Satellite Geodesy Unveils a Decade of Summit Subsidence at Ol Doinyo Lengai Volcano, Tanzania

The processing of hundreds of Synthetic Aperture Radar (SAR) images acquired by two satellite systems: Sentinel‐1 and COSMO‐SkyMed reveals a decade of ground deformation for a ∼0.5 km diameter area around the summit crater of the only active carbonatitic volcano on Earth: Ol Doinyo Lengai in Tanzania. Further decomposing ascending and descending orbits when the appropriate SAR data sets overlap allow us to interpret the imaged deformation as ground subsidence with a significant rate of ∼3.6 cm/yr for the pixels located just north of the summit crater. Using geodetic modeling and inverting the highest spatial resolution COSMO‐SkyMed data set, we show that the mechanism explaining this subsidence is most likely a deflating very shallow (≤1 km depth below the summit crater at the 95% confidence level) magma reservoir, consistent with geochemical‐petrological and seismo‐acoustic studies.


Introduction
Although crucial for eruption forecasting, imaging localized summit deformation at active volcanoes is challenging with ground-based geodetic methods due to potential accessibility issues and the occurrence of vigorous eruptive phases that could damage instruments.Interferometric Synthetic Aperture Radar (InSAR) is a good alternative to obtain satellite geodetic measurements with a cm or even better accuracy (Hanssen, 2001).In particular, the processing of hundreds of SAR images in InSAR time series can uncover previously unknown deformation processes over several years to a decade, as shown recently in active volcanic areas in Guatemala (Gonzalez-Santana & Wauthier, 2021), the East African Rift (Albino & Biggs, 2021;Wauthier et al., 2018), and Indonesia (Zorn et al., 2023).However, the spatial resolution can be a limiting factor to obtain enough deformation measurements for highly localized deformation processes around or inside summit craters at volcanoes worldwide (Bemelmans et al., 2023;Pritchard et al., 2022;Richter et al., 2013).Localized deformation around or inside craters can thus easily be missed using >10 m spatial resolution for typical SAR sensors, especially when high multilooking factors (factors increasing the pixel size by averaging multiple pixels in range or azimuth direction) or strong smoothing and filters are being used (Pritchard et al., 2022).
Measuring deformation at Ol Doinyo Lengai, Tanzania, highlights these challenges due to remote access and vigorous summit activity, preventing the easy installation and maintenance of GNSS stations near the summit crater (Daud et al., 2023;Jones et al., 2019), as well as the steep slopes and relatively small (∼0.2 km across) active summit crater.A new cone developed following a rather unusual explosive-effusive sequence in 2007-2008 (Biggs et al., 2013;Keller et al., 2010) and preliminary InSAR results shown in (Tournigand et al., 2023) suggest that this cone might experience subsidence.In this study, we closely imaged the spatial and temporal evolution of the ground deformation on the upper slopes of Ol Doinyo Lengai around the summit crater using decadal SAR observations from two satellites and interpret it as likely deflation of a very shallow magma reservoir beneath Ol Doinyo Lengai's summit crater.• The modeled reservoir is less than 1 km deep below the summit crater, consistent with geophysical and geochemical studies

Supporting Information:
Supporting Information may be found in the online version of this article.

Background
Located in the Natron Basin in the Eastern Branch of the East African rift, Ol Doinyo Lengai (Figure 1) is the only active carbonatitic volcano on Earth experiencing effusive and explosive eruptions (Keller et al., 2010).The Natron Basin sustained an unusual period of intense seismic and magmatic unrest between 2007 and 2008 and extensive ground deformation associated with faulting and diking across the dormant Gelai volcano and the Naibor Soito volcanic field (Baer et al., 2008;Biggs et al., 2013;Calais et al., 2008).Likely related to this intrusive activity, Ol Doinyo Lengai suddenly began to experience episodic explosive eruptions on 4 September 2007, with a maximum volcanic explosivity index of 3 after 25 years of effusive eruptions (Kervyn et al., 2010).The explosive eruptions blasted a new crater over 100 m deep and 300 m wide in an area north of the Ol Doinyo Lengai summit, where there was previously a platform of solidified lava that started to fill the old crater in 1983 (Laxton, 2020).The explosive activity continued until the end of April 2008, and returned to effusive activity at the end of 2008 (Keller et al., 2010;Kervyn et al., 2010).Since then, several effusive eruptions have been confirmed between 2008 and 2023.Furthermore, a fissure below the western rim of the summit crater (Figure 1c) was first discovered in 2013, observed to contain lava, and measured to be 30 m long and 1 m wide on 4-5 July 2014, and about 100 m long and 1-10 m wide in 2021 (Tournigand et al., 2023).
The geometry and characteristics of the shallow magma plumbing system below Ol Doinyo Lengai remain elusive with contradictions or large uncertainties in the literature, especially for the horizontal location of a shallow (∼3 km depth) reservoir (Figure 3).Recent geophysical seismic studies (Roecker et al., 2017;Weinstein et al., 2017) suggest deep reservoirs probably corresponding to a complex network of sills extending in the upper crust (between 5 and 20 km depth) below Ol Doinyo Lengai and extending to the east under the Nabor Soito volcanic field.An offset magma reservoir slightly to the East of the Ol Doinyo Lengai summit crater at ∼3 km depth has also been proposed by a geodetic study using InSAR acquired in 2007-2008 (Biggs et al., 2013) and more recently at a depth between ∼3 and 5 km, but offset ∼4 km to the northeast of the Ol Doinyo Lengai summit crater by using subtle GNSS and InSAR signals spanning 2016-2021 (Daud et al., 2023).A new seismic and acoustic study infers a shallow magma source right under the Ol Doinyo Lengai cone (Reiss et al., 2023), in agreement with an older petrological study that infers a very shallow ∼0.6 km deep magma reservoir right below the Ol Doinyo Lengai edifice connected to a deeper ∼3.3 km one (Petibon et al., 1998).

Geodetic Analysis
We measured ground deformation at Ol Doinyo Lengai using multitemporal InSAR data processed with the GAMMA software (Werner et al., 2000).We used a 12-m Tandem-X Digital Elevation Model, concatenating geocoded tiles acquired between 2010 and 2015 (B.Wessel, 2016), to remove the topographic contributions.Note that no atmospheric correction was applied given their current inability to fully characterize atmospheric signals, especially at steep volcanoes (Stephens et al., 2020;Sun et al., 2020).Finally, we applied the small baseline analysis subset method (Berardino et al., 2002) to process InSAR time series.We processed an X-band (wavelength ∼3.1 cm) SAR data set acquired by the COSMO-SkyMed constellation made of 97 SAR images covering 2013-2014 (Table 1).We also used 214 C-band (wavelength ∼5.6 cm) Sentinel-1 SAR images covering 2015-2023 in ascending orbits, as well as 88 Sentinel-1 SAR images covering 2018-2021 along descending orbits (Table 1).
For the period 2018-2021, for which we have both ascending and descending Sentinel-1 data, we furthermore used the extra functionalities of the Multidimensional Small Baseline approach (Samsonov & d'Oreye, 2017) that allows the two Sentinel-1 Line-Of-Sight (LOS) geometries (ascending and descending) to be decomposed into vertical and East-West horizontal displacements, assuming a negligible North-South component of displacements.
For all InSAR time series results, A 10 by 10 pixel relative reference region (assuming zero displacements) centered over the location of the OLO3 station belonging to the TZVOLCANO GNSS Network (Daud et al., 2023;Stamps et al., 2016aStamps et al., , 2016b) is chosen (Figure 1c).The processed GNSS time series were obtained from the Nevada Geodetic Observatory database, which uses MIDAS (Blewitt et al., 2018) to detrend the data to remove any plate motion signal, consistent with the relative reference of InSAR data sets also being inside the rift and thus insensitive to plate motion.The location of the OLO3 station was chosen as the time series reference point for having a relatively stable location (see Figure S1 in Supporting Information S1) and offering enough Geophysical Research Letters 10.1029/2023GL107673 selected pixels while not being too isolated and disconnected from the deformation area of interest around the summit crater.
InSAR time series LOS displacement maps show a concentric zone of ∼0.5 km diameter at around the summit crater of Ol Doinyo Lengai (starting on the northern upper slopes above ∼2,730 m of elevation) with ground moving away from the satellite with a steady rate of displacement over time of ∼ 4.1, 2.6, and 3 cm per year, for the COSMO-SkyMed, Sentinel-1 descending, and Sentinel-1 ascending InSAR time series, respectively (Figure 2).The decomposition into vertical and East-West displacements using the common timespan of both Sentinel-1 data sets show further that most of these LOS displacements correspond to vertical ground subsidence with an average subsidence rate of 3.6 cm/yr.

Discussion
Although we only have a limited number of reliable (coherent in time) pixels around the summit crater to the north and south of it, the subsidence signal is clear, steady over a decadal period, and correlated spatially with the post-2007 summit cone (Figure 1c).Biannual cyclic patterns are also observed in the InSAR time series and are  (Tournigand et al., 2023).Blue star shows the location of the deformation time series plotted in Figure 2.  especially recognizable when the displacements are split into vertical and East-West components (Figures 2d and  2e), as observed in other areas for example, in Central America (Gonzalez-Santana et al., 2024).The oscillating displacement patterns correspond to periods of high precipitation in the Natron Basin.The two wet seasons occur in March-May and October-December (Rey et al., 2021).There is more water vapor in the atmosphere and the ground swells with rainwater during these months, which explains the observed cyclicity in the deformation data.Similar seasonal effects have been identified at other East African rift volcanic targets and show that typical InSAR time series uncertainties for Sentinel-1 on long-term multi-year deformation trends are of the order of ∼0.2-0.8 cm/year (Albino & Biggs, 2021).
Active volcanic summit areas may subside through a variety of processes, such as volumetric changes in shallow magma reservoirs, lava flow, or other surface deposits cooling and compaction (Lu et al., 2005;Wauthier et al., 2013), magma intrusion cooling and compaction (Turcotte & Schubert, 2002), and/or normal slip along ring faults, which are found at the edge of calderas (Jónsson et al., 2005).
In this case, lava flow subsidence can be ruled out given the lack of recent lava flows on the upper flanks of the volcano.The subsidence pattern includes and extends beyond the spatial extent of the 2007-2008 summit cone.This observation coupled with the high and sustained subsidence rates of ∼3.6 cm/yr for at least 10 years, make cone compaction or loading effects unlikely.
To generate about 0.36 m of subsidence over 10 years, a cooling layer or intrusions would need to be of the minimum following thickness (Turcotte & Schubert, 2002): Using general values of a coefficient of thermal expansion (κ) of 5 * 10 5 K 1 (Huppert & Sparks, 1988) and a temperature change (ΔT) of 1,150°C-27.5°C(corresponding to a depth of ∼0.5 km).A cooling sill of ∼6 m thick is on the upper bound of observed East African sills but is plausible (Acocella, 2021).
Given the significant volcanic activity of Ol Doinyo Lengai, another more likely possibility is that the ground is sinking due to a deflating shallow magma reservoir (Figure 3) as suggested by geochemical (Petibon et al., 1998) and new recent seismic-acoustic data (Reiss et al., 2023).We performed geodetic inversions (Fukushima et al., 2005;Sambridge, 1999aSambridge, , 1999bSambridge, , 1999c;;Wauthier et al., 2012) of geometrical and overpressure or volume which likely feeds a 3 km depth reservoir, whose exact horizontal location is highly debated in the literature (Biggs et al., 2013;Calais et al., 2008;Daud et al., 2023;Petibon et al., 1998), that then allows magma transport into a very shallow (≤1 km depth, This study; Petibon et al., 1998;Reiss et al., 2023) reservoir located inside the upper volcanic edifice and feeding eruptions in the summit crater.
parameters minimizing the misfit (details in the Supporting Information S1) of the observed COSMO-SkyMed data set against modeled displacements using a point-source (Mogi, 1958) and a penny-shaped crack embedded in an elastic half-space (Battaglia et al., 2013a;Fialko et al., 2001).The lowest misfit model is a pointsource (Mogi, 1958) simulating a deflating magma reservoir experiencing a volume decrease of ∼19,000 m 3 at a very shallow depth of ∼200 m below the summit crater (see more details in the Text and Table S1 of Supporting Information S1).
A fissure at the western edge of the summit crater grew from 30 m long in 2013 to over 100 m long as of 2021 (Tournigand et al., 2023).The subsidence around the summit crater could be related to the growth of the fissure.Similar features also exist at other volcanoes, such as at Okmok (Johnson et al., 2010), Tendürek (Bathke et al., 2013), and Sierra Negra volcanoes (Jónsson et al., 2005).This fissure could further elongate as Ol Doinyo Lengai continues to erupt and subside.Therefore, it is important to keep investigating this area for ground deformation and future geodetic modeling efforts should account for the effect of ring faults and other crustal heterogeneities on observed surface deformation patterns, or else geodetic models can be biased to locate shallower magma sources and underestimate volume changes (Beauducel et al., 2004;Lu et al., 2005;Masterlark, 2007).A more thorough investigation of magma reservoir shapes, depths, volume or pressure change, and interactions with a possible ring fault system at Ol Doinyo Lengai could be undertaken using more complex numerical modeling methods (e.g., Jónsson et al., 2005) but goes beyond the scope of this study.
The lack of ground deformation monitoring around and inside summit craters at volcanoes can prevent accurate eruption and related hazards forecasting and the characterization of very shallow magma reservoirs.We show here that using multitemporal decadal InSAR time series data offers a good opportunity to image a localized subsidence deformation pattern of about 0.5 km in diameter around an active summit crater with InSAR satellite geodesy.The presence of a very shallow (≤1 km at the 95% confidence interval) deflating magma reservoir is inferred with geodetic modeling for the first time at Ol Doinyo Lengai.It is particularly crucial to image and model summit deformation processes at active volcanoes at risk of caldera collapse or subject to flank instability that can lead to catastrophic collapse.
diameter subsiding area around the summit crater at Ol Doinyo Lengai • The subsidence is likely due to a deflating shallow magma reservoir (point-source)

Figure 1 .
Figure 1.(a) East African rift with Holocene volcanoes as red triangles (Global Volcanism Program, 2024).Red square shows the location of the Lake Natron area.(b) Natron area including the active Ol Doinyo Lengai carbonatitic volcano.GNSS stations from the TZVOLCANO GNSS network (Stamps et al., 2016b) are marked with black circles with the area of 10 by 10 pixels used as our Interferometric Synthetic Aperture Radar time series reference centered on OLO3.(c) Close-up on Ol Doinyo Lengai summit area showing a major fissure that opened since 2013 in red and a high-resolution hill shade topography of the summit cone that formed during the 2007-2008 eruption(Tournigand et al., 2023).Blue star shows the location of the deformation time series plotted in Figure2.

Figure 2 .
Figure 2. Displacement maps in centimeters of the Ol Doinyo Lengai summit area (geographical extent is the same as Figure 1c) and corresponding time series of the deformation for a group of coherent pixels centered on the location of the summit cone (see blue star location on Figure 1c).(a) Line-Of-Sight (LOS) displacement map for the COSMO-SkyMed descending data set spanning February 2013-November 2014.(b) LOS displacement map for the Sentinel-1 descending data set spanning July 2018-November 2021.(c) LOS displacement map for the Sentinel-1 ascending data set spanning May 2015-May 2023.Note that a decrease in LOS displacements corresponds to a range (satellite-Earth distance) increase.(d) Vertical (Up-Down) displacement map and (e) Horizontal East-West displacement map for the period when both Sentinel-1 data sets overlap.Note that on all displacement maps, the deformation is only shown at reliable pixels (Interferometric Synthetic Aperture Radar phase signal is coherent through time).Zero displacement in all time series plots corresponds to the first Synthetic Aperture Radar image acquisition.

Figure 3 .
Figure 3. Conceptual illustration of the shallow magma plumbing system beneath Ol Doinyo Lengai volcano.Magma is transported from a deep reservoir (≥17 km depth,Roecker et al., 2017) to a sill complex (≥10 km depth,Reiss et al., 2022) which likely feeds a 3 km depth reservoir, whose exact horizontal location is highly debated in the literature(Biggs et al., 2013;Calais et al., 2008;Daud et al., 2023;Petibon et al., 1998), that then allows magma transport into a very shallow (≤1 km depth, This study;Petibon et al., 1998;Reiss et al., 2023) reservoir located inside the upper volcanic edifice and feeding eruptions in the summit crater.

Table 1
Synthetic Aperture Radar (SAR) Images Used in This Study 3Note.LOS Stands for Line-Of-Sight.