically converts earthquake source parameters into spatial tsunami hazard and exposure metrics; (2) results are organized consistently across administrative levels, from region to province and district, enabling analysis at multiple decision scales; (3) multiple geospatial outputs, such as arrival-time isochrones, inundation footprint, coastal points, and summary tables, are integrated into a single framework that supports automated reporting and interactive decision support. While the present implementation focuses on hazard and exposure, the modular framework can be extended to estimate building and infrastructure damage as well as fatalities and casualties.

Some limitations should be acknowledged. Tsunami hazard is highly sensitive to fault geometry and slip heterogeneity, which can significantly affect wave heights, inundation extent, and their spatial distribution [26,27,28], introducing epistemic uncertainty in the results. Therefore, the use of a simplified rectangular fault with uniform slip based on empirical scaling relationships may lead to underestimation or misrepresentation of localized impacts. However, these assumptions are adopted to enable a rapid and automated assessment.

Likewise, the use of coarse-resolution elevation data (15 arc-seconds) and a uniform bottom-roughness parameter represents a strong simplification for nearshore inundation modeling [29]. At this resolution, coastal topography and land-surface variability (e.g., urban areas and wetlands) are not adequately represented, which can lead to significant misrepresentation of run-up and inundation extent, particularly in densely populated coastal areas, and consequently introduce substantial uncertainty in the estimated population and infrastructure exposure. However, this resolution is adopted to maintain feasible simulation times within the automated workflow; as a result, a detailed spatial representation of bottom roughness is not supported at this scale, and a constant value is used.

Accordingly, the results should be interpreted as first-order estimates for situational awareness rather than as detailed local hazard quantification. Future work will focus on reducing these sources of uncertainty and improving confidence in local-scale estimates. In addition, the integration of other types of critical infrastructure (e.g., ports or water systems) would further enhance the comprehensiveness of the exposure assessment. Nonetheless, the Lima 1746 example demonstrates the practical value of an automated, integrated workflow for producing hazard and exposure information to support more targeted tsunami alerting and emergency response.

4. Conclusions

This paper presented an automated system for generating tsunami hazard and exposure reports that integrates seismic event screening, numerical tsunami simulation, geospatial exposure analysis, and WebGIS-based visualization. By automating the full workflow from seismic data acquisition to the generation of hazard and exposure reports, the proposed framework addresses key limitations of conventional tsunami warning approaches that rely on generalized alerts and lack spatially explicit impact information.

Application of the system to the historical 1746 Lima earthquake scenario demonstrates its capability to produce coherent, multi-scale information for decision support, including tsunami arrival times, inundation heights, alert levels, and population and infrastructure exposure at the national, regional, provincial, and district levels. Under the current computational configuration, a 6-hour tsunami simulation can be completed in approximately 45 min, with hazard and exposure reports generated and published within 3 min.

Overall, the proposed framework provides a practical foundation for enhancing tsunami impact assessment and risk communication in Peru and similar subduction-zone settings. Future developments will focus on integrating real-time data, improving coastal data resolution, and expanding scenario analyses to strengthen tsunami risk management.