Solar eclipses offer unique opportunities to understand how the ionosphere reacts when solar radiation is interrupted during daylight conditions. A “small night” generated by a total solar eclipse produces disturbances in the ionosphere, both directly by the suppression of incident radiation or induced by chemical and transport processes. As solar radiation decreases, the electron concentration of the ionospheric layers E and F1, which mainly depend on the production and loss terms, decrease considerably or even disappear throughout the eclipse. Other layers, such as the F2 layer, for which the electron concentration depends significantly on plasma transport by neutral winds and electrodynamics, react with delays and are difficult to predict (Le et al., 2009; Hoque et al., 2016; Zhang et al., 2024). Furthermore, the ionosphere's response depends on regional conditions and external factors such as space weather and lower atmospheric coupling. Additionally, changes in the virtual and real heights of ionospheric layers, such as h'F1, h'F/F2, and hmF2, have been frequently reported, generally showing an upward motion followed by a post-maximum descent as the ionosphere recovers (Le et al., 2008; Chuo, 2013; Zhang et al., 2024).

Historically, from 1920 onward, studies of solar eclipses have progressively revealed how reduced solar radiation affects the ionosphere and its coupled electrodynamic processes (Mitra et al., 1933; Ratcliffe, 1956; Rishbeth, 1968). Early radio observations established solar radiation as the main ionization source and highlighted layer-specific density and plasma transport effects (Higgs, 1942; Evans, 1965a, b). These early eclipse experiments also contributed to the development of ionospheric recombination theories and to the identification of characteristic temporal delays in ionospheric recovery processes. From 1960–2015, hundreds of studies based on observations with instruments such as Very Low Frequency (VLF) receivers, Global Navigation Satellite System (GNSS) arrays, ionosondes, riometers, incoherent scatter radars and Doppler systems have determined eclipse-induced variations in Total Electron Content (TEC), critical frequency (foF2) and height of the maximum electron density of the F2 layer (hmF2), electron temperature, and ion velocities (Cheng et al., 1992; Afraimovich et al., 2002; Jakowski et al., 2008; Le et al., 2008; Momani et al., 2010; Kumar et al., 2013; Pezzopane et al., 2015). These studies have consistently shown that during eclipses there are delayed responses, latitude-dependent effects, and evidence of acoustic gravity waves (AGWs) and traveling ionospheric disturbances (TIDs) associated with eclipse conditions (Cheng et al., 1992; Jakowski et al., 2008; Kumar et al., 2013). In several cases, eclipse-induced perturbations were found to differ substantially between the E, F1, and F2 regions, reflecting the distinct chemical and dynamical processes governing each ionospheric layer. Major events like the 2017 Great American Eclipse provided unprecedented high-resolution data, allowing detailed modeling and confirmation of earlier findings (Huba and Drob, 2017; Reinisch et al., 2018; Lei et al., 2018; Aryal et al., 2019), while recent studies emphasize the role of geomagnetic activity and AGW generation in modulating post-eclipse ionospheric dynamics. Recent studies have also emphasized the importance of combining long-term ionosonde records with modern GNSS observations to better characterize regional ionospheric responses to eclipse forcing. A detailed review of these studies is compiled in Appendix A of Bravo et al. (2020).

Early days determination of ionospheric responses during solar eclipses were made mainly from ionosonde (vertically incidence HF radar) observations. There are long time series of these observations, allowing the study of ionospheric long-term trends. Early works are, for example, those of Smith and King (1981), Bremer (1992), Ortiz de Adler et al. (1997), Jarvis et al. (1998), Foppiano et al. (1999). Later work has been reviewed by Laštovicka et al. (2017), and references therein, and recent progress reported by Laštovička (2023). One of these long-term series also offers a unique dataset to analyze both short-term eclipse-induced ionospheric variations and broader temporal trends in the South American sector (Bravo et al., 2020).

A vast historical record of ionograms is preserved on physical media, the recovery of which is essential for constructing long-term time series. This information is key to conducting long-term trend studies that contribute to a better understanding of the behavior and evolution of the regional ionosphere. Therefore, this work aims to demonstrate the scientific value of rescuing this analog material by digitizing it and correctly extracting the relevant ionospheric parameters.

The purpose of this work is to characterize the response of the Concepción (36.79° S, 73.03° W)/Chillán (36.64° S, 71.99° W) ionosphere under solar eclipse conditions, so that its response can be associated with parameters such as the maximum obscuration level or time of day in order to predict the response for future eclipses.

The diurnal variations of critical frequencies and virtual heights observed during the days in which the 16 selected eclipses occur are shown in Figs. 1 to 3. Reference curves are also shown for comparison. As already indicated, observed values are given at different time intervals as appropriate to the available ionograms for the different eclipses. Diurnal variations are arranged according to the obscuration level, regardless of the time-of-day, month, year or solar activity epoch.

The clearer ionospheric effects of eclipses are seen, as expected, during eclipses occurring just before or around local noon. The maximum level of obscuration ranges from 31 % to 92 % for these eclipses, thus allowing to determine a fairly dependence of ionospheric effects on darkness level. The critical frequencies of the F1 and E layers show a typical dip variation, while the critical frequency of the F2 layer has more complex variations, probably showing two different stages, before and after the maximum darkness level, somehow following the variations of h'F/F2. The E layer virtual height does not seem to significantly change during these eclipses.

Figure 4 presents the foE, foF1, and foF2 deviations, calculated as the difference between the minimum value observed during the eclipse interval and the corresponding reference value. The adjusted F10.7 solar flux corresponding to each event is indicated by the color scale shown in the right-hand bar. Clearer variations can be identified for foE and foF1 during specific daytime intervals, whereas the foF2 response appears less well defined. To obtain the most suitable set of differences for the linear regression analysis, only eclipse events with maximum obscuration occurring between 08:30 and 16:00 LT (75° W) were selected. These correspond to six events, listed in Table 2. When no differences are observed between the reference curve and the day of the eclipse, or due to a lack of data or other reasons mentioned below, a zero value is recorded for the frequency deviations

These six selected events (29 %) exhibited clear ionospheric signatures and sufficient data coverage during the eclipse period for the regression analysis (Fig. 5). The remaining 10 events (48 %) were not considered due to one or more of the following limitations: (1) missing observations during critical eclipse phases (particularly around maximum obscuration), (2) eclipse occurrence near sunrise or sunset when ionospheric conditions are rapidly changing, making it difficult to isolate eclipse effects from diurnal variations, (3) severe film degradation preventing reliable parameter extraction despite multiple scaling attempts, or (4) obscuration levels below the detection threshold for significant ionospheric perturbations. It should be noted that the regression analysis was performed only on critical frequencies and not on virtual heights. Changes in virtual heights (h'E, h'F1, h'F/F2) were observed to be inconsistent – sometimes increasing, sometimes decreasing – making prediction difficult. These variations are likely influenced by additional factors such as neutral winds, plasma transport, and other dynamical processes, which complicate their response to eclipse conditions.

Analyzing Fig. 5, the E-layer critical frequency (foE) exhibited a robust linear response to solar obscuration, with absolute deviations showing the strongest correlation among all analyzed parameters (r2=0.78, n=5 events, upper left panel). The regression yielded a slope -0.01 of MHz per percent obscuration and an intercept of -0.03 MHz, indicating that foE decreases nearly proportionally to eclipse magnitude. This relationship implies that a hypothetical total eclipse (100 % obscuration) would reduce foE by approximately 1.0 MHz relative to unperturbed reference conditions. The p-value =0.05 is in the limit of the conventional significance threshold (α=0.05), indicating that the relationship is statistically significant (*). In relative terms (%), the relationship is even stronger (r2=0.83, p=0.03).

The high proportion of explained variance (r2=0.78) reflects the rapid photochemical equilibrium characteristic of the E-region, where recombination timescales (20–40 min) are comparable to or shorter than typical eclipse durations. Consequently, E-layer ionization density responds almost instantaneously to variations in solar EUV flux, with minimal influence from dynamic processes such as neutral winds or plasma diffusion that complicate interpretation of higher-altitude layers.

When analysing foF1 response, in MHz values, the fit is linearly significant (r2=0.71, p=0.034), demonstrating a direct dependence on solar obscuration. However, when normalizing the data as a percentage (%), the linear relationship becomes outstanding and exhibits the most robust coupling in the entire study (r2=0.93, p=0.002). This remarkable improvement when using relative values suggests that normalization successfully eliminates the variability of the baseline (seasonal and solar cycle) conditions for each eclipse, revealing that the behavior of the F1 layer is governed almost purely by photochemical control proportional to the radiation loss.

Regarding the response of the F2 layer, there is no evidence of a linear dependence on the darkening. The r2 values are practically zero (0.02) and the p values are extremely high (0.88 and 0.85). This leads to accepting (not rejecting) the null hypothesis: the variation of ΔfoF2 during the analyzed eclipses appears to be dominated by external dynamic factors, such as thermospheric winds or plasma transport, rather than by instantaneous local photoionization.

An additional source of complexity in the observed ionospheric response arises from the varying trajectories of the analyzed solar eclipses. Specifically, eclipse paths that intersect the Equatorial Ionization Anomaly (EIA) region can significantly alter regional electrodynamics rather than just local photochemistry. Obscuration over low latitudes reduces the E-region conductivity, which subsequently weakens the equatorial E × B vertical drift and modulates the equatorial fountain effect (e.g., Bravo et al., 2020). Consequently, this disruption in the poleward plasma transport introduces further variability in the observed foF2 measurements over mid-latitude stations like Concepción/Chillán. For such events, the ionospheric depletion is a complex superimposition of local photochemical loss and altered remote dynamical transport, which may explain the high dispersion observed in the F2-layer data points across different eclipses.

Based on the established relationship, the 6 February 2027 eclipse (70 % obscuration at Chillán) is predicted to induce a foE reduction of 0.95 MHz (orange star, Fig. 5, upper left panel). This forecast will enable direct validation of the regression model and assessment of its predictive skill.

Several campaigns were conducted historically to obtain ionograms at higher temporal resolution during eclipses, with observations every 5 min for significant events. However, most of these high-cadence data were never published. We noted that cases with higher temporal resolution provided a significantly clearer depiction of the ionospheric response to the lunar shadow, whereas 1 h cadence datasets typically captured only one or two measurements during the obscuration period. This difference in temporal resolution may affect the regression analysis, as the exact moment of maximum obscuration is not always captured by the low-cadence measurements.

For the 2 July 2019 solar eclipse, there was a unique opportunity to measure the ionospheric response at two locations in Chile: Chillán (36.64° S, 71.99° W) and La Serena (29.9° S, 71.3° W). However, the ionosonde in Chillán failed during the event, so only the response at La Serena was published (Bravo et al., 2020). One of the analyzed events corresponds to the 14 December 2020 solar eclipse over Chillán, for which ionospheric responses were also studied by de Haro Barbás et al. (2022). Their results included calculations of the alpha and beta recombination coefficients, which were found to be consistent with values reported by previous authors, confirming the reliability of the ionospheric observations in this region. Furthermore, prior to the 2020 eclipse, a prediction of the ionospheric response over the Chillán station had been performed using the SUPIM-INPE model, estimating the expected variations in the different ionospheric layers during the event (Martínez-Ledesma et al., 2020). This prediction was later validated using the observed ionospheric data, as reported in Bravo et al. (2022), showing good agreement between the modeled and measured responses.

It is important to highlight the dedicated effort of ionosonde operators and technicians, particularly during the 1958–1994 period, who ensured continuous monitoring and undertook the considerable effort of recording frequent ionograms during eclipses. Recovering historical data from analog ionograms stored unprocessed for decades presents a significant methodological and technical challenge. The first obstacle lies in the state of preservation of the physical medium (typically film reels), whose natural degradation necessitates an extremely rigorous scanning process to preserve information and capture optimal contrast without damaging the medium. Subsequently, the workflow requires digital clipping to extract and isolate each ionogram from the continuous record, a task that involves meticulously correcting optical and non-linear distortions along the frequency and virtual height axes. Finally, parameter extraction is extremely complex: although scaling systems (such as DISS software) are used, internal imperfections in the old recording – such as high background noise, physical rays on the film, and diffuse traces – severely limit the effectiveness of modern pattern recognition algorithms. Consequently, the software often requires constant and thorough manual intervention by an expert operator to validate, correct, or redraw the traces, making the recovery of these time series a highly labor-intensive task.

The present study relied on a historical dataset that represents a significant rescue of scientific heritage. Many records were on obsolete 35 mm film, degraded, or even potentially flammable, and have now been digitized and standardized for analysis. Similar conditions exist at other older ionospheric stations, emphasizing the importance of preserving long-term ionospheric observations and fully exploiting their scientific value.

This work analyzed the response of the Concepción/Chillán ionosphere to six selected solar eclipses, out of a total of 21 identified events (29 %) over the period 1958–2024, using a long-term historical ionogram dataset. Critical frequencies and virtual heights were extracted from scaled ionograms, and regression analysis was performed to quantify the relationship between solar obscuration and ionospheric parameter deviations. The study demonstrates that the E and F1 layers respond nearly linearly to eclipse-induced reductions in solar radiation, while higher layers, particularly the F2 layer, exhibit more complex and variable behavior due to additional dynamical processes. High-resolution observations, when available, provided insights into short-term responses and enabled predictions for future eclipses.

Regression analysis focused exclusively on critical frequencies, as virtual heights often exhibited inconsistent behavior, reflecting the influence of neutral winds, plasma transport, and other dynamical factors that complicate their interpretation. It is important to note that only the ionospheric responses measured at Chillán during the 2 July 2019 and 14 December 2020 eclipses were published; no data from the remaining eclipse events have been published. The success of this study relied heavily on the dedication of ionosonde operators and technicians, particularly during the 1958–1994 period, who ensured frequent and reliable observations during solar eclipses. Moreover, this work represents a significant rescue of scientific heritage, digitizing and standardizing records that were previously on fragile or potentially hazardous 35 mm film. Recovering decades-old analog ionogram data presents major technical challenges across three stages: (1) carefully scanning fragile, degrading film reels; (2) digitally clipping and correcting non-linear distortions on the axes; and (3) scaling the parameters, where heavy background noise and film damage severely limit semi-automatic software (like DISS), requiring constant, labor-intensive manual corrections by experts. Preserving and exploiting these long-term datasets is crucial for advancing the understanding of ionospheric dynamics.

Predictions for the upcoming 6 February 2027 eclipse, with an expected 70 % obscuration at Chillán, indicate a foE and foF1 decrease of 0.95 MHz (%ΔfoE =28%) and 1.56 MHz (%ΔfoF1 =29%), respectively, providing a clear opportunity to validate the regression models and assess their predictive skill across different solar cycle conditions.