Monitoring Ionospheric and Atmospheric Conditions During the 2023 Kahramanmaraş Earthquake Period

Abstract

Recent advancements have led to a growing prevalence of studies examining ionospheric and atmospheric anomalies as potential precursors to earthquakes. In this context, the study involved analyzing variations in ionospheric total electron content (TEC), investigating anomalies, assessing space weather conditions, and examining changes in atmospheric parameters to evaluate potential precursors and post-seismic effects related to the Mw 7.7 and Mw 7.6 earthquakes that struck Kahramanmaraş consecutively in 2023. To compute the total electron content (TEC) values, data from 29 GNSS receivers covering a period of approximately 49 days were processed. In addition, since identical code signals were not available among all receiver stations, the study conducted an analysis of TEC estimations applying different GPS codes. To analyze space weather conditions, which are considered the main source of changes in the ionosphere, variations in sunspot number, solar activity index, magnetic activity indices (Kp and Dst), and geomagnetic field components were examined across the relevant period. To assess the potential presence of a distinct relationship between seismic activity at the Earth’s surface and ionospheric conditions, atmospheric parameters including temperature, relative humidity, and pressure were meticulously monitored and evaluated. As a result of the study, it was determined that TEC anomalies that could be evaluated as earthquake precursors independent of space weather conditions were observed starting from the 3rd day before the earthquake, and high positive TEC anomalies occurred immediately after the earthquakes. In atmospheric parameters, the change in behavior, particularly in temperature value, 10 days before the earthquake, is noteworthy.

1. Introduction

Investigating earthquake mechanisms, identifying their impacts, and detecting potential precursors are crucial steps toward developing effective prediction systems. The precursor research initiated by Leonard and Barnes’s [1] study of the relationship between the ionosphere and the M = 9.2 earthquake that struck Alaska on March 27, 1964, has since been extended to numerous other earthquakes around the world, serving as a model for understanding and modeling the connection between seismic events, the atmosphere, and ionospheric disturbances [2,3,4,5,6,7]. The key parameters of the LAIC (Lithosphere–Atmosphere–Ionosphere Coupling) model, developed by Pulinets and Ouzounov [8], begin with the emission of aerosols, particularly radon gas, from the Earth’s surface. These emissions lead to ionization in the atmospheric electric field, followed by anomalies in thermal emissions and fluctuations in atmospheric variables such as temperature, humidity, and pressure. Subsequently, these processes manifest as anomalies in the total electron content (TEC) within the ionosphere and the propagation of electromagnetic waves from the lithosphere through the atmosphere into the ionosphere. Collectively, these inputs form the LAIC model, providing a framework for understanding how these interconnected processes link seismic activity to atmospheric and ionospheric disturbances. Many studies related to the Kahramanmaraş earthquakes, which are the subject of this research, have contributed to the enhancement of the LAIC (Lithosphere–Atmosphere–Ionosphere Coupling) model in the region.

Kherani et al. [9] investigate traveling ionospheric disturbances (TIDs) associated with the 2011 Tohoku-Oki tsunami, focusing on co-tsunami TIDs (CTIDs) and ahead-of-tsunami TIDs (ATIDs). Using GNSS observations and simulations based on Navier-Stokes and hydromagnetic equations, the study examines acoustic-gravity waves (AGWs) generated by the tsunami and their effects on the ionosphere. Results reveal that ATIDs, occurring 20–60 min before the tsunami, can be detected as early warning indicators due to their long wavelengths and rapid propagation. The findings suggest that ionospheric monitoring through GNSS networks can effectively complement tsunami early warning systems. Riabova and Shalimov [10] investigate geomagnetic field variations caused by the 6 February 2023, earthquakes in Türkiye and Syria, focusing on signals from both the mainshock (magnitude 7.8) and the strongest aftershock (magnitude 7.5). Using data from ground-based magnetometers at observatories in Türkiye, Romania, Bulgaria, and Serbia, the study analyzes geomagnetic responses at distances of 700–1600 km from the epicenters. Spectral and wavelet analyses were employed to distinguish between seismic Rayleigh waves, atmospheric acoustic-gravity waves, and other geomagnetic disturbances. Key findings include the identification of geomagnetic anomalies linked to Rayleigh waves and internal gravity waves propagating from the epicenters, with periods varying over a broad range. The study highlights the importance of separating seismic-induced signals from solar and external geomagnetic influences for accurate interpretation of ionospheric responses to earthquakes. Rolland et al. [11] focus on detecting and modeling ionospheric disturbances caused by Rayleigh waves generated by major earthquakes, utilizing dense GPS networks like GEONET in Japan. It demonstrates how seismic surface waves trigger atmospheric acoustic waves, which propagate upwards and create detectable ionospheric electron density perturbations. The study combines GPS total electron content (TEC) data, spectral filtering, and numerical modeling to analyze ionospheric responses to two significant seismic events: the 2008 Wenchuan earthquake (China) and the 2003 Tokachi-Oki earthquake (Japan). Findings reveal that low satellite elevation angles and dense GPS arrays enhance the detection of these disturbances, which are influenced by seismic source characteristics, geomagnetic field effects, and observation geometry. The results provide insights into ionospheric-seismic coupling and demonstrate the potential for ionospheric data to contribute to earthquake source characterization and monitoring.

Haider et al. [12] investigated various potential precursors for the 2023 Kahramanmaraş earthquakes, including TEC fluctuations, land surface temperature, sea surface temperature, air pressure, relative humidity, outgoing longwave radiation, and air temperature, using statistical and machine learning (ML) methods. The behavior of these same parameters during 2021 and 2022 was also analyzed for comparison. Additionally, the study evaluated Dst, Ap, and Kp indices to interpret the anomalies. The results indicated that many of the examined precursor parameters exhibited abnormal fluctuations 6–7 days prior to the earthquake’s occurrence, suggesting their potential as earthquake precursors. Eroglu and Basciftci [13] utilized TEC maps obtained from the CODE (Center for Orbit Determination in Europe) analysis center and applied Fourier Transform (FT) to convert time-domain signals into the frequency domain to detect TEC anomalies related to the 2023 Kahramanmaraş earthquakes. Anomalies over the epicenter were identified through grid interpolation, and a threshold of 1.34σ was used in the sliding window method for anomaly detection. The study also investigated solar activity, geomagnetic storms, interplanetary magnetic field variations, as well as volcanic and anthropogenic effects. Anomalies occurring three days prior to the earthquake were identified as potential precursors. Cianchini et al. [14] conducted a comprehensive study into the 2023 Kahramanmaraş earthquakes and investigated several parameters as potential precursors, examining them over a long-term period. They evaluated the variations of the b value, revised accelerating moment release, and a range of atmospheric parameters (including outgoing longwave radiation (OLR), skin temperature (SKT), CO2, CO, and SO2 gases). They also assessed ionospheric critical frequencies (foF2 and foEs), electron density (Ne), magnetic field components, and electron loss data, along with their cumulative totals over time, within the framework of the earthquake preparation process. They noted that many anomalies progress from the lithosphere to the ionosphere through a series of sequential processes. However, some anomalies reacted differently. The study identified two distinct types of behavior of anomalies: one is thermodynamic, characterized by a diffusive or delayed nature, while the other is potentially electromagnetic, exhibiting an oscillating and sporadic pattern. Riabova et al. [15] investigated geomagnetic field variations and ionospheric fluctuations after the 2023 Kahramanmaraş earthquakes using ground-based magnetometers and GPS data. The results indicated that post-seismic ionospheric disturbances were observed at distances of 1200–1600 km in the lower ionosphere and 500 km in the upper ionosphere from the epicenter. These findings were evaluated in the context of seismic and atmospheric wave propagation. Vesnin et al. [16] examined ionospheric effects caused by two major earthquakes in Türkiye on 6 February 2023 (Mw 7.8 and Mw 7.5). GNSS and ionosonde data reveal circular ionospheric disturbances, with the daytime event (Mw 7.5) producing a stronger response (0.5 TECU/min) than the nighttime event (0.1 TECU/min), based on the rate-of-TEC index (ROTI). Disturbances propagated up to 750 km from the epicenters at velocities of ~2000 m/s (ROTI) and 1500–900 m/s (TEC variations). The study highlights asymmetrical propagation dominated by Rayleigh and acoustic modes, with no evidence of acoustic gravity modes. Haralambous et al. [17] explored ionospheric disturbances over Europe caused by the 6 February 2023, Türkiye earthquake, using Doppler sounding systems, ionosondes, and GNSS receivers. It identified diverse disturbances propagating via different mechanisms and at varying velocities. Beyond the typical focus on total electron content (TEC) variations, this work examined disturbances at multiple ionospheric altitudes. Notably, it highlighted “multiple-cusp signatures” in ionograms, linked to electron density irregularities caused by Rayleigh surface waves generating acoustic waves. The study demonstrated the value of multi-instrument approaches in tracing earthquake-induced ionospheric effects across altitudes and distances. Maletckii et al. [18] examined the ionospheric response to a series of major earthquakes in Türkiye and Northern Syria on 6 February 2023, using GNSS data from Türkiye, Israel, and Cyprus. The events were divided into two sequences: the first (01–02 UTC) and the second (10–11 UTC). During the first sequence, an N-shaped total electron content (TEC) disturbance was detected following the Mw 7.8 mainshock and Mw 6.7 aftershock, with a smaller response attributed to the Mw 5.6 earthquake, marking the smallest event detected using ionospheric GNSS data. Co-seismic ionospheric disturbances (CSID) propagated southwest at 750–830 m/s velocities. In the second sequence, CSID linked to the Mw 7.5 and Mw 6.0 aftershocks propagated southwest and northwest at 950–1100 m/s.

This study aims to investigate the changes in the ionosphere and atmosphere during the preparation time and after the 7.7 and 7.6 magnitude earthquakes that occurred in Kahramanmaraş, Türkiye in 2023. GPS data from 29 stations were utilized to analyze ionospheric TEC variations, and an evaluation of TEC calculations using different GPS codes was performed. To interpret anomalies in TEC values, parameters related to space weather conditions, including sunspot number (R), solar activity index (F10.7), magnetic storm and activity indices (Kp and Dst), and geomagnetic field components (Bx, By, and Bz), were assessed. Furthermore, to evaluate changes in atmospheric conditions in the region, data from the LTAU sounding station, located near the earthquake epicenter, were analyzed, including atmospheric pressure, temperature, and relative humidity information, with data from 2023 as well as the years 2020, 2021, and 2022 included in the analysis. Section 2 provides an overview of the study area, details the methodologies employed, and describes the parameters used. Section 3 presents the results of the study, including VTEC variation and anomaly graphs, space weather conditions, and changes in atmospheric parameters with their corresponding analyses. Finally, Section 4 discusses the conclusions of the study and outlines recommendations and future research objectives.

2. Materials and Methods

Kahramanmaraş province and its surroundings, which are under the influence of the Dead Sea and Eastern Anatolia faults, have long been recognized as high-risk zones since the region is continuously accumulating energy, and there are seismic gaps in the active faults that influence it [19,20]. On 6 February 2023, two earthquakes struck Kahramanmaraş/Türkiye, nine hours apart, with moment magnitudes of 7.7 and 7.6, respectively [21].

Table 1. Earthquake information.

Location of the Earthquake	Earthquake Date and Time	Moment Magnitude (Mw)	Depth of Focus (km)
Pazarcık, 37.288° N 37.043° E	06.02.2023, 01:17:34 UTC	7.7	8.6
Elbistan, 38.089° N 37.239° E	06.02.2023, 10:24:47 UTC	7.6	7.0

provides detailed information regarding the earthquakes. The earthquake preparation area for the earthquake Mw 7.7 was found to be 2046.445 km using the Dobrovolsky formula ρ = 10^0.43*Mw [22]. In the study, data from 29 GNSS stations, including 6 IGS and 23 CORS-TR, within this earthquake preparation area were used. Figure 1 demonstrates the study area, delineating the earthquake preparation zone and the GNSS stations, while Figure 2 provides a detailed view of the CORS-TR stations.

2.1. TEC Calculation from GPS Signals

TEC values are calculated by the slowing/accelerating effects of the total electron content of the ionospheric layer on GPS signals. When the differences are taken using the dual-frequency signal characteristics of the receivers, creating geometry-free combinations $\left(P_{4,r}^s\right)$ will eliminate all frequency-independent error sources. The STEC value of the ionospheric pierce point between receiver and satellite (IPP) can be computed by inserting the values of the remaining frequency-dependent effects into the equation. VTEC values were obtained from STEC values using the Modified Single Layer Model (MSLM) mapping function M(z) in this study, and 2nd order polynomial interpolation was used to calculate the VTEC value in the zenith direction of the receiver. The polynomial coefficients were determined using the least squares approach. Comprehensive details on this method and its procedural steps can be found in [23].

The procedural steps of calculating VTEC from GPS observations are provided in Equations (1)–(5).

$${\overline{P}}_F={\rho }_F+c\left(\Delta t_r-\Delta t^s\right)+T+I_F+c\left(DC{B}^s+DC{B}_r\right)+\alpha$$

(1)

where ${\overline{P}}_F$ represents the code observation for frequencies F = 1 and F = 2, ${\rho }_F$ denotes the geometric distance between the receiver and satellite, c is the speed of light in a vacuum, $\Delta t_r$ and $\Delta t^s$ are receiver and satellite clock errors, $T$ represents the tropospheric effect, $I_F$ is the ionospheric effect, DCB^s and DCB_r are differential code biases of the satellite and receiver, respectively. The relevant DCB values that are systematic errors and occur due to hardware delays in satellites and receivers have been obtained from the daily ionosphere files of the CODE analysis center. DCB values not found in the ionosphere files have been calculated with the help of M_DCB software developed by Jin et al. [24]. $\alpha$ indicates noise of the signal. Code observations in GPS are subject to signal reflections and noise, resulting in lower accuracy compared to phase observations. Using phase observations instead of code observations for TEC estimation allows for the computation of highly accurate TEC values. However, due to the complexity of resolving phase ambiguities, they are not typically preferred. Instead, a smoothed code observation approach, which combines phase observations with code observations, is employed to achieve higher accuracy than using code observations alone for TEC calculations.

$${\tilde{P}}_F\left(t\right)={\varphi }_F\left(t\right)+{\overline{P}}_F-{\overline{\varphi }}_F+2{\displaystyle \frac{f_2^2}{f_1^2-f_2^2}}\left(\left({\varphi }_1\left(t\right)-{\overline{\varphi }}_1\right)-\left({\varphi }_2\left(t\right)-{\overline{\varphi }}_2\right)\right)$$

(2)

where ${\tilde{P}}_F\left(t\right)$ is the smoothed code observation (epoch t and frequency F = 1, 2) [25]. For the step of smoothing the code observations, the RNXSMT command was applied in the Bernese v5.4 program. The objective is to improve the accuracy of TEC values by smoothing the code observations (${\overline{P}}_F$) with phase observations ${(\overline{\varphi }}_F)$.

$$P_{4,r}^s={\tilde{P}}_1-{\tilde{P}}_2=40.3\left(1/\left(f_1^2\right)-1/\left(f_2^2\right)\right)STE{C}_r^s+DC{B}^s+DC{B}_r$$

(3)

By rearranging Equation (3) to isolate the STEC variable and incorporating smoothed code observations along with frequency-dependent error sources, the STEC equation transforms into Equation (4).

$$STE{C}_r^s={\displaystyle \frac{1}{40.3}}\left({\displaystyle \frac{f_1^2{f}_2^2}{f_2^2-f_1^2}}\right)\left(P_{4,r}^s-c\left(DC{B}^s+DC{B}_r\right)\right)$$

(4)

After calculating the STEC values at each IPP point, these values are first converted to VTEC values. This conversion employs the mapping function $M\left(z\right)$ from the modified single-layer model (MSLM), as suggested in the IONEX files provided by the CODE analysis center, as shown in Equation (5).

$$VTEC={\displaystyle \frac{STEC}{M\left(z\right)}}, M\left(z\right)={\displaystyle \frac{1}{\sqrt{1-{\mathit{sin}}^2{z}^{\prime }}}}, z^{\prime }={\mathit{sin}}^{-1}\left({\displaystyle \frac{Re}{Re+H}}\mathit{sin}\left({\alpha }_{SLM}z\right)\right)$$

(5)

Re represents the Earth’s radius as 6371 km, H denotes the orthometric height of the single-layer ionospheric model (450 km, as determined by the CODE analysis center), ${\alpha }_{SLM}$ indicates the geocentric angle in the single-layer ionospheric model, and z refers to the zenith angle of the relevant satellite.

Once the VTEC values for all IPP points are determined, the zenith direction VTEC values at the receiver for each epoch are computed in this study using a second-order polynomial interpolation method with Equation (6). ${\phi }_{IPP}, {\lambda }_{IPP}$ refer to the latitude and longitude, respectively, of the ionosphere piercing point in the solar-fixed reference system, and $a_i$ represents the coefficients of the polynomial surface.

$$TEC\left({\phi }_{IPP},{\lambda }_{IPP}\right)=a_0+a_1{\phi }_{IPP}+a_2{\lambda }_{IPP}+a_3{\phi }_{IPP}^2+a_4{\phi }_{IPP}{\lambda }_{IPP}+a_5{\lambda }_{IPP}^2$$

(6)

In GPS technology, it is known that P code observations have a higher resolution than C code observations. However, it was observed that the GPS code observations were not the same across all stations within the study area. For the majority of stations, TEC values were calculated based on C1 and P2 code observations, while P1-P2 code observations were only applicable to RAMO and BHR4 stations. To assess the differences, the daily TEC values in the zenith direction for four sample stations, derived from different code observations and using different sources of DCB values, were compared with those from the CODE-IONEX file, and the RMS (root mean square) errors (m₀) were computed. The results of these comparisons are presented in

Table 2. Accuracies of TEC values calculated based on different code observations.

Station Name	Code Information Used	Source of DCB Values	m0 (TECU)
RAMO	Raw P1-P2	CODE-IONEX P1-P2	6.26
	Raw C1-P2	M_DCB C1-P2	10.04
	Raw C1-P2	CODE-IONEX P1-P2	6.71
	Smoothed P1-P2	CODE-IONEX P1-P2	3.89
	Smoothed P1-P2	M_DCB P1-P2	6.27
BSHM	Raw P1-P2	CODE-IONEX P1-P2	5.77
	Raw C1-P2	M_DCB C1-P2	7.10
	Raw C1-P2	CODE-IONEX P1-P2	5.28
	Smoothed C1-P2	M_DCB C1-P2	5.10
ADN2	Raw C1-P2	M_DCB C1-P2	7.11
	Smoothed C1-P2	M_DCB C1-P2	5.32
ONIY	Raw C1-P2	M_DCB C1-P2	6.07
	Smoothed C1-P2	M_DCB C1-P2	4.34

.

As shown in Table 1, the GPS-TEC values obtained from smoothed code observations are more reliable than those derived from raw code observations, and the use of P1-P2 codes yields more accurate results compared to the use of C1-P2 codes. In evaluating the effects of different sources of DCB values on TEC calculations, it was observed that the root mean square error is smaller when using the DCB values from the IONEX files of the CODE analysis center. Because the reference values are also the TEC values from the CODE analysis center, this outcome is expected.

In order to identify potential anomalies in the ionosphere in the zenith direction of the station, the moving median method on the interquartile range approach was employed: Equation (7) provides the boundary values (Upper Boundary, UB, and Lower Boundary, LB). After the limit boundaries determined by Liu et al. [26] were put into the equation, a new TEC value between LB and UB has a probability of almost 65%.

$$\begin{array}{c}UB=M+1.5\left(UQ-M\right)\\ LB=M-1.5\left(M-LQ\right)\end{array}$$

(7)

Here, UB denotes the upper bound, M represents the median value, LB indicates the lower bound, and UQ and LQ refer to the upper and lower quartiles, respectively. TEC levels that exceed or fall below the upper or lower limits are identified as anomalies. The study aimed to calculate a 49-day TEC for 29 stations, although some days had data gaps. The first 15 days of a 49-day station data were utilized as input for the moving median method using the interquartile range approach. This led to the calculation of the boundary values for the 16th day, which in turn yielded the first day of the anomaly graphs.

Table 3. Station information and metrics of processed days of stations.

Station Name	Network	Number of Days	Minimum m0 (TECU)	Maximum m0 (TECU)	Median m0 (TECU)
ARUC	IGS	49	2.46	8.40	3.26
BHR4	IGS	48	3.37	11.01	5.61
BSHM	IGS	48	14.36	17.83	16.34
RAMO	IGS	49	3.57	6.38	4.45
TUBI	IGS	49	2.73	122.03	8.29
ZECK	IGS	49	11.48	34.80	13.69
ADN2	CORS-TR	49	3.03	9.05	5.06
ADY1	CORS-TR	33	1.51	7.76	3.16
AKLE	CORS-TR	46	1.90	92.43	9.57
ANTE	CORS-TR	48	2.7	10.19	3.69
EKZ1	CORS-TR	43	2.99	30.25	4.38
ELAZ	CORS-TR	49	2.56	13.03	3.67
ERGN	CORS-TR	49	3.38	70.18	7.44
FEEK	CORS-TR	48	2.74	167.29	4.64
GURU	CORS-TR	48	2.73	23.54	4.27
HAT2	CORS-TR	36	1.04	8.00	4.04
KAY1	CORS-TR	49	2.42	23.78	3.93
KLS1	CORS-TR	49	5.75	36.21	8.24
MAR1	CORS-TR	47	2.84	103.4	4.28
MGOS	CORS-TR	49	2.47	7.45	4.05
MLY1	CORS-TR	44	2.57	15.58	4.36
MRSI	CORS-TR	49	0.93	23.03	5.63
NIGD	CORS-TR	49	2.52	9.54	4.48
ONIY	CORS-TR	49	0.52	11.09	3.90
POZA	CORS-TR	49	2.86	15.81	4.19
SILF	CORS-TR	49	2.83	9.70	5.56
SIV1	CORS-TR	47	2.08	10.90	4.34
TUF1	CORS-TR	49	2.92	54.83	3.75
VIR2	CORS-TR	46	2.42	18.27	3.85

presents the stations utilized in the study along with the statistical metrics of the accuracies of the calculated TEC values, using the relevant CODE-IONEX files as a reference.

When examining Table 3, it is notable that the maximum m₀ values at certain stations, such as MAR1 and EKZ1, are relatively high. However, the small median values indicate that these high maximum values are likely due to abrupt peaks. Additionally, stations like BSHM and ZECK, which exhibit high median m₀ values, are identified. The TEC values calculated at these stations, which have lower accuracy, also affect the accuracy of the calculated threshold values (LB and UB). Therefore, in such studies, the accuracy of the TEC values should be considered when interpreting the TEC graphs produced for the relevant stations.

2.2. Space Weather Conditions

The primary source of ionospheric fluctuations is variations in space weather conditions induced by the Sun’s impact. Today, since the precise timing and magnitude of the Sun’s impacts on ionospheric TEC values cannot be differentiated, anomalies during periods of calm space weather conditions are assessed to explore a potential connection between TEC variations and earthquakes. For this purpose, the F10.7 index, which measures the amount of solar flux on the 2800 MHz frequency band at 10.7 cm wavelength, and sunspot number were used to assess solar activity. To evaluate geomagnetic field activities, variations in the Bx, By, and Bz geomagnetic field components, as well as geomagnetic storm (Kp) and geomagnetic activity index (Dst), were examined. These data were obtained from the National Aeronautics and Space Administration’s Goddard Space Flight Center (NASA/GSFC) website [27].

2.3. Atmospheric Data

The interaction between earthquakes and the ionosphere occurs through atmospheric processes. Therefore, to investigate potential anomalies, a comprehensive analysis was conducted on various atmospheric parameters (atmospheric pressure, temperature, relative humidity) utilizing meteorological data provided by the University of Wyoming [28]. To provide a baseline for normal conditions during non-earthquake periods and make comparisons with the 2023 earthquake period, data from 2022, 2021, and 2020 were also included. These parameters were recorded twice daily at 00:00 and 12:00 UTC, with measurements indicating an altitude of 1096 m for the year 2023 and 1094 m for the years 2022, 2021, and 2020. The data were obtained from the LTAU sounding station, located near the earthquake epicenters (Figure 2).

3. Results

Fluctuations and anomalies in the calculated VTEC values are illustrated as time series in Figure 3, Figure 4, Figure 5 and Figure 6. The graphs presented here represent the data from the four stations, which are EKZ1, MAR1, ANTE, and ONIY, nearest to the earthquake epicenters. Graphs of the remaining stations and analysis of them were provided in Appendix A (Figure A1–Figure A4). Due to the lack of data at some stations, gaps in the graphs and short period time series were formed.

When examining Figure 3, Figure 4, Figure 5 and Figure 6, it becomes apparent that similar anomaly trends, with minor variations, are observed across all four stations. At all four stations, positive anomalies were observed between the 16th and 25th GPS days. Notably, on the 18th GPS day, intense positive anomalies reaching approximately 10 TECU were particularly striking. These anomalies were followed by negative anomalies that persisted until around the 33rd or 34th GPS day. Notably, the negative anomalies showed a consistent pattern approximately between the 27th and 33rd GPS days. After the negative anomalies, positive anomalies began appearing periodically from the 34th to 37th GPS day, the day of the earthquake, with small quantities of approximately 4–5 TECU. On the day of the earthquake and especially for 4 days afterward in the EKZ1, ANTE, and ONIY stations, very high and intense positive anomalies, reaching up to 20 TECU, were observed. At the MAR1 station, this period of heightened anomalies persisted until the 43rd GPS day. The increase in TEC anomalies observed after the earthquakes can be associated with the effects of crustal fractures and slip processes in the lithosphere, which influence atmospheric electric fields and enhance ionospheric charge transport. Additionally, heat generation and gas emissions (e.g., radon gas) from the Earth’s crust may alter the ionization levels in the atmosphere. The propagation of seismic and electromagnetic waves generated by the earthquakes into the atmosphere can trigger acoustic-gravity waves, leading to variations in plasma density within the ionosphere. Collectively, these processes contribute to the observed increase in TEC anomalies. These processes, as described in the LAIC model proposed by Pulinets and Ouzounov [8], can be observed as a result of the lithospheric disruptions caused by the earthquake. At the MAR1 station, no data were available from the GNSS receiver for the 38th and 39th GPS days following the earthquake. The intense positive anomalies subsequently diminished, giving way to smaller and less frequent positive anomalies. Additionally, persistent positive anomalies observed at the MAR1, ANTE, and ONIY stations during the late 46th GPS day and early 47th GPS day were also noteworthy. In contrast, data from the EKZ1 station could only be processed up to the 44th GPS day.

For the analysis of space weather conditions, graphs of sunspot number (R), solar flux index (F10.7), geomagnetic storm (Kp) and activity (Dst), and geomagnetic field indices (Bx, By, Bz) on the relevant days were obtained and are demonstrated in Figure 7.

It is observed that the solar activity index is high between the 16th and 26th GPS days. Since solar activity is considered the primary cause of ionospheric alterations, it is unclear whether the earthquake had any effect on the anomalies during this period. The subsequent fall in solar activity, followed by a rise as the earthquake date approaches, is quite similar to the trend in the earthquake anomaly graphs. However, despite the positive anomaly values observed 2–3 days before the earthquake, solar activity remained low during that period. Therefore, it can be inferred that the anomaly changes detected on the 34th–36th GPS days are unlikely to be caused by solar activity. During these days, no observations reached moderate or high activity levels in the geomagnetic storm and geomagnetic activity indices. However, data gaps in the geomagnetic field indices are noticeable on the 34th to 36th GPS days. Solar activity increased in the days following the earthquake until the 40th GPS day, after which it began to decrease, but it remained at a high activity level until the 50th GPS day. The geomagnetic storm index reached minor activity between GPS days 38 and 42, and moderate activity on the 46th and 47th GPS days. Similarly, moderate activity was observed in the geomagnetic activity index on the 46th GPS day. This effect is clearly apparent in geomagnetic field indices, especially in the By and Bz components.

To examine the relationships between TEC anomalies and space weather conditions, cross-correlation analysis graphs have been generated and provided in Figure 8. In the graphs, the horizontal axis represents the time delay in hours, while the vertical axis indicates the magnitude and direction of the relationship. The 0-h point corresponds to the simultaneous correlation between the TEC anomaly time series in the zenith direction of the respective station and the space weather index. By keeping the time series of TEC anomalies fixed, the time series of space weather conditions were sequentially shifted by +1, +2, +3, …, +24 and −1, −2, −3, …, −24 h, and the corresponding correlation values were calculated. The generation of these time-lagged graphs provides insight into how long it takes for space weather conditions to impact the ionospheric layer. However, it should be noted that space weather conditions are not the sole reasons for changes in the ionosphere. The presence of earthquake-induced ionospheric TEC anomalies, as discussed in this study, is expected to influence the calculated correlation coefficients between TEC anomalies and space weather conditions.

An analysis of the cross-correlation graphs for the stations reveals generally low correlations between TEC anomalies and space weather conditions. For all four stations, the cross-correlation graphs corresponding to the same space weather condition index exhibit similar trends despite differing correlation magnitudes. In the EKZ1, ANTE, and ONIY stations, the correlation values in the sunspot number and solar activity index graphs increase as the delay approaches +24 h, reaching approximately 0.2, indicating a positive relationship. In contrast, at the MAR1 station, correlation values also increase toward +24 h but approach only about 0.1. This suggests that the effects of solar activity on ionospheric TEC anomalies may take several hours to appear. While the rising trend in correlations is observed across all four stations, the overall correlation values remain quite low, between 0.1 and 0.2.

When examining the correlation graphs between TEC anomalies and the Kp index, the graphs for the ANTE and ONIY stations are similar, with the highest correlations occurring at delays of −9 and −7 h, respectively. For the ANTE station, the correlation value approaches 0.3, while for the ONIY station, it slightly exceeds 0.3. This suggests that the Kp index may be influenced by the occurrence of TEC anomalies. Similarly, the EKZ1 and MAR1 stations show the highest correlations at delays of −3 and −1 h, respectively, with correlation values of approximately 0.2 and 0.1.

For the Dst index, most time-delayed correlations across all four stations show a negative relationship. The Dst time series and TEC anomalies exhibit inverse relationships at delays of +15 and +16 h, with correlation values exceeding −0.2 for the EKZ1, ANTE, and ONIY stations. A similar inverse relationship is observed for the MAR1 station, where the highest correlation, at −18 h, is approximately −0.1.

Examining the correlation graphs for the Bx, By, and Bz indices, the graphs for the EKZ1, ANTE, and ONIY stations are quite similar, showing the highest positive correlations at delays of −5, −6, and +14 h, with correlation values ranging between 0.1 and 0.2. The occurrence of higher correlation values at both −5, −6, and +14 h delays raises ambiguity about whether these indices influence the ionospheric anomalies or if the ionospheric anomalies first emerge and subsequently affect these indices. At the MAR1 station, however, the correlation values between TEC anomalies and the Bx, By, and Bz indices remain close to zero for nearly all delays. This indicates that no significant relationship exists between the TEC anomalies occurring in the zenith direction of the MAR1 station and these indices.

The time series of atmospheric parameters were acquired from the LTAU sounding station and are illustrated in Figure 9. When Figure 9 is examined, it is seen that the atmospheric pressure values in 2023 are fluctuating as in previous years. However, sharp changes in the fluctuation in 2023 are striking. The notable drop in pressure values, followed by a sudden increase after the earthquakes, suggests that this variation may be related to seismic activity. In the temperature graph, it is observed that, unlike previous years, the day-night temperature difference in 2023 remained around 10 °C for an extended period. However, starting from the 27th GPS day, this difference suddenly and significantly decreased, indicating a notable change in trend. This pattern persisted for two days following the earthquake. Although it is not very noticeable in the relative humidity values, the amount of the day-night change difference altered and lost its pattern after the 28th GPS day for this parameter as well.

To examine the relationships between TEC anomalies and atmospheric parameters, cross-correlation analysis graphs have been generated and provided in Figure 10. Since the data interval of atmospheric parameters is 12 h, TEC anomaly values with interval of same 12 h are used to create cross-correlation graphs.

Upon examining Figure 10, the pressure parameter reveals notable trend similarities in the correlation graphs for the EKZ1, ANTE, and ONIY stations. When considering a lag of +5 (equivalent to 60 h), the correlations between the pressure parameter and TEC anomalies exhibit the highest positive relationships within their respective graphs. Specifically, the correlation values are approximately 0.5 for the EKZ1 station, 0.4 for the ANTE station, and 0.3 for the ONIY station. In contrast, at the MAR1 station, while the directions of the correlations at different lags are similar to those observed for the EKZ1, ANTE, and ONIY stations, the magnitudes differ. At this station, the highest positive correlation of 0.55 is observed at a lag of +1 unit (12 h).

For the temperature parameter, the graphs, particularly for the ANTE and ONIY stations, demonstrate similar correlation trends across the lagged time series. The highest correlations for the EKZ1 station are observed at a lag of +3 units, showing a negative correlation of −0.3, while the MAR1 station exhibits a similar negative correlation of approximately −0.3 at a lag of +1 unit. For the ANTE and ONIY stations, the correlation trends across all lagged time series are notably consistent. The strongest correlations for these two stations are observed at a lag of +7 units, with approximately −0.4 negative correlations. It is noteworthy that, for this atmospheric parameter, the correlations alternate sequentially between positive and negative across different lagged times. This can be attributed to the periodic behavior of temperature values resulting from the diurnal variations between day and night.

In the relative humidity graphs, similar to those of the temperature parameter, the lagged correlation values alternate sequentially between negative and positive. The highest correlations are observed as follows: for the EKZ1 station, the strongest negative correlation of −0.25 occurs at a lag of +6 units; for the MAR1 station, the strongest positive correlation of approximately 0.25 is observed at a lag of −5 units; for the ANTE station, a positive correlation of approximately 0.3 is observed at a lag of −3 units; and finally, for the ONIY station, the strongest positive correlation of approximately 0.3 is observed at a lag of +1 unit.

The cross-correlation graphs between TEC anomalies and space weather conditions, as well as atmospheric parameters, provide insights into the strength, direction of correlation, and time lag at which the relationships appear most prominent. The correlations between TEC anomalies and space weather conditions generally fall within the range of −0.1 to +0.1 or −0.2 to +0.2. Only the Kp index shows a correlation value reaching up to 0.3. On the other hand, the cross-correlation graphs for atmospheric parameters predominantly fall within the ranges of −0.3 to +0.3 or −0.4 to +0.4. Notably, for the pressure parameter, a correlation value approaching 0.6 was observed at the MAR1 station. These findings suggest that the relationships between TEC anomalies and atmospheric parameters used in this study are stronger than those between TEC anomalies and space weather conditions.

The analysis of TEC anomaly graphs, following the examination of space weather conditions and atmospheric parameters, demonstrates that the positive anomalies observed on GPS days 34, 35, and 36 may be associated with earthquake-related processes. Similar to the findings of Eroglu and Basciftci [10] in their study on the same earthquake, this conclusion was reached after accounting for and removing the effects of solar activity and magnetic anomalies. Additionally, according to the LAIC model described by Pulinets and Ouzounov [8], the release of radon from underground sources is suggested to lead to a significant increase in surface temperature. It has been stated that the temperature difference between regions distant from faults and the earthquake-affected area causes horizontal air movements and air mixing, resulting in a temperature rise throughout the entire earthquake preparation zone. The attachment of water vapor to ions reduces the amount of free water vapor in the air, which is why the observed temperature increase is often accompanied by a decrease in relative humidity. Rising warm air is expected to cause a drop in atmospheric pressure. However, these expected observations were not distinctly evident in the atmospheric parameters analyzed in this study. Only a sharp drop in atmospheric pressure was noticeable during the three days preceding the earthquake. The measurement altitude, being considerably distant from the Earth’s surface (1096 m above sea level in 2023), is considered a potential factor that may have attenuated or masked the observable impacts of variations in these parameters, thereby limiting the detection of more pronounced effects that could be associated with the earthquake preparation process. It has been stated that Rayleigh waves generated at the surface during the earthquake induce atmospheric oscillations, as observed in ground pressure measurements, and that these oscillations can be linked to the propagation of atmospheric waves from the lower atmosphere to the ionosphere as a mechanism for ionospheric perturbations figure [11,29]. The co-seismic and post-seismic changes observed in atmospheric parameters and TEC variations in this study could also be associated with these atmospheric waves.

4. Conclusions

The absence of P1 code in most station data in the study necessitated the completion of the study with C1-P2 codes. Consequently, this study first investigated the accuracies of TEC values obtained with different code measurements (P1-P2 or C1-P2), and smoothed or raw code measurements were investigated. It was observed that TEC values obtained from P1-P2 code observations demonstrated higher accuracy compared to those derived from C1-P2 values. Additionally, TEC values obtained using smoothed codes, in both P1-P2 and C1-P2 combinations, exhibited greater accuracy than those derived from raw code measurements. The use of different sources for DCB information is also a factor affecting the results. The trends in ionospheric TEC anomaly graphs were found to be similar across multiple stations, including both CORS-TR and IGS networks. After eliminating the days influenced by space weather conditions, positive anomalies observed on the 34th–36th GPS days could be interpreted as potential earthquake precursors. The intense and high number of positive anomalies that occur immediately after the earthquakes, reaching up to 20 TECU, are quite striking. In atmospheric parameters, especially in temperature values, it was observed that the trend of day-night differences deteriorated after the 27th GPS day, 10 days before the earthquake. Although it is not very noticeable in the relative humidity value when compared to previous years, it can be said that there was a change in its behavior after the 27th day in this parameter as well. The atmospheric pressure values exhibited a fluctuating behavior similar to the years 2020–2022, but the sudden drop starting on the 34th GPS day as the earthquake approached, followed by an abrupt increase after the earthquakes, is particularly noteworthy.