LAGO: Distributed Astroparticle Observatory

• LAGO is a distributed astroparticle observatory network featuring water-Cherenkov detectors deployed across diverse geographic sites to capture cosmic and atmospheric signals.
• Its scientific program targets high-energy gamma-ray bursts, space weather modulation, and atmospheric radiation using techniques like Single Particle and multi-spectral analysis.
• The observatory leverages simulation frameworks, VEM calibration, and decentralized data repositories to enhance reproducibility, capacity building, and open-science.

LAGO usually denotes the Latin American Giant Observatory, an extended, non-centralized astroparticle observatory built from a distributed network of autonomous water-Cherenkov detectors and associated local data repositories deployed across Latin America, Antarctica, and, in later accounts, Spain. Established in 2005, it was conceived to study the high-energy component of gamma-ray bursts, space weather, and atmospheric radiation at ground level, while also developing a continent-scale infrastructure for simulations, data preservation, reproducibility, and training (Mora et al., 2017, Asorey et al., 2017).

1. Origins, collaboration model, and geographic logic

LAGO was organized as a distributed collaboration rather than a single-site instrument. A 2017 overview described it as a non-centralized collaborative union of more than 30 institutions from ten countries, while later work described a non-centralized alliance of 30 institutions in 11 countries operating detectors in 10 Latin American countries and beyond (Sidelnik et al., 2017, Rubio-Montero et al., 2022). Earlier collaboration-oriented work reported more than 80 scientists from more than 25 institutions spanning Argentina, Bolivia, Brazil, Colombia, Ecuador, Guatemala, Mexico, Peru, Venezuela, and Spain (Asorey et al., 2016).

The network’s defining feature is geographic heterogeneity. LAGO sites extend from Mexico to the Antarctic Peninsula and from sea level to more than 5500 m above sea level, thereby sampling very different atmospheric depths and geomagnetic rigidity cutoffs (Rubio-Montero et al., 2021). This spatial distribution was not treated as a mere deployment constraint; later accounts explicitly describe it as a scientific asset, because simultaneous measurements at distinct latitudes and altitudes expose different primary-rigidity thresholds and different secondary-particle mixtures at ground level (Sarmiento-Cano et al., 15 Jan 2026).

That logic is central to the observatory’s design. A site at high altitude experiences reduced atmospheric overburden and therefore enhanced electromagnetic secondaries at ground level, whereas lower-altitude or higher-cutoff sites emphasize more penetrating components and longer-term modulation signatures. LAGO’s continent-scale architecture therefore functions as a comparative instrument, in which local environmental diversity is deliberately converted into differential sensitivity for astrophysical and heliophysical studies (Sidelnik et al., 2017, Rubio-Montero et al., 2022).

2. Scientific program and measurement principles

LAGO’s scientific program has consistently been organized around three lines: the search for high-energy components of gamma-ray bursts and related extreme-universe phenomena, the study of space weather through modulation of galactic cosmic rays, and the characterization of atmospheric radiation at ground level (Mora et al., 2017, Sidelnik et al., 2017). Later summaries retain those goals while emphasizing regional-scale, simultaneous monitoring of secondary cosmic radiation for transient heliospheric disturbances and longer-term solar modulation (Alberto et al., 2019).

The physical rationale is governed by geomagnetic filtering, atmospheric development of extensive air showers, and Cherenkov detection in water. The rigidity of a charged primary is written as

$R=\frac{pc}{Ze},$

and LAGO explicitly exploits the fact that different sites admit different primary populations because their cutoff rigidities differ with geomagnetic location and disturbed-field conditions (Rubio-Montero et al., 2022). Atmospheric depth is represented conceptually as

$X(h)=\int_h^\infty \rho(h')\,dh',$

so changes in altitude directly alter the attenuation and composition of secondaries reaching the detector (Rubio-Montero et al., 2022).

At detector level, the relevant emission process is Cherenkov radiation. The condition for emission in water is

$\beta>\frac{1}{n},$

with Cherenkov angle

$\cos\theta=\frac{1}{n\beta}.$

These relations underpin the use of water-Cherenkov detectors as broadband monitors of charged secondaries from air showers (Rubio-Montero et al., 2022).

Operationally, LAGO uses two closely related modes. In scaler mode, or the Single Particle Technique (SPT), detectors count threshold crossings in fixed time bins to search for statistically significant departures from background; this is the core mode for gamma-ray burst searches and for space-weather monitoring (Sidelnik et al., 2017, Mora et al., 2017). In addition, LAGO developed a multi-spectral analysis technique (MSAT) in which deposited-charge bands are associated with different secondary components, enabling component-resolved studies of solar modulation and Forbush decreases (Rubio-Montero et al., 2022, Rubio-Montero et al., 2021). Environmental corrections are integral to that program; one standard form reported in the simulation literature is

$$\frac{\Delta R}{R}=\alpha_P\,\Delta P+\alpha_T\,\Delta T,$$

with site-specific pressure and temperature coefficients (Rubio-Montero et al., 2022).

3. Detector architecture, acquisition systems, and calibration

The canonical LAGO detector is a light-tight water-Cherenkov detector with a diffusely reflective inner lining and a top-mounted photomultiplier tube. One overview describes plastic tanks filled with purified water and instrumented with one to four large-area PMTs, while the later standardized simulation framework describes cylindrical WCDs typically in the $1\!-\!4\,\mathrm{m}^3$ range with a single PMT pointing downward (Sidelnik et al., 2017, Rubio-Montero et al., 2022). Inner reflective materials include Tyvek-like diffusive surfaces or optimized reflective coatings, and water treatment is emphasized because optical transparency directly affects calibration stability and signal yield (Mora et al., 2017, Sidelnik et al., 2017).

Front-end electronics evolved over time but retained the same design priorities: autonomous operation, environmental monitoring, timing synchronization, and remote transfer. A 2017 overview describes a custom 40 MHz electronics board controlled by a Digilent Nexys2 FPGA, with GNSS timing and Arduino-based environmental sensing (Sidelnik et al., 2017). By 2019, newer systems based on RedPitaya STEMLab boards were demonstrated, with automatic baseline correction, GPS timing, HV monitoring, and networked storage; CAEN-based digitizers also appear in site-specific deployments such as Chiapas (Alberto et al., 2019, Mora et al., 2017). In the cloud-integration literature, the acquisition and monitoring layer is formalized as ACQUA, LAGO’s custom hardware and firmware system for detector telemetry, local atmospheric conditions, and secondary-particle fluxes (Rubio-Montero et al., 2021).

Calibration is organized around the Vertical Equivalent Muon (VEM). The detector response is tracked through charge histograms in which a muon-dominated feature serves as a fiducial reference for gain and stability (Sidelnik et al., 2017). Geant4-based detector studies reported that, in a standard LAGO WCD, a 3 GeV vertical muon yields the most probable signal of approximately 100 photoelectrons, corresponding to an energy deposition of about 180 MeV (Alberto et al., 2019). A later overview adds a complementary in-situ method based on Michel electrons from stopped-muon decays, described as improving calibration stability and signal-to-noise ratio without external instrumentation (Sarmiento-Cano et al., 15 Jan 2026).

The Chiapas-Mexico node illustrates how this general detector concept is instantiated locally. There, UNACH assembled a prototype WCD with Tyvek lining, multilayer external shielding, a Photonis 9-inch XP1805 PMT, a CAEN V1718/V6533/V1720 stack, and an Escaramujo scintillator telescope for PMT characterization and muon-based calibration, with the short-term goal of deployment on Tacaná volcano at 4092 m a.s.l. (Mora et al., 2017). The same report explicitly frames the site as a training platform for students and early-career researchers in detector construction, DAQ, calibration, and analysis (Mora et al., 2017).

4. Simulation, analysis, and computational frameworks

LAGO’s simulation program is organized around an end-to-end chain that links geomagnetic access, atmospheric transport, primary spectra, and detector response. In the 2021 and 2022 infrastructure papers, that chain is formalized as ARTI, which couples GDAS atmospheric profiles, MAGCOS geomagnetic transmission using IGRF13 and Tsyganenko TSY2001, CORSIKA air-shower simulation, and a Geant4 detector model including geometry, reflective coating, water properties, PMT response, and electronics (Rubio-Montero et al., 2021, Rubio-Montero et al., 2022). Outputs are organized into three levels: S0 for raw CORSIKA shower outputs, S1 for analyzed ground-level secondaries, and S2 for detector-response simulations and calibration histograms (Rubio-Montero et al., 2021).

The same framework is used to compute site-specific directional rigidity cutoffs and select only primaries that exceed local transmission thresholds. In the 2022 standardization paper, the geomagnetic filtering is written in terms of a directional cutoff tensor $R_C(\theta,\varphi;\phi,\lambda)$, and primary cosmic-ray species from hydrogen to iron are injected over

$1\,\mathrm{GeV}\le E_p \le 10^6\,\mathrm{GeV},$

with atmospheric development handled by CORSIKA and detector response by Geant4 (Rubio-Montero et al., 2022). Typical runs for a 1 m$^2$ detector integrated over one hour require on the order of $(2\!-\!5)\times 10^7$ primaries, and compressed CORSIKA output is reported at roughly 25 GB per site-hour configuration (Rubio-Montero et al., 2022).

LAGO’s space-weather simulation chain predates the cloud standardization effort and already emphasized direction- and time-resolved geomagnetic effects. In the 2018 study of Bucaramanga and Bariloche during the May 2005 geomagnetically active period, MAGNETOCOSMICS and CORSIKA were used to model cutoff variations, penumbral structure, and ground-level secondaries; the analysis concluded that neutrons dominate the geomagnetic sensitivity in the relevant momentum band, with reductions of order $X(h)=\int_h^\infty \rho(h')\,dh',$0, while effects above approximately 10 GeV become negligible (Durán et al., 2018).

The analysis layer is formalized as ANNA, which consolidates measured and simulated data and implements MSAT-oriented processing (Rubio-Montero et al., 2021). A later overview describes ARTI-MEIGA as a key development extending the full-chain simulation concept into a broader open-science and application-oriented framework, including muography, radiation-dose estimation, and environmental sensing (Sarmiento-Cano et al., 15 Jan 2026). This suggests a progressive shift from site-specific detector modeling toward a reproducible digital observatory in which simulation products are treated as first-class scientific objects.

5. Distributed repositories, persistent identification, and open-science infrastructure

LAGO’s data architecture is explicitly decentralized. The repository paper defines a continent-spanning distributed network in which each site preserves, catalogs, and generates data locally, including raw detector output and derived products from analyses and simulations (Asorey et al., 2017). This site-centric design was motivated by Data Accessibility, Reproducibility, and Trustworthiness (DART), as promoted by CHAIN-REDS, and contrasted with centralized observatory models in which data flow from a single place into a repository system (Asorey et al., 2017).

The core software stack initially centered on DSpace, with OAI-PMH for metadata harvesting, SWORD for automated deposit, and GRNET PID services for persistent identifiers (Asorey et al., 2017). The repository paper also notes a practical limitation of native DSpace—its inability to upload or download multiple records at once—which LAGO addressed through a custom batch-ingest script (Asorey et al., 2017). Earlier collaboration-wide documentation provides the corresponding metadata context: measured, simulated, and calibration datasets were curated in LAGOData, using Dublin Core plus custom fields for detector electronics, site coordinates, altitude, voltage, pressure, temperature, CORSIKA input files, included libraries, and computational environment descriptors such as uname -a, lsb_release -a, free, and gcc -v (Asorey et al., 2016).

The same papers frame the repository system as part of a broader reproducibility pipeline. Around 2016, LAGO reported data volumes of approximately 150 GB per month per detector, around 1.5 TB per month for the collaboration, and roughly 10 GB per site for CORSIKA particle-flux datasets, which made persistent cataloging and programmatic reuse operational necessities rather than archival conveniences (Asorey et al., 2016). The repository strategy supported not only detector data, but also the complete simulation workflow needed to characterize each site (Asorey et al., 2016).

From 2021 onward, this ecosystem was extended into the EOSC-Synergy cloud environment. There, certified Docker images for ARTI stages run on EGI FedCloud resources provisioned through IM or EC3, usually under Slurm, while storage is handled by EGI DataHub (OneData) (Rubio-Montero et al., 2021, Rubio-Montero et al., 2022). Metadata are expressed in JSON-LD 1.1 and DCAT-AP2; identities are federated through EduTeams and EGI Check-in; persistent identifiers are minted via Handle.net through B2HANDLE; and discovery is exposed through B2FIND (Rubio-Montero et al., 2021, Rubio-Montero et al., 2022). In practical terms, the decentralized DSpace-plus-PID model was not abandoned; it was generalized into a cloud-native FAIR workflow with stronger provenance capture and standardized simulation publication.

6. Results, extensions, and terminological scope

LAGO’s observational and computational programs have produced both methodological and scientific results. A 2017 overview reports a possible gamma-ray burst candidate at the Chacaltaya station on 7 December 2011 at 15:45:49.675 UTC with duration 5.5 s, identified through the Single Particle Technique and adaptive moving-window analysis (Sidelnik et al., 2017). The same paper describes a Forbush decrease on 8 March 2012 detected in the electromagnetic-dominated band by a 1.8 m$X(h)=\int_h^\infty \rho(h')\,dh',$1 WCD in Bariloche and compared against the Rome neutron monitor, illustrating the practical value of MSAT for low-cost space-weather monitoring (Sidelnik et al., 2017).

Subsequent work deepened the site-specific and regional interpretation of such phenomena. The May 2005 simulation study showed larger storm-time depressions at Bucaramanga than at Bariloche and identified neutrons as the most affected component, with the timing of the simulated decreases matching neutron-monitor observations at similar cutoff rigidities (Durán et al., 2018). More recent collaboration summaries extend the scientific scope further, describing applications in volcano muography, industrial muon imaging, chloride-enhanced neutron sensitivity for soil-moisture sensing, and space-weather campaigns in the South Atlantic Magnetic Anomaly (Sarmiento-Cano et al., 15 Jan 2026). In those accounts, LAGO is no longer only a gamma-ray burst and space-weather network; it is a multidisciplinary platform linking astroparticle physics, environmental monitoring, instrumentation, and regional capacity building.

Capacity building is not incidental to that trajectory. The Chiapas development report explicitly presents LAGO as a mechanism for introducing hands-on astroparticle instrumentation in regions where activity had previously been restricted to analysis of remote detectors (Mora et al., 2017). Later overviews connect that educational role to ERASMUS+ capacity-building projects such as LA-CoNGA and EL-BONGÓ, and to open-hardware, edge-computing, and data-integrity efforts (Sarmiento-Cano et al., 15 Jan 2026). A plausible implication is that LAGO’s distributed architecture has been as important pedagogically as it has been scientifically.

Within arXiv more broadly, however, LAGO is not unique to the observatory. The acronym is also used for unrelated frameworks in public-health trial design, online reinforcement learning, optimization, robotic manipulation, cross-lingual embedding inversion, and temporal community detection (Bing et al., 2023, Liu et al., 23 Jun 2026, Dieren et al., 3 Mar 2026, Shi et al., 16 Jun 2026, Yu et al., 21 May 2025, Brabant et al., 1 Oct 2025). In astroparticle physics and related data-infrastructure work, by contrast, LAGO conventionally denotes the Latin American Giant Observatory and its associated detector, simulation, and repository ecosystem.