Advanced IoT Monitoring for Cultural Heritage: Validation of a Threat-Oriented Monitoring Framework Across Five European Heritage Sites

1. Introduction

Cultural heritage assets constitute an irreplaceable record of human history, identity, and collective memory [1]. However, their long-term preservation is increasingly challenged by a combination of environmental, structural, geological, and anthropogenic pressures [2,3]. Climate change has intensified the frequency and severity of environmental stressors, including extreme temperatures, prolonged droughts, intense rainfall events, flooding, and accelerated weathering processes [1,4,5]. At the same time, many historic structures are affected by ageing materials, ground instability, moisture ingress, pollution, biological colonization, and growing tourism pressures [3,4]. These factors often act simultaneously and cumulatively, accelerating deterioration processes and increasing the vulnerability of heritage assets [1,2,5].

One of the major challenges in heritage conservation is that deterioration is rarely sudden. Instead, damage often develops progressively through a series of interconnected mechanisms that remain invisible until advanced stages are reached [6,7]. Moisture accumulation within masonry may precede salt crystallization and material loss by several years; seasonal thermal cycles may gradually contribute to crack propagation; and environmental fluctuations may lead to structural deformations long before visible damage appears [8,9,10].

As a result, conservation interventions are often undertaken only once signs of deterioration have already appeared, leading to increased expenses, more extensive loss of original materials, and less effective preservation outcomes [6]. Therefore, preventive conservation has become a cornerstone of cultural heritage management, focusing on the early identification and reduction of potential risks to avoid irreversible damage before it occurs [11,12].

The growing availability of digital technologies has created new opportunities for implementing preventive conservation approaches [13,14]. In particular, Wireless Sensor Networks (WSNs), Internet of Things (IoT) technologies, Structural Health Monitoring (SHM) systems, and environmental monitoring platforms have enabled the continuous observation of heritage assets at temporal and spatial scales that were previously unattainable [15,16]. Such technologies allow the acquisition of long-term datasets describing environmental conditions, structural responses, and operational parameters, supporting evidence-based conservation decisions through predictive maintenance and digital-twin integration [16,17,18]. Recent developments in cloud computing, edge computing, and low-power communication protocols have further facilitated the deployment of monitoring systems in remote and complex heritage environments while minimizing intervention in protected structures [15,16,18,19].

Within the cultural heritage domain, IoT technologies offer several advantages. They enable continuous and non-invasive monitoring of assets [12,20,21,22], support long-term observations of slowly evolving deterioration processes [20,23,24], improve the understanding of relationships between environmental forcing and structural response [20,24,25], and provide the basis for early-warning systems capable of identifying abnormal behaviors before visible damage develops [26,27]. Furthermore, the integration of IoT monitoring with digital twins and data-driven decision-support systems is progressively shifting conservation practice from reactive maintenance models toward predictive and risk-informed management strategies [12,17,22,28].

Numerous studies have demonstrated the value of monitoring individual heritage monuments using wireless sensing networks [12,14,20]. Previous research has successfully applied IoT technologies to monitor historical buildings, archaeological sites, museums, monuments, and cultural landscapes [12,29,30]. However, most existing studies are characterized by a strong site-specific focus, addressing individual hazards or evaluating the performance of a limited number of sensors [14,23,31]. As a result, the knowledge generated by these deployments often remains difficult to transfer to other heritage contexts [23,31,32].

A significant gap therefore remains between individual monitoring applications and the development of generalized methodologies that can guide monitoring strategies across different heritage typologies [31,33]. Existing approaches frequently start from the nature of the monument itself, i.e. its architectural or construction typology, rather than from the deterioration mechanisms affecting those assets [33]. Consequently, there is still limited understanding of how monitoring systems can be designed according to identified risks, how sensor selection can be systematically linked to deterioration processes, and how monitoring results can be generalized across heterogeneous cultural heritage contexts [34]. Moreover, few studies have validated monitoring methodologies simultaneously across multiple sites affected by different environmental, structural, and operational threats [3,34].

To address these limitations, this paper presents and validates the ARGUS framework, a threat-oriented IoT monitoring methodology for preventive conservation of cultural heritage [34]. Rather than focusing on the characteristics or typology of a specific heritage asset, the methodology identifies dominant deterioration threats, establishes their associated monitoring requirements, and maps them to appropriate sensing technologies. This approach enables the deployment of scalable and transferable monitoring strategies capable of adapting to diverse heritage environments.

The framework was validated across five European cultural heritage sites representing highly different geographical, environmental, and conservation contexts: the archaeological site of Delos (Greece), the underground wine cellars of Baltanás (Spain), the archaeological and architectural heritage of Monti Lucretili (Italy), the Abbey of Sant’Antonio di Ranverso (Italy), and Schenkenberg Castle (Switzerland). These sites were selected because they encompass a broad range of deterioration mechanisms. By monitoring diverse threats within a common methodological framework, the study provides a unique opportunity to evaluate the transferability of risk-based monitoring strategies across heterogeneous heritage environments.

The objectives of this work are therefore fourfold. First, to develop a risk-based methodology linking deterioration mechanisms with monitoring technologies through a threat-oriented sensor selection process. Second, to validate this methodology across five culturally and geographically diverse heritage sites. Third, to identify common deterioration patterns, transferable monitoring strategies, and effective sensor combinations capable of supporting preventive conservation. Finally, to derive practical recommendations for future heritage monitoring programs, including guidelines for sensor deployment, risk prioritization, stakeholder involvement, and conservation decision-making.

2. Materials and Methods

2.1. ARGUS Threat-Oriented Monitoring Framework

The ARGUS monitoring framework supports preventive conservation through a threat-oriented approach that prioritizes deterioration mechanisms over monument typology, as these mechanisms are often independent of asset type. Thus, heritage assets of different types may require similar monitoring strategies when exposed to comparable threats, while assets within the same category may demand substantially different approaches. Based on this principle, a seven-stage workflow was established consisting of: (i) heritage asset characterization; (ii) threat identification; (iii) risk assessment; (iv) threat-to-sensor mapping; (v) sensor deployment; (vi) data acquisition and interpretation; and (vii) preventive conservation decision-making. Figure 1 illustrates the Transferable ARGUS workflow.

The framework begins with the characterization of the heritage asset and its physical, environmental, and operational context. Based on this assessment, potential deterioration mechanisms are identified and prioritized according to their expected impact and likelihood of occurrence. Appropriate monitoring parameters are subsequently selected and linked to specific sensor technologies. After deploying the sensor network and monitoring the heritage site, the collected data is interpreted in relation to the identified risks to support the definition of preventive conservation measures and early-warning strategies. The resulting workflow provides a transferable methodology that can be adapted to different cultural heritage contexts while maintaining a common conceptual structure. The validation presented in this work focuses on the first five stages, which are described in greater detail in the following subsections.

2.1.1. Assets Characterization: The ARGUS Pilot Sites

Prior to sensor deployment, a structured characterization was carried out for each pilot site, covering its heritage typology, construction materials and age, physical and environmental setting, and available conservation history. This information was compiled through site visits, consultation with heritage managers and conservators, and review of existing conservation reports and, where available, prior structural documentation. The characterization established the baseline context required to interpret the subsequent threat assessment, sensor selection and monitoring results.

The five ARGUS pilot sites were deliberately selected to cover a broad range of cultural heritage contexts and deterioration conditions. They include an extensive archaeological landscape, a subterranean heritage complex, a rural cultural landscape combining natural and built heritage, a medieval monastic complex, and a fortified hilltop ruin. Their geographical distribution across Greece, Spain, Italy and Switzerland also exposes the monitoring framework to markedly different climatic and environmental conditions.

Delos, Greece, represents an extensive archaeological heritage site located on an island in the Aegean Sea. The archaeological complex includes remains of the ancient settlement together with mosaics, frescoes and architectural elements exposed to a combination of marine and climatic influences. The site provides a particularly demanding monitoring context because deterioration is influenced by environmental exposure, including wind, solar radiation, rainfall, salinity and marine pollution, while the archaeological remains are distributed across a large and predominantly open landscape.

Baltanás, Spain, represents a substantially different heritage context: an extensive underground cellar town composed of hundreds of cellars excavated into clay soil. Abandonment and inadequate interventions have contributed to structural deterioration, including partial collapses and landslides. Its underground configuration creates a distinctive microclimatic environment, characterized by very high relative humidity and limited ventilation, while ground movements and water-related processes constitute important structural risks.

Monti Lucretili, Italy, represents a cultural landscape in which natural and built heritage are closely interconnected. The pilot encompasses several areas, including the Villa of Horace, the castle of Montefalco and associated dry-stone wall terraces. The sites are located within a mountainous environment characterized by complex geomorphological conditions, vegetation, erosion processes and difficult accessibility. The pilot is therefore particularly relevant for testing monitoring approaches that operate across different spatial scales and combine environmental, geomorphological and structural information.

Sant'Antonio di Ranverso, Italy, represents a medieval monastic heritage complex in which the conservation of architectural and artistic elements is strongly influenced by environmental conditions. The monitoring context includes the interaction between moisture, temperature, biodeterioration and visitor-related features. These elements affect the condition of the built fabric and decorated surfaces.

Schenkenberg Castle, Switzerland, represents medieval fortified heritage situated in a hilltop environment and exposed to strong seasonal climatic variations. The castle provides a case in which environmental fluctuations, particularly temperature cycling and freeze–thaw processes, interact with structural and material vulnerability.

Taken together, the five pilot sites provide a deliberately heterogeneous test set for evaluating the ARGUS threat-oriented monitoring framework. They differ in heritage typology, scale, materials, accessibility, environmental exposure and dominant deterioration processes. At the same time, they present several recurring conservation challenges, notably moisture-related deterioration, climatic exposure, geomorphological and structural instability, biological processes and anthropogenic pressure. This combination of diversity and recurring deterioration mechanisms provides the basis for evaluating whether a common risk-oriented methodology can be adapted to different heritage contexts and whether comparable monitoring requirements can be identified across sites.

2.1.2. Threat Identification

A preliminary risk assessment was conducted for each pilot site to identify the principal factors contributing to material degradation and structural deterioration. Threat identification was carried out through collaboration between heritage managers, conservators, structural engineers, archaeological experts, and monitoring specialists. Historical records, previous conservation reports, visual inspections, and local environmental conditions were used to define the deterioration mechanisms relevant to each site.

To ensure consistency across all pilot sites, identified threats were grouped into five categories: environmental and climatic, geological and geomorphological, structural, biological, and anthropogenic threats. Environmental threats comprise factors such as moisture, temperature fluctuations, precipitation, solar radiation, wind, air pollution, and other climatic processes that may accelerate material decay. Geological and geomorphological threats include ground instability phenomena, such as erosion, subsidence, landslides, and coastal degradation. Structural threats refer to mechanisms affecting the stability and mechanical behavior of heritage assets, including cracking, displacement, tilting, dynamic movements and soil moisture. Biological threats encompass the activity of organisms capable of degrading heritage materials, including molds, vegetation, insects, and wood-boring organisms. Anthropogenic threats represent human-related factors that may adversely influence conservation conditions, including visitor pressure, vandalism, abandonment, inadequate maintenance, and traffic-induced impacts.

Figure 2 summarizes the threat classification adopted within the ARGUS framework and illustrates how the identified threat categories support risk assessment, early warning, preventive intervention, and long-term conservation planning.

2.1.3. Risk Assessment

Following threat identification, each identified threat was prioritized using the ABC risk assessment methodology previously developed by ICCROM and the Canadian Conservation Institute (CCI) [35], adapted for the ARGUS pilots through the project preparation framework described in Deliverable D6.1 [36] and applied within the ARGUS project [33,37]. The ABC method provides a structured approach for evaluating risks to cultural heritage by considering three components: A, the frequency or time of accumulation of the risk factor; B, the loss of value of the portion of cultural heritage potentially affected; and C, the extent, i.e. the number or proportion of heritage assets that could be affected. Each component is assigned a score from 1 to 5, and the resulting risk magnitude is calculated as (MR = A + B + C), producing a value between 3 and 15.

The use of structured risk-assessment methodologies is increasingly recognized as a key component of preventive conservation, as it enables heritage managers to systematically identify, compare, and prioritize threats before damage becomes irreversible. Beyond serving as a technical evaluation tool, risk assessment functions as a decision-support mechanism for conservation planning, resource allocation, and the definition of monitoring and mitigation strategies under conditions of uncertainty [38].

The application of the ABC methodology provided a standardized framework for comparing threats across the five ARGUS pilot sites and, where appropriate, within specific areas or heritage elements of a site. The assessment was based on the identification of risk scenarios following site inspection and the collection of available contextual and historical information. In particular, the methodology was used to translate qualitative knowledge about the frequency of hazards, the heritage value potentially affected, and the spatial extent of the impact into a structured prioritization of risks.

The assessment also highlighted an important methodological consideration for ARGUS. The ABC method was originally developed primarily for heritage assets such as buildings and collections, whereas several ARGUS pilots involve extensive archaeological and landscape contexts. Consequently, assigning value and estimating spatial extent can be challenging where heritage assets are heterogeneous, dispersed, or characterised by different scales and materials. The risk assessment therefore also explored the application of the method to specific spots and feature classes, while considering GIS-based landscape analysis as a complementary approach to address intra-site variability [38].

Within the ARGUS framework, risk assessment is not an end in itself but rather a mechanism for guiding monitoring design. The prioritization of threats provides a rational basis for determining which processes should be observed, the level of monitoring required, and the allocation of sensing resources across the heritage site. This approach is consistent with the broader ARGUS threat-oriented methodology, where risk identification and prioritization constitute the foundation for subsequent threat-to-sensor mapping and monitoring deployment.

The resulting scores established a hierarchy of threats for each pilot site and provided the basis for defining the monitoring priorities. Rather than being considered as a purely numerical exercise, the ABC assessment was therefore used as a decision-support mechanism to identify the threats for which continuous or periodic monitoring could provide the greatest conservation value. This prioritization directly informed the subsequent definition of monitoring requirements and sensor deployment. The results of the asset identification, threat categorization, and risk assessment for the five ARGUS pilot sites are presented in Section 3.1

2.1.4. Threat to Sensor Mapping

The resulting risk scenarios were subsequently translated into monitoring requirements through a threat–parameter–sensor mapping procedure. The purpose of this step was to establish a direct link between the risks identified through the ABC assessment and the physical variables or observable indicators that could provide evidence of their occurrence, evolution, or potential impact on the heritage asset.

For each priority threat, the monitoring process therefore followed three successive steps. First, the threat identified in the risk assessment was defined in terms of the deterioration or hazard process to be monitored. Second, one or more measurable parameters or indicators were identified that could provide information about that process. Third, the most appropriate sensing technology was selected according to the characteristics of the parameter, the required spatial and temporal resolution, and the conditions under which measurements could be obtained at the site. This provided a common framework for translating risk assessment results into technically implementable monitoring requirements.

The mapping was not intended to establish a one-to-one relationship between a threat and a single sensor. Several threats can require multiple complementary measurements, while the same parameter can contribute to the monitoring of more than one threat. For example, environmental parameters such as temperature and relative humidity can be relevant to several deterioration processes, whereas tiltmeters, crackmeters, GNSS or vibration measurements provide more direct information on structural or ground movements. The sensor planning developed reflects this multimodal approach, combining common environmental sensors with more specialised technologies such as tiltmeters, crackmeters, GNSS, acoustic-emission sensors, accelerometers, fibre-optic Bragg sensors, and multispectral-thermal cameras.

The threat–parameter–sensor mapping consequently provided the basis for defining the minimum monitoring requirements at each pilot site. The final selection also considered practical deployment constraints, including accessibility, network coverage, power supply, and the relevance of the selected location to the monitoring objectives. The detailed threat-to-parameter-to-sensor mappings developed for the ARGUS pilot sites are presented in Section 3.2.

2.1.5. Monitoring Deployment

Sensor deployment translated the threat–parameter–sensor mapping into an operational monitoring network at each pilot site. The objective was not simply to install the largest possible number of sensors, but to establish a monitoring configuration capable of collecting meaningful observations at locations relevant to the identified risks while remaining compatible with the physical and operational constraints of each heritage context.

The initial sensor configurations were therefore treated as adaptive deployment plans rather than fixed hardware lists. The sensors deployed included both commercial devices and custom-developed solutions, while some devices remained subject to technical validation or confirmation of their suitability for final installation [38]. This flexibility allows for adjustments to be made according to actual site conditions and preliminary test results.

The final positioning of the monitoring devices was determined through a combination of risk relevance and practical deployment constraints. In particular, the planning considered accessibility, network coverage, power supply, and the relevance of the monitoring location to the objectives of the project. Consequently, the most appropriate sensor location was not necessarily the location where a threat was expected to be strongest, but rather the location where the relevant phenomenon could be measured reliably without compromising installation and operation.

The resulting monitoring networks combine environmental, structural and, where relevant, anthropogenic measurements, representing the final translation of the risk assessment into field-based monitoring: priority threats determine the parameters to be observed; the parameters determine the sensing technologies; and the physical and operational conditions of each site determine how and where those technologies can ultimately be deployed. The overall sensor networks for the five pilot sites are presented in Section 3.3, while the evaluation of the resulting IoT monitoring performance is presented in Section 3.4.

2.2. Individual and Cross-Site Validation and Transferability

The first five stages of the ARGUS threat-oriented monitoring framework, namely asset characterization, threat identification, risk assessment, threat-to-sensor mapping, and monitoring deployment, provide the basis for designing monitoring strategies tailored to specific heritage sites. However, to demonstrate the robustness and broader applicability of the framework, it is necessary to assess both its effectiveness at individual sites and its capacity to support monitoring decisions across different heritage contexts. Validation was therefore undertaken not only to determine whether the monitoring systems successfully captured the deterioration processes identified during the preliminary assessment, but also to evaluate whether recurring deterioration mechanisms could be addressed through comparable monitoring approaches. This latter aspect is particularly important for assessing the transferability of the framework, as one of its core premises is that monitoring strategies should be driven by deterioration mechanisms rather than by the specific characteristics of individual heritage assets.

The ARGUS framework was therefore evaluated at two complementary levels: (i) individual-site validation and (ii) cross-site analysis and transferability assessment. These levels were applied sequentially to assess whether the preliminary risk assessment was supported by monitoring evidence at each pilot site and whether the deterioration processes identified across the sites showed sufficient recurrence to support the development of transferable monitoring strategies.

Individual-site validation was performed by comparing, for each pilot site, the priority threats identified during the preliminary ABC risk assessment (Section 2.1.3) with the evidence obtained from the monitoring campaign. The analysis examined whether the threats identified during the preliminary assessment were effectively captured by the monitoring system and whether the corresponding measurements provided evidence of the expected deterioration processes. Particular attention was given to determining whether the threats initially classified as priorities were confirmed as significant following the monitoring campaign, as well as whether other threats emerged from the measurements as more critical than initially anticipated. This site-level analysis therefore provided a direct validation of the initial threat assessment and allowed the relative importance of the identified deterioration mechanisms to be reassessed on the basis of measured evidence.

Cross-site analysis and transferability assessment were conducted together to identify recurring deterioration processes and monitoring requirements across the five pilot sites [34]. The monitoring results were compared to identify common deterioration patterns and, where similar processes occurred at two or more sites, to determine whether comparable sensor types or combinations had been used. Recurring deterioration processes and associated monitoring requirements were considered indicative of potential transferability, resulting in the identification of threat-oriented monitoring packages. These packages were not intended as fixed configurations, but as adaptable monitoring strategies that can be applied to different heritage sites presenting comparable deterioration mechanisms.

Overall, the validation procedure followed a progressive logic: the individual-site analysis assessed whether the preliminary risk assessment was confirmed or revised by monitoring evidence, while the combined cross-site and transferability analysis identified recurring deterioration processes and examined whether these could be addressed through comparable monitoring configurations. The results of the individual-site validation and the cross-site analysis and transferability assessment are presented in Section 3.5 and Section 3.6.

3. Results

3.1. Assets, Threats and Risk Assessment of the ARGUS Pilot Sites

Applying the characterization and risk-assessment procedure described in Section 2.1.1, Section 2.1.2 and Section 2.1.3, the principal threats affecting each of the five pilot sites were identified and ranked prior to monitoring deployment. Table 1 summarizes the pilot sites, their heritage characteristics, the identified threats, and the resulting risk-assessment ranking based on [35]. The ranking provides the basis for defining the priority threats that subsequently guided the monitoring strategy.

3.2. Threat-to-Sensor of the ARGUS

Applying the mapping procedure described in Section 2.1.4, each priority threat identified through the risk assessment was translated into one or more measurable parameters and linked to a corresponding sensor type. Table 2 presents representative examples of these threat–parameter–sensor relationships. The resulting mappings provided the basis for selecting the sensor categories deployed at each pilot site, with the final configurations adapted to site-specific conditions and constraints, as described in Section 3.3 (Table 3).

3.3. Monitoring Deployment

The ARGUS framework was implemented across the five pilot sites through the deployment of environmental, structural, and operational monitoring systems. Sensor selection was guided by the threat assessment process described in Section 2, ensuring that each deployment targeted the dominant deterioration mechanisms identified for the site. In total, 80 sensors were installed, covering a broad range of environmental, structural, and operational variables. Table 3 summarizes the deployed monitoring infrastructure at each pilot site.

The deployed sensor networks reflect the diversity of conservation challenges represented by the pilot sites. Environmental monitoring constituted the largest component of the infrastructure, while structural and operational sensors were deployed according to site-specific monitoring requirements.

3.4. IoT Operational Performance

To evaluate the operational performance of the monitoring infrastructure, deployment and communication metrics were compiled across the five pilot sites. In total, 80 sensing systems were deployed to monitor environmental, structural, and operational parameters under substantially different heritage and environmental conditions. Monitoring periods varied according to site-specific deployment schedules. Monitoring began in September 2025 at Ranverso, October 2025 at Delos and Monti Lucretili, December 2025 at Baltanás, and May 2026 at Schenkenberg. Monitoring continued until the end of September 2026, except at Baltanás, where restoration works beginning on 15 April 2026 affected the monitoring deployment.

The monitoring architecture combined LoRaWAN and Wi-Fi communication technologies according to site characteristics, connectivity requirements, and power availability. Cloud services included Worldsensing [40] and Shelly (Shelly Europe Ltd., Sofia, Bulgaria), while MQTT was used for data integration. LoRaWAN supported low-power, long-range communication at geographically distributed and difficult-to-access monitoring locations, whereas Wi-Fi was used where local connectivity and power infrastructure were available.

The monitoring infrastructure operated under markedly different site conditions, including marine exposure, underground environments, mountainous terrain, historic masonry, and exposed ruins. These environments introduced different operational constraints. At Delos, high salinity contributed to corrosion of sensors and protective enclosures. At Baltanás, the underground environment reduced communication coverage and required periodic maintenance of connectivity and power systems. At Monti Lucretili, dense vegetation, difficult accessibility, and occasional vandalism affected sensor access and communication reliability. At Ranverso, administrative and installation requirements constrained deployment locations and extended installation times. At Schenkenberg, the location of monitoring points required specialized access procedures, including climbers, while low temperatures and snow imposed additional installation constraints.

To provide a consistent comparison of LoRaWAN communication performance among the five deployments, gateway records were evaluated over the common observation window from 10 July to 19 August 2026. Hourly gateway availability was calculated as the proportion of expected hourly records for which gateway observations were available. Across the five gateways, aggregate hourly availability was 80.45%. Performance was heterogeneous: two gateways achieved 100% availability, while two others achieved 98.65% and 90.57%, respectively. The remaining gateway showed substantially lower availability (13.06%), accounting for most of the missing gateway-level observations and markedly reducing the aggregate network value.

Radio-link measurements provided additional information on communication performance. Across the available gateway observations, the mean received signal strength indicator (RSSI) was approximately −91.1 dBm, with a median of −89 dBm and an observed range from −126 to −53 dBm. The corresponding mean signal-to-noise ratio (SNR) was approximately 8.3 dB, with a median of 9 dB and an observed range from −19 to 15 dB. These values demonstrate substantial variability among individual links and monitoring locations, consistent with the heterogeneous propagation environments encountered across the pilot sites.

Gateway restart records contained 25 distinct reboot episodes across the five deployments. These events were considered indicators of operational interruptions rather than communication failures because the available records do not establish the cause of each restart. Furthermore, the gateway restart exports retain only a limited number of historical events. Consequently, a formal mean time between failures (MTBF) was not calculated. The mean interval between the recorded reboot episodes was 9.02 days, but this value should be interpreted only as a descriptive metric for the available restart records rather than as a long-term reliability parameter. Similarly, packet-loss rate was not estimated because the gateway dataset contains received-packet counts but does not provide an independent reference for the total number of packets transmitted.

Table 4. LoRaWAN gateway operational performance and maintenance indicators.

Metric	Value
LoRaWAN gateways evaluated	5
Common observation window	10 July–19 August 2026
Mean hourly gateway data availability	80.45%
Gateway availability range	13.06–100%
Gateway reboot episodes recorded	25
Mean interval between recorded reboot episodes*	9.02 days
Mean RSSI	−91.1 dBm
Median RSSI	−89 dBm
RSSI range	−126 to −53 dBm
Mean SNR	8.3 dB
Median SNR	9 dB
SNR range	−19 to 15 dB
Sensor maintenance interventions	21
Battery replacements	7
Vandalism incidents	5
Animal-related incidents	2
LoRaWAN gateway uptime	135d 9h 20min (3249.333h)

*Calculated from the available gateway restart records. The restart exports contain a limited number of historical events and therefore do not provide a complete failure history; this value should not be interpreted as formal MTBF. The gateway availability and radio-link statistics were calculated from the LoRaWAN gateway dataset.

Table 4. LoRaWAN gateway operational performance and maintenance indicators.

Metric	Value
LoRaWAN gateways evaluated	5
Common observation window	10 July–19 August 2026
Mean hourly gateway data availability	80.45%
Gateway availability range	13.06–100%
Gateway reboot episodes recorded	25
Mean interval between recorded reboot episodes*	9.02 days
Mean RSSI	−91.1 dBm
Median RSSI	−89 dBm
RSSI range	−126 to −53 dBm
Mean SNR	8.3 dB
Median SNR	9 dB
SNR range	−19 to 15 dB
Sensor maintenance interventions	21
Battery replacements	7
Vandalism incidents	5
Animal-related incidents	2
LoRaWAN gateway uptime	135d 9h 20min (3249.333h)

*Calculated from the available gateway restart records. The restart exports contain a limited number of historical events and therefore do not provide a complete failure history; this value should not be interpreted as formal MTBF. The gateway availability and radio-link statistics were calculated from the LoRaWAN gateway dataset.

Although interruptions and maintenance requirements occurred during the monitoring campaign, the results indicate that the communication infrastructure provided reliable operation at most deployments, while individual site-specific connectivity limitations could substantially affect aggregate performance. The strong contrast between the four higher-performing gateways and the lowest-performing gateway also demonstrates the importance of evaluating communication performance at deployment level rather than relying exclusively on an overall network average. In complex heritage environments, gateway positioning, topography, physical obstructions, accessibility, and local environmental conditions should therefore be considered integral components of monitoring-system design.

Overall, the operational results demonstrate the feasibility of maintaining an interoperable IoT monitoring infrastructure across heterogeneous cultural heritage environments. The combination of LoRaWAN and Wi-Fi communication, together with cloud-based data management, enabled the acquisition of environmental, structural, and operational observations despite substantial differences in accessibility and environmental exposure among the pilot sites. The observed variability in gateway availability further highlights the need for site-specific communication planning, gateway optimization, and routine maintenance when implementing long-term IoT monitoring for preventive conservation.

3.5. Validation of the Threat-Oriented Framework

To evaluate the ARGUS framework, the dominant threats identified during the preliminary ABC risk assessment were compared with monitoring observations collected during the deployment period. The objective was to determine whether the threat-oriented methodology successfully identified the deterioration processes governing each pilot site. To illustrate the capability of the deployed monitoring systems to capture deterioration processes, representative datasets from each pilot site are presented in Figure 3, Figure 4, Figure 5, Figure 6 and Figure 7. The selected examples correspond to the dominant threats identified during the risk assessment stage and demonstrate the relationship between environmental forcing, material response, and structural behavior.

At the Baltanás pilot site (Figure 4), vertical movements were recorded at the micrometre scale, indicating measurable responses of the cellars to changes in vertical loading or pressure conditions. Accelerometers recorded vibration outliers coinciding with reported collapses, demonstrating their potential for detecting structural events. Soil moisture increased rapidly by approximately 5% over two weeks, suggesting significant changes in subsurface moisture conditions. Relative humidity was consistently high, with a mean above 90% and frequent values approaching 100%, representing potentially critical conditions for material deterioration.

Figure 5 illustrates the influence of anthropogenic pressure at the highly visited site of Monti Lucretili, with up to 30 visitors recorded within a single hour. Although humidity fluctuated considerably, UV radiation remained relatively low during the monitoring period. However, as the data correspond to winter conditions, with UV indices around 3, higher exposure during summer should be considered. Short-duration rainfall events also produced accumulations of approximately 3 mm over consecutive hours. Tiltmeter measurements further recorded displacements of up to 0.2°, highlighting the relevance of structural movement as a potential threat.

At Ranverso Abbey (Figure 6), UV levels inside the abbey remained below 0.025, suggesting extremely low, almost absent, UV exposure to the paintings under the monitored conditions. Relative humidity, however, remained relatively high, with a mean of approximately 70%, which may be relevant for the conservation of painted surfaces. Particulate matter concentrations reached values of approximately 70, likely associated with visitor activity and indicating sustained particle deposition that may, over time, contribute to optical degradation, color shifting, and structural damage to the paint layers. Rapid increases and decreases in soil moisture may also indicate short-term responses to heavy rainfall or localized flooding events.

Finally, Figure 7 presents the monitoring results for Schenkenberg, characterized by substantial daily temperature variations exceeding 10°C, mean wind speeds of approximately 7 m/s, and recorded accelerations predominantly within the 0–10 Hz range. These measurements indicate that thermal and wind-induced environmental fluctuations, together with low-frequency structural vibrations, should be considered among the relevant threats at the site during the summer months.

Across the five pilot sites, monitoring observations consistently confirmed the deterioration mechanisms predicted during the preliminary risk assessment. Moisture-related processes dominated at Baltanás and Ranverso, climate-driven deterioration was predominant at Delos, active structural instability characterized Monti Lucretili, and climate-induced structural stress was the principal concern at Schenkenberg. Table 5 compares the threats identified during the preliminary risk assessment with the evidence obtained during the monitoring campaign. For each site, threats were reassessed according to their observed significance in the monitoring data. A monitoring-based priority ranking was assigned, where 1 represents the threat showing the strongest evidence or greatest apparent significance during the monitoring period, with progressively higher numbers indicating lower observed priority. Threats that could not be adequately evaluated during the monitoring period, either because the relevant conditions did not occur or because sufficient monitoring evidence was unavailable, were left unranked. The monitoring-based ranking is therefore intended as a relative, site-specific prioritization based on the observations collected during the monitoring campaign, rather than as a replacement for the original ABC risk score.

These results provide qualitative validation of the threat-oriented methodology and demonstrate that the preliminary risk assessment successfully guided monitoring design and sensor deployment. The ability to correctly identify the dominant deterioration process across all pilot sites highlights the applicability of the ARGUS framework under diverse environmental and conservation conditions. Ranking threats by observed priority at each site provides the basis for identifying which deterioration mechanisms recur across multiple heritage contexts, examined in Section 3.6.

3.6. Cross-Site Comparability and Transferable Monitoring Packages

Aggregating the priority rankings established in Section 3.5 across the five pilot sites shows that, although the specific dominant threat differs by site, a consistent pattern nevertheless emerges beneath this diversity: environmental conditions were repeatedly associated with measurable material or structural responses, although the strength of this relationship varied among sites. At Delos, strong climatic exposure, including wind gusts and rainfall events, occurred alongside measurable structural inclination. At Baltanás, rainfall preceded rapid increases in soil moisture, while structural movements and vibration outliers were recorded in association with reported cellar collapses. At Monti Lucretili, rainfall and humidity variations occurred alongside sudden crack and inclination movements, with tiltmeter variations reaching approximately 0.2°. At Ranverso, elevated relative humidity and rapid soil-moisture variations demonstrated the response of the site to changing moisture conditions, although a corresponding structural response could not be clearly established. At Schenkenberg, large daily temperature oscillations and moderate-to-high wind exposure occurred alongside measurable low-frequency structural vibrations.

This recurring relationship between environmental forcing and material or structural response, rather than the identity of the threat itself, is transferable across heritage typologies. Across the pilot sites, environmental monitoring provided contextual or leading information for interpreting subsequent changes in material conditions and, where structural measurements were available, structural behavior. This observation supports the central hypothesis explored in this section: that environmental variables can provide early or contextual information for interpreting deterioration risk, independently of monument typology or threat category. It is this relationship, rather than the individual threats themselves, that underpins the transferable monitoring packages presented below.

Analysis of the monitoring deployments carried out across the five pilot sites revealed recurring combinations of sensors associated with specific deterioration mechanisms. These sensor combinations emerged independently of heritage typology, geographical location, and construction characteristics, suggesting that monitoring requirements are primarily determined by conservation threats rather than monument categories. As a result, a series of transferable monitoring packages were identified, each targeting a specific group of deterioration processes. Table 6 shows the five proposed threat packages, the sensors and the representative ARGUS pilot sites for package validation.

The recurrence of these packages across different pilot sites demonstrates that similar deterioration processes generate comparable monitoring requirements, even when heritage assets differ substantially in typology, age, materials, and environmental context. For example, Baltanás and Ranverso represent fundamentally different heritage assets, yet both required nearly identical monitoring configurations due to the predominance of moisture-related deterioration mechanisms. Likewise, Delos and Schenkenberg exhibited comparable requirements for monitoring climate exposure despite their distinct geographical and cultural characteristics. These results provide empirical support for the central hypothesis of the ARGUS framework: heritage assets affected by similar deterioration mechanisms can be monitored using comparable sensor configurations regardless of monument typology. The identified threat packages therefore constitute reusable monitoring solutions that can facilitate future deployments, simplify sensor selection, and improve the transferability of preventive conservation strategies across diverse cultural heritage contexts. This finding represents one of the principal outcomes of the ARGUS validation process and provides the basis for the broader threat-oriented monitoring approach discussed in the following section.

4. Discussion

4.1. From Monument-Based to Threat-Based Monitoring

The comparison of the five ARGUS pilot sites revealed a consistent and significant finding: monitoring requirements were more strongly associated with deterioration mechanisms than with monument typology. Although the pilot sites differed substantially in age, construction materials, architectural characteristics, environmental conditions, and conservation status, sites affected by similar deterioration processes required comparable monitoring strategies.

Traditional cultural heritage monitoring programs are commonly organized according to asset typologies. However, the results obtained during the ARGUS validation demonstrate that deterioration mechanisms constitute a more reliable basis for monitoring design than monument classification. Sites affected by comparable degradation processes consistently required similar monitoring configurations despite substantial differences in function, historical context, and geographical location.

For example, Baltanás and Ranverso represent fundamentally different heritage assets, yet both exhibited moisture-related deterioration processes associated with groundwater influence, capillary rise, and high relative humidity. Consequently, both sites required similar monitoring systems based on temperature, relative humidity, moisture content, airflow, and hydrological measurements. Likewise, Delos and Schenkenberg, despite their geographical and cultural differences, shared monitoring requirements related to climatic exposure, weathering, and environmental forcing.

These observations support the central hypothesis of the ARGUS framework: monitoring strategies should be defined according to deterioration mechanisms rather than monument typology. The successful validation of this approach across five contrasting pilot sites demonstrates its applicability to a wide range of cultural heritage environments.

Furthermore, the transferability of the framework extends beyond the methodological level. By integrating commercial monitoring systems and custom-developed low-cost sensor nodes through a common MQTT-based architecture, ARGUS demonstrates that threat-oriented monitoring can remain independent of specific hardware implementations and communication technologies. This interoperability facilitates the deployment of reusable monitoring solutions across diverse heritage contexts.

This shift from monument-based to threat-based monitoring represents the principal scientific contribution of the ARGUS framework and provides a practical foundation for risk-informed heritage monitoring and preventive conservation.

4.2. Transferable Monitoring Packages

One of the most relevant outcomes of the cross-site comparison was the identification of recurring sensor combinations associated with specific deterioration mechanisms. These recurring configurations emerged independently of heritage typology and therefore provide the basis for transferable monitoring packages applicable to a wide range of conservation scenarios.

The Moisture Monitoring Package combines temperature sensors, relative humidity sensors, soil and wall moisture sensors, airflow measurements, and water-level observations. This package proved particularly effective at Baltanás and Ranverso, where moisture accumulation, capillary rise, flooding risk, and groundwater dynamics represented the dominant deterioration mechanisms. The integration of these variables allows moisture sources to be identified and drying conditions to be evaluated more accurately than with isolated measurements.

The Structural Stability Package integrates tiltmeters, crackmeters, GNSS systems, and vibration sensors. This configuration was especially valuable at Monti Lucretili and Schenkenberg, where structural behaviour constituted a major conservation concern. The combination of geometric, displacement, and vibration measurements provides complementary information for distinguishing environmentally related responses from potentially progressive structural changes.

The Climate Exposure Package combines weather stations with measurements of rainfall, temperature, wind, solar radiation, UV exposure, and illuminance. Deployments at Delos and Schenkenberg demonstrated the effectiveness of this package for characterizing environmental forcing and evaluating conservation risks associated with weathering, thermal stress, and freeze-thaw processes.

Perhaps the most important observation is that reliable diagnosis rarely resulted from individual sensors considered in isolation. Instead, deterioration mechanisms became significantly easier to interpret when environmental variables and structural responses were analyzed together. Interpretation of crack development was improved when evaluated alongside changes in moisture, rainfall, temperature, and wind conditions. Similarly, the interpretation of tilt measurements improved considerably when climatic and hydrological information was incorporated.

These findings indicate that future heritage monitoring systems should prioritize integrated sensing networks capable of simultaneously capturing deterioration drivers and the responses they generate.

4.3. Implications for Preventive Conservation

The results obtained from the ARGUS monitoring campaign have important implications for preventive conservation. Traditional conservation strategies often remain reactive, focusing on the treatment of visible damage after deterioration has already occurred. While corrective interventions are sometimes unavoidable, they generally involve higher costs, greater material loss, and more invasive actions than preventive approaches.

The threat-oriented methodology demonstrated by ARGUS provides a mechanism for identifying deterioration drivers before significant damage develops, allowing interventions to be implemented at an earlier stage.

Moisture-related processes emerged as one of the most recurrent deterioration mechanisms across the pilot sites, particularly at Baltanás and Ranverso, while also contributing to climate-driven deterioration processes at other sites. Consequently, interventions aimed at improving drainage, reducing water infiltration, enhancing ventilation, controlling groundwater influence, and limiting humidity accumulation are likely to provide the greatest conservation benefits.

A second priority concerns structural instability. Sites exhibiting progressive crack development, displacement trends, or persistent geometric changes require detailed structural assessment and stabilization before deterioration reaches critical levels. The observations obtained at Monti Lucretili illustrate the importance of monitoring structural responses alongside environmental forcing.

A third priority relates to environmental and climatic exposure. Highly exposed sites such as Delos and Schenkenberg are particularly vulnerable to weathering processes associated with solar radiation, wind, extreme temperatures, and freeze-thaw cycles. Protective measures, improved drainage, vegetation management, and climate adaptation strategies may therefore contribute significantly to reducing long-term deterioration rates.

Beyond individual sites, the ARGUS framework also provides a practical mechanism for translating climate-related risks into measurable monitoring variables. The integration of environmental, structural, and operational datasets enables the development of evidence-based conservation strategies that support adaptation to changing environmental conditions.

Overall, the results demonstrate that preventive conservation can be strengthened by prioritizing deterioration drivers rather than deterioration symptoms. By linking risk assessment, monitoring, and conservation decision-making within a common framework, ARGUS provides a transferable methodology for supporting long-term, evidence-based management of cultural heritage assets.

4.4. Limitations and Future Work

Several limitations of the present study should be acknowledged. First, monitoring periods varied among the pilot sites and, in some cases, were limited to less than one annual climatic cycle. Consequently, the long-term evolution of certain deterioration mechanisms, particularly those influenced by seasonal variability or low-frequency events, could not be comprehensively assessed. Second, while the framework was validated across a diverse set of environmental and heritage contexts, some deterioration processes, such as biological colonization and material-specific decay mechanisms, were not systematically monitored or validated at all sites. Third, the validation focused primarily on the ability of the framework to identify and monitor dominant deterioration mechanisms rather than on quantifying deterioration rates or predicting future damage. As a result, the monitoring thresholds used for alerts and preventive interventions should be considered preliminary and require further calibration through longer-term observations and larger datasets.

In addition, the current validation involved five European pilot sites representing a limited range of climatic, geological, and management conditions. Further testing in other geographical regions and heritage contexts would be necessary to evaluate the broader transferability of the framework. Future research should also explore the integration of predictive analytics, deterioration forecasting models, and data-driven decision-support tools capable of transforming monitoring observations into quantitative risk predictions and conservation recommendations. Finally, the long-term assessment of framework performance would benefit from direct evaluation of how monitoring-informed interventions influence conservation outcomes, maintenance costs, and asset resilience over time.

5. Conclusions

This study validated the ARGUS framework across five European cultural heritage pilot sites, demonstrating the feasibility of a transferable threat-oriented monitoring approach that integrates risk assessment, IoT-based sensing technologies, and preventive conservation principles. Unlike conventional monitoring strategies organized according to monument typology, the proposed framework was designed around deterioration mechanisms, enabling monitoring requirements to be defined according to conservation risks rather than asset categories.

The results showed a high degree of correspondence between the threats identified during the preliminary ABC risk assessment and the deterioration processes observed through monitoring. Across all pilot sites, the dominant deterioration mechanisms predicted during the assessment stage were confirmed by sensor observations, demonstrating the effectiveness of the threat-oriented methodology for guiding sensor selection and deployment. Environmental variables provided important contextual and, in some cases, leading information for interpreting material and structural responses, while moisture-related processes emerged as a recurrent conservation risk across multiple sites.

The comparative analysis further revealed that monuments exposed to similar deterioration mechanisms required comparable monitoring configurations despite substantial differences in architecture, materials, geographical setting, and historical function. This finding led to the identification of a set of transferable monitoring packages associated with recurring threats, including moisture-related deterioration, climate exposure, and structural instability. These packages provide practical and reusable monitoring solutions that can support future heritage monitoring deployments.

The principal contribution of the ARGUS framework is the demonstration that heritage monitoring can be organized and standardized around deterioration processes rather than monument categories. By linking threat assessment, sensor selection, monitoring deployment, and preventive conservation within a common methodology, the framework provides a scalable foundation for risk-informed conservation management, predictive maintenance, and climate-resilient heritage preservation.

Overall, the results support a transition from monument-based monitoring strategies to threat-oriented approaches that focus on the drivers of deterioration. This shift has the potential to improve the efficiency, transferability, and long-term sustainability of monitoring programs, while strengthening the role of digital technologies in preventive conservation and evidence-based heritage management.