Upgrading a Low-Cost Seismograph for Monitoring Local Seismicity

Abstract

The use of a dense network of commercial high-cost seismographs for earthquake monitoring is often financially unfeasible. A viable alternative to address this limitation is the development of a network of low-cost seismographs capable of monitoring local seismic events with a precision comparable to that of high-cost instruments within a specified distance from the epicenter. The primary aim of this study is to compare the performance of an advanced, contemporary low-cost seismograph with that of a commercial, high-cost seismograph. The proposed system is enhanced through the integration of a 24-bit analog-to-digital converter board and an optimized architecture for a low-noise signal amplifier employing active components for seismic signal detection. To calibrate and assess the performance of the low-cost seismograph, an installation was deployed in a region of high seismic activity in Evgiros, Lefkada Island, Greece. The low-cost system was co-located with a high-resolution 24-bit commercial digitizer, equipped with a broadband (30 s—50 Hz) seismometer. An uninterrupted dataset was collected from the low-cost system over a period of more than two years, encompassing 60 local events with magnitudes ranging from 0.9 to 3.2, epicentral distances from 5.71 km to 23.45 km, and focal depths from 1.83 km to 19.69 km. Preliminary findings demonstrate a significant improvement in the accuracy of earthquake magnitude estimation compared to the initial configuration of the low-cost seismograph. Specifically, the proposed system achieved a mean error of ±0.087 when benchmarked against the data collected by the high-cost commercial seismograph. These results underscore the potential of low-cost seismographs to serve as an effective and financially accessible solution for local seismic monitoring.

1. Introduction

Monitoring local seismicity with high accuracy requires the integration and deployment of advanced technologies and methodologies. A dense network of high-sensitivity seismometers enables the precise detection of earthquakes within local and regional seismic monitoring areas [1,2,3]. Seismometers are broadly categorized into two types: velocimeters and accelerometers [4,5], each designed with distinct bandwidths and detection techniques to measure ground motion. Furthermore, as highlighted in [6], the amplification of ground vibrations can significantly influence the sensitivity and accuracy of seismic measurements, particularly in rocky environments. These factors are critical when deploying low-cost systems in such terrains during earthquake monitoring. Seismic sensors quantify ground motion by measuring both velocity and acceleration, converting these measurements into electrical signals for analysis [7]. In contrast to standard 4.5 Hz, 380 ohm geophones, broadband seismometers feature a wider frequency passband, ranging from 0.001 Hz to 500 Hz. This broader range provides the capability to capture a more extensive frequency spectrum, yielding higher-resolution information about seismic waves and their characteristics.

Numerous studies on seismological networks have focused on monitoring seismic activity using high-cost seismometers and broadband sensors distributed across large areas. While dense networks equipped with high-cost broadband sensors provide highly accurate seismic data, their implementation costs are typically affordable only for specific seismic projects and over limited time periods [8,9,10,11]. Transitioning to dense seismological networks composed of low-cost systems offers a more sustainable and cost-effective solution, enabling continuous monitoring over extended periods within the area of interest.

In recent years, significant efforts have been directed toward minimizing the cost of datalogger systems by leveraging state-of-the-art electronic components. These include microcomputer boards, low-noise, high-precision 24-bit analog-to-digital conversion boards, low-noise signal preamplifiers, as well as low-cost seismic velocity sensors (short-period geophones) and low-cost accelerometer sensors (MEMS accelerometers). Many researchers have adopted this approach, developing their own low-cost systems. For instance, starting with [12], a software component was created to convert analog seismic waves into a digital format utilizing a computer sound card. Additionally, [13,14] introduced a simple data-gathering method to capture seismic data, initially sensed by a vertical geophone and later by a three-component low-cost seismic recorder.

In the studies conducted by [5,15], a low-cost data acquisition and analysis system designed for microtremor measurements was proposed. Additionally, several other studies have focused on utilizing MEMS technology [16] for recording accelerations, among other applications [17,18,19]. For example, [20,21] introduced a low-cost accelerograph that utilizes the Atmel (ATmega328P) microcontroller and an acceleration sensor. The primary objective of this device is to accurately record the distribution of intense ground motion in metropolitan areas during seismic sequences. Furthermore, [22] assessed the effectiveness of a cost-efficient network comprising private Raspberry Shake (RS) seismographs equipped with vertical geophones to capture ground movements caused by induced microseismicity. Similar systems, such as those in [23], were used to monitor seismic activity in Haiti. Additionally, [24,25,26] successfully presented their own low-cost seismic recording systems. In a previous study [27], an earlier version of the proposed system hardware was developed to record local seismicity, with a magnitude estimation deviation of ±0.2 to ±0.4 compared to a broadband conventional seismometer.

The current study focuses on the successful completion of the hardware implementation and builds upon the initial results from the prior version presented in [27,28]. The objective is to propose an updated and more innovative version of the low-cost system. To investigate local seismicity, records of low-magnitude earthquakes within a radius of 5.71 km to 23.45 km were analyzed. The study begins by describing and simulating the fundamental hardware components of the proposed seismograph, followed by outlining the specifications of the sensors and the general capabilities of the system. Furthermore, details regarding the research area and the methodology employed are provided. Subsequently, a seismic catalog comprising 60 representative and typical events was compiled, and signal processing analysis was conducted. Ultimately, the performance efficiency of the proposed low-cost seismograph was evaluated by comparing the results with those obtained from a high-cost broadband conventional seismometer (Figure 1).

The methodology involved analyzing amplitude measurements in conjunction with magnitude calculations [29] developed a robust method for calculating local earthquake magnitudes by incorporating zero-to-peak amplitude measurements and distance correction factors. This method was chosen for its high accuracy in deriving local magnitudes, particularly for near-field seismic events. Its key advantage lies in its ability to minimize systematic errors caused by variations in local geological conditions, making it especially suitable for comparing data from low-cost and high-cost systems.

The primary approach involved the accurate measurement of the zero-to-peak amplitude for each of the sixty analyzed seismic events recorded by the low-cost system. The primary focus was to determine whether the low-cost system could reliably estimate the magnitude of seismic events using a derived mathematical equation described in [29], while also considering the deviation from the reference magnitude values obtained from the high-cost seismograph. To address this, two cases were investigated, examining the system’s efficiency for both epicentral and hypocentral distances of the recorded events. The first case analyzed the zero-to-peak amplitude for epicentral distances, and the second case examined the zero-to-peak amplitude for hypocentral distances. Magnitudes for the low-cost system were calculated for both approaches.

Given that the sixty analyzed seismic events occurred at close distances and shallow focal depths relative to the installation point, analyzing both epicentral and hypocentral distances was deemed essential to enhance the accuracy of the results. Using the relation for reliable local magnitude calculation proposed by [29], the magnitudes derived from the low-cost system’s data were compared with those obtained from the high-cost seismograph operated by the Aristotle University of Thessaloniki (EVGI). This comparison allowed for the determination of magnitude performance for both systems (low-cost and high-cost) across epicentral and hypocentral distances. Finally, a new catalog was compiled, incorporating the results of the comparison between the two systems.

2. Specifications of an Upgraded Low-Cost Seismograph

The primary sensor in the recorder is a low-cost velocity sensor with a linear response range from 4.5 Hz to 200 Hz and a sensitivity of 32 V/m/s in undamped mode. The sensor selected for this proposed recorder is the Geophone GS-11D, which features a linear response at 4.5 Hz and a coil resistance of 380 ohms. According to [30], this sensor is particularly effective in areas with small ray distances and for recording local earthquakes. By integrating a powerful stand-alone microcomputer with a low-noise 24-bit sigma-delta analog-to-digital converter board and combining it with a state-of-the-art operational amplifier as the preamplifier circuit, we have developed a robust, accurate, low-noise, low-cost seismic recorder. This datalogger can reliably record seismic events and magnitudes of 0.9 or greater at epicentral distances of up to 10 km and focal depths of up to 15 km from the system’s installation point.

The system operates with a maximum sampling rate of 200 Hz (200 samples per second), with a timestamp recorded for each data measurement. To ensure precise and accurate timestamps in the system’s records, a real-time clock (RTC) board is connected to the microcomputer. This board is synchronized hourly using time corrections from a Global Positioning System (GPS) board and a Network Time Protocol (NTP) server. The GPS board also provides precise geolocation information for the installed system.

Internet connectivity is achieved via the microcomputer board’s built-in Wi-Fi adapter or its onboard Ethernet connector. To prevent disruptions caused by an unstable Wi-Fi connection and to ensure the reliability of the acquired data, the system exclusively utilizes the Ethernet port for Internet access. The system is powered by a battery in combination with a step-down power supply converter. A battery charger is employed to maintain the battery in an optimally charged state, ensuring reliable and continuous system operation.

2.1. Low-Cost Hardware and Software Analysis

The low-cost system is built around the popular Raspberry Pi 3 B+ microcomputer board. Since this microcomputer does not include built-in analog-to-digital input interface pins, an external 24-bit analog-to-digital converter board is used instead. For this purpose, a Waveshare High-Precision AD/DA Expansion Board for Raspberry Pi 3 B+ is integrated into the system. The ADS1256 A/D-D/A board, based on sigma-delta conversion technology, converts the analog signal from the sensor (geophone) into 24-bit resolution digital records. Additionally, a precision real-time clock (RTC) module, equipped with the DS3231 chip, ensures stable timestamping of the output data stream files, including date, time, and voltage counts.

When the low-cost system is connected to the Internet via its Ethernet port and the connection is stable, the system maintains continuous synchronization of its real-time clock (RTC). Additionally, a Network Time Protocol (NTP) server updates the RTC every 30 min to ensure precise timekeeping. In the event of an Internet connection loss, a Global Positioning System (GPS) board, specifically the NEO-6M GPS module with an external antenna, has been integrated into the system to synchronize the RTC every hour. The primary source of timestamp error in the recorded data arises from the RTC module’s time drift. The maximum recorded timestamp drift introduced by the RTC module is approximately ±2 min per year under operational temperatures ranging from −40 °C to +85 °C. This equates to a drift rate of 3.8 × 10⁻⁶ seconds per second without any time error correction. However, corrections applied by the NTP server and GPS module every 30 min and one hour, respectively, effectively minimize this error. Such a level of accuracy is considered satisfactory for recording seismic events [27,30]. Additionally, a low-noise signal preamplifier is required to transmit the seismic signal from the 4.5 Hz geophone to the A/D converter. Ensuring that weak signals are sufficiently amplified for accurate processing while maintaining a high signal-to-noise ratio.

To address the requirements for seismic applications, a custom-designed low-noise preamplifier was developed using state-of-the-art components, as specified by the manufacturer (https://www.analog.com/en/resources/analog-dialogue/articles/low-noise-and-power-daq-solution-for-seismology-and-energy-exploration-application.html, accessed on 10 October 2024). According to the internal Table 1 at page 2 of the ADA4522-2 (https://www.analog.com/media/en/technical-documentation/data-sheets/ada4522-1_4522-2_4522-4.pdf, accessed on 10 October 2024.) datasheet, the ADA4528-1 Precision, Ultralow Noise, RRIO, Zero-Drift Operational Amplifier is recommended as an equivalent component for 5-volt applications. Based on this information, the new upgraded architecture of the low-noise preamplifier is built around the ADA4528-1 (https://www.analog.com/en/products/ada4528-1.html#part-details, accessed on 10 October 2024). This preamplifier ensures the effective capture and amplification of weak seismic signals, including those from events with magnitudes below 1.0, without introducing significant noise distortion [31,32]. This capability is particularly crucial for accurate earthquake monitoring in local seismic networks, where small-magnitude events are predominant.

The upgraded architecture of the low-noise preamplifier is designed to perfectly match the specifications of the sensor. It features a low voltage offset of 2.5 µV and a low voltage offset drift of 0.015 μV/°C, along with a maximum noise level of 5.6 nV/√Hz at a frequency of 1 kHz and a peak-to-peak noise of 97 nV within a bandwidth ranging from 0.1 Hz to 10 Hz. The preamplifier provides an amplification gain exceeding 100 times and exhibits key characteristics such as a minimum open-loop gain of 130 dB, a common-mode rejection ratio (CMRR) of at least 135 dB, and a power supply rejection ratio (PSRR) of at least 130 dB [33]. To optimize the preamplifier design, simulations were conducted using LTSpice XVII (https://ez.analog.com/design-tools-and-calculators/ltspice/f/q-a/579849/download-link-of-last-ltspice-17-version accessed 15 September 2024, Analog Devices, USA) [34], an open-source simulation software, as illustrated in Figure 2. Following the simulation phase, the preamplifier was successfully developed in-house to meet the stringent requirements of seismic applications.

A high-pass filter was designed and implemented at the input of the preamplifier to eliminate dc signal spikes from the geophone, with a cutoff frequency of f_highpass = 0.5 Hz (−3db). Additionally, a low-pass filter was incorporated at the output to remove unwanted higher frequencies, with a cutoff frequency of f_lowpass = 28.5 Hz (−3db). Together, these filters create a bandpass filter with a linear frequency response ranging from 0.5 Hz to 28.5 Hz, as shown in the block diagram of Figure 2. The geophone sensor is connected to the input of the low-noise preamplifier. The geophone setup, consisting of the geophone housing, sensor, and connection cable (Figure 3a), is installed on the ground floor. The geophone is stabilized on the floor using its needle to ensure optimal sensing of vertical ground vibrations along the z-axis.

The system is powered by a 12 V/55 Ah battery pack, coupled with a step-down converter circuit board that reduces the voltage from 12 V to 5 V at a current of 3A DC. A cost-effective commercial battery charger and maintainer are used to ensure the battery remains fully charged, even in the event of a power grid failure. The entire setup is housed within an IP67-rated general-purpose plastic electrical enclosure, providing protection against environmental factors. The system is designed to operate both on-grid and off-grid. When connected to the power grid, it draws power from a 220 V/50 Hz electrical network. For off-grid operation, a simple modification involving the addition of a solar panel and a solar charge controller circuit enables autonomous functionality. The system’s field deployment is straightforward, and with the integration of a GSM 4G modem, it achieves continuous 24/7 Internet connectivity to transmit data to the server.

The system’s software is built on open-source platforms, using Raspbian Buster (32-bit) as the operating system. Python scripts were developed to program the various boards, ensuring seamless integration and enabling the system to operate continuously as a recorder with a sampling rate of 200 Hz [35]. The software generates data files (chunks) with a duration of 5 min, which are then transmitted over the Internet to the server.

Additionally, open-source virtual remote desktop software has been installed, allowing for 24/7 remote maintenance and monitoring of the system. This setup ensures the system remains functional and accessible at all times.

2.1.1. The Geophone Sensor

The primary sensor used in the system is the GS-11D geophone [36], with its characteristics presented in Figure 3b. The geophone’s output velocity versus frequency response varies based on the damping factor, which is determined by the value of the external resistor connected to its two pins. For the low-cost system setup, the damping factor is set to ζ = 50% by connecting a 4420 ohm external resistor in parallel with the geophone pins. As shown in the diagram in Figure 3b, this configuration results in an output sensor sensitivity of 29.4 V/m/s at 4.5 Hz and higher. The electrical equivalent circuit of the geophone has been modeled and analyzed in detail in [37].

2.1.2. The Custom-Made Low-Noise Preamplifier

The weak signals (voltages) generated by the geophone sensor must be amplified to ensure they are recognized, digitized, and recorded by the 24-bit high-precision sigma-delta analog-to-digital converter board. Consequently, a low-noise preamplifier circuit with high sensitivity is required. This preamplifier is designed to detect extremely low seismic signals, including shocks with magnitudes as low as M < 0.5, at epicentral distances of up to 15 km or more from the system’s installation point.

The schematic diagram of the upgraded low-noise preamplifier, used in the low-cost system (recorder), is presented in Figure 4a. The corresponding simulation output Bode diagram (Figure 4b) shows the signal-to-noise output level response for frequency ranges from 100 mHz to 1 kHz. The simulation indicates a maximum output noise response of 2.29 μV/√Hz at approximately 3.71 Hz.

The signal preamplifier is configured in differential mode, resulting in a gain determined by the ratio of the resistors, expressed as Gain = ${\displaystyle \frac{\raisebox{1ex}{$R_3$}\!\left/ \!\raisebox{-1ex}{$R_5$}\right.}{2}}$ or ${\displaystyle \frac{\raisebox{1ex}{$R_7$}\!\left/ \!\raisebox{-1ex}{$R_5$}\right.}{2}}$. Since the two inputs are wired in differential mode with a single +5V power supply, the total output amplitude is calculated as $A={\displaystyle \frac{550}{2}}=275$ within the bandwidth limits of fc₁ up to fc₂.

The high-pass filter circuit is formed by the C₃–R₆ resistor-capacitor network, with a −3 dB cutoff frequency of f_highpass = 0.5 Hz. Similarly, the C₄–R₅ network exhibits the same high-pass behavior. The low-pass filter of the system is created by the C₅–R₃ and C₆–R₇ resistor-capacitor networks, with a -3 dB cutoff frequency of f_lowpass = 28.5Hz.

The total frequency bandwidth response of the preamplifier is calculated as BW = f_lowpass − f_highpass = 28 Hz. Consequently, the preamplifier operates linearly within a frequency spectrum of 28 Hz, with a lower frequency limit of 0.5 Hz and an upper-frequency limit of 28.5 Hz (Figure 5).

2.1.3. The Datalogger

The core of the proposed low-cost system (recorder) is the Raspberry Pi 3 B+ microcomputer board, combined with the Waveshare high-precision 24-bit sigma-delta analog-to-digital converter board. This converter board, which is compatible with the 40 GPIO pins of the Raspberry Pi 3 B+, supports a sampling frequency rate of up to 30 ksps (kilosamples per second), enabling the development of a powerful datalogger-digitizer capable of recording, storing, and transferring data continuously (24/7). A 32 GB SD card is used as the microcomputer’s primary storage, containing the operating system (Raspbian Buster release) and the system firmware, including Python scripts, libraries, and other necessary components. The SD card is also employed to store collected data files, which are subsequently transmitted over the Internet to the CMODLab server at Ionian University.

The microcomputer board is connected to the Internet via either the onboard Wi-Fi controller or the onboard Ethernet NIC port. To avoid potential distortion of the collected data—comprising 200 data lines per second, each containing digitized voltage values and timestamps—the system relies exclusively on the Ethernet connection, using a UTP Cat6 cable, instead of the Wi-Fi connection. Time synchronization is maintained through a real-time clock (RTC) module, complemented by an NTP (Network Time Protocol) server and a GPS board, ensuring minimal error in the printed timestamps of recorded data.

Similar low-cost systems have been utilized in related works, such as those described in [38,39]. In the present work, the analog-to-digital converter board is based on the ADS1256 chip, offering 24-bit accuracy and a quantization error equal to ±${\displaystyle \frac{1}{2}}$ of the least significant bit (LSB). This quantization error corresponds to ±${\displaystyle \frac{1}{2}}$ of the minimum input voltage divided by 2²³. The converter board can be configured to operate with a maximum reference input voltage of 3.3 V or 5 V. In the proposed low-cost recorder setup, the reference input voltage is set to 3.3 V (±1.65 V)

The board includes eight analog inputs, which can operate in either single-ended mode (eight separate input channels) or differential mode (four separate input channels), as used in this setup. It supports sample rates of up to 30 ksps in single-channel mode and features an internal programmable gain amplifier (PGA) with amplification factors up to 64, adjustable in multiplier steps of 2.

The 24-bit ADC was selected for its capability to provide high-resolution data while maintaining low noise levels, a critical feature for capturing precise amplitude measurements. A sampling rate of 200 Hz was chosen to achieve a balance between capturing detailed seismic event information and ensuring efficient data storage and transmission [38,39]. This sampling rate satisfies the Nyquist theorem, allowing the low-cost system to accurately record seismic signals with frequencies up to 100 Hz.

To achieve the desired sampling rate of 200 Hz, a higher internal sampling frequency was configured in the analog-to-digital board’s programming setup. A frequency of 3750 samples per second (sps) was selected to compensate for delays introduced by the board’s internal multiplexer circuitry. The effective sampling frequency can be maintained when the analog-to-digital board operates in single-channel mode with only one data channel and without incorporating timestamps in the recorded files.

According to the ADS1256 datasheet (ADS1256 Datasheet—https://www.ti.com/lit/ds/symlink/ads1256.pdf accessed on 15 October 2024.), the noise-free resolution of the board is up to 18.1 bits, while the effective number of bits (ENOB), with the buffer disabled, is approximately 20.8 bits. Given that the amplification factor of the internal programmable gain amplifier (PGA) is set to 1 and the low-noise preamplifier provides an amplification factor of A = 275, the total output signal amplification equals 275. The differential maximum output voltage of the preamplifier is configured to ±1.65 V, matching the input voltage range of the A/D converter board. The input voltage measurable by the A/D converter cannot exceed ±1.65 V, resulting in each count having a value of ±1.65 V/2²³. The differential maximum output voltage of the preamplifier is configured to ±1.65 V, matching the input voltage range of the A/D converter board. The input voltage measurable by the A/D converter cannot exceed ±1.65 V, resulting in each count having a value of ±0.1967 μV at the Last Significant Bit (LSB—minimum count step).

According to the A/D converter board manual, the ENOB reaches 20.8 bits during operation at a sampling rate of 3750 samples per second. This yields an LSB value of ± 1.65 V/2^20.8, equivalent to ± 0.904 μV. Additionally, considering the noise-free resolution of 18.1 bits, the noise-free level for recorded signals is calculated as ± 1.65 V/2^18.1, which corresponds to ± 5.8727 μV. Finally, considering the preamplifier’s amplification factor of A = 275, the smallest input signal that the low-cost system can detect, and record is calculated as ± 5.8727 μV/275, resulting in ± 21.355 nV. Interpreting the recorded voltage values into velocity (based on the geophone sensor response curve at 4.5 Hz, 380 ohms, as shown in Figure 3b), for a linear frequency response starting from 4.5 Hz and above, the smallest velocity that can be recorded by the low-cost system is calculated as ± 21.355 nV/29.2 V/m/s, resulting in ± 0.7313 nm/s.

The recorded data are stored internally in the system in 5 min chunks, formatted as .csv files with unique filenames based on the date-time (timestamp) of their creation for easy identification. Each file spans a duration of 300 s (5 min) and contains two rows of data: one row records the count values corresponding to the measured voltage output, and the other row records the separate timestamps for each value. A new data file with a unique filename is generated by the system every 5 min, saved to the SD card, and then transmitted to the server of Ionian University via the Ethernet connection of the low-cost system.

2.1.4. The Power Supply

The low-cost system (recorder) is powered by a +12 V, 55 Ah battery pack. A step-down voltage converter reduces the input voltage from 12 V DC to +5 V DC, providing up to 3A of current to meet the system’s voltage and current requirements. A charger-maintainer ensures that the battery remains fully charged and in optimal condition, allowing uninterrupted operation in the event of a power grid failure.

For outdoor and autonomous operation, the system can be connected to a solar panel via a solar charger-maintainer, providing a sustainable energy source. Additionally, a GSM 4G modem is integrated to establish Internet connectivity, enabling the transmission of data to the CMODLab server at Ionian University.

3. Low-Cost Seismograph Installation

The island of Lefkada is located in the central Ionian Sea, on the western side of Greece, which is the region with the highest seismic activity in the country [40,41,42,43]. On the island, five seismological stations with online connectivity are operated by the Geophysics Department of the Aristotle University of Thessaloniki (AUTH). These stations were installed in collaboration with the Regional Union of Municipalities of the Ionian Islands. Additionally, several accelerometers, maintained by various seismological institutes, are also in operation. One of the permanent seismological stations is situated in an abandoned old school building (38.621° N, 20.656° E) in the village of Evgiros, located in the southeastern part of the island. This station is equipped with a high-resolution (24-bit) digitizer (RefTek 130) and a Guralp CMG40T broadband seismometer (30 s—50 Hz), as illustrated in Figure 6a.

To minimize local biases, the low-cost system was installed within 1 m of the high-cost system, ensuring identical exposure to ground motion. Differences in sensor housing and coupling were carefully assessed and determined to have minimal impact on data comparability. For these reasons, the proposed new version of the low-cost system was installed adjacent to the AUTH seismograph. The 4.5 Hz geophone sensor was positioned in close proximity to the broadband sensor, as shown in Figure 6a.

For comparison purposes, the magnitudes calculated by the low-cost system were evaluated against those recorded by the high-cost system. A catalog of 60 seismic events, with magnitudes ranging from M = 0.9 to M = 3.2 and epicentral distances between 5.71 km and 23.45 km, was utilized. The magnitudes were calculated using a formula proposed by [29]. Details of all the seismic recordings used in this study are presented in Appendix A—

Table A1. Catalog of the 60 different compared earthquake recordings from the low-cost and the high-cost reference seismometer.

E/V No	EVENT DATE (EVGI)	EVENT TIME (EVGI)	MAGNITUDE (R) (EVGI)	EPICENTRAL DISTANCE (Km) (EVGI)	FOCAL DEPTH (Km) (EVGI)	HYPOCENTRAL DISTANCE (Km) (EVGI)	AMPLITUDE VELOCITY) (mm/s) (LOW COST)	EVENT NAME (EVGI)
1	17-05-22	01:46:03.812	1.65	14.42	6.467	15.80374921	13.4964	auth2022jopx
2	18-05-22	00:51:38.290	1.66	12.35	10.092	15.94901138	18.6442	auth2022jqjq
3	20-05-22	01:41:57.601	1.91	14.37	6.173	15.63978353	26.1602	auth2022juch
4	23-05-22	13:13:00.052	1.89	14.70	9.112	17.29504391	33.8431	auth2022kalp
5	27-05-22	10:18:43.042	2.23	11.82	5.536	13.05219123	55.9847	auth2022khnw
6	31-05-22	21:45:57.839	1.74	12.01	5.977	13.4150896	28.2974	auth2022kpso
7	05-06-22	01:28:38.571	1.35	15.42	6.957	16.91674463	6.0631	auth2022kxhx
8	06-06-22	10:49:50.653	2.43	14.89	6.173	16.11887183	68.021	auth2022kzvx
9	09-06-22	02:41:23.928	1.05	11.98	4.752	12.88805276	4.8685	auth2022lesi
10	09-06-22	04:23:26.335	1.05	16.99	5.389	17.82418079	3.3824	auth2022levr
11	12-06-22	03:15:01.573	1.86	11.52	8.9	14.55748605	23.0425	auth2022lkfy
12	13-06-22	11:33:23.287	2.24	12.2	7.3	14.21724305	54.7547	auth2022lmrw
13	23-06-22	11:49:14.078	2.19	12.65	3.8	13.20842534	59.7676	auth2022mezl
14	01-07-22	10:56:36.404	1.76	11.06	2.7	11.38479688	32.2817	auth2022mtns
15	05-07-22	06:46:27.858	2.63	14.86	8.54	17.13917151	82.807	auth2022nanm
16	07-07-22	02:29:20.686	2.27	13.86	10.778	17.55747374	41.0083	auth2022ndvz
17	10-07-22	11:29:27.081	2.1	7.22	7.251	10.23256571	87.4194	auth2022nkai
18	14-07-22	22:41:08.174	1.99	17.9	3.1	18.1664526	25.8898	auth2022nsem
19	18-07-22	10:05:45.219	1.61	18.25	4.3	18.74973333	9.0717	auth2022nynp
20	11-08-22	17:13:21.639	1.99	11.95	8.328	14.56564739	35.5793	auth2022pqxt
21	12-08-22	00:55:14.752	1.48	17.54	2.733	17.75164468	9.1228	auth2022prna
22	15-08-22	22:04:13.527	2.3	13.19	8.524	15.70460684	73.2183	auth2022pypk
23	18-08-22	10:32:02.622	2.11	18.91	11.562	22.16456505	26.5712	auth2022qdfb
24	22-08-22	11:25:21.130	2.74	13.62	3.625	14.09414861	146.6278	auth2022qkov
25	26-08-22	21:33:47.603	2.37	10.52	11.366	15.4872966	98.033	auth2022qsqx
26	27-08-22	08:41:01.634	1.81	9.31	11.268	14.61656334	26.9657	auth2022qtmx
27	27-08-22	21:33:40.467	1.69	17.47	8.916	19.61366758	12.6236	auth2022qumk
28	16-09-22	03:33:54.103	1.97	10.76	2.646	11.08056479	42.7446	auth2022sdqw
29	13-10-22	02:16:21.907	1.42	10.58	5.585	11.96363762	13.2762	auth2022uaww
30	26-10-22	10:38:15.036	2.74	16.63	19.694	25.77616217	125.3326	auth2022uzgz
31	31-10-22	01:01:21.256	2.5	23.45	6.173	24.24888511	62.8158	auth2022vhrl
32	01-11-22	12:38:49.545	2.36	9.39	6.124	11.21050739	141.7646	auth2022vkjz
33	03-11-22	15:03:00.502	1.27	10.22	9.014	13.62720059	9.9755	auth2022vofs
34	07-11-22	16:56:19.215	2.35	12.52	4.801	13.40895227	70.6975	auth2022vvrm
35	21-11-22	22:12:09.494	2.46	14.47	8.916	16.99635126	62.5576	auth2022wvqy
36	06-12-22	05:06:13.167	2.35	11.93	5.6	13.179	117.1648	auth2022xvtq
37	25-12-22	11:41:50.889	2.84	12.07	5.683	13.34096657	186.113	auth2022zezg
38	01-01-23	12:28:01.829	1.97	8.79	6.614	11.00041345	55.5318	auth2023aayr
39	06-01-23	20:55:36.347	2.97	13.97	5.144	14.88696195	223.1903	auth2023aksy
40	03-02-23	05:30:17.541	2.03	14.24	9.112	16.90580208	46.8723	auth2023cism
41	10-02-23	23:31:48.461	1.67	13.11	10.092	16.54450253	15.8004	auth2023cwws
42	16-02-23	18:48:42.426	3.22	15.36	4.44	15.98884611	222.2577	auth2023dhmj
43	19-02-23	17:38:00.606	2.02	14.53	9.308	17.25571685	42.2152	auth2023dmwo
44	31-03-23	02:13:51.707	0.93	10.5	6.761	12.48843949	5.9021	auth2023ggue
45	19-04-23	20:18:39.725	1.98	11.03	2.225	11.25217868	43.6222	auth2023hqwn
46	20-04-23	17:31:06.999	2.93	14.12	4.507	14.82185714	166.4831	auth2023hsmn
47	28-04-23	17:52:53.444	2.13	15.28	10.288	18.42067708	28.6996	auth2023ihdg
48	06-05-23	23:52:27.083	1.64	14.55	2.358	14.73983256	14.5944	auth2023iwfd
49	16-05-23	14:25:58.561	2.28	9.42	10.582	14.16739651	83.8643	auth2023jntm
50	18-05-23	07:18:09.696	2.26	8.24	8.328	11.7155104	124.9216	auth2023jqwj
51	07-06-23	04:44:40.063	2.39	9.48	12.248	15.4882	121.3965	auth2023lbfj
52	30-06-23	03:29:39.357	2.62	12.65	8.916	15.47635474	133.6742	auth2023mrdm
53	11-07-23	20:47:13.691	1.25	8.88	10.778	13.96494483	11.6498	auth2023nmog
54	12-07-23	10:40:45.087	2.59	11.55	9.112	14.71159556	113.6208	auth2023nnps
55	13-07-23	15:28:53.338	1.6	12.47	1.83	12.60356299	16.3874	auth2023nput
56	19-07-23	22:22:33.409	1.42	9.42	9.994	13.73376991	10.5396	auth2023obhk
57	22-07-23	13:08:41.715	2.07	14.79	3.437	15.1841058	49.6679	auth2023ogbq
58	25-07-23	03:26:32.890	2.35	12.88	9.504	16.00688652	103.1323	auth2023okuy
59	28-07-23	13:27:02.579	1.52	12.79	4.148	13.44581734	17.5778	auth2023orbg
60	09-08-23	12:32:00.243	2.03	5.71	9.308	10.91984267	80.3042	auth2023pmxm

. Signal analysis for both systems (low-cost and high-cost) was conducted exclusively for the vertical axis. On the maps shown in Figure 6b,c, blue dots represent the epicenters of the 60 seismic events, while the red dot indicates the location of both the low-cost and high-cost systems.

The location of Evgiros village on Lefkada Island was chosen for its high surrounding seismic activity and favorable geological structure [42]. The rocky bedrock in the area minimizes signal attenuation [6], while the village’s sparse population and low levels of human activity—such as noise from vehicles and machinery—reduce non-seismic noise sources [44]. This combination of factors enhances the system’s reliability for local seismic monitoring. Additionally, the high seismicity of the area is visualized in Figure 6d, which presents data recorded from January 2019 to December 2024.

4. Data Acquisition and Analysis

Data analysis in this study is based on recorded voltage values translated into velocities using the linear frequency response of the sensor for frequencies of 4.5 Hz and above, with a sensitivity of 29.4 V (ζ = 0.5) m/s. The conversion is performed using the formula Velocity = 29.4/Output Voltage. The maximum zero-to-peak velocity amplitude measurements recorded by the low-cost system, along with the magnitude values estimated by the reference high-cost seismograph for each seismic event, are presented in Appendix A—Table A1.

The primary focus is on the maximum zero-to-peak amplitude values for each independent seismic event. These amplitude values are extracted from the low-cost system’s seismograms (Figure 7) using data analysis tools such as Octave. Records for 60 earthquakes—including magnitude, epicentral distance, hypocentral distance, focal depth from the high-cost reference seismograph (AUTH Network), and maximum zero-to-peak velocity amplitudes from the low-cost system—are summarized in Appendix A—

Table A2. Catalogue of 60 compared magnitude earthquake recordings and magnitude differences calculated for Epicentral and Hypocentral Distances for high-cost reference commercial seismograph versus the low-cost system.

E/V No	MAGNITUDE (EVGI)	M-CALCULATED EPICENTRAL	M-CALCULATED HYPOCENTRAL	M-DIFFERENCE EPICENTRAL	M-DIFFERENCE HYPOCENTRAL	Ci EPICENTRAL	Ci HYPOCENTRAL	EVENT NAME (EVGI)
1	1.65	1.6571	1.6163	0.0071	0.0337	1.8225	1.7669	auth2022jopx
2	1.66	1.704	1.7622	0.0440	0.1022	1.7856	1.631	auth2022jqjq
3	1.91	1.9424	1.8974	0.0324	0.0126	1.7972	1.7458	auth2022juch
4	1.89	2.0680	2.0706	0.1780	0.1806	1.6516	1.5526	auth2022kalp
5	2.23	2.1552	2.1184	0.0748	0.1116	1.9044	1.8448	auth2022khnw
6	1.74	1.8684	1.8386	0.1284	0.0986	1.7012	1.6346	auth2022kpso
7	1.35	1.3502	1.2364	0.0002	0.1136	1.8294	1.8468	auth2022kxhx
8	2.43	2.3789	2.3602	0.0511	0.0698	1.8807	1.803	auth2022kzvx
9	1.05	1.1026	1.1855	0.0526	0.1355	1.7770	1.5977	auth2022lesi
10	1.05	1.1558	0.8919	0.1058	0.1581	1.7238	1.8913	auth2022levr
11	1.86	1.7542	1.7991	0.1058	0.0609	1.9354	1.7941	auth2022lkfy
12	2.24	2.1645	2.1614	0.0755	0.0786	1.9051	1.8118	auth2022lmrw
13	2.19	2.2243	2.1543	0.0343	0.0357	1.7953	1.7689	auth2022mezl
14	1.76	1.8763	1.7972	0.1163	0.0372	1.7133	1.696	auth2022mtns
15	2.63	2.4631	2.4533	0.1669	0.1767	1.9965	1.9099	auth2022nanm
16	2.27	2.1158	2.1632	0.1542	0.1068	1.9838	1.84	auth2022ndvz
17	2.1	2.0559	2.1661	0.0441	0.0661	1.8737	1.6671	auth2022nkai
18	1.99	2.0717	1.9844	0.0817	0.0056	1.7479	1.7388	auth2022nsem
19	1.61	1.6281	1.5464	0.0181	0.0636	1.8115	1.7968	auth2022nynp
20	1.99	1.9649	1.9877	0.0251	0.0023	1.8547	1.7355	auth2022pqxt
21	1.48	1.6062	1.4424	0.1262	0.0376	1.7034	1.7708	auth2022prna
22	2.3	2.3376	2.5588	0.0376	0.2588	1.7920	1.4744	auth2022pypk
23	2.11	2.1167	1.841	0.0067	0.2690	1.8229	2.0022	auth2022qdfb
24	2.74	2.6586	2.64	0.0814	0.1000	1.9110	1.8332	auth2022qkov
25	2.37	2.3288	2.43	0.0412	0.0600	1.8708	1.6732	auth2022qsqx
26	1.81	1.6955	2.0492	0.1145	0.2392	1.9441	1.4940	auth2022qtmx
27	1.69	1.7449	1.3732	0.0549	0.3168	1.7748	2.0500	auth2022qumk
28	1.97	1.9818	1.9488	0.0118	0.0212	1.8178	1.7544	auth2022sdqw
29	1.42	1.4639	1.9119	0.0439	0.4919	1.7857	1.2413	auth2022uaww
30	2.74	2.7116	2.8485	0.0284	0.1085	1.8580	1.6247	auth2022uzgz
31	2.5	2.6239	2.0771	0.1239	0.4229	1.7057	2.1561	auth2022vhrl
32	2.36	2.4213	2.5478	0.0613	0.1878	1.7683	1.5454	auth2022vkjz
33	1.27	1.3191	1.3855	0.0491	0.1155	1.7805	1.6177	auth2022vofs
34	2.35	2.2910	2.3799	0.0590	0.0299	1.8886	1.7033	auth2022vvrm
35	2.46	2.3252	2.173	0.1348	0.287	1.9644	2.0202	auth2022wvqy
36	2.35	2.4814	2.4523	0.1314	0.1023	1.6982	1.6309	auth2022xvtq
37	2.84	2.6894	2.5375	0.1506	0.3025	1.9802	2.0357	auth2022zezg
38	1.97	1.9751	2.1943	0.0051	0.2243	1.8245	1.5089	auth2023aayr
39	2.97	2.8564	2.6095	0.1136	0.3605	1.9432	2.0937	auth2023aksy
40	2.03	2.1902	2.1981	0.1602	0.1681	1.6694	1.5651	auth2023cism
41	1.67	1.6680	1.7127	0.0020	0.0427	1.8316	1.6905	auth2023cwws
42	3.22	2.9120	2.84	0.3080	0.38	2.1376	2.1132	auth2023dhmj
43	2.02	2.1569	2.1652	0.1369	0.1452	1.6927	1.588	auth2023dmwo
44	0.93	1.1073	1.1147	0.1773	0.1847	1.6523	1.5485	auth2023ggue
45	1.98	2.0054	1.9209	0.0254	0.0591	1.8042	1.7923	auth2023hqwn
46	2.93	2.7355	2.6685	0.1945	0.2615	2.0241	1.9947	auth2023hsmn
47	2.13	2.0199	2.0376	0.1101	0.0924	1.9397	1.8256	auth2023ihdg
48	1.64	1.6965	1.6079	0.0565	0.0321	1.7731	1.7653	auth2023iwfd
49	2.28	2.1952	2.2295	0.0848	0.0505	1.9144	1.7837	auth2023jntm
50	2.26	2.2890	2.6954	0.0290	0.4354	1.8006	1.2978	auth2023jqwj
51	2.39	2.3596	2.504	0.0304	0.114	1.8600	1.6192	auth2023lbfj
52	2.62	2.5739	2.4314	0.0461	0.1886	1.8757	1.9218	auth2023mrdm
53	1.25	1.3029	1.5401	0.0529	0.2901	1.7767	1.4431	auth2023nmog
54	2.59	2.4487	2.5288	0.1413	0.0612	1.9709	1.7944	auth2023nnps
55	1.6	1.6537	1.6256	0.0537	0.0256	1.7759	1.7076	auth2023nput
56	1.42	1.2945	1.4654	0.1255	0.0454	1.9551	1.6878	auth2023obhk
57	2.07	2.2383	2.0453	0.1683	0.0247	1.6613	1.7579	auth2023ogbq
58	2.35	2.4721	2.5073	0.1221	0.1573	1.7075	1.5759	auth2023okuy
59	1.52	1.6994	1.6332	0.1794	0.1132	1.6502	1.62	auth2023orbg
60	2.03	1.8812	2.168	0.1488	0.138	1.9784	1.5952	auth2023pmxm

.

The figure panel presents ten (10) representative seismic events sampled from the 60 analyzed seismograms, recorded by the low-cost system. For each event, plots of velocity versus time and the FFT acquisition process output (frequency spectrum response: power versus frequency) are provided [45]. The major frequency peaks (F_c1 = 0.5 Hz, F_c2 = 28.5 Hz, and F_cc = 4.5 Hz) are annotated on each seismic event plot in Figure 7, highlighting the dominant frequencies observed during the recordings.

In addition, the spectrum response plots include vertical lines indicating key frequencies: the cutoff frequencies F_c1 = 0.5 Hz, F_c2 = 28.5 Hz, and F_cc = 4.5 Hz of the low-noise signal preamplifier are marked with blue lines, while the resonance frequency of the GS-11D geophone (F_cc = 4.5 Hz) is marked with a red line. It is evident that the recorded signal events (a) lie within the frequency response characteristics of the system (0.5 Hz to 28.5 Hz), and (b) exhibit an almost flat (linear amplitude) frequency response between 4.5 Hz and 28.5 Hz. Furthermore, the system demonstrates significant linear frequency response from 3.5 Hz and above in nearly all the analyzed signals. A comparative analysis with other low-cost systems, such as Raspberry Shake, Smart Solo, Guralp, and MEMS-based accelerometers, reveals that the proposed system achieves superior accuracy in magnitude estimation, with mean errors reduced to ±0.087, compared to the ±0.2 to ±0.4 range reported in previous studies [20]. Additionally, the hardware design of the proposed system provides improved durability and cost-effectiveness.

All data recorded by the low-cost system underwent rigorous procedures to identify and handle abnormal data points, ensuring their accuracy and reliability. The procedures employed included (a) visual inspection of seismograms to identify noise artifacts, (b) cross-referencing events with known seismic occurrences from the AUTH network, and (c) automated filtering to remove outliers using amplitude thresholds and noise suppression algorithms. A comparative analysis with other similar low-cost seismograph systems is presented in

Table A4. A comparative analysis with other similar low-cost seismograph systems with a comprehensive comparison of their characteristics together with the low-cost system of that scope.

System	Technology	Performance	Cost Category	Applications	Advantages
Raspberry Shake	Geophones + Raspberry Pi	Good for local/regional seismic monitoring	Medium	Educational, citizen science, research	Open source, user-friendly, global community
Geophonino	Arduino + Geophones	Limited to local seismic detection	Low	Microseismic studies, education	Low-cost, Arduino-based customization
SmartSolo	MEMS-based geophone system	High sensitivity, suitable for small earthquakes	High	Passive surveys, oil exploration	Lightweight, autonomous operation
Güralp CMG-6TD	Compact broadband seismometer	Excellent for weak signals and broad frequency	Very High	Research-grade seismic monitoring	High sensitivity, integrated digitizer
MEMS Accelerographs	MEMS accelerometers	Strong motion detection but limited weak signals	Low	Structural health monitoring, strong motion	Compact, cost-effective
Kronos	Custom-built DAQ system	High accuracy for microseismic monitoring	Medium	Research, microseismic studies	Reliable, research-grade data acquisition
DigiSeis	PC sound card-based	Basic performance, limited dynamic range	Very Low	Educational, small-scale projects	Extremely low-cost, PC-based simplicity
System Before Upgrade	Geophone, Ceramic Accelerometer, Raspberry Pi, and Arduino	Local–regional seismic monitoring with a maximum estimation error of +/− 0.4 R	Medium	Microseismic studies, education	Reliable, research-grade data acquisition
Upgraded System (This work)	Geophone and Raspberry Pi, ADS1256 24bit A/D board, Custom build preamplifier	Local–regional seismic monitoring with a maximum estimation error of +/− 0.087 R	Low	Microseismic studies, education	Compact, cost-effective

of the Appendix, providing a detailed comparison of their characteristics alongside the proposed low-cost system.

The frequency distributions depicted in Figure 8a–e reveal distinct trends: epicentral distances correlate with amplitude velocities, and magnitude values show close alignment between the systems for the majority of events. The data used for these analyses were sourced from Appendix A—Table A2.

4.1. Methodology

The calibration process involved the simultaneous recording of seismic events by both the low-cost and high-cost systems. By aligning time-series data and comparing peak amplitudes, correction factors were derived to address discrepancies arising from sensor and system differences. These corrections were applied to ensure comparable magnitude estimates across both systems.

To compare the zero-to-peak amplitude velocity values recorded by the Geospace GS-11D 4.5 Hz, 380 ohm geophone of the low-cost system with the magnitudes calculated from the high-cost reference seismometer, the method proposed by [22] for accurate local magnitude calculations was adopted. Using the data from the catalog (Appendix A—Table A1) of 60 compared seismic recordings from both the high-cost seismograph and the low-cost system, a modified Hutton–Boore equation is proposed [27],

$$C_i=M_i-log{A}_i-1.319log\left({\displaystyle \frac{R_i}{100}}\right)-0.00226\left(R_i-100\right)$$

(1)

where M_i represents the earthquake magnitude value of each recorded event, expressed in local magnitude, as determined by the high-cost seismograph; A_i represents the maximum zero-to-peak velocity amplitude (in mm/s) for the same event, recorded by the low-cost system; and R_i is the distance from the seismic event to the installation point (i.e., the old school in Evgiros, Lefkada), measured in kilometers.

In the initial calculations, the distance R_i is the epicentral (R_epc,i). The epicentral distance is calculated individually for each seismic event using data from the high-cost seismic network operated by the Aristotle University of Thessaloniki (AUTH). The high-cost seismograph EVGI is an integral part of the AUTH network. Therefore, the Equation (1) becomes:

$$C_i=M_i-log{A}_i-1.319log\left({\displaystyle \frac{R_{epc,i}}{100}}\right)-0.00226\left(R_{epc,i}-100\right)$$

(2)

After calculating the C_i value for each of the sixty (60) events, the mean value of C_i (denoted as C_i,avg) is computed using the following equation:

$$C_{i,avg}={\displaystyle \frac{1}{N}}\sum _{i=1}^{N=60}{C}_i$$

(3)

Subsequently, the earthquake magnitudes based on the epicentral distances (M_calc,epc,i) are calculated by solving Equation (2) with respect to M, incorporating C_i,avg derived from Equation (3). Thus, Equation (2) is reformulated as follows:

$$M_{calc,epc,i}=C_{i,avg,epc}+log{A}_i+1.319log\left({\displaystyle \frac{R_{epc,i}}{100}}\right)+0.00226\left(R_{epc,i}-100\right)$$

(4)

The magnitudes of the 60 seismic events, calculated using this method, are presented in Appendix A—Table A2.

Subsequently, instead of using the epicentral distance (R_epc,i) in Equation (4), the hypocentral distance (R_hpc,i) is determined using the Pythagorean theorem. This calculation is performed using the following equation:

$$R_{hpc,i}=\sqrt{{R_{epc,i}}^2+D^2}$$

(5)

where R_epc,i is the epicentral distance and D is the focal depth taken from the data of the catalog (Appendix A—Table A1). The epicentral distance (R_epc,i) and focal depth (D) are given in Kilometers (Km) for all 60 earthquakes (Appendix A—Table A1).

Then, the C_i,avg,hpc is calculated, but now for the hypocentral distance, by using again Equations (1) and (2). The M_calc,hpc,i is calculated by replacing in Equation (4) the epicentral distance with the hypocentral distance, and the equation now becomes:

$$M_{calc,hpc,i}=C_{i,avg,hpc}+log{A}_i+1.319log\left({\displaystyle \frac{R_{hpc,i}}{100}}\right)+0.00226\left(R_{hpc,i}-100\right)$$

(6)

4.2. Results

The mean values of C_i,avg,epc for the epicentral distance and C_i,avr,hpc for the hypocentral distance were calculated from Equation (3) as 1.8296 and 1.6639, respectively. These values were then applied to Equations (4) and (6) to calculate the magnitudes for each of the 60 earthquake events recorded by the low-cost system, using both the epicentral and hypocentral distances. A new catalog was created, including an additional column with data representing the magnitude differences for the same 60 earthquake events, comparing the low-cost system to the high-cost system (Appendix A—Table A2).

Statistical metrics, including mean, maximum, and minimum error values, along with scatter plots, were employed to evaluate the reliability and effectiveness of the low-cost system compared to the reference commercial seismograph. The scatter plots presented in Figure 9 and Figure 10 illustrate the relationship between the local magnitudes calculated from the low-cost seismograph (on the y-axis) and the corresponding local magnitudes recorded by the broadband seismometer (EVGI) (on the x-axis) for epicentral and hypocentral distances, respectively.

Additionally, estimated or found instead of calculated error propagation was analyzed by combining uncertainties from sensor sensitivity, ADC resolution, and preamplifier amplification. The cumulative error was estimated or found instead of calculated to be ±0.087 for epicentral distance estimates, aligning with the observed systematic errors. This analysis confirms the system’s suitability for local seismic monitoring within the specified range of magnitudes from 0.9 to 3.2, epicentral distances from 5.71 km to 23.45 km, and focal depths from 1.83 km to 19.69 km.

The analysis demonstrates a strong correlation between the calculated magnitudes using epicentral distance and those using hypocentral distance. Furthermore, the mean systematic error is low, approximately 0.1, with standard deviation values ranging from −0.0869 to 0.11, as detailed in Appendix A—Table A2.

The epicentral distance was ultimately selected as the primary metric due to its direct relevance in local seismicity studies, where depth variations are less significant for smaller seismic events. The analysis revealed that epicentral distance produced lower mean errors compared to hypocentral distance, establishing it as a more reliable parameter for the scope of this study.

y = 0.8797x + 0.245
R² = 0.9608

(7)

To ensure greater accuracy in calculations, we use magnitude sizes with two decimal places instead of the usual single decimal point. This adjustment applies to all computations and outputs, ensuring precision and consistency across all results.

The linear regression equations for Figure 9 and Figure 10 are provided, along with the R² values to quantify the accuracy of the fitted regression lines. These metrics indicate a strong correlation for epicentral distances (R² = 0.96), while also strong correlation is observed for hypocentral distances (R² = 0.86)). This analysis demonstrates the system’s reliability for local magnitude estimation when using epicentral or hypocentral distances, confirming its suitability for seismic monitoring within the defined parameters.

Conversely, when using the hypocentral distance, both the mean error and standard deviation increased. This observation is supported by Figure 10, while Appendix A—Table A2 quantifies the statistical metrics, showing a mean systematic error of approximately 0.143, with values ranging from 0.0023 to 0.4919. Based on these results, as summarized in

Table 1. Overall statistics for the 60 events in the dataset include maximum, minimum, and mean errors in local magnitude, representing the magnitude differences between the high-cost and low-cost seismographs for both epicentral and hypocentral distances.

	MAGNITUDE DIFFERENCE (EPICENTRAL)	MAGNITUDE DIFFERENCE (HYPOCENTRAL)
MAX ERROR	0.3080	0.4919
MIN ERROR	0.0002	0.0023
MEAN ERROR (60 EVENTS)	0.0871	0.1433

, the hypocentral distance approach is excluded from further analysis. The data analysis continues using only the epicentral distance, which consistently demonstrated superior accuracy, as evidenced by the higher R² value.

Figure 11 presents the data from Table 1 as a line plot, comparing the maximum, minimum, and mean magnitude errors for epicentral and hypocentral distances as calculated in this study.

Using data from Appendix—Table A2, the calculated magnitude error is found to be smaller for events with epicentral distances up to 10 km and focal depths up to 15 km, as shown in

Table 2. The calculated magnitude difference error, in absolute values, between the high-cost and low-cost seismographs is presented for various ranges of epicentral distances and focal depths.

	EPICENTRAL DISTANCE 0–10 Km	EPICENTRAL DISTANCE 10–23.45 Km
FOCAL DEPTH 5–10 Km	0.0792	0.093
FOCAL DEPTH 10–15 Km	0.0587	0.068

.

Figure 12 presents the data from Table 2 as a chart comparing the magnitude mean errors for focal depths of 5–10 km and 10–15 km, combined with epicentral distances of 0–10 km and 10–23.45 km, as calculated in this study.

5. Conclusions and Future Work

In this study, an upgraded low-cost seismograph system was designed and implemented using cost-effective hardware and software, primarily based on open hardware and open software platforms. The low-cost system was installed near the high-cost seismograph (EVGI) at a high-seismicity location in Greece (Lefkada Island, Evgiros village). By continuously recording seismic events from both systems simultaneously, a comprehensive data catalog was compiled for further analysis. Specifically, a dataset of 60 seismic events was created and analyzed to evaluate the magnitudes recorded by the low-cost datalogger system in comparison with the high-cost commercial broadband seismograph. The statistical metrics yielded highly promising results, with the mean systematic error for the low-cost seismograph using epicentral distances calculated to be approximately 0.1 when compared to the reference high-cost system. Additionally, for varying focal depth ranges, the mean error was observed to range between 0.06 and 0.1.

The results indicate that while the low-cost seismograph exhibits small systematic errors in magnitude estimation (mean error ±0.087), its cost-effectiveness makes it a practical solution for local seismic monitoring. The system performs optimally within a 10 km epicentral distance, up to a 15 km focal depth, and for magnitudes up to 3.2, demonstrating accuracy comparable to high-cost systems under these conditions. Its affordability enables deployment in dense seismic networks, which is often impractical for high-cost systems due to budget constraints. The low-cost system achieved high accuracy in local magnitude estimation, with errors well within acceptable limits for seismic monitoring. The comparative analysis confirms its suitability for monitoring small-magnitude events in local networks, particularly where cost considerations are critical.

The low correlation observed between the low-cost and high-cost seismographs for hypocentral magnitude estimation (R² = 0.04) can be attributed to several factors. Unlike epicentral distances, hypocentral calculations rely on both horizontal distance and focal depth, introducing additional uncertainty due to errors in depth estimation. The low-cost system employs a 24-bit analog-to-digital converter, and a geophone sensor optimized for surface wave detection, which may lack the sensitivity required to accurately capture signals from deeper seismic events.

Additionally, local geological conditions and the damping characteristics of the geophone (e.g., the ζ\zetaζ-factor) may disproportionately affect lower-frequency components associated with deeper events. These limitations emphasize the need for hardware upgrades, such as a higher-resolution analog-to-digital converter or enhanced pre-processing circuitry, to improve the system’s accuracy in depth-related seismic calculations. Future work could also focus on refining the frequency response of the low-cost system to account for signal acquisition fluctuations, thereby enhancing its accuracy in hypocentral distance estimation.

The performance of the low-cost datalogger system can be further enhanced by upgrading the analog-to-digital converter to a 32-bit model, replacing the 24-bit converter currently used in the proposed system. Additionally, incorporating external electronic circuits to overdamp the ζ-factor of the geophone sensor would allow it to operate linearly at lower cutoff frequencies, thereby improving signal fidelity.

The results obtained from the analyzed seismic event recordings demonstrate that the low-cost system achieves a level of accuracy comparable to that of the high-cost seismograph for earthquake magnitudes ranging from 0.9 to 3.2, within an epicentral radius of up to 10 km and focal depths of up to 15 km.

Future work will aim to enhance the system’s performance through the integration of advanced sensors and the exploration of applications in diverse geological environments. Expanding the deployment of the low-cost system to varied geographical and geological settings will allow for the analysis of broader datasets. Additionally, testing the system’s performance for higher-magnitude seismic events and over longer distances will help confirm its robustness and applicability in a wider range of seismic monitoring scenarios.