Vertical Ground-Motion Effects in Base-Isolated Buildings: Preliminary Observations from Twin Fixed-Base and Base-Isolated Structures During the 18 March 2025 Potenza Sequence

Abstract

On 18 March 2025, a moderate earthquake with moment magnitude Mw 4.2 struck the Basilicata region in Southern Italy. The event occurred at 09:01:25 UTC with an epicentre located approximately 4 km northeast of the city of Potenza (PZ). The earthquake was clearly felt across the urban area and followed by a sequence of low-magnitude aftershocks. A few hours after the main shock, researchers from the University of Basilicata installed a temporary structural monitoring network to check the structural conditions of several buildings located in Potenza. This installation enabled the acquisition of accelerometric recordings of several aftershocks, providing a valuable dataset for preliminary observations on structural seismic response. The monitoring campaign focused on two adjacent twin buildings with similar geometry and structural layout but different seismic design strategies: one conventionally fixed at the base and the other equipped with seismic base isolation made by rubber bearings. Comparative analyses revealed distinct differences in dynamic response. The results highlight the need for refined regulatory tools to address near-epicentral conditions, particularly potential dynamic interactions among the vertical ground-motion component, the vertical vibration frequencies of the superstructure, and floor-system resonance. While not critical for ultimate limit states, these effects may influence comfort and performance in operational and damage limit states.

1. Introduction

Earthquakes occurring in near-epicentral conditions are increasingly recognised as a challenging scenario for the seismic performance assessment of structures, even when associated with moderate magnitudes. It has been demonstrated that shallow hypocentral depths, short source-to-site distances, and non-standard ground-motion characteristics may result in seismic actions that are not adequately represented by conventional design spectra derived from probabilistic seismic hazard analysis [1,2,3,4,5,6]. In such conditions, ground motions may exhibit pronounced high-frequency content, significant vertical components, and atypical vertical-to-horizontal ratios, potentially activating dynamic mechanisms that are not explicitly addressed by current seismic codes. Recent numerical investigations have highlighted the potential significance of these features for structures equipped with seismic protection systems. As demonstrated by several authors [7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29], near-fault and pulse-type ground motions have the capacity to induce unexpected response phenomena in both fixed-base and base-isolated buildings. These include torsional effects, pounding between adjacent structures and soil–structure interaction effects, even when collapse prevention requirements are satisfied [30,31,32,33]. The findings of the study highlighted that base isolation, while being highly effective in reducing horizontal seismic demand, may introduce sensitivity to specific characteristics of the seismic input that are not fully accounted for in conventional design approaches. Recent observational studies on low-to-moderate-magnitude seismicity in Italy have yielded supplementary evidence. Analyses of recent seismic sequences have demonstrated that, despite their limited magnitude, shallow earthquakes have the capacity to generate non-negligible structural actions in close proximity to the epicentre, particularly during periods of short vibration and for the vertical ground-motion component [34]. Furthermore, site-specific ground motion models calibrated for shallow and volcanic seismic sources indicate systematically larger spectral amplitudes at short periods and non-standard attenuation trends. This raises questions regarding the adequacy of current ground-motion prediction models and design spectra in near-source conditions [35].

The objective of this paper is to report preliminary observations on the seismic response that was recorded during the 18 March 2025 Mw 4.2 Potenza earthquake sequence. The study utilised accelerometric data, which was obtained through the rapid post-event deployment of a structural health monitoring network on a selection of reinforced concrete buildings. The focus of the study was on two structures with similar geometry and structural layout but different seismic design strategies: one fixed-base and one equipped with seismic base isolation based on rubber bearings. This configuration facilitates a direct comparison of their dynamic behaviour under real near-epicentral shaking. The considerations presented herein are preliminary; however, it is believed that they will provide potentially significant technical and scientific insights. In particular, the observations highlight the possible need to improve current design criteria adopted in Italian and international seismic codes, such as the Italian NTC 2018 [36] and Eurocode 8 [37], which are primarily calibrated to ensure structural safety with respect to ultimate limit states under horizontal seismic actions. The findings of this study indicate that, in the context of low-to-moderate-intensity earthquakes occurring in near-epicentral conditions, there is a necessity for heightened focus on serviceability and operational limit states, particularly with regard to comfort performance. The vertical component of ground motion, and the dynamic interaction effects among the vertical vibration modes of the superstructure, the resonance characteristics of floor systems, and the membranal frequencies of slabs, are of particular concern. Whilst the effects described above are not critical in terms of preventing collapse or causing structural damage, as addressed by current provisions of NTC 2018 and Eurocode 8, they have been shown to generate perceptible vibrations and discomfort, particularly in base-isolated buildings. The issue is not yet explicitly treated within existing normative frameworks. Further comprehensive analyses of the seismic input and structural response, incorporating refined numerical modelling currently under development, are underway and will be the subject of forthcoming scientific publications. The objective of the forthcoming studies is to provide more detailed quantitative assessments and practical indications for structural designers. In addition, robust benchmarks will be provided for the scientific community to support future comparisons and potential updates of seismic design criteria at both national and international levels. This is particularly important in view of the substantial economic and intellectual resources invested in seismic protection strategies and approaches [38,39,40,41]. These strategies are intended not only to safeguard human life, but also to preserve economic assets. It should be noted that a significant share of a building’s overall value is often associated with non-structural and architectural/decorative components. It is therefore essential that these components are explicitly protected and maintained [42,43,44].

2. Analysis of the Mainshock and Comparison with NTC2018 Provisions

The earthquake under scrutiny in this study occurred on 18 March 2025 at 09:01:25 UTC (10:01:25 local Italian time, UTC +01:00), with moment magnitude Mw 4.2. The epicentre of the earthquake was located approximately 4 km northeast of the city of Potenza (PZ), at geographical coordinates 40.6632° N latitude and 15.8432° E longitude, and at a focal depth of about 13 km. The occurrence was located and officially documented in the Italian Seismic Bulletin by the Istituto Nazionale di Geofisica e Vulcanologia (INGV, [45]), based on data recorded by the national seismic monitoring network. The macroseismic intensity map (see Figure 1) illustrates the spatial distribution of perceived shaking and reported effects.

Maximum macroseismic intensities in the epicentral area range between V and locally VI on the EMS/OFM scale, while a rapid attenuation is observed with increasing distance from the source. The nearly circular symmetry of the intensity field is consistent with a moderate-magnitude tectonic event characterized by limited rupture dimensions and relatively shallow depth. The absence of severe damage reports and the prevalence of light effects and human perception further confirm the moderate nature of the event.

The Peak Ground Acceleration (PGA) ShakeMap (Figure 2) shows maximum PGA values on the order of a few hundredths of g in the near-field, with tightly spaced contour lines indicating a steep decay of seismic demand away from the epicentre. These values are well below those associated with structural damage or ultimate limit states but are representative of seismic actions relevant to serviceability and operational performance.

The recorded three-component accelerograms (Figure 3) display a short-duration signal, with most of the energy concentrated within the first 5–7 s [46,47,48].

A salient feature is the comparable amplitude of the vertical component with respect to the horizontal components, thus emphasising the non-negligible role of vertical ground motion in near-epicentral conditions. This aspect is further emphasised by the acceleration response spectra (see Figure 4), which demonstrate peak spectral ordinates within a narrow range of short vibration periods (approximately 0.1–0.3 s).

This range of frequencies is coincident with the fundamental vibration periods of low- to mid-rise buildings and with the vertical and membranal vibration frequencies of typical floor systems. This suggests the possibility of dynamic interaction and resonance phenomena affecting comfort and operational limit states.

The maximum peak ground accelerations that were recorded during the earthquake measuring 4.2 on the Richter scale that occurred on 18 March 2025 were of the order of approximately 0.02 g, i.e., about two hundredths of gravitational acceleration. When considered in isolation, this value suggests a relatively low level of seismic demand. A comparison with the seismic hazard parameters stipulated in current design standards [36] for standard buildings (Use Class II, with a nominal life of 50 years) provides a clearer engineering interpretation of the event. In particular, with regard to the operational limit state (SLO), Figure 5 presents the code-based seismic hazard parameters at the site in accordance with the Italian design provisions.

The corresponding design horizontal acceleration is equal to 0.055 g, with a return period of 30 years. This value is more than double the maximum acceleration measured during the earthquake, thereby confirming that the event falls well within the range of low-intensity seismic actions when evaluated in terms of peak ground acceleration and ultimate structural safety.

The combined interpretation of seismological parameters, macroseismic observations, instrumental recordings, and spectral response indicates that, although the event does not raise concerns in terms of structural safety, it provides a meaningful case study for investigating vertical motion effects and serviceability-related issues that are not explicitly addressed by current seismic design provisions.

3. Description of the Temporary Network for Structural Monitoring

In the preliminary phase of the monitoring campaign, a temporary seismic monitoring network composed of several three-directional force-balance sensors was deployed on selected structures in Potenza. The aim of this deployment was to record the structural response during the aftershock sequence following the 18 March 2025 earthquake. The network was configured with the specific purpose of enabling a direct comparison to be made between different structural typologies that were subjected to an identical seismic input. The temporary monitoring network employed Lunitek TRITON triaxial force-balance accelerometers (FBA) with a full-scale range of ±1 g and 24-bit A/D conversion. The acquisition chain provides a high dynamic range (136 dB at 100 SPS), multi-stage anti-aliasing (fixed 5th-order sinc stage, FIR low-pass with 140 dB attenuation at Nyquist, and analog 2-pole LPF), and GNSS-based timing with sub-microsecond accuracy. The selected processing bandwidth (of the order of 30 Hz) is well within the instrument flat frequency response (from DC to at least 80–100 Hz, depending on configuration), and the observed amplitudes are orders of magnitude below the clipping level, thus excluding any saturation artefact. One triaxial accelerometric station was installed at ground level inside a fixed base structure, providing a reference measurement representative of the input ground motion at the site. A secondary station was installed at the uppermost level of a building, which is characterised by a conventional fixed-base structural system, with the objective of capturing the global response of a non-isolated structure. A third triaxial accelerometric station was installed at the uppermost level of a building equipped with a seismic base-isolation system made by 36 rubber bearings. This enabled the dynamic behaviour of an isolated structure to be directly compared with that of its fixed-base counterpart. In Figure 6, a map of the monitored area is provided, with the locations of the installed accelerometric stations indicated.

For each station, the direction and orientation of the red arrow are consistent with the local y-axis of the accelerometric sensor. This provides a clear reference point for the interpretation of the recorded horizontal components, thereby ensuring consistency in the analysis of directional seismic response across the different measurement points. All accelerometric stations were triaxial in nature, with the capacity to record motion in two horizontal directions and the vertical direction. Furthermore, the stations were synchronised via GPS in order to ensure accurate time alignment of the recorded signals (Figure 7).

This configuration enables reliable comparative analyses of amplitude, frequency content, and phase relationships among the different measurement points, providing a robust experimental basis for preliminary assessments of structural response under low-amplitude seismic excitation. Assuming a conservative dynamic range of 136 dB, referred to full-scale sinusoidal RMS, the equivalent instrumental RMS noise is approximately 10⁻⁷ N (sub-micro-g). It is evident that the 24-bit quantization step for a ±1 g range is approximately 2 g/2²⁴ ≈ 1.2 × 10⁻⁷ g, signifying that quantization noise does not act as a limiting factor. Consequently, even for weak events (PGA ≈ 0.003 g), the expected instrumental SNR remains high (approximately 1.9 × 10⁴).

4. Analyses of the Structural Response During the Two Main Aftershocks

In the aftermath of the Mw 4.2 mainshock of 18 March 2025, two aftershocks, categorised by the INGV as the most significant in the sequence, were selected as engineering-relevant low-amplitude inputs for the interpretation of the monitored structural response [45]. The first of these events occurred on 19 March 2025 at 19:13:16 (local time), with an initial magnitude of 2.3, epicentral coordinates 40.6457° N–15.8562° E, and a hypocentral depth of approximately 10 km. The second occurrence was observed on 23 March 2025 at 20:38:07 (local time), with a magnitude of 2.7, located at coordinates 40.6625° N–15.8493° E, and with an estimated depth of approximately 11 km. A thorough analysis of the INGV locations reveals a striking spatial correlation between the two events, with both exhibiting a concentration within a few kilometres of the epicentre of the primary seismic event. Furthermore, the hypocentral depths of both events are found to be shallow, which is consistent with an aftershock sequence generated within the same activated seismogenic volume (or closely associated fault segment) as the main event.

For both aftershocks, the peak ground acceleration (PGA) recorded at ground level by the PTZ station [45,46] was lower than 3 cm/s², i.e., approximately 0.003 g. From an engineering perspective, the input level in question is distinctly lower than the design action associated with the operational limit state (SLO) at the site (order of a few hundredths of g). This finding suggests that the two aftershocks should be interpreted as operational-level excitations. Consequently, the primary value of the aforementioned elements lies in their capacity to support serviceability-oriented analyses. Such analyses may include the identification of modal properties, the assessment of frequency-dependent amplification, and the preliminary comparison between structural configurations. It should be noted that these elements are not intended for use in any evaluation related to ultimate limit states or damage.

Figure 8 and Figure 9 present the three-component acceleration time histories measured at the top level of the fixed-base building and of the base-isolated building, where X, Y, and Z denote the transverse, longitudinal, and vertical components, respectively. Beyond the anticipated similarity in the horizontal response to such low-amplitude input, the engineering-relevant and somewhat unexpected outcome concerns the vertical response: the base-isolated building exhibits a more pronounced amplification of the Z component compared to the fixed-base counterpart. This aspect is of particular significance given that the two structures are essentially twin buildings, located only a few metres apart and therefore subjected to nearly the same input motion, with the main distinguishing feature being the adoption of base isolation in one case and a conventional fixed-base configuration in the other. The observed vertical amplification in the isolated structure, even under very modest PGA levels, suggests that the isolation system and the associated global dynamic characteristics may influence the transmission and/or redistribution of vertical vibration components. It is imperative to consider the potential implications of this phenomenon on serviceability and comfort performance, in addition to the interpretation of operational-level monitoring data.

In Figure 10, a comprehensive overview of the maximum absolute accelerations recorded at the uppermost level of the two observed twin buildings during the aftershocks that occurred on 19 and 23 March 2025 is provided. This summary is presented separately for the transverse (X), longitudinal (Y), and vertical (Z) components.

The results of the study indicate a component-dependent trend. In the horizontal directions (X and Y), the base-isolated building consistently exhibits lower peak accelerations than the fixed-base building for both events, with the difference becoming more evident for the stronger 23 March aftershock. In contrast, the vertical component (Z) displays a different pattern of behaviour. The base-isolated structure exhibits higher peak vertical accelerations than the fixed-base counterpart in both records. It is evident that the two buildings are very similar and located a few dozen meters apart from each other. This outcome serves to reinforce the engineering relevance of the observed vertical amplification in the isolated configuration. This may be critical for serviceability and comfort performance, even under very low-amplitude seismic inputs.

In order to achieve a more profound engineering interpretation of the recorded response, the analysis was expanded to the frequency domain. This was achieved by implementing a consistent pre-processing and spectral estimation workflow in a MATLAB R2025b routine. Subsequently, each component was subjected to band-pass filtering within the 0.1–25 Hz range, with the objective of removing quasi-static trends and high-frequency noise. Subsequently, an automatic time window was extracted around the strongest portion of motion. In the Welch PSD estimation, the key parameters were explicitly set to ensure a stable averaged periodogram despite the short, low-amplitude records. The PSD was computed using a Tukey window with segment length = 1024 samples, overlap = 900 samples, and FFT length = 2048, after applying the same preprocessing adopted throughout the study (band-pass filtering within 0.1–25 Hz and linear detrending on the selected analysis window). The selection of a high overlap and a moderate segment length has been demonstrated to increase the number of segments that effectively contribute to the average within the available record length. This, in turn, has been shown to reduce the variance of the spectral estimate when compared to a single-periodogram approach. The selection was centred on the peak of the longitudinal component (Y). The process under scrutiny incorporated both a concise pre-event segment and an extensive post-peak portion. Prior to the implementation of spectral analysis, the signals were subjected to linear detrending in order to circumvent the occurrence of bias in the low-frequency content. As demonstrated in Figure 11 and Figure 12, the subsequent Welch power spectral density (PSD) estimates for the two buildings are reported, with separate analyses conducted for the transverse (X), longitudinal (Y), and vertical (Z) components for the 19 March and 23 March aftershocks, respectively.

The Welch method [49] provides a robust estimate of the distribution of vibration energy over frequency under these low-amplitude excitations. From an engineering perspective, the key finding from both figures is the significant increase in spectral content in the vertical component of the base-isolated structure relative to the fixed-base twin. This is evident in a pronounced and narrow-band amplification in the Z-direction PSD. In contrast, the horizontal components demonstrate comparable or reduced spectral levels in the isolated building, which is consistent with the intended mitigation of horizontal accelerations. The concurrence of this vertical amplification trend across Figure 11 and Figure 12 indicates that the phenomenon is not event-specific, but rather a reproducible attribute of the isolated configuration under operational-level seismic inputs.

It is possible to offer a concluding engineering observation by interpreting the spectral features visible in Figure 11 and Figure 12 directly. The predominant peaks detected in the horizontal components (X and Y) are in alignment with the fundamental lateral frequencies of the fixed-base structure and, more broadly, with the lateral dynamic characteristics of the superstructure. Conversely, the distinctive peak linked to the horizontal isolation mode is not distinctly apparent in the PSDs. This outcome is technically coherent with the very low input levels of the two aftershocks. It is expected that, under such modest excitation, the isolation system will not enter a significant large-displacement regime. Therefore, the isolation mode is not effectively mobilised in the recorded response. In contrast, the vertical component (Z) displays a clearly identifiable and reproducible spectral peak at approximately 12.6 Hz, which is significantly more pronounced in the base-isolated structure. This peak is attributable to the fundamental vertical vibration frequency of the isolated system, governed primarily by the vertical stiffness of the isolation devices and the mass of the superstructure. In engineering terms, as illustrated in Figure 11 and Figure 12, even in circumstances where the isolation system is not significantly activated in the horizontal direction due to the low intensity of the input motion, the vertical dynamics of the isolation system can be distinctly expressed and may result in measurable amplification at the superstructure level. This has direct implications for serviceability and comfort-related performance.

The vertical transfer function was computed to experimentally characterise the superstructure dynamics while minimising the influence of the base input. This was achieved by using the triaxial accelerometric station installed at the top of the base-isolated building superstructure and the reference station. The transfer function was estimated using the same Welch-based spectral estimation settings that were adopted for the power spectral density calculations, ensuring methodological consistency in terms of windowing, overlap, and averaging. The resulting spectral ratio, as illustrated in Figure 13, displays a prominent amplification peak at approximately 12.9 Hz, which is in alignment with the previously obtained engineering estimate and substantiates the experimental identification of the fundamental vertical oscillation frequency of the base-isolated building superstructure.

As illustrated in Figure 14, a concise sensitivity-based, back-of-the-envelope consistency check is employed to position the experimentally identified vertical response peak within a physically meaningful range of parameters, as opposed to serving as the principal foundation for frequency identification. Specifically, the figure illustrates how the vertical period and the corresponding vertical frequency vary when the horizontal isolation period is varied between 1 and 3 s and the vertical-to-horizontal stiffness ratio is varied between 800 and 1000. Representative values commonly adopted in practice are shown as curves in the figures, which demonstrate that the resulting vertical characteristic frequency falls in the order of tens of hertz. For instance, for a horizontal period of 2 s and a stiffness ratio of 800, the vertical period is on the order of 0.07 s, corresponding to a vertical frequency of about 14 Hz. This is consistent with the experimentally observed peak identified from the vertical transfer function of the base-isolated building superstructure. This simplified reasoning implicitly relies on the assumption that the effective participating mass associated with the isolation-dominated horizontal response is comparable, in order of magnitude, to that governing the vertical response of the isolated configuration. Such an assumption is plausible in the context of seismic base isolation because the isolation strategy is intended to produce a dynamically distinct response of the isolated system with respect to the corresponding fixed-base configuration.

From an engineering standpoint, the frequency range highlighted by Figure 14 is also relevant because it overlaps with the typical range of characteristic frequencies associated with stiff reinforced-concrete floor systems, including Predalles-type slabs. Therefore, the observed concentration of response around this band is more appropriately discussed in terms of physically plausible dynamic interaction between the superstructure vertical response and floor-system dynamics, with potential implications for perceived vibrations, comfort, and performance at operational and serviceability limit states.

Figure 15 and Figure 16 provide a concise, engineering-oriented synthesis of the frequency-domain evidence by translating the Welch spectra into energy-type indicators over the bandwidth of interest (0–20 Hz). As illustrated in both figures, the left panel reports the bandpower that has been computed by integrating the PSD of each component (X transverse, Y longitudinal, Z vertical) over 0–20 Hz, while the right panel reports the corresponding bandpower ratio (base-isolated/fixed-base). This may be interpreted in one of two ways. Firstly, the vibration energy in the isolated configuration may be said to be amplified (ratio > 1) or attenuated (ratio < 1) relative to the fixed-base twin.

As illustrated in Figure 15, the bandpower levels indicate that the base-isolated structure demonstrates comparable or reduced energy in the horizontal components. In contrast, the vertical component exhibits a significant increase in comparison with the fixed-base building. As demonstrated in the ratio plot, this component-dependent behaviour is accentuated by the fact that the horizontal ratios remain below unity, while the vertical ratio is significantly greater than one. This is coherent with the pronounced vertical spectral peak that was observed in the PSDs.

This same pattern becomes even clearer for the 23 March aftershock (see Figure 16). Despite the fact that the overall input remains negligible, the base-isolated edifice once again exhibits a reduction in horizontal vibration energy, concomitant with a substantial increase in vertical energy. The Z-component ratio attains the most substantial values among the three components. From an engineering perspective, Figure 15 and Figure 16 are of particular relevance, as they demonstrate that the vertical amplification is not limited to peak response metrics, but persists in an integrated energy sense across the frequency band under consideration. This lends further support to the hypothesis that the isolated configuration is effective in limiting horizontal demand, and that it may systematically exhibit enhanced vertical response under operational-level excitations. This is an aspect that is directly relevant to serviceability and comfort, as well as to potential interaction with floor-system dynamics in the 10–20 Hz range.

The analysis introduces the Arias intensity as an integral parameter with a view to complementing peak-based metrics (for example, maximum peak acceleration) and frequency-domain indicators (for example, PSD and bandpower). The Arias intensity is a widely adopted metric for quantifying the severity of ground shaking in terms of cumulative energy content. From an engineering perspective, the Arias intensity is particularly useful because it accounts not only for the amplitude of acceleration but also for its duration, by integrating the squared acceleration over time [50,51]. Consequently, it can be regarded as a robust descriptor for the purpose of comparing seismic demand and structural response under low-to-moderate excitations. This is especially the case in instances where differences between records are not fully captured by peak values alone.

As illustrated in Figure 17, the Arias intensity has been calculated for the top-level responses of the fixed-base and base-isolated twin buildings.

This has been achieved through separate analysis of the transverse (X), longitudinal (Y), and vertical (Z) components, as well as the aftershocks that occurred on both 19 March and 23 March. The results provide an energy-based confirmation of the trends previously observed in the time and frequency domains. In the horizontal directions (X and Y), the base-isolated structure generally exhibits a lower Arias intensity than the fixed-base structure, indicating a reduced accumulation of vibration energy in the superstructure. This approach aligns with the prescribed horizontal mitigation strategy. In contrast, the vertical component shows a distinct behavior that plays a pivotal role. The base-isolated edifice evinces considerably elevated air intensity in Z, with the discrepancy becoming distinctly more pronounced for the 23 March occurrence.

Figure 18 provides an engineering-meaningful synthesis of the observed amplification trends by reporting Base-Isolated to Fixed-Base (BI/FB) ratios obtained from both bandpower (0–20 Hz) and Arias intensity.

The results consistently indicate that the vertical response of the base-isolated configuration is markedly amplified at the superstructure level, with Z-direction ratios well above unity for both aftershocks and for both metrics, yielding an overall mean vertical amplification of approximately 5.95 across the reported cases. Conversely, the horizontal components do not exhibit amplification in the isolated configuration, as the BI/FB ratios in X and Y remain below unity on average, with mean values of about 0.65 and 0.29, respectively. Taken together, Figure 18 generalizes the key outcome that the observed response differences between the two buildings are strongly direction-dependent: the isolated configuration shows a pronounced increase in vertical cumulative demand, whereas the horizontal response is, on average, reduced relative to the fixed-base counterpart.

5. Discussion and Conclusions

The present paper proffers a preliminary, engineering-focused set of observations on the operational-level seismic response of two essentially twin buildings located a few metres apart and differing primarily in their foundation condition (fixed-base versus base-isolated). Despite the fact that the selected aftershocks (19 and 23 March 2025) were characterised by very modest ground shaking, the monitoring campaign proved to be sufficiently sensitive to highlight a clear and repeatable component-dependent behaviour. In particular, the base-isolated configuration demonstrated the anticipated diminution of horizontal response, as substantiated by peak metrics and frequency-domain indicators. However, an unexpected and systematic amplification emerged in the vertical component at the superstructure level. This vertical amplification is consistently confirmed by the BI/FB ratios summarised in Figure 18, which show Z-component values markedly above unity for both aftershocks and for both integral metrics, with an overall mean vertical amplification of approximately 5.95 across the reported cases. Conversely, the horizontal components do not demonstrate amplification in the base-isolated configuration, with mean BI/FB ratios falling below unity (approximately 0.65 in X and 0.29 in Y), indicating an average reduction in the horizontal demand relative to the fixed-base counterpart. The results obtained demonstrate a strongly direction-dependent response, characterised by a pronounced vertical demand increase at the superstructure level and a concurrent reduction in the horizontal response. These results remain consistent with the identified spectral concentration in the frequency band compatible with the fundamental vertical dynamics of the base-isolated building superstructure.

Moreover, it has been established that this concentration coincides with the standard membrane vibration frequencies of prevalent concrete floor systems. From an engineering perspective, these findings suggest that, even in circumstances where ultimate limit states under horizontal actions are not threatened, the vertical response may govern serviceability and comfort during frequent low-intensity earthquakes. Furthermore, the vertical response may contribute to adverse demand on non-structural components during medium-intensity events, especially when resonance-type interactions are possible. The study also highlights practical barriers that currently limit rigorous back-analysis, most notably the difficulty in promptly retrieving complete design and as-built documentation for existing structures. This motivation is driving ongoing efforts to reconstruct structural details and calibrate numerical models that are currently being implemented. Subsequent research will utilise these refined models, in conjunction with additional recordings, to quantify the mechanisms behind the observed vertical amplification. This will facilitate an assessment of the role of isolation-device vertical stiffness and floor-system dynamics. This will facilitate a more extensive technical and scientific discourse on the necessity of complementing current provisions in NTC 2018 and the Euro-codes with explicit design-oriented checks addressing vertical stiffness selection for base-isolation systems. The objective of these inspections would be twofold: to ensure safety, and to evaluate operational performance and occupant comfort.