The research, conducted by Nihal Tekin Ünlütürk of Erciyes University and Uğur Doğan of Yıldız Technical University, evaluated the effect of tropospheric corrections derived from Global Navigation Satellite System (GNSS) observations against those produced by the Generic Atmospheric Correction Online Service for InSAR, widely known as GACOS. Tropospheric delay is one of the dominant error sources in InSAR processing, and if left uncorrected it can distort the interpretation of earthquake-induced deformation fields—the very patterns scientists rely on to infer fault slip, rupture extent, and seismic hazard. The team’s central question was deceptively simple: which correction approach actually improves the displacement maps, and by how much?

To find out, the researchers processed both ascending and descending Sentinel-1A radar images from the European Space Agency’s Copernicus mission using the Sentinel Application Platform (SNAP) within a standard Differential InSAR (DInSAR) workflow. This involved generating interferometric pairs spanning the earthquake, applying phase filtering to suppress noise, and performing phase unwrapping with the Statistical-Cost, Network-Flow Algorithm for Phase Unwrapping, or SNAPHU—a computational step that recovers continuous displacement values from the wrapped fringes that radar phase measurements naturally produce. Working with both ascending and descending orbits matters because each viewing geometry samples the ground differently along the satellite line-of-sight, providing complementary constraints on the three-dimensional deformation field.

The GNSS-based correction strategy exploited Turkey’s dense TUSAGA-Active continuous GNSS network, whose data were provided by the General Directorate of Mapping. Because GNSS signals also pass through the troposphere, each station continuously estimates the zenith tropospheric delay—the total delay a signal experiences traveling vertically through the atmosphere. By regressing these delays against station elevation, the researchers derived parameters that allowed them to predict the topography-correlated component of the atmospheric delay across the entire interferogram, interpolating between stations and removing the estimated delay from the radar phase. GACOS, by contrast, is an online service that generates correction maps by combining global weather model data with an iterative tropospheric decomposition model, separating the stratified, elevation-dependent part of the delay from turbulent components.

The results are a sobering reminder that atmospheric correction is not a solved problem. Both GNSS- and GACOS-based corrections clearly influenced the spatial distribution of the line-of-sight (LOS) displacement fields, reshaping the apparent deformation patterns in ways that could alter scientific conclusions. Yet their effects were not spatially uniform across the study area. In some regions the corrections agreed closely; in others, the corrected products diverged noticeably, leaving analysts with genuinely different pictures of how the ground moved. The study found that neither correction strategy produced a spatially uniform improvement across all interferometric pairs—a finding that challenges the common assumption that applying any atmospheric correction is automatically better than applying none.

The local discrepancies between the corrected products, the authors suggest, likely stem from a combination of factors: regional atmospheric variability, the rugged topography of the East Anatolian Fault Zone, the acquisition geometry of the satellite passes, and the uneven distribution of GNSS stations. Eastern Anatolia is a region of sharp elevation contrasts, where stratified tropospheric delay can change rapidly over short horizontal distances. Where GNSS stations are sparse, interpolation of zenith delays becomes uncertain, and turbulent water vapor variations—which are poorly correlated with elevation—can dominate. Weather-model-based corrections like GACOS, meanwhile, are limited by the spatial and temporal resolution of the underlying meteorological data, which may not capture localized convective moisture features at the moment of a satellite overpass.

Rather than declaring a single winner, the study advocates a more rigorous, diagnostic approach. The researchers emphasize the need to carefully evaluate atmospheric correction performance using spatial statistics, profile comparisons across the deformation field, and displacement-difference analyses between corrected and uncorrected products. In other words, the choice of correction should be treated as a hypothesis to be tested for each interferogram and each tectonic setting, not a routine checkbox. This methodological caution has practical consequences: deformation signals from the Elazığ–Sivrice rupture feed into models of fault slip and earthquake source parameters, and atmospheric artifacts of comparable magnitude could bias estimates of how much the fault moved and where.

The 2020 Elazığ–Sivrice earthquake itself makes a compelling case study. The event struck the East Anatolian Fault Zone, a major strike-slip boundary accommodating the westward extrusion of the Anatolian plate between the Arabian and Eurasian plates. Previous studies have used InSAR and GNSS data to characterize its rupture behavior, and the region’s tectonic importance has motivated extensive geodetic monitoring. The fault zone’s combination of strong topographic relief, semi-arid climate with pronounced seasonal atmospheric variation, and a growing but still uneven GNSS infrastructure makes it an ideal proving ground for atmospheric correction techniques that must ultimately work worldwide.

The broader significance of the study lies in its framing of GNSS- and GACOS-based corrections as complementary rather than competing tools. Because they draw on fundamentally different information sources—direct physical measurements of atmospheric delay at ground stations versus model-derived estimates from meteorological data—their residuals behave differently. Comparing them side by side provides a way to identify atmospheric contributions embedded in interferometric phase and to assess the uncertainty of InSAR-derived earthquake deformation fields. For hazard assessment, volcano monitoring, and land subsidence studies, where InSAR increasingly underpins operational decision-making, this kind of uncertainty quantification is becoming essential.

The work also fits into a vibrant international research effort. Recent studies have compared GNSS local observations with global weather-based models over tropical volcanoes, applied GNSS spatial interpolation for atmospheric correction in New Zealand, and used neural networks trained on GNSS delays for terrain-challenging regions. The Turkish study adds a critical data point from an active continental strike-slip setting, demonstrating that the performance hierarchy of correction methods is context-dependent. As Sentinel-1 and future radar missions deliver ever more frequent global coverage, and as GNSS networks densify, the message from Elazığ–Sivrice is clear: the atmosphere will always leave its fingerprint on radar interferograms, and taming it demands both better correction products and smarter validation of when, where, and how much those products actually help.

Subject of Research: Comparative assessment of GNSS- and GACOS-based tropospheric corrections for InSAR-derived surface deformation from the 2020 Elazığ–Sivrice earthquake

Article Title: Comparative assessment of GNSS- and GACOS-based tropospheric corrections for InSAR-derived surface deformation: a case study of the 2020 Elazığ–Sivrice earthquake

Article References: Ünlütürk, N. T., & Doğan, U. (2026). Comparative assessment of GNSS- and GACOS-based tropospheric corrections for InSAR-derived surface deformation: a case study of the 2020 Elazığ–Sivrice earthquake. Earth Science Informatics, 19(10), Article 182. https://doi.org/10.1007/s12145-026-02236-1

Image Credits: AI Generated

DOI: 10.1007/s12145-026-02236-1

Keywords: InSAR, GNSS, GACOS, tropospheric correction, surface deformation, Elazığ–Sivrice earthquake, Sentinel-1A, DInSAR, East Anatolian Fault Zone, SNAPHU, atmospheric delay, Earth Science Informatics