Ancient Sea Levels Leave Fingerprints in Japan’s Soft Riverbeds, Refining Erosion Models

Rivers carving through soft sedimentary rock are among the most powerful sculptors of landscapes in tectonically active regions like Japan, yet the mathematical rules that govern their erosion have long been calibrated almost exclusively on harder bedrock. A new study of the Kamikita Coastal Plain in northeast Japan has now filled one of the most conspicuous gaps in the global dataset, estimating the key parameters of the widely used stream-power incision model for soft rocks and revealing that the story they tell is more complicated, and more interesting, than a simple physical law.

The research, led by Shizuka Takai of the Nuclear Safety Research Center at the Japan Atomic Energy Agency together with Tomoji Sanga, Taro Shimada, and Seiji Takeda, was published in the journal Earth Surface Dynamics. The team focused on six small rivers that dissect a staircase of marine terraces on a coastal plain roughly 30 by 50 kilometres in size, where the underlying rocks range in age from the Miocene to the Pleistocene and the land has been rising steadily at about 0.2 millimetres per year for the past 300,000 years. That combination of uniform lithology and uniform uplift makes the plain an unusually clean natural laboratory for isolating the physics of river incision.

At the heart of the study lies the detachment-limited stream-power model, a deceptively simple equation that expresses the long-term incision rate E as the product of an erosion coefficient K, the upstream drainage area A raised to a power m, and the channel slope S raised to a power n. The ratio of the two exponents, known as the concavity index, describes the characteristic downward flattening of a river profile, while K bundles together everything the equation does not explicitly resolve: the erodibility of the bedrock, the climate, and the hydraulic geometry of the channel. If the erosion rate can be measured independently, all three parameters can be recovered from the shape of the river profile as recorded in digital elevation data.

Measuring erosion rates, however, is where most previous studies have stumbled. The standard tool is the concentration of beryllium-10, a cosmogenic radionuclide that accumulates in quartz grains as they are exhumed from the landscape. That method works beautifully in quartz-rich granites but fails in the humid, tectonically restless terrain of Japan, where more than half of the surface geology consists of Paleogene and Neogene sedimentary rocks that contain little quartz and have been deformed and reworked in complicated ways. Cosmogenic clocks also integrate erosion over timescales that shrink as erosion accelerates, spanning roughly a hundred thousand years in slow eroding terrain but only centuries to millennia where rivers cut at several metres per thousand years.

Takai and colleagues sidestepped the problem by reading the erosion record directly from the landscape itself. The Kamikita Plain preserves a sequence of marine terraces formed during the high sea levels of Marine Isotope Stages 5e, 7, 9, and 11, whose ages have been pinned down by sediment-rate chronology, optically stimulated luminescence dating, and tephra and phytolith stratigraphy. Because the terraces were cut flat by the sea and then uplifted, the difference between a terrace’s reconstructed summit level and the modern riverbed, corrected for the thickness of overlying tephra and loess, gives the depth of rock the river has removed since the terrace formed. Dividing that depth by the terrace age yields an incision rate averaged over hundreds of thousands of years, exactly the timescale that long-term landscape forecasts require.

The analysis relied on high-resolution topographic data, combining airborne LiDAR and photogrammetric elevation models at five-metre resolution, smoothed with a 500-metre moving window to suppress measurement noise. The team identified the reaches where the rivers behave as detachment-limited bedrock channels, confirmed that these reaches lie upstream of the alluvium deposited as sea level rose during the Holocene, and extracted the channel steepness index and concavity using both slope-area regression and the integral chi-transform method. A statistical segment-fitting algorithm, evaluated with the corrected Akaike Information Criterion and validated through a thousand bootstrap trials, located the knickpoints, the abrupt changes in channel gradient that record disturbances migrating upstream through the profile.

The results were striking. The reference concavity came out at 0.44 plus or minus 0.10, squarely within the range expected for channels near steady state. Yet the slope exponent n, estimated across a range of assumed concavities from 0.4 to 0.6, consistently exceeded one, falling between 1.14 and 1.34. Nonlinearity of this kind implies that incision rate grows faster than proportionally with channel steepness, a behaviour consistent with global compilations that typically report n between one and two. The erosion coefficient K, by contrast, spanned a much wider range, from about four times ten to the minus seven to just over ten to the minus five, reflecting the sensitivity of this parameter to local conditions.

The most intriguing finding concerns why n exceeds one. All the knickpoints the team identified were slope-break features rather than waterfalls, and nearly all of them cluster at similar elevations of roughly 25 and 50 metres above sea level, independent of lithological boundaries. That pattern points to knickpoints generated by past falls in sea level, which propagated upstream through the channels during successive glacial cycles, much like features documented on the Sanriku Coast some 50 kilometres to the south. When the researchers averaged erosion rates and steepness over entire channel reaches, the apparent nonlinearity weakened toward n of about one, suggesting that the observed exponent is largely a transient artefact of sea-level-driven knickpoint migration rather than an intrinsic property of the incision process. Physical mechanisms such as shear-stress thresholds and abrasion by sand derived from the soft bedrock may still contribute, and the authors note that disentangling these contributions will require further work.

The erosion coefficient told a cleaner story. Converting standard penetration N-values to unconfined compressive strength, the team found that the rocks underlying most of the study rivers have strengths of only about 1.2 to 3.1 megapascals, far softer than any bedrock in previous global compilations, which began at roughly 15 megapascals. Despite falling well outside that range, the estimated K values landed almost within the 90 percent confidence interval of the global relationship in which K scales inversely with the square of rock strength, a correlation originally established for much harder rocks. The agreement, with a coefficient of determination of 0.84 when the new data are combined with the global compilation, provides strong support for the idea that bedrock lithology exerts a first-order control on how fast rivers can cut.

Some systematic underestimation of K relative to the global trend remains, and the authors attribute it to three plausible causes: the transient steepening from migrating sea-level knickpoints, the use of present-day channel steepness as a proxy for long-term averages in rivers that have only gradually incised flat terraces, and the cohesive resistance of the semi-consolidated sandstones and siltstones, which standard geotechnical indices do not fully capture. Beyond its scientific interest, the work has a practical edge. Safety assessments for radioactive waste repositories, carbon storage sites, and landfills demand erosion forecasts over ten thousand to a hundred thousand years, and the parameters reported here, integrated over multiple glacial-interglacial cycles, offer a calibrated baseline for simulating future landscape evolution in the soft-rock terrains where much of Japan’s infrastructure sits. The team’s next step is to reproduce the plain’s past topography with coupled detachment- and transport-limited landscape evolution models, testing whether the numbers carved into its terraces can also predict what comes next.

Subject of Research: Parameter estimation of stream-power river incision models for soft sedimentary rocks using marine terraces in northeast Japan

Article Title: Parameter estimation of river incision models of soft sedimentary rocks – a case study on the Kamikita Coastal Plain, northeast Japan

Article References: Takai, S., Sanga, T., Shimada, T., & Takeda, S. (2026). Parameter estimation of river incision models of soft sedimentary rocks – a case study on the Kamikita Coastal Plain, northeast Japan. Earth Surface Dynamics, 14(3), 417-432. https://doi.org/10.5194/esurf-14-417-2026

Image Credits: AI Generated

DOI: 10.5194/esurf-14-417-2026

Keywords: river incision, stream-power model, soft sedimentary rocks, marine terraces, Kamikita Coastal Plain, knickpoints, sea-level change, erosion coefficient, concavity index, cosmogenic nuclides, landscape evolution, Japan