Sinking Rift: Satellites Reveal Kenya’s Nakuru County Is Slowly Collapsing

Deep in Kenya’s Great Rift Valley, the ground beneath one of the country’s fastest-growing cities is quietly giving way. A decade-long satellite study of Nakuru County has now put hard numbers on the problem, and the results are striking: parts of the county are subsiding at rates approaching 29 millimeters per year, more than three times the speed at which many tectonically active regions creep. The research, published in the open-access journal Discover Geoscience, combined European radar satellite observations with ground-based GPS measurements to produce the most detailed picture yet of land deformation in this volcanically active, water-stressed corner of East Africa.

The study, led by Turyamureeba Esau of Dedan Kimathi University of Technology with colleagues Arthur W. Sichangi and Moses K. Gachari, processed 124 radar images from the Sentinel-1A satellite, covering the period from October 2014 to December 2024. Rather than relying on a single radar technique, the team merged two complementary approaches known as Persistent Scatterer and Small Baseline Subset interferometry, or PS–SBAS InSAR. The persistent scatterer method excels at tracking stable reflectors such as buildings and rocky outcrops in urban environments, while the small baseline approach performs better across farmland and semi-vegetated terrain. Fusing the two produced a dense network of measurement points spanning the entire county of roughly 7,500 square kilometers.

The technical workflow behind the deformation maps is considerable. Each radar image was corrected for orbital errors using precise ephemerides from the European Space Agency, stripped of thermal noise, radiometrically calibrated, and carefully aligned with its neighbors. Phase unwrapping, the mathematical step that converts wrapped radar phase differences into continuous displacement measurements, was handled by the widely used SNAPHU algorithm. Perhaps most critically, the team applied atmospheric correction based on ERA5 reanalysis data, a crucial step in a rift valley where sharp elevation gradients and variable water vapor can otherwise masquerade as ground movement. Radar signals travel down through kilometers of turbulent atmosphere, and without such corrections, a hum of humidity can be mistaken for a sinking city.

The headline finding is that subsidence in Nakuru is real, widespread, and accelerating. Annual vertical deformation rates ranged from about minus 28.5 millimeters per year in the fastest-sinking zones to localized uplift of around 20.5 millimeters per year in more stable areas. The most severe downward motion clustered in the central and southeastern parts of the county, near Nakuru Town and the corridor toward Naivasha and Gilgil. Just as concerning was the temporal trend: deformation hotspots that were localized patches in 2014 had grown larger, more intense, and more interconnected by 2024, with the years 2020 through 2024 showing the greatest spatial extent of subsidence of the entire decade.

Satellite radar alone can never be more than a sophisticated estimate, so the researchers turned to ground truth. Four GNSS stations, at Subukia, Berea, Egerton, and Gilgil, provided independent vertical velocity measurements against which the radar-derived rates could be tested. The comparison was remarkably tight: a mean bias of just minus 0.61 millimeters per year and a root mean square error of 0.99 millimeters per year. In practical terms, the satellite measurements and the ground-based positions agree to within about a millimeter per year, a level of precision that transforms the deformation maps from interesting pictures into actionable engineering data. Small residual differences are expected, since GPS samples a single point while each radar pixel averages an area of several meters across, and lingering atmospheric effects can never be entirely eliminated.

To determine whether the sinking was randomly scattered or geographically organized, the team applied the Getis-Ord Gi* hotspot statistic and an Optimized Hot Spot Analysis, the latter of which automatically selects the most appropriate spatial scale for cluster detection. Both methods told the same story with high statistical confidence, up to the 99 percent level. Statistically significant subsidence hotspots dominated the central and southeastern sectors throughout the study period, while cold spots, zones of stability or even uplift, persisted in the north and west. The consistency of these patterns across ten years and across two independent statistical approaches suggests that the deformation is driven by stable underlying conditions rather than by transient events such as earthquakes or individual storm seasons.

The search for those underlying conditions led the researchers to groundwater. Using NASA-derived groundwater storage data, they found that storage across the county fluctuated dramatically over the decade, peaking near 2,056 kilograms per square meter in 2020 and plunging to a minimum of about 455 kilograms per square meter in 2022. When they overlaid subsidence rates on groundwater storage zones, a clear pattern emerged: mean subsidence values rose from 2.31 in the driest storage class to 4.10 in the wettest, an increase of roughly 77 percent. The association makes physical sense, because areas of high groundwater storage are typically underlain by thick, porous, compressible sediments that compact when pore-water pressure drops. Regions underlain by shallow volcanic bedrock, by contrast, resist compaction and remained comparatively stable.

There is an important caveat to that interpretation, and the authors are careful to draw it. Because groundwater abstraction records and long-term borehole level data were unavailable for the county, the study demonstrates a spatial association between groundwater-rich environments and subsidence, not a proven causal chain. Nakuru sits within the East African Rift System, one of the most geologically restless regions on Earth, where crustal extension, faulting, volcanism, and geothermal fluid circulation all contribute to surface deformation. Disentangling how much of the sinking reflects human water extraction, how much reflects tectonic lowering, and how much reflects natural sediment compaction will require the systematic groundwater monitoring networks the study explicitly recommends.

Classifying the deformation into five severity bands sharpened the picture further. Moderate subsidence proved to be the dominant category across much of the county, meaning measurable ground lowering is the norm rather than the exception, while the very high and high subsidence classes, reaching down to minus 28.5 millimeters per year, stayed tightly concentrated in the hotspot zones. Year over year, the moderate and high-severity zones expanded while stable and uplifted areas shrank. For a county whose population, agriculture, industry, and geothermal ambitions are all growing rapidly, the trajectory matters: cumulative subsidence threatens buildings, roads, water conveyance systems, and drainage networks, and it lowers land closer to flood levels.

The implications reach well beyond Nakuru. Groundwater-driven subsidence has already sunk parts of Jakarta by more than 20 centimeters per year, warped Xi’an in China, and eaten away at the Nile Delta and Lagos, and previous InSAR work in the neighboring Naivasha Basin had detected vertical displacement exceeding a centimeter per year. What the Nakuru study adds is a calibrated, GNSS-validated, decade-long template that other rift communities can replicate with freely available Sentinel-1 data. The authors recommend continuous integrated monitoring, expansion of groundwater observation networks, collection of abstraction records, and the use of both ascending and descending satellite orbits to resolve full three-dimensional motion. They also urge county planners to fold subsidence maps directly into infrastructure and land-use decisions. In a rapidly urbanizing rift valley where the ground itself is measurably sinking, knowing exactly where, how fast, and why may be the difference between a manageable hazard and a slow-motion crisis.

Subject of Research: Land subsidence monitoring using integrated GNSS and InSAR in Nakuru County, Kenya

Article Title: Quantifying spatiotemporal land subsidence rates and deformation hotspots using integrated GNSS and InSAR in Nakuru County, Kenya

Article References: Esau, T., Sichangi, A. W., & Gachari, M. K. (2026). Quantifying spatiotemporal land subsidence rates and deformation hotspots using integrated GNSS and InSAR in Nakuru County, Kenya. Discover Geoscience, 4(1), Article 379. https://doi.org/10.1007/s44288-026-00731-y

Image Credits: AI Generated

DOI: 10.1007/s44288-026-00731-y

Keywords: land subsidence, InSAR, GNSS, Sentinel-1, Nakuru County, Kenya Rift, groundwater, deformation hotspots, PS-SBAS, East African Rift, remote sensing, geohazards