Prior to establishing the observatory buildings, a magnetic survey was conducted in November 2012 over an area of 200 m × 200 m with 20 m intervals, between the main building and the southern side of the campus, marked by the pink rectangle in Fig. 1b, which had the smallest anomalies as per the earlier survey. This area was sufficiently far away from the boundary of the campus to ensure that local activities outside the campus may not have significant contribution to the measurements. The total magnetic anomaly range (Fig. 2) was ∼ 150 nT with changes in magnetic field within ∼ 20 nT around proposed locations of the old PVR and ABS rooms.

In this central location the surface topography is low, and a shallow water channel (nalla) flows SE–NW through the area between the old PVR and ABS. The old PVR construction (with double-walled, 3.5 m underground vault) started in June 2013 using non-magnetic sandstone. The ABS room was constructed on slightly elevated ground with two pillars inside and four pillars outside. In April 2014, CPL observatory was commissioned and equipped with a tri-axial fluxgate magnetometer and Zeiss single-axis fluxgate theodolite for declination–inclination measurements. The XVII IAGA Observatory Workshop was held on these premises in October 2014 with 93 international participants from 33 countries (Arora et al., 2016). The definitive data from CPL are published at INTERMAGNET (https://intermagnet.org/data_download.html) for 2015 onwards.

The region of CPL observatory has a semi-arid climate with average annual maximum temperatures ranging as 28–45 °C. The mean annual rainfall is around 751 mm, which ranges from 2 mm in February to 171 mm in July. Water levels are highly variable depending on the monsoon and usually range between 2 and 26 m (metres below ground surface). Water level measurements at the northern part of the Choutuppal campus have been monitored since 1999 in two dozen boreholes (some of them are marked in Fig. 1b) by the Indo-French Center for Groundwater Research (IFCGR) (Maréchal et al, 2018) to study the hydrodynamic properties and associated hydrological processes in crystalline aquifers. As part of a governmental scheme of strengthening groundwater through recharge state-wise MAR projects, an infiltration basin was dug in Choutuppal (marked in black outline as MAR in Fig. 1b) in 2015 to meet the demands of farmers in the area facing water scarcity. The basin has dimensions of 120 m × 40 m, with a depth of about 2 m, effectively removing the regolith layer and extending into the saprolite. The basin is mainly supplied by a canal which diverts water from the Musi River (Nicolas et al., 2019; Maurya et al., 2021). Nicolas et al. (2019) have shown that intra-seasonal groundwater fluctuations only have a moderate response to rainfall patterns, which could be due to usage trends as well as hydraulic permittivity parameters. After the MAR basin filling, groundwater levels rose by 6 m in 1 year. Figure 3a shows the water level changes of five boreholes (CH5, CH6, CH4, CH7 and CH9) before and after the monsoon from 2011 to 2023, with a data gap during 2021–2022. While it is clear that most of the time water level lies at intermediate depth of ∼ 10 to 30 m, individually, CH5 and CH7 show the least seasonal fluctuations over the years, CH6, CH4 and CH9 show variations of 20 m or more; possibly very local fracture properties facilitate vertical flow of water. Starting from April 2017, the water levels in the boreholes rose significantly, coming almost to the surface in September 2017 and reducing a little by July 2018. In December 2023, the water levels recorded are nearly similar to those of September 2017. It can be inferred that sustained water in the MAR basin has allowed the shallow aquifers to be permanently recharged. In September 2017, the rainfall of the monsoon combined with prevalent saturated condition led to the flooding of the old PVR vault (Fig. 3c). The water level receded by a few metres the following summer but did not fall to earlier levels. While this was good news for recharge, the old PVR was now unusable.

With the need for a new variometer building, a fresh survey was conducted in May 2019 using GSM19 Overhauser magnetometers, marked by the yellow rectangle in Fig. 1b. This area was some tens of metres west of the earlier survey location, still central to the campus, on ground which is about 2–3 m higher, which could avoid the shallow groundwater conditions. This total field magnetic survey was carried out during 6 geomagnetically quiet days in May 2019 over a 240 × 240 m² area with 10 m spacing. The magnetic anomaly of the region shows total amplitude variation of ∼ 300 nT (Fig. 4a) with ∼ 10 nT anomaly north of the ABS room; the proposed location is marked as new PVR in Fig. 4a. This time a raised building with double walls and a double roof was constructed of non-magnetic limestone to ensure no groundwater incursion issues for the foreseeable future (Fig. 4b). The new PVR was commissioned in January 2021.

Recurrent malfunctions of the electronics (two to three times a month) of the digital fluxgate magnetometer (DFM) recording electronics were noticed in the second half of 2021 and 2022, a new phenomenon. After rounds of thorough inspections, it was suspected that lightning activity in the vicinity of the Choutuppal campus was causing the damage in spite of installation of six lightning arrester rods and four earthing pits. Being an open ground with no tall obstructions, the lightning activity in this area was found to be more frequent than around HYB observatory in Hyderabad. The earthing pits and lightning arrester system were strengthened further but only marginally countered the effects on the electronics.

ISRO-National Remote Sensing Center (NRSC) has a network of 46 radio frequency lightning detection sensors, covering the entire part of the India (shown in Fig. 5a). The lightning detection sensor network monitors the cloud to ground lightning occurrences by virtue of emitted waves in the 5 Hz–30 MHz range, and geolocation is calculated using the time of arrival method. The detection range is up to 800 km, with more than 98 % confidence within 300 km range, with 50 % overlap to maintain high geo-location accuracy of lightning occurrences (Taori et al., 2022, 2023). We have examined the lightning data in a radius of 10 km around CPL observatory, marked as + in Fig. 1a. Figure 5b shows the shows the occurrence frequency of lightning over the months from January to August 2022; + symbols denote the instances when the lightning caused damage in the fluxgate electronics. Figure 5b shows that substantially lower intensity lightning activities are recorded during January, April, May and June. Surprisingly July had no lightning in the area in 2022, and in August, the higher-intensity lightning was more frequent. Two instances of failure of the instrument electronics occurred during the higher-intensity lightning of April and August, whereas a disturbance in May was associated with lower-amplitude lightning, discussed in the next section.

Examination of the H (horizontal), D (declination) and Z (vertical) components of 1 s data in LT (local time) for 4 May 2022 data from the new PVR shows continuous spikes from 7–8.2 LT in all the components, followed by a shift of ∼ 70, 20 and 150 nT in H, D and Z respectively (Fig. 6a). After rebooting the instrument, the data came back to their normal range. Comparison with lightning data established that this disturbance was due to the lightning effect (correlated red mark). It is noticed that at the time of 7.94 LT there was a shift in the data that correlates with the lightening intensity of 8879 and 14 243 A (ampere) at the same time, which struck the ground within 10 km radius of CPL campus.

Further, we examined the H, D and Z components of 1 s data in LT for the 30 April and 26 May 2017 data from the old PVR, on days which had some weather activities. From the data, it is clear that there were no shifts in the data, but some continuous spikes were observed from 18.0–18.8 LT (Fig. 7). The spikes are more prominent in the Z component (>0.5 nT). Though lighting data were not available in this duration, the general conditions lead us to believe that these minor signatures were lightning induced.

It was suspected that because the new PVR is constructed on the surface and the cables were laid in the surface layer, instead of the vault configuration as in the one which was flooded, the effects of lightning activity on the data have been amplified. Given the fact that in future years more uncertainty and swings in climatic parameters are anticipated due to the effects of climate change, it was deemed necessary to conduct further studies to find a location based on optimal ranges of a variety of parameters like topography, distance from MAR lake and the configuration of near-surface regolith and saprolite layers, along with groundwater conditions.

A 1 s total magnetic field survey was conducted during February 2024 (4 d) at 10 m sampling intervals as same done in 2019 survey. The data of 2019 and 2024 were combined, thus covering a total of ∼ 400 ×320 m². The diurnal and International Geomagnetic Reference Field (IGRF 14) corrections are applied to the F data, so that the residuals only reflect the local crustal contributions. Finally, we applied the kriging interpolation method to generate the grid and produce the magnetic anomaly of this area (Fig. 8a). Further, this was converted by a reduction-to-equator (RTE) computation to remove ambiguities in the location of causative sources of magnetic anomalies, at low and high latitudes. In this study, we chose the values of delineation =0° and inclination =24° and estimate the RTE-generated magnetic anomaly map of the study region. The RTE-filtered map centres anomalies over their sources and removes the asymmetry of the magnetic anomaly due to nonzero magnetic inclination and helps in magnetic data interpretation (Fig. 8b).

The RTE magnetic anomaly shows a variation of ∼ 260 nT in the region. The low anomaly is dominant in the central part, followed by the high anomaly near the main building and solar panel. The new PVR lies in the low-anomaly region where the three-axis flux magnetometer is installed to record 1 s H, D and Z component data of Earth's magnetic field. The ABS room location is set up in the low anomaly zone to measure the Delineation-Inclination using the Mag-01 DI-fluxgate magnetometer mounted on theodolite.

We have considered four profiles aa', bb, cc' and dd' along EW in the RTE magnetic anomaly map to characterise the subsurface susceptibility model (Fig. 8b). The magnetic data show their importance in characterising the shallow subsurface structures, which would be further beneficial for the selection of new location to install a magnetic observatory on the campus. The lightning data are considered from the National Remote Sensing Centre (NRSC), Hyderabad, India, to investigate the failure of the DFM (digital fluxgate magnetometer) electronics during the lightning strike. The high-resolution topography data are obtained from the Shuttle Radar Topography Mission (SRTM) Global 30 (https://earthexplorer.usgs.gov/, last access: 16 October 2024) to plot the elevation of the region.

The ERT surveys for different profiles (AA', DD', EE', FF' and GG') were carried out using Wenner and Schlumberger arrays with 48 electrodes in the campus. The profiles AA', DD' and EE' with 5 m unit electrode spacing achieved a maximum length of 360 m, whereas FF' and GG' cover 240 m (Maurya et al., 2021). The vertical cross-section along these profiles and a horizontal depth slice at a depth of 30 m below ground level (b.g.l.) derived from a 3D model reflect few linear conductive features and surficial resistivity heterogeneities (Fig. 13a).

The fast Fourier transform (FFT) is a robust technique used by several researchers to calculate the mean depth of layered interfaces of the potential field datasets (Spector and Grant, 1970; Chamoli et al., 2011, 2023; Dwivedi et al., 2019). The power spectrum analysis gives the average depth of the sources with an error limit of 10 % (Mishra and Pedersen, 1982). The 2D radially averaged power spectrum of the RTE magnetic anomaly data shows two linear slope segments corresponding to the average depth of two interfaces around 12±1.2 and 1±0.1 m (Fig. 9).

The tilt depth is an effective method in characterising the location of source edge as well as magnetic source depth using the tilt angle (TDR) approach (Salem et al., 2007). First, the TDR is described by the following equation (Miller and Singh, 1994): 1θ=tan⁡-1∂M∂Z/∂M∂x2+∂M∂y2, where ∂M∂x, ∂M∂y and ∂M∂z are the first derivative of the magnetic field M in the x, y and z directions. The zero contour (θ=0°) demarcates the spatial location of the magnetic source, and tilt amplitudes are restricted in the range of +90° to -90°. Salem et al. (2007) introduced the tilt depth technique using the relationship among tilt angle (θ), depth (Zc) and horizontal location (h) of a contact as 2θ=tan⁡-1hZc. In Eq. (2), the contact location (h=0) lies at the zero values of the contour, and depth relates to the horizontal distance between the 0° and ±45° contour in the TDR map. We apply the technique to generate the tilt angle (TDR), tilt depth solution and histogram plot on the magnetic gridded data. Figure 10 presents the TDR map displaying contours of -45, 0 and +45° (dashed black lines). The TDR shows the short-wavelength anomalies and closely spaced contours that correspond to shallow subsurface sources. The zero contour values of the TDR show the location of the source, whereas the half of the distance between ±45° contours demarcates the depth of the sources. It can be seen that the distance between 0 and +45° is demarcated by relative closeness over the shallow sources and wide expanses over the deeper sources.

The tilt depth solutions of the RTE magnetic anomaly data with depth vary from ∼ 2 to 45 m. Most of the sources lie in shallow depths of ∼ 2 to 20 m and are extended in one direction. The histogram plot shows the variation between the number of depth solution and the depth (Fig. 10c). It is clear from the plot that the number of depth solution is maximum in the depth range of ∼ 2 to 20 m, which corresponds to the shallow source in the subsurface.

We invert the RTE magnetic anomaly to estimate the depth variation of the interface with strong susceptibility contrast. We use the method of Parker (1973), which works in the Fourier domain to estimate the depth variation of an undulated interface. The depth of interface can be computed from the magnetic anomaly due to an uneven, uniform magnetised layer by inversion procedure. The method is improved by incorporating a high-cut filter to ensure the convergence of series and to avoid instability at high wavenumber (Pham et al., 2020). Based on the power spectrum characteristics, we have chosen the cut-off wavenumber between 0.038–0.13 m⁻¹ to remove the high frequencies. The resultant map shows that this interface is deepest in the centre of the survey area and shallowest towards the edges; then the present PVR and the ABS Room are in the zones where this interface is deep, whereas the old PVR which was flooded was in the zone of shallow interface (Fig. 11a). The calculated depth is inferred as the saprolite layer, which varies from ∼ 12 to 16 m with a mean depth of ∼ 14 m. Figure 11b shows the variation of the root mean square (RMS) error against the iteration number. In this case, the inversion process performed 175 iterations, and the RMS error between two successive approximations was reduced from 0.0680 m to 9.9584×10-5 m.

Further, we model the RTE-derived magnetic anomaly data along four profiles, aa', bb', cc' and dd' (Fig. 8b) to delineate the details of the subsurface structures. The forward modelling along these profiles is carried out using the IGMAS+ software, a tool for forward and inverse modelling of potential field datasets (Anikiev et al., 2023). The total length of theses profiles is ∼ 310 m, which shows the depth variations up to 50 m from the surface (Fig. 12a). The 2D models along these profiles explain the presence of three layered structures from the surface up to a depth of 50 m as sandy regolith (∼ 0.3 susceptibility in SI), saprolite (∼ 3 susceptibility in SI) and fissured granite (∼ 2.5 susceptibility in SI). The average susceptibility value for the sandy regolith layer is measured in the field using a KT-10 magnetic susceptibility meter, whereas others are referenced from Telford et al. (1990). The saprolite interface is incorporated in the models based on previous results of ERT data (Nicolas et al., 2019). The average thickness of the sandy regolith layer is ∼ 3 m, whereas the saprolite layer is ∼ 10 m in the models. Both the saprolite and regolith layers show undulations in the all models.

The cc' magnetic profile is close to the vertical resistive cross-section along AA' of Maurya et al. (2021), where resistivity increases from surface to depth (Fig. 13a). The old surface and apparent resistivity data of the region are digitised from the Sanker Narayan et al. (1967) and overlaid over the topography (Fig. 13b, c). The apparent resistivity shows lateral varying high resistivity outward and vice versa inside the Choutuppal campus (Fig. 13b). This indicates that conductivity decreases outward, highlighting the presence of conductive soil layers within the campus.

The saprolite depth calculated by inverting magnetic data at different locations is ∼ 15.5 m (new PVR), ∼ 15 m (ABS Room and CH9), ∼ 13 m (MB and SP), ∼ 14 m (old PVR, CH5, CH6, CH7) and ∼ 14.5 m (CH4) (Fig. 11a), whereas depth estimation by the ERT data is ∼ 13.5 m (new PVR), ∼ 18 m (ABS Room), ∼ 15 m (CH9), ∼ 14 m (MB, SP and CH4), ∼ 22 m (old PVR and CH6), ∼ 29 m (CH5) and ∼ 17 m (CH7) respectively (Fig. 13d). The calculated depth from these two different datasets shows a discrepancy of ∼ 2 m, except at the locations of CH5, CH6 and old PVR. The ERT survey estimates the upper fissured layer depth (∼ 9 to 33 m) of the Choutuppal campus and shows the undulated interface creating compartmentalised aquifers (Nicolas et al., 2019).