Jeju rainfall isotopes reveal strong elevation and seasonal signals

A historical monitoring network across Jeju Island, South Korea, has produced a spatially resolved record of stable isotopes in monthly precipitation, revealing strong seasonal cycles and measurable changes with elevation. The dataset contains 462 paired measurements of oxygen-18 and deuterium from samples collected at 15 stations between September 2000 and December 2003. Stations ranged from 5 to 1,230 metres above sea level, allowing the researchers to examine seasonal variability and isotope–elevation relationships within the same island-wide network. Across the common 2002–2003 period, unweighted station means showed that δ18O declined by −0.146 ± 0.013‰ per 100 metres of elevation, while δ2H declined by −1.043 ± 0.128‰ per 100 metres. The reported uncertainties are standard errors. The authors present the observations as a historical baseline for future precipitation-isotope monitoring, hydrological studies and comparisons with isotope-enabled climate-model output, while emphasizing that the 39-month record is too short to establish a long-term climatology.

The recovered measurements cover a broad elevation range on an island shaped largely by fractured basalt. Jeju lies about 100 kilometres south of the Korean Peninsula, covers 1,845 square kilometres and rises to the 1,950-metre summit of Mount Halla. Its permeable volcanic rocks promote infiltration and restrict surface runoff, contributing to groundwater-dominated hydrology. The 15 sampling stations were distributed among northern, southern, eastern and western sectors and were also divided into coastal, lowland, mid-elevation and high-elevation zones. This arrangement was intended to capture coastal-to-inland transitions, slope exposure and topographic gradients. The study does not treat volcanic geology as a direct control on precipitation-isotope composition. It notes, however, that precipitation-isotope observations can contribute to groundwater-recharge studies on the island when combined with independent evidence about infiltration, subsurface mixing and other hydrological processes.

Laboratory analyses found that individual δ18O values ranged from −13.2 to −2.0‰ and δ2H values from −93.3 to −5.8‰. The scientists measured δ18O through carbon dioxide–water equilibration and cryogenic condensation, and δ2H through chromium-reduction pyrolysis to produce hydrogen gas. Analytical precision was ±0.1‰ for δ18O and ±1.0‰ for δ2H, based on repeated internal standards calibrated against VSMOW2, SLAP2 and GISP reference waters. Deuterium excess, or d-excess, was calculated as δ2H minus eight times δ18O and used to examine departures from the isotope relationship between the two primary measurements. One sample identified in earlier work as ST-3 was excluded because it was interpreted as having undergone substantial evaporation before analysis. No additional observations were removed merely for having high or low isotope values, and site-months without a valid archived isotope measurement were left missing rather than reconstructed or statistically imputed.

Seasonality was clear, although its timing and magnitude varied among years. Monthly δ18O and δ2H values were generally lowest from June through September and relatively high during parts of the cool season from November through February. The summer minima broadly coincided with the main summer-monsoon precipitation period, while the largest values generally occurred during the relatively dry winter period. The researchers considered the depleted summer signatures consistent with the combined influence of intensified convection, long-distance moisture transport and rainout along air-mass pathways. Winter enrichment was interpreted as consistent with lower atmospheric humidity, stronger sub-cloud evaporation and a larger contribution from dry continental air. These are process-based interpretations rather than event-specific findings: the monthly samples do not include atmospheric trajectories or other diagnostics that would allow each event to be assigned to a unique moisture source or mechanism.

Deuterium excess followed a less symmetrical pattern than the two primary isotope values. It ranged from 3.3 to 29.1‰ across individual samples and showed its strongest inverse relationship with air temperature, with an R² of 0.77 and p < 0.0001, compared with R² values of 0.07 for δ18O and 0.29 for δ2H. The researchers therefore considered d-excess more sensitive than the other isotope measures to changes in humidity and evaporation conditions. Summer months generally had lower d-excess, while winter values were higher, a pattern consistent with seasonal shifts in moisture-source humidity. The authors stopped short of assigning low and high values to a strict maritime-versus-continental binary because several processes—including source-region evaporation, cloud processes and sub-cloud exchange—can affect the parameter. Calendar-month empirical fits gave δ18O and δ2H baselines of −7.09 and −43.86‰, with respective amplitudes of 1.60 and 17.42‰ and R² values of 0.40 and 0.63. These fits were presented as descriptive summaries rather than mechanistic models or monthly climatological normals.

Seasonal regressions also showed that the relationship between δ18O and δ2H changed through the year. Across all 462 samples, the annual local meteoric water line was δ2H = 8.17 × δ18O + 15.72, with an R² of 0.89. The June-to-September line had a relatively higher slope and markedly lower intercept, whereas the November-to-February line had a reduced slope and elevated intercept. The researchers interpreted these contrasts as consistent with differences in humidity, evaporation and transport pathways between summer maritime and winter continental-influenced regimes. March-to-May and October formed transitional periods with intermediate characteristics. Because d-excess showed a quadratic rather than sinusoidal annual shape, the analysis did not force it into the same periodic form used for δ18O and δ2H. Instead, it was summarized with a quadratic calendar-month curve, preserving the observed asymmetry while avoiding the claim that a simple oscillation fully describes its behaviour.

The clearest spatial result was the depletion of δ18O and δ2H with increasing elevation. When all available observations were considered together, δ18O decreased at an average of about −1.5‰ per 1,000 metres with an R² of 0.88 and p < 0.005, while δ2H decreased by about −11‰ per 1,000 metres with an R² of 0.77. To reduce unequal-record effects, the principal gradient analysis used unweighted station means from the common 2002–2003 network period. The resulting network-wide relationships were stronger, with R² values of 0.913 for δ18O and 0.837 for δ2H, and both were significant at p < 0.001. D-excess showed a weaker elevation relationship, increasing by 0.126 ± 0.040‰ per 100 metres. Separately fitted slopes in the island’s four sectors ranged from −0.105 to −0.155‰ per 100 metres for δ18O and from −0.682 to −1.158‰ per 100 metres for δ2H. Those sector fits remain exploratory because each contains only three or four stations over a different elevation range, and they do not demonstrate significant differences among directional slopes.

Without sample-resolved wind and trajectory observations, the directional patterns cannot be assigned uniquely to marine influence, prevailing winds, topographic shielding or windward–leeward position. The network-wide gradients nevertheless provide an empirical basis for a local precipitation isoscape, a surface that estimates expected δ18O and δ2H at locations where direct measurements are unavailable. The authors say such estimates could offer an initial constraint on modern precipitation inputs for cave and speleothem studies on Jeju. They would not, by themselves, calibrate a paleoclimate archive, because the isotope signal is also modified by seasonal recharge weighting, evaporation, soil and epikarst storage, cave hydrology, water–rock interaction and calcite formation. Comparisons with dripwater or speleothem records would therefore need independent hydrological evidence and could only test plausible transfer pathways rather than establish recharge seasonality or elevation on their own.

The researchers also placed the observations beside output from IsoGSM4, an isotope-enabled atmospheric general circulation model run at approximately 50-kilometre horizontal resolution and nudged by ERA5 reanalysis. Three grid points—a central point on Jeju and offshore points to the west and east—were averaged to provide an island–nearshore approximation, with model months receiving less than 0.2 millimetres per day of precipitation excluded. The model-minus-observation mean differences from May through October were +0.3‰ for δ18O, +1.2‰ for δ2H and −1.2‰ for d-excess. For November through April, the corresponding differences were +1.4‰, +5.9‰ and −5.1‰. The smaller May-to-October differences indicate broad seasonal consistency, but the authors do not treat the comparison as model validation or process attribution. The coarse grid cannot resolve Jeju’s topography, station-scale isotope gradients or local-to-mesoscale rainfall processes, and the available comparison lacks the diagnostics needed to distinguish among unresolved topography, land–sea contrasts, spatial representativeness and cloud or precipitation parameterization.

The dataset’s principal value lies in documenting temporal and spatial variability simultaneously. Its monthly structure preserves recurring seasonal behaviour across 15 stations, while the elevation range reveals a strong first-order relationship between altitude and isotope depletion. Supporting information includes station and sampling metadata, 488 station-month records of precipitation and air-temperature estimates—including 26 records without corresponding isotope measurements—and documentation of the single excluded sample. The precipitation and temperature records were derived from nearby meteorological stations and were not used as numerical criteria for screening isotope observations. By making the underlying measurements, calculations and methodological decisions openly available, the data description allows independent researchers to reuse the record, compare later monitoring with this earlier interval and construct local isotope–elevation relationships. Its limits are equally clear: it cannot characterize interannual-to-decadal variability, assign individual monthly samples to unique atmospheric pathways, independently validate groundwater-recharge models or serve as a stand-alone calibration for paleoclimate reconstructions.

Subject of Research: Spatial and seasonal variability of stable water isotopes in monthly precipitation across Jeju Island

Article Title: A spatially resolved dataset of stable water isotopes in monthly precipitation over Jeju Island, Korea (2000–2003)

Article References: Lee, K.-S., Koh, D.-C., Kim, Y.-H., Park, W.-B., Kim, S., Jung, H., Bong, Y.-S., Shin, W.-J., Kim, C., & Lee, J. (2026). A spatially resolved dataset of stable water isotopes in monthly precipitation over Jeju Island, Korea (2000–2003). Earth System Science Data, 18(10), 7645-7659. https://doi.org/10.5194/essd-18-7645-2026

Image Credits: AI Generated

DOI: 10.5194/essd-18-7645-2026

Keywords: stable water isotopes, precipitation, Jeju Island, elevation gradient, δ18O, δ2H, deuterium excess, monsoon climate, isotope-enabled model, historical dataset, spatially, resolved