The problem begins with scale. To study how climate change has reshaped Alpine catchments since the mid-twentieth century, researchers often use the Weather Research and Forecasting model, or WRF, to dynamically downscale coarse global reanalysis data to a much finer grid. In this study, the team took the Twentieth Century Reanalysis Version 3, a global dataset with a spatial resolution of roughly 100 kilometers, and downscaled it to 2 kilometers with hourly output for the period 1969 to 2015. That is a remarkable achievement, but even 2 kilometers is too coarse to capture the fine-grained reality of a glacierized catchment, where snowfields, rock debris, and ice interact with the atmosphere in ways that vary from one slope to the next.

When the researchers evaluated the WRF output against observations from four meteorological stations spanning nearly 900 meters of elevation in the upper Kauner Valley, they found a pronounced cold bias in the diagnosed 2-meter air temperature at elevations above roughly 2600 meters. The culprit was the model’s coarse representation of snow- and ice-covered surfaces. In the WRF land surface scheme, grid cells above about 2900 meters were classified as permanent Snow or Ice, and the cooling influence of those bright, frozen surfaces bled into the simulated near-surface air temperature. The result was air that was too cold in exactly the places where snowmelt and glacier melt are most sensitive to temperature, a distortion that would ripple through every downstream calculation.

Conventional bias correction, such as empirical quantile mapping, adjusts model output by matching its statistical distribution to that of observations. But that approach requires long, high-quality observational records, and such records are notoriously scarce at high elevations in the Alps. Stations above 2700 meters are few, gappy, and prone to sensor failures, and their statistical properties may not be stationary over decades of rapid warming. Rather than lean on sparse data, Hofmeister and his colleagues turned to the model’s own atmosphere for the answer.

Their method exploits a fundamental property of the atmosphere: temperature generally decreases with height according to a lapse rate. In the WRF simulations, the researchers found that at the lowest atmospheric model levels, temperatures were strongly influenced by the underlying land cover, with snow- and ice-covered grid cells showing markedly colder values and considerable scatter. But higher up, between the first and twentieth vertical levels, the temperature-elevation relationship became almost perfectly linear, with very little scatter. At these heights, the model atmosphere was essentially free of surface contamination, governed by a uniform domain-wide lapse-rate structure. The lapse rate at the first model level averaged minus 0.78 degrees Celsius per 100 meters in a test period, close to the free-atmospheric value of minus 0.65, and it decreased steadily with height until it approached zero at the highest levels.

The correction workflow is a three-step procedure applied to hourly time series. First, the team computed spatial averages of hourly temperature and geopotential height across the 49 grid cells covering roughly 200 square kilometers of the Kauner Valley domain, at the first and twentieth vertical levels. Averaging in space reduces the dependence of the lapse rate on land cover type. Second, they calculated an hourly temperature lapse rate from the difference between the two levels, capturing the sub-daily variability of the atmospheric temperature gradient rather than assuming a fixed value. Third, they used that hourly lapse rate to extrapolate the temperature at the first atmospheric model level, roughly 17 meters above the ground on average, down to any station or catchment elevation. Because the reconstruction starts from the free atmosphere rather than the surface, it sidesteps the surface-related biases entirely. The mean lapse rate derived between the two levels over 1973 to 2015 was minus 0.57 degrees Celsius per 100 meters, slightly steeper in summer than in winter.

Temperature alone was not enough. Because snowmelt and glacier melt models rely on energy balance calculations in which latent heat fluxes play a substantial role, the team also applied a sequential, thermodynamically consistent correction of relative humidity, adjusting moisture to match the corrected temperatures so that the physical relationship between the two variables was preserved. Notably, they left shortwave radiation and wind speed untouched, since physics-based hydrological models such as WaSiM, the Water Flow and Balance Simulation Model used in this study, already account for topographic effects on radiation and can represent wind-driven snow redistribution on their own.

The validation was rigorous and multi-layered. Against station observations, the corrected temperatures showed substantially better agreement than the uncorrected output, with the improvement most dramatic at the highest station, Vernagt Pegel at 2640 meters, where the maximum distance between observed and modeled temperature distributions dropped from 0.13 to 0.03. The consequences for hydrological modeling were striking. When WaSiM was forced with uncorrected WRF temperatures, it could not reproduce the observed dynamics of snow accumulation and melt, glacier mass balance, snow cover, or streamflow: the cold bias suppressed melt so severely that simulated annual streamflow was nearly 50 percent too low, and the simulated glacier mass balance stayed unrealistically positive during a period when Alpine glaciers were actually shrinking. With the corrected temperatures, the model matched the observed mean annual streamflow almost exactly, and the Kling-Gupta efficiency of daily streamflow improved from 0.42 to 0.71. Adding the humidity correction pushed performance further, raising the efficiency to 0.78 and improving the representation of late-summer flows during peak glacier melt, when the Fagge carries its maximum meltwater load.

Crucially for climate research, the correction preserved the long-term warming signal. Comparing the corrected product with homogenized observations at Obergurgl, a station in a neighboring valley with records stretching back to 1851, the team found that both the observations and the corrected WRF temperatures showed a significant warming trend of about 0.05 to 0.06 degrees Celsius per year over 1973 to 2015, and both detected the well-documented abrupt Alpine warming shift in May 1987. A 42-year simulation of the catchment reproduced the observed transition from near-equilibrium glacier conditions in the 1980s to continuous mass loss thereafter, along with a significant shortening of snow cover duration after 1985. Because the method relies on lapse rates derived from the climate model itself rather than on local observations, it is transferable to other glacierized mountain regions where ground data are lacking. As the authors note, future work should test the approach in catchments with different climatic and topographic settings, but the message is already clear: physically consistent corrections of model output can rescue hydrological simulations in the world’s most water-critical landscapes, without erasing the climate signal that makes those simulations worth running in the first place.

Subject of Research: Physically based bias correction of regional climate model temperatures for hydrological simulation in a glacierized Alpine catchment

Article Title: Technical note: A physically based correction of WRF temperature using atmospheric lapse rates for high-resolution hydrological simulations in a glacierized catchment

Article References: Hofmeister, F., Fan, X., Pfeiffer, M., Marzeion, B., Schaefli, B., & Chiogna, G. (2026). Technical note: A physically based correction of WRF temperature using atmospheric lapse rates for high-resolution hydrological simulations in a glacierized catchment. Hydrology and Earth System Sciences, 30(19), 6265-6288. https://doi.org/10.5194/hess-30-6265-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6265-2026

Keywords: WRF model, temperature lapse rate, bias correction, glacierized catchment, hydrological modeling, Alps, snowmelt, glacier mass balance, streamflow, regional climate model, relative humidity, climate change

The research, led by Rizwan Mansha and Muhammad Azmat of the National University of Sciences and Technology in Islamabad, together with colleagues in Pakistan, China, and Saudi Arabia, focused on the Gilgit River basin in Northern Pakistan. This catchment is a textbook example of what hydrologists call a sparsely gauged basin: a landscape where weather stations are few, far between, and often unreliable, yet where the river itself sustains communities downstream that depend on its flow for agriculture, hydropower, and daily life. In basins like this, conventional hydrological modeling is crippled at the first step, because models require continuous records of temperature, precipitation, and evapotranspiration that simply do not exist on the ground.

To circumvent the data drought, the team turned to remotely sensed and reanalysis products. They used daily temperature, precipitation, and evapotranspiration data from the European Centre for Medium-Range Weather Forecasts ERA-Interim reanalysis, a globally consistent dataset produced by assimilating observations into a physical model of the atmosphere. They supplemented this with daily solar radiation, wind speed, and relative humidity fields from the Coupled Model Intercomparison Project Phase 3 archive, the same multimodel dataset that underpinned major international climate assessments. The approach effectively replaces broken ground networks with a coherent, physically constrained picture of the atmosphere above the basin, delivered through a geographic information system framework that organizes the spatial data the models need.

With the forcing data in place, the researchers ran a head-to-head comparison of two widely used but structurally different hydrological models. The first was the Soil and Water Assessment Tool, or SWAT, a physically based, spatially distributed model that divides a watershed into sub-basins and further into hydrological response units, simulating processes such as infiltration, evaporation, snow accumulation and melt, and channel routing in considerable detail. The second was the HBV model, originally developed by the Swedish Meteorological and Hydrological Institute, a conceptual rainfall-runoff model that represents the catchment with a compact set of stores and routines, including a dedicated snow module, soil moisture accounting, and response functions that convert water into runoff.

The models were calibrated against thirteen years of observed daily streamflow from 1990 to 2002, and then validated against an independent five-year window from 2003 to 2007. The team evaluated performance using the Nash-Sutcliffe efficiency, a standard metric that measures how well simulated flows match observed ones relative to the mean of the observations, with a value of one indicating a perfect match, alongside the percent difference in total volume between simulated and observed flows. For the critical annual snowmelt season, HBV achieved a Nash-Sutcliffe efficiency of 0.93 during calibration with a volume difference of just 0.23 percent, and 0.75 with a volume difference of 26.30 percent during validation. SWAT recorded 0.60 and 34.26 percent during calibration, improving to 0.74 and 10.73 percent in validation.

Those numbers tell a nuanced story. Both models captured the broad rhythm of the basin’s hydrology, in which summer flow is dominated not by rain but by the melting of the winter snowpack and the slow release of water from glaciers. But HBV’s near-perfect calibration performance and its more consistent behavior across both periods suggest that its simpler structure, which depends on fewer parameters, is better suited to the peculiar hydrology of a cryosphere catchment when the only available inputs come from satellites and reanalysis. SWAT, with its many more parameters to constrain, appears more vulnerable to the uncertainties that pervade data-scarce mountain environments, where every parameter must be inferred from limited information.

Having established which model could be trusted, the researchers then used both to ask the question that matters most for the future: what happens to the river as the climate changes? Rather than relying on a single climate projection, they applied a systematic set of perturbations to the observed climate, shifting temperature across a range from minus two to plus four degrees Celsius and precipitation from minus ten to plus twenty percent. This scenario-based approach, sometimes called a sensitivity analysis, deliberately brackets the plausible future rather than betting on one trajectory, and it reveals how the basin’s hydrology responds to each driver independently.

The results were striking and, in places, sobering. Under the temperature scenarios, HBV projected annual streamflow changes ranging from a decrease of 16.92 percent at the coldest perturbation to a decrease of 6.6 percent at the warmest, while SWAT projected changes from minus 28.71 percent to minus 2.96 percent. For precipitation shifts, HBV indicated variations between minus 6.65 percent and 13.3 percent, and SWAT between minus 18.13 percent and 27.5 percent. The direction and magnitude of these responses reflect the delicate balance of a snow- and glacier-fed system: warmer temperatures alter the timing and volume of snowmelt, shifting the seasonal peak earlier and reducing the frozen storage that sustains summer flows, while changes in precipitation control how much snow accumulates in the first place. The divergence between the two models also carries a practical warning, because the choice of hydrological model can change the projected impact by a factor of several, meaning that model selection is not a technical footnote but a first-order source of uncertainty in water planning.

The broader significance of the study lies in its demonstration that meaningful hydrological assessment is possible in places where the traditional infrastructure of hydrology, the dense networks of rain gauges and stream gauges that underpin water management in wealthy countries, simply does not exist. By relying solely on remotely sensed and reanalysis data, the researchers showed that the Gilgit basin, and by extension hundreds of similar catchments across the Hindu Kush, Karakoram, and Himalaya region, can be brought within the reach of quantitative water resource planning. This matters because the region is often described as a water tower for Asia: the ice and snow stored in its mountains feed the Indus and other river systems on which hundreds of millions of people depend, and the region’s cryosphere is known to be changing, with glacier behavior varying dramatically from one valley to the next.

The authors position the work as a baseline for selecting appropriate hydrological models and datasets for hydrologically peculiar catchments, and they highlight the potential of HBV for operational use in data-scarce cryosphere environments. For water managers in Pakistan and across the greater Himalayan region, the message is twofold. First, the data needed to understand and forecast these rivers is already overhead, streaming down from satellites and reanalysis systems, waiting to be exploited. Second, when it comes to translating that data into reliable streamflow projections, the simplest tool that faithfully captures snow and glacier melt may be the most trustworthy guide through an uncertain climate future. As warming accelerates across the world’s high mountains, the basins that can least afford to be blind to their own hydrology may finally have a way to see.

Subject of Research: Comparative hydrological modeling of climate-induced streamflow variability in the snow- and glacier-fed Gilgit River basin using GIS-based SWAT and HBV models with remotely sensed data

Article Title: Climate-Induced Streamflow Variability Assessment Using GIS-Based Hydrological Models in a Snow-Glacier-Fed Sparsely Gauged River Basin

Article References: Mansha, R., Azmat, M., Arshad, M. Z., Tariq, A., Faqeih, K. Y., Alamri, S. M., & Alamery, E. R. (2026). Climate-Induced Streamflow Variability Assessment Using GIS-Based Hydrological Models in a Snow-Glacier-Fed Sparsely Gauged River Basin. Water Resources Management, 40(13), Article 535. https://doi.org/10.1007/s11269-026-04889-5

Image Credits: AI Generated

DOI: 10.1007/s11269-026-04889-5

Keywords: hydrological modeling, HBV, SWAT, streamflow, climate change, snowmelt, glacier-fed rivers, Gilgit River basin, remote sensing, ERA-Interim, Pakistan, water resources

The study, led by Ying Yan, Hamed Sharif, and Ali A. Ameli of the University of British Columbia, rests on an extraordinary volume of evidence. The researchers identified and analyzed approximately 1.7 million runoff-producing events across 4,838 catchments spanning six continents. Each event was classified according to two descriptors that together define its hydro-meteorological character: the type of water input, whether rainfall, snowmelt, or rain falling on snow, and the wetness of the soil before the event began, sorted into dry, moderate, or wet antecedent conditions. That ninefold classification scheme allowed the team to ask, in a standardized way, whether a given basin treats a rainstorm on dry soil the same way it treats snowmelt on saturated ground.

The scale of the computation behind this work is itself a story. Identifying individual runoff-producing events in millions of streamflow and precipitation records is a task far beyond what a desktop workstation can handle, so the bulk of the analyses ran on the Compute Canada network of parallel high-performance computing clusters. The effort was supported by grants to Ameli through the Natural Sciences and Engineering Research Council of Canada’s Discovery Grants program. The result is a dataset that turns a question once explored catchment by catchment, in scattered field campaigns and modeling studies, into a globally quantified phenomenon.

The first major finding concerns the diversity of what the authors call dominant hydro-meteorological event types. For each catchment, the team summarized which combination of water-input type and antecedent soil moisture most frequently produced runoff, doing so separately for the dormant season and the growing season. Rather than converging on a handful of universal regimes, the global catchment population displayed a wide diversity of dominant types. Some basins are, in effect, rain-on-dry-soil systems; others are dominated by snowmelt onto wet ground; still others by rain on moderately moist soils. That heterogeneity alone is a caution against one-size-fits-all hydrology, but it is not the paper’s central surprise.

The central surprise is the pervasiveness of time-variance. The researchers quantified, for every catchment, how much the precipitation-to-streamflow transfer function, the mathematical representation of catchment hydrologic function, varied from event to event. The verdict was stark. In the dormant season, 95.6 percent of catchments showed moderate to strong event-to-event variation in their hydrologic function. In the growing season, that figure rose to 99.0 percent. In other words, virtually every river basin on Earth examined in this study behaves differently from one runoff event to the next, at least to a degree that the authors classify as meaningful.

What makes this result more striking is that the variation is not simply a byproduct of events arriving under obviously different weather. Even catchments that experienced highly similar hydro-meteorological conditions across their events, the same input types striking similarly wet or dry soils, remained functionally time-variant. Something beyond the coarse descriptors of input type and antecedent wetness is modulating how these landscapes convert water into runoff. The study points to one candidate mechanism with particular force: within-catchment spatial heterogeneity in hydro-meteorological conditions. A single basin is rarely uniform. One hillslope may be drenched while another stays dry; snow may linger on shaded aspects while sun-facing slopes shed meltwater days earlier. Because different parts of a catchment are activated under different conditions, the integrated streamflow response at the outlet shifts depending on which portions of the landscape are doing the work during any given event.

The implications ripple outward quickly. The precipitation-streamflow transfer function is not an academic abstraction; it is the backbone of how hydrologists translate meteorological forecasts into flood warnings, how engineers size culverts and reservoirs, and how water managers estimate how much of a season’s precipitation will actually arrive in rivers and aquifers. If that transfer function is stable, a model calibrated on past events can be trusted, within limits, to anticipate future ones. If it is time-variant, as this study suggests it almost universally is, then confidence in flood-risk and water-security assessments, the very applications the authors cite as motivation, is directly limited by how well those shifting dynamics are captured. A forecasting system tuned to wet-season events may misjudge a dry-season storm in the same basin, and a model built on rain-dominated events may falter when a rain-on-snow episode arrives.

The study also establishes a link between variability in function and variability in outcome. Greater time-variance in a catchment’s hydrologic function was associated with greater event-to-event variation in the catchment’s runoff response. That connection matters because runoff response is what communities experience: how fast a river rises, how high it crests, how much water is available downstream. Basins whose internal translation rules fluctuate strongly are also basins whose rivers behave less predictably. Identifying such basins, and understanding what drives their internal heterogeneity, could become a priority for flood-risk mapping and infrastructure design in a warming climate, where rain-on-snow events and abrupt shifts between dry and saturated soil conditions are expected to become more frequent in many regions.

Methodologically, the paper demonstrates the power of treating the planet’s river basins as a single statistical population rather than as a collection of idiosyncratic case studies. By imposing a common event classification, rain, snowmelt, or rain-on-snow crossed with dry, moderate, or wet antecedent soil moisture, and by applying it uniformly across 1.7 million events, the researchers could measure time-variance on a comparable scale everywhere. The seasonal split, analyzing dormant and growing seasons separately, adds another layer of nuance, revealing that functional variability is even more widespread when vegetation is active, a hint that evapotranspiration and plant-driven soil dynamics may deepen the nonstationarity of catchment behavior during the growing season.

The broader message is one of humility and opportunity for hydrology. The assumption that a catchment has a single, learnable personality, a fixed transfer function waiting to be calibrated, appears to be wrong for nearly every basin on Earth. That does not mean prediction is hopeless; it means prediction must embrace variability as the norm rather than the exception. Models may need to represent within-catchment heterogeneity explicitly, and forecasting frameworks may need to condition their parameters on event type and antecedent state rather than assuming a single set of values. For a world facing intensifying storms, shrinking snowpacks, and rising demands on freshwater, knowing that the planet’s river basins rarely behave the same way twice is not a discouraging finding. It is, instead, an accurate map of the challenge ahead, drawn from nearly two million events and nearly five thousand landscapes, and it gives hydrologists a far clearer picture of where and why their forecasts are most likely to bend.

Subject of Research: Time-variance in catchment hydrologic function across global river basins

Article Title: Catchment hydrologic function is widely time-variant

Article References: Yan, Y., Sharif, H., & Ameli, A. A. (2026). Catchment hydrologic function is widely time-variant. Communications Earth & Environment. https://doi.org/10.1038/s43247-026-04079-6

Image Credits: AI Generated

DOI: 10.1038/s43247-026-04079-6

Keywords: hydrology, catchments, streamflow, runoff events, soil moisture, snowmelt, rain-on-snow, flood risk, water security, transfer function, global analysis, Communications Earth & Environment

The new technique, called the Robust Variance-based Event Identification Method, or RVEIM, was developed by Mohammad Masoud Mohammadpour Khoie, Danlu Guo, and Conrad Wasko and published in the journal Hydrology and Earth System Sciences. Its central insight is that a runoff event is best detected not by hunting for peaks in the streamflow record, but by watching for sudden surges in the variance of that record. When streamflow variance briefly and clearly exceeds what is normal for a catchment, the stream is doing something unusual, and that something is almost always a response to rain. By anchoring event detection to this statistical signature of physical process, RVEIM needs only two user-defined parameters, a dramatic reduction from the half-dozen or more that conventional methods demand.

The problem RVEIM addresses is one that has quietly undermined event-based hydrology for years. Because no instrument directly measures where a runoff event begins or ends, researchers must rely on rules: a threshold to filter out trivial flows, a minimum spacing to keep consecutive events independent, a search window to match each flood peak with its causative storm. None of these rules has a ground truth to validate against, so their values are chosen by expert judgment or local calibration. Sensitivity analyses have shown that even within accepted ranges, changing these values can alter the number, length, and volume of identified events by amounts comparable to the climatic differences between regions. In one striking example cited by the team, simply reversing the direction of the pairing search, from runoff to rainfall rather than rainfall to runoff, introduced as much variability as differences in climate itself.

Worse, the conventional two-step workflow, detecting runoff events first and pairing them to rainfall afterward, can produce physically impossible results. When a mismatched pair is created, the calculated runoff coefficient, the ratio of runoff volume to rainfall volume, can exceed one, implying that more water left the catchment than fell on it as rain. Such events are clear flags of identification error, yet some methods generate them routinely. RVEIM sidesteps this by detecting and pairing simultaneously: it identifies independent rainfall events first, then searches for the corresponding streamflow response within a window whose length is not fixed by the user but computed from the catchment’s characteristic rainfall-to-runoff lag plus the duration of each individual storm. This time-variant window mirrors the causal chain of runoff generation and allows the method to capture both brief, sharp responses and long, multi-peak events without extra tuning.

The variance threshold at the heart of the method is also derived rather than imposed. The team first applies the widely used Lyne and Hollick digital filter to estimate baseflow, the slow groundwater-fed component of streamflow. They then use a median absolute deviation test to exclude baseflow values inflated by event conditions, and take the variance of the remaining, baseflow-dominated records as the threshold below which streamflow behavior counts as normal. Because this threshold emerges from each catchment’s own data, it transfers across climates without recalibration. In tests across eight representative catchments spanning desert, grassland, temperate, subtropical, tropical, and equatorial zones, the variability of event characteristics under RVEIM stayed within roughly fifteen percent of the mean, while two benchmarking methods, a local maxima approach and the more elaborate Detrending Moving-average Cross-correlation Analysis, exhibited uncertainties of forty-three to ninety-three percent.

Physical plausibility improved as well. Across the representative catchments, the percentage of events with runoff coefficients exceeding one remained below about four percent under RVEIM, and was essentially zero for the DMCA benchmark, whereas the local maxima method produced consistently higher and more variable rates. The differences mattered most in wet catchments such as those in western Tasmania, where frequent storms and high soil moisture make event separation genuinely difficult. There, the three methods diverged sharply in their pictures of catchment behavior: RVEIM attributed far fewer events to near-zero runoff coefficients than its competitors, a result consistent with the hydrological reality of a saturated, fast-responding landscape.

With its robustness established, the team unleashed RVEIM on all 467 unregulated Hydrologic Reference Stations monitored by the Australian Bureau of Meteorology, catchments with at least thirty years of quality streamflow record and minimal human interference. The resulting atlas reveals a continent organized by water. Runoff coefficients climb toward the coasts and reach their highest values in Tasmania, where abundant rainfall, low evapotranspiration, and saturated soils conspire to send most event rainfall straight into streams. In the arid interior, by contrast, most storms vanish into infiltration and evaporation, leaving runoff coefficients clustered near zero. Northern Western Australia presents an intriguing anomaly: despite modest rainfall event volumes, its catchments yield substantial runoff, a consequence of low evapotranspiration and high antecedent soil moisture that leave little room for losses.

The most consequential finding is a systematic climate gradient in the shape of the runoff coefficient distributions. In desert and grassland catchments, the empirical distribution rises steeply at near-zero values, indicating that the vast majority of events produce negligible runoff, a signature pointing to infiltration-excess processes in which rain falls faster than parched soils can absorb it only during intense bursts. In temperate, subtropical, and equatorial regions, the distributions rise gradually toward higher coefficients, the hallmark of saturation-excess runoff from already-wet catchments. Tropical catchments occupy an intermediate, mixed regime. Moving from dry to wet climates, the distributions also become less skewed and more uniform, echoing patterns previously documented in Austria and the European Alps and suggesting the gradient is a global hydrological regularity rather than an Australian curiosity.

The implications reach well beyond academic taxonomy. Flood estimation, infrastructure design, and projections of how catchments will respond to a changing climate all depend on event-scale runoff coefficients, and biased event identification can distort all of them. By capturing the full spectrum of event magnitudes rather than privileging large floods or short responses, RVEIM offers a consistent yardstick for comparing catchments across regions and for tracking how runoff generation shifts as droughts, land-use change, and warming reshape the water cycle. The method has been released as a function in the open-source hydroEvents R package, and its developers argue that because its parameters are unlikely to require adjustment between climates, it should transfer readily to continents beyond Australia. If it does, hydrologists worldwide may finally agree on where one storm ends and the next river begins.

Subject of Research: A parsimonious rainfall-runoff event identification method applied across 467 Australian catchments to characterize event-scale runoff processes

Article Title: Characterising runoff processes for Australia: insights from a parsimonious rainfall-runoff event identification method

Article References: Mohammadpour Khoie, M. M., Guo, D., & Wasko, C. (2026). Characterising runoff processes for Australia: insights from a parsimonious rainfall-runoff event identification method. Hydrology and Earth System Sciences, 30(18), 5947-5970. https://doi.org/10.5194/hess-30-5947-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-5947-2026

Keywords: hydrology, rainfall-runoff, runoff coefficient, event identification, catchment response, Australia, baseflow separation, flood estimation, climate gradient, streamflow, RVEIM, large-sample hydrology

The research, led by Binyam Workeye Asfaw of Virginia Tech’s Department of Biological Systems Engineering with colleagues at Virginia Tech, Morehead State University, and Virginia Tech’s School of Animal Sciences, tested whether satellite soil moisture observations could serve as a calibration target in their own right. The team worked in the upper Stroubles Creek watershed, a 14.5-square-kilometer headwater catchment in Montgomery County, Virginia, monitored by the Virginia Tech StREAM Lab. The watershed sits in the Valley and Ridge physiographic province, underlain by dolomite and limestone geology riddled with springs and sinkholes. It receives roughly 1,200 millimeters of precipitation annually, about 760 millimeters of which returns to the atmosphere through evapotranspiration. Crucially, the catchment is dominated by saturation excess runoff, meaning that water only runs off the land where the soil has already filled to capacity, a process controlled largely by topography and shallow soil storage.

That hydrologic regime made the site an ideal proving ground for the Soil and Water Assessment Tool with a Variable Source Area modification, known as SWAT-VSA. Unlike conventional model setups that lump land areas into uniform hydrologic response units, SWAT-VSA redistributes soil and hydrologic properties according to classes of the topographic index, a measure of how strongly each landscape position accumulates moisture from upslope. In the Stroubles Creek application, the watershed was divided into three topographic index classes covering 56, 38, and 6 percent of the area, from driest ridge tops to wettest convergent hollows. This structure matters because in saturation excess landscapes, runoff is wildly disproportionate: the wettest class, just 6 percent of the watershed, generated roughly 30 percent of surface runoff on a per-area basis, while the largest, driest class contributed only about 9 percent. Capturing that spatial organization is essential if a model is to represent where water actually moves.

The satellite data at the heart of the study came from NASA’s Soil Moisture Active Passive mission, or SMAP, which measures surface soil moisture globally at a native resolution of 36 kilometers, far too coarse for a watershed of 14.5 square kilometers. The researchers turned to an enhanced 10-kilometer SMAP product and then downscaled it further to 500 meters using a pretrained quantile random forest model from the mlhrsm package in R. The downscaling algorithm fuses an impressive array of auxiliary data: Sentinel-1 radar backscatter, MODIS land surface temperature, Landsat-derived vegetation indices, a 10-meter digital elevation model, POLARIS soil properties including clay, sand, and bulk density, and National Land Cover Database classes. The result was a daily volumetric soil moisture estimate for every 500-meter cell in the watershed from April 2015 through August 2022, spatially averaged into a single watershed-mean time series for calibration.

There was a catch. When compared against ground measurements, the downscaled product showed a compressed dynamic range: it peaked at roughly 32 percent volumetric water content while field observations reached about 45 percent, and it systematically muted the wetting pulses that follow heavy rain. Rather than discarding the product, the team applied a deliberately parsimonious, event-based bias correction. Whenever daily effective precipitation exceeded a threshold of 5 millimeters, the depth needed to bring the top 50 millimeters of soil from field capacity to saturation, the satellite estimate was nudged upward by a multiple of the uncertainty bound reported by the downscaling model itself. Importantly, none of the correction parameters were derived from the in-situ field data, preserving the independence of the evaluation dataset. The correction amplified event-scale wetting responses while leaving dry-period behavior and long-term means untouched.

For ground truth, the researchers relied on an unusually dense field campaign: 25 measurement locations spread across a 4.2-hectare mixed-grass pasture within the watershed, sampled on 20 dates between March 2023 and January 2024, one to three days after precipitation events. The observations spanned nearly the full moisture range, from near wilting point to near saturation, between 0.15 and 0.45 cubic meters of water per cubic meter of soil. Point measurements were aggregated to a field-scale mean using topographic index weighting consistent with the model’s internal structure, a scaling choice justified by prior work showing that in this landscape, spatial moisture patterns are governed more by terrain than by local soil variation.

The team then ran three calibration experiments: one targeting streamflow only, one targeting satellite soil moisture only, and a multi-objective version optimizing both simultaneously. Single-objective runs used a differential evolution algorithm evaluating 650 parameter vectors, while the multi-objective case employed the non-dominated sorting genetic algorithm NSGA-II to map out a Pareto front of optimal trade-off solutions. The results were striking. All three models performed comparably on streamflow, with Nash-Sutcliffe efficiency values between 0.54 and 0.58 during calibration and 0.55 to 0.56 during the independent evaluation period. But the soil moisture picture diverged dramatically. The streamflow-calibrated model explained only 50 percent of observed field-scale soil moisture variability, while both the soil-moisture and multi-objective calibrations explained 88 percent. The multi-objective model cut the root mean squared error from 0.05 to 0.03 cubic meters per cubic meter and reduced bias to a negligible minus 0.8 percent.

Yet the study’s most important contribution may be its cautionary lesson about statistics versus physical realism. The soil-moisture-only calibration achieved excellent agreement with moisture observations, but it did so partly through compensatory parameter adjustments: it converged on lower available water capacity and evapotranspiration compensation parameters that dampened vertical moisture redistribution, producing systematically overestimated low flows and an altered water balance partitioning. Calibrated soil parameters deviated from database values by as much as 155 percent for available water capacity and 136 percent for saturated hydraulic conductivity, a reminder that regional soil datasets often misrepresent local conditions. The multi-objective approach, by contrast, narrowed the range of acceptable parameter values for the most sensitive soil and evapotranspiration parameters, reducing equifinality, the persistent problem in hydrology where many different parameter combinations produce equally good fits. Fewer equally acceptable solutions mean more confidence that the model is right for the right reasons.

The authors are careful to delineate the limits of their findings. The study watershed is small, humid, and strongly terrain-controlled, and the SWAT-VSA framework applies specifically to saturation excess environments; infiltration-excess or arid catchments may not benefit from the topographic index structure. The bias correction relies on watershed-average precipitation and may fail where rainfall varies sharply across space, and the relative contributions of the native SMAP retrieval, the machine learning downscaling, and the bias correction were not disentangled. Still, the implications are considerable. Satellite soil moisture products offer global coverage at essentially no marginal cost, making them a promising calibration resource for the many watersheds that lack gauges altogether. Combined with terrain-informed model structures and multi-objective optimization, they appear capable of constraining the internal states of hydrological models, not just the water leaving them, bringing field-scale predictions of soil wetness, and the agricultural and water quality decisions that depend on them, within reach.

Subject of Research: Using downscaled and bias-corrected satellite soil moisture data to calibrate watershed hydrological models

Article Title: Calibration using downscaled and bias-corrected satellite soil-moisture data can improve watershed model representation of soil-moisture variability

Article References: Calibration using downscaled and bias-corrected satellite soil-moisture data can improve watershed model representation of soil-moisture variability. (n.d.). https://doi.org/10.5194/hess-30-5999-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-5999-2026

Keywords: soil moisture, SMAP, satellite remote sensing, watershed modeling, SWAT-VSA, model calibration, hydrology, downscaling, bias correction, saturation excess runoff, multi-objective optimization, streamflow

The core of the problem lies in a long-standing gap in flood data. Floods triggered by intense convective storms can develop in a matter of hours, especially in small catchments, meaning that the true instantaneous peak flow, the maximum flow occurring at a single moment, can be far higher than the average flow recorded over a whole day. Yet sub-daily streamflow records are far less common than daily ones, so hydrologists have spent more than a century trying to estimate instantaneous peak flows from daily data. The classic approach, dating back to Fuller’s work in 1914, relates the ratio between daily mean flows and instantaneous peaks, known as the peak ratio, to catchment characteristics such as area, elevation, and slope. Small catchments typically show a large gap between hourly and daily peaks, while in large basins the two converge.

What has been missing from this picture, the researchers argue, is the role of lakes and reservoirs. Both types of water bodies are known to modulate streamflow: reservoirs dampen floods depending on their storage capacity, management practices, and the return period of the flood, while lakes stabilize baseflow and reduce streamflow variability. Both buffer short, intense floods better than long-lasting ones, because the volume of a brief event is limited. But no one had systematically assessed how this attenuation differs between daily and hourly flood peaks, and therefore how water bodies reshape the peak ratio itself across a large spatial domain.

To close this gap, the team combined two complementary approaches. The first was a set of four local case studies with streamflow gauges both upstream and downstream of a water body, allowing the effect of the water body to be isolated from other catchment influences. Three of these were reservoirs in Bavaria, Germany, namely Perlsee, Vilstalsee, and Mertsee, used primarily for flood protection and recreation, and the fourth was the Walensee in Switzerland, a large natural and unregulated lake. The Bavarian records, available at 15-minute resolution, span 38 to 56 years after reservoir construction, while the Walensee data cover 49 years at hourly resolution. The second approach was a large-sample analysis of 183 Swiss catchments drawn from the CAMELS-CH dataset, each with hourly streamflow records extending from the 1981 to the 2024 hydrological year, a maximum of 44 years of data.

The case studies delivered a striking result. When the researchers fitted generalized extreme value distributions to annual maximum flows and estimated flood quantiles for return periods of 2 to 25 years, they found that sub-daily peak discharge is attenuated far more strongly than daily peak discharge. For floods with a 10-year return period, the downstream-to-upstream ratio was between 0.12 and 0.33 lower for sub-daily flows than for daily flows, corresponding to reductions in sub-daily peaks of up to 70 percent. Daily peaks, by contrast, changed by amounts ranging from a 55 percent decrease at Perlsee to a 9 percent increase at Walensee, the latter explained by the fact that the downstream catchment is 77 percent larger than the upstream one. Even in the Walensee case, where tributaries complicate the picture, the lake’s suppression of sub-daily peaks was unmistakable.

The consequence of this differential dampening is remarkable: downstream of reservoirs, daily and hourly flood peaks become nearly identical. For three of the four case studies, the difference between 10-year daily and sub-daily flood estimates was less than 10 percent downstream of the water body, compared with much larger gaps upstream. In the Mertsee case, the daily-to-sub-daily ratio still rose from 0.5 to 0.7, a substantial convergence even if not complete. In practical terms, a reservoir acts like a filter that strips out the sharp, short-lived spikes in flow while leaving the slower, day-scale component of the flood largely intact, flattening the hydrograph into a form where the distinction between hourly and daily maxima almost disappears.

The large-sample analysis across Switzerland confirmed that this is not a local curiosity. The ratio between daily and hourly 10-year flood peaks, the D/H ratio, increases with catchment area, as expected from classical flood scaling, but it also approaches 1 in catchments strongly influenced by water bodies. Of the 29 catchments in which more than 70 percent of the area drains into a standing water body, 28 had a D/H ratio above 0.85, indicating great similarity between daily and hourly high flows. To disentangle the effects of area and water-body influence, the team trained a random forest model to predict the D/H ratio from seven catchment characteristics. The best-performing model, explaining 67 percent of the variability, used just four predictors: the contributing area percentage, catchment area, geological permeability, and biogeographical region, with the first two dominating in importance.

Partial dependence analysis revealed a clear threshold. Below a contributing area percentage of 60 percent, meaning less than 60 percent of the catchment lies above the water body, the predicted D/H ratio changes only marginally. Above that threshold, the ratio rises sharply, indicating that water bodies only exert their full harmonizing effect when they command a large share of the catchment’s flow-generating area. The team then trained a model on weakly influenced catchments and applied it to strongly influenced ones: while it predicted the weakly influenced catchments well, with errors below 0.1 for 91 percent of them, it systematically underestimated the D/H ratio in strongly influenced catchments, with 80 percent of predictions off by more than 0.1 and a median error of minus 0.20. Ignoring water bodies, in other words, produces systematically wrong estimates of how similar daily and hourly floods are.

The implications for flood estimation are profound. In strongly regulated catchments, daily flood peaks turn out to be a good proxy for hourly peaks, meaning that instantaneous peak flows can plausibly be estimated from daily data where sub-daily records are unavailable. Conversely, in near-natural catchments, especially small ones, hourly peaks can substantially exceed daily peaks, and applying relationships calibrated elsewhere could badly misestimate design floods. The study also exposes a blind spot in large-sample hydrology. Commonly used catchment descriptors aggregate water-body information into lumped statistics such as total reservoir volume or the percentage of land covered by water, but ignore the spatial organization of the water bodies within the catchment. A large reservoir high in the headwaters may have enormous storage capacity yet limited buffering power if its inflow area covers only a small fraction of the catchment. The authors argue that the fraction of the catchment above the water body, a metric that can be derived even when detailed management information is lacking, should become a standard descriptor, and that lakes and reservoirs, whose attenuation effects they show to be very similar, should be treated consistently rather than one being included and the other excluded as unnatural.

The researchers acknowledge limitations. A few catchments with small hydropower reservoirs and limited storage showed low peak ratios despite high contributing area percentages, suggesting that active storage capacity should be considered alongside spatial position. The FARL index, developed in the United Kingdom, performs poorly for deep Alpine reservoirs whose depth-to-area ratios make surface area a poor proxy for storage. And because Switzerland’s largest catchments all lie downstream of Lake Geneva or Lake Constance, the sample lacks large near-natural basins, complicating the separation of area and water-body effects. Nevertheless, the central pattern is robust: standing water bodies dampen hourly flood peaks far more than daily ones, and where they dominate a catchment, they erase the very distinction that a century of peak-ratio research has tried to quantify. As climate change intensifies Alpine storms and flood defenses are redesigned, the still waters behind dams and in mountain lakes may deserve far more credit, and far more careful accounting, than they have traditionally received.

Subject of Research: The scale-dependent attenuation of daily versus hourly flood peaks by lakes and reservoirs in Alpine catchments

Article Title: The influence of lakes and reservoirs on estimated flood peaks at hourly vs. daily timescale in the Alps

Article References: Götte, J., Astagneau, P. C., & Brunner, M. I. (2026). The influence of lakes and reservoirs on estimated flood peaks at hourly vs. daily timescale in the Alps. Hydrology and Earth System Sciences, 30(19), 6235-6248. https://doi.org/10.5194/hess-30-6235-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6235-2026

Keywords: flood peaks, lakes, reservoirs, Alps, hydrology, instantaneous peak flow, flood frequency analysis, catchment area, random forest, Switzerland, flood attenuation, streamflow

The research team, led by Muhammad Abrar Faiz and Liangliang Zhang of Northeast Agricultural University in Harbin, China, together with colleagues across China, the United Arab Emirates, Pakistan and Vietnam, confronted a long-standing uncertainty in hydrology. Scientists have known for decades that rivers are fed by two distinct pathways: quickflow, the rapid runoff generated when rain falls faster than the ground can absorb it, and baseflow, the delayed discharge of water that infiltrated the landscape weeks, months or even years earlier. What remained unclear was how strongly baseflow buffers rivers during droughts, and whether that buffering behaves consistently across different climates, geologies and continents. Without that knowledge, drought forecasts and water management plans risked misjudging how long rivers can keep flowing once the precipitation stops.

To resolve the question, the team assembled three complementary streams of evidence: observations from a global network of streamflow gauges, satellite-derived estimates of terrestrial water storage, and simulations from the CMIP6 generation of climate models used in international assessments. From these they constructed two new metrics. The first, the drought-sustained fraction, or DSF, expresses the share of drought-period streamflow that is contributed by baseflow. The second, the additive baseflow buffering index, or BBI, measures how much the baseflow fraction during drought departs from the climatological baseflow index, the long-term average proportion of river flow that comes from slow storage release. Together, the two indices allow researchers to distinguish between rivers that lean harder on groundwater when drought bites and rivers whose slow-flow contribution collapses along with everything else.

The results paint a coherent global picture. Under severe meteorological drought, 74.2 percent of the stations analyzed showed a positive buffering index, meaning the relative contribution of baseflow to streamflow strengthened as conditions dried. A further 12.7 percent of stations showed little change, while 13.1 percent actually registered a decline in the baseflow fraction, a signal that in some catchments even the stored water reserves become depleted or disconnected from the channel. In other words, at most of the world’s monitored rivers, drought does not simply shrink the stream; it transforms the character of the water in it, shifting the balance decisively toward old, slowly released groundwater rather than fresh surface runoff.

The mechanics behind this shift are rooted in the different response times of the two flow pathways. Quickflow depends on recent rainfall and near-surface saturation, so it collapses quickly once precipitation ceases. Baseflow, by contrast, is governed by the gradual drainage of water stored in soils, shallow aquifers and deeper geological formations, which continues to release water to channels long after the storm clouds have gone. The study found that the buffering index increased with low-flow severity and with the magnitude of the streamflow deficit, while baseflow retention declined. That combination indicates that quickflow generally decreased more rapidly than baseflow during droughts, so even as the absolute volume of baseflow diminished, its share of the shrinking river grew. The river becomes, in effect, a progressively more groundwater-dominated system as the drought deepens.

Perhaps the most consequential finding concerns the future. By comparing present-day flow partitioning with CMIP6 climate projections, the researchers discovered that how a catchment divides its water between quickflow and baseflow today constrains how its slow-flow behavior will unfold under warming. Catchments that currently rely heavily on storage release are likely to maintain that character, and the vulnerabilities embedded in present-day storage dynamics will carry forward into a hotter, more drought-prone world. This persistence means that the baseflow buffering observed now is not a transient curiosity but a structural property of each landscape, one that water managers can measure today and use to anticipate tomorrow’s low-flow regimes.

The implications ripple outward well beyond hydrology. Environmental flows that sustain fish, mussels and riverine vegetation during dry seasons depend almost entirely on baseflow in most of the world’s rivers, so the study provides a physical basis for setting drought-era flow thresholds. Water utilities drawing from rivers during droughts are, often unknowingly, harvesting groundwater discharged to the surface, which links surface-water allocation decisions directly to aquifer management. Agricultural regions that irrigate from shrinking streams are tapping the same slow-release reserves, and the 13 percent of stations where the baseflow fraction declined serve as a warning: where storage is exhausted or the connection between aquifer and channel is broken, even the groundwater safety net can fail.

The work also carries a methodological lesson for the modeling community. Because terrestrial water storage observations from satellite gravimetry were woven together with gauge records and climate model output, the study demonstrates how multi-source data can constrain processes that no single dataset captures alone. Climate models have historically struggled to represent groundwater discharge and storage-release dynamics realistically, and the finding that present-day partitioning predicts future slow-flow behavior offers a benchmark against which model simulations can be evaluated. Regions flagged by the analysis as having weak buffering, where baseflow declines during drought, deserve particular scrutiny in projections, since rivers there may lose flow faster than average models suggest.

As climate change intensifies the hydrological cycle, droughts are expected to become more frequent, longer and more severe in many regions, placing unprecedented demands on the storage reservoirs that sustain baseflow. This global inventory of drought buffering gives scientists and policymakers a baseline against which to track whether that resilience is eroding. It also elevates groundwater from an invisible afterthought to the central actor in drought narratives: the quiet, slow, subterranean release of stored water is what keeps rivers flowing, wetlands wet and ecosystems breathing when the sky withholds its share. Protecting recharge areas, managing aquifer withdrawals and preserving the landscape’s capacity to store water are therefore not peripheral concerns but the front line of drought adaptation for the majority of the world’s rivers.

Subject of Research: Global quantification of baseflow contribution to river flow during hydrological droughts

Article Title: Baseflow dominates river flow during droughts across the global streamflow network

Article References: Faiz, M. A., Zhang, L., Liu, D., Fu, Q., Li, M., Zhang, Y., Ma, N., Qi, X., Li, T., Cui, S., Baig, F., Naz, F., & Viet Ha, T. T. (2026). Baseflow dominates river flow during droughts across the global streamflow network. Communications Earth & Environment. https://doi.org/10.1038/s43247-026-04107-5

Image Credits: AI Generated

DOI: 10.1038/s43247-026-04107-5

Keywords: baseflow, hydrological drought, groundwater, streamflow, CMIP6, terrestrial water storage, river flow, low-flow, catchment storage, climate change, hydrology, drought buffering

The research, led by Felicia Yeboah of the University of Ghana with colleagues at the International Water Management Institute and Kenya’s State Department for Irrigation, focused on the Central Highlands, a 52,064-square-kilometer region spanning eleven counties around Mount Kenya. The team evaluated five widely used satellite precipitation products, CHIRPS, PERSIANN-CCS, PERSIANN-PDIR, TAMSAT, and MSWEP, against ground observations from eight rain gauge stations between 2012 and 2021. They also tested a model-derived river discharge dataset, VegDischarge, which simulates streamflow across more than 650,000 African river segments by routing runoff from the VegET agro-hydrologic model through the mizuRoute routing framework.

The validation relied on a battery of standard hydrological statistics. Root mean square error, or RMSE, quantifies the magnitude of deviations; the Nash-Sutcliffe Efficiency, or NSE, measures predictive skill relative to simply using the observed mean; the correlation coefficient captures how well timing and variability align; and relative mean bias reveals systematic over- or underestimation. Categorical metrics, the probability of detection and the false alarm ratio, assessed how well each product identified rainy days. The verdict was sobering: no product was uniformly reliable, and every one of them overestimated rainfall at most stations, a pattern the authors attribute to orographic processes that satellite algorithms struggle to resolve in complex terrain.

Still, two products stood out. CHIRPS achieved monthly RMSE values between 34.85 and 134.26 millimeters, NSE values from minus 0.94 to 0.44, and correlations of 0.52 to 0.87, while MSWEP recorded RMSE of 24.17 to 125.66 millimeters, NSE from minus 0.91 to 0.65, and correlations of 0.45 to 0.84. The two excelled in complementary ways: MSWEP generally reduced magnitude errors, whereas CHIRPS captured temporal variability more faithfully, posting stronger correlations at five of the eight stations. TAMSAT and the PERSIANN variants fared considerably worse, with TAMSAT producing NSE values as low as minus 4.56 at Munyaka during the long rains and RMSE figures approaching 1,000 millimeters at the annual scale.

Performance also shifted with season and elevation, a finding with direct operational consequences. During the March-to-May long rains, MSWEP was the most dependable product; during the October-to-December short rains, PERSIANN-PDIR took the lead; and across full years, CHIRPS proved most consistent. At wet, high-elevation sites such as Naro Moru and Nyahururu, all products degraded sharply, while in drier, semi-arid locations like Archer’s Post and Wamba they aligned far better with gauges. The lesson, the authors stress, is that product selection must be matched to the timescale, season, and location of the intended application, because no single dataset reproduces rainfall magnitudes reliably everywhere.

The discharge side of the analysis told a similar story of promise undermined by bias. VegDischarge successfully reproduced the seasonal rhythm of river flow, with peaks in April and May and a secondary rise in November tracking the bimodal rainfall regime. But the quantitative agreement was weak, with NSE values ranging from minus 2.28 to just 0.23 and correlations between 0.46 and 0.71. The model underestimated peak flows at the high-yielding Naro Moru stations, overestimated discharge elsewhere, and systematically underrepresented low flows across nearly all sites, a shortcoming traced to the underlying VegET model’s tendency to simulate baseflow at or near zero.

To rescue these datasets, the team applied four bias correction techniques: linear scaling, quantile mapping, the delta method, and a Random Forest machine learning model. The three statistical approaches improved matters to varying degrees, but the Random Forest method was in a class of its own. Trained on the relationship between satellite estimates and gauge observations, the ensemble of decision trees captured the complex, nonlinear error structures that simple scaling and distribution-based adjustments could not. After correction, monthly precipitation NSE rose to a consistently positive 0.85 to 0.95 across all eight stations, correlations climbed to 0.93 to 0.98, and RMSE was roughly halved at most sites, falling to between 9.05 and 46.23 millimeters.

The same technique worked equally dramatic improvements on streamflow. Corrected discharge reached NSE values of 0.80 to 0.88 and correlations of 0.92 to 0.95 at five of the six gauging stations, while preserving the variance structure of the hydrograph rather than smoothing it away. One station, A5 Naro Moru, exposed a structural limitation: Random Forest predictions are bounded by the range of their training data, so the method could not extrapolate to that catchment’s extreme flood peaks, and linear scaling actually performed better there. The authors note this weakness matters most precisely for flood applications, and they point toward hybrid, physics-informed correction approaches as the next frontier.

The implications extend well beyond Kenya. Sparse observational networks plague much of Africa, where global hydrological products are often adopted without local validation, and errors in rainfall inputs propagate through every downstream model that depends on them. An annual water-balance check lent physical credibility to the corrected datasets, yielding plausible positive storage changes of 2.8 to 8.1 cubic kilometers per year between 2018 and 2021 for this recharge-dominated water tower. For a region where agriculture, biodiversity, and growing urban populations all compete for water, the bias-corrected CHIRPS and VegDischarge datasets offer something previously scarce: defensible, high-accuracy estimates on which allocation decisions and flood planning can credibly rest.

Challenges remain. With only eight rainfall stations and six discharge gauges, the study’s corrections may not generalize to underrepresented peripheral areas, and the transferability of machine learning models across regions, time periods, and hydroclimatic regimes is still unproven. The authors also call for formal uncertainty quantification and for testing the newly released CHIRPS version 3 in similar environments. Even so, the study delivers a clear and consequential message: in data-scarce mountain regions, systematic bias, not random noise, is the dominant error in global hydrological products, and machine learning has matured into an essential tool for turning satellite eyes and model simulations into water management instruments that communities can actually trust.

Subject of Research: Bias correction of satellite precipitation products and model-derived river discharge in Kenya's Central Highlands

Article Title: Toward reliable water resource assessment: Correcting bias in satellite-based rainfall and model-derived discharge in Kenya’s Central Highlands

Article References: Yeboah, F., Owusu, A., Leh, M., Akpoti, K., Mekonnen, K., Adamseged, E. M., Odera, E., & Velpuri, N. (2026). Toward reliable water resource assessment: Correcting bias in satellite-based rainfall and model-derived discharge in Kenya’s Central Highlands. PLOS Water, 5(9), e0000623. https://doi.org/10.1371/journal.pwat.0000623

Image Credits: AI Generated

DOI: 10.1371/journal.pwat.0000623

Keywords: satellite precipitation, bias correction, machine learning, Random Forest, CHIRPS, MSWEP, VegDischarge, Kenya, Central Highlands, hydrology, streamflow, water resources

The stakes are high in snow-influenced regions. Floods there arise from two fundamentally different processes. Rainfall floods typically strike in summer or autumn, responding within hours or days to intense precipitation. Snowmelt floods, by contrast, unfold in spring or early summer as weeks of accumulated snowpack release water in a slow, sustained surge. A model that handles one flood type well may fail on the other, and most large-scale evaluations of neural networks have quietly averaged over this distinction, because snow-dominated catchments form a minority in global datasets. By splitting flood events by their generating process, the Norwegian team exposed performance differences that conventional evaluations would have masked.

The LSTM architecture is well suited in principle to the challenge. Unlike traditional recurrent neural networks, which suffer from vanishing gradients and forget information over long sequences, LSTMs use memory cells with gates that selectively retain or discard information across hundreds of time steps. In hydrological terms, this means the network can encode winter precipitation and temperature conditions, hold them in an internal state through months of cold, and then use that stored memory to anticipate the timing and volume of spring runoff. Earlier work has shown that LSTM cell states track snow accumulation and melt with correlations above 0.8 against snow depth reanalysis products, suggesting the network builds an internal analogue of a snowpack without ever being told one exists.

To test this rigorously, the researchers trained a single LSTM model on daily discharge records from 200 gauging stations across mainland Norway, a country spanning 58 to 71 degrees north with annual precipitation ranging from under 300 millimetres in the sheltered east to more than 6000 millimetres on the Atlantic-facing coast. The model received four dynamic forcing series per catchment, derived from the 1-kilometre SeNorge_2018 gridded dataset: catchment-averaged precipitation, catchment-averaged mean temperature, and the catchment-minimum and catchment-maximum daily temperatures. Twenty-one static attributes describing each catchment’s geometry, physiography and climate completed the input. Training covered the fifteen hydrological years from September 2009 to August 2024, with hyperparameters tuned through five-fold cross-validation and the catchment-averaged Nash-Sutcliffe efficiency used as the loss function.

The benchmark was no straw man. The HBV model, a semi-distributed conceptual rainfall-runoff model dividing each catchment into ten elevation zones with explicit snow, soil and runoff storages, has underpinned Norwegian flood forecasting for decades. Its twelve parameters were calibrated catchment by catchment over the same period using the same loss function, ensuring a fair comparison in which both models saw identical data. Evaluation was then performed over a completely unseen fifteen-year window from September 1994 to August 2009, a period deliberately chosen to precede training and to avoid artefacts from changes in the precipitation observing network.

On overall streamflow simulation, the LSTM was emphatically strong. It achieved an average Nash-Sutcliffe efficiency of 0.84 and an average Kling-Gupta efficiency of 0.86 across the 103 evaluated catchments, with NSE exceeding 0.7 in 100 of them. Compared with HBV, the LSTM posted higher NSE in 96 percent of catchments and higher KGE in 77 percent, with the largest gains concentrated where the operational model performed worst. Mean absolute errors were lower for the LSTM in all but three catchments. The only metric where HBV prevailed was the variability ratio component of KGE, a known consequence of training against NSE, which mathematically rewards a slight underestimation of flow variability.

The flood-specific results revealed a genuine asymmetry. The LSTM simulated the correct peak day for 77 percent of rainfall-generated floods but only 50 percent of snowmelt-generated floods, a 27 percentage point gap that mirrored a similar 29 point gap in HBV. The researchers attribute this not to any failure to learn snow processes, but to the physics of the events themselves: snowmelt floods last days to weeks and lack the sharp, unambiguous peak of a rainfall-driven flood, so discharge on adjacent days differs little and hitting the exact day is intrinsically harder. Supporting this interpretation, peak magnitude errors were nearly identical between flood types, with close to half of all events, whether snowmelt or rainfall driven, simulated within 20 percent of the observed maximum. False alarm rates and probabilities of detection were likewise similar for the two pure flood types and slightly worse for mixed events.

Against the operational benchmark, the deep learning model won broadly. LSTM outperformed HBV in 70 to 91 percent of catchments depending on the flood type and metric, with median timing improvements of 14 percentage points for snowmelt floods and 11 for rainfall floods, and improvements exceeding 30 points in several catchments. The largest gains in peak magnitude came for rainfall-generated events, particularly in catchments where HBV’s errors exceeded 40 percent. Notably, the LSTM achieved its improved bias ratios without the precipitation correction factors HBV requires to balance its water budget, suggesting that the data-driven model’s freedom from hard physical constraints can, in this case, yield a more faithful representation of average streamflow.

The study’s authors are careful about what the comparison does and does not prove. By constraining the LSTM’s forcing data and training period to match HBV’s, they guaranteed fairness but also left obvious headroom: longer training records, richer atmospheric inputs, and per-catchment fine-tuning could all push performance further. They also note that the preferred model sometimes depends on the application, since flood warning prizes timing while inundation mapping prizes magnitude, and in roughly a fifth of catchments the two metrics pointed toward different models. Still, the overall message is one of confidence. A neural network trained on data alone, with no snow routine written into its code, can match or beat a physically structured operational model on both of the distinct flood-generating processes that define hydrology in snowy latitudes. For national hydrological services weighing the future of flood forecasting, that is a result worth taking seriously.

Subject of Research: Evaluating LSTM deep learning models for simulating snowmelt- and rainfall-generated floods in Norwegian catchments

Article Title: The ability of LSTM to model snowmelt versus rainfall generated floods

Article References: The ability of LSTM to model snowmelt versus rainfall generated floods. (n.d.). https://doi.org/10.5194/hess-30-6207-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-6207-2026

Keywords: LSTM, deep learning, flood prediction, snowmelt, rainfall-runoff modelling, hydrology, HBV model, Norway, streamflow, Nash-Sutcliffe efficiency, catchments, flood forecasting

The research team, led by Yuhan Guo and Jinghua Xiong of Tsinghua University together with Yuting Yang and Dawen Yang, compiled streamflow observations from 29 global and regional hydrometric networks and databases, assembling one of the largest observational samples ever brought to bear on the question of how drought propagates from sky to stream. Their study window spans 1981 to 2019, four decades during which climate warming has measurably intensified the atmosphere’s evaporative pull. To quantify atmospheric drought, the team used the Standardized Precipitation Evapotranspiration Index, or SPEI, computed from multiple combinations of observational precipitation and evaporation datasets, including CHIRPS, MSWEP, GLEAM4 and hPET. To quantify hydrological drought, they used the Standardized Streamflow Index, or SSI, computed directly from measured river discharge at each catchment.

The contrast between the two indices is the heart of the finding. When the researchers mapped trends in SPEI across the globe, they saw a broad spatial predominance towards drying: across most of the land surface, the atmosphere has indeed been growing more drought-prone over the past forty years. Streamflow drought trends, by contrast, were substantially more heterogeneous. Some catchments dried in step with their climate; others showed no significant change in low flows at all; and a considerable fraction became wetter even as the atmospheric drought index above them worsened. At the global scale, the authors describe the streamflow response as highly spatially variable, a pattern that stands in stark contradiction to the assumption, embedded in many drought assessments, that atmospheric drying marches in lockstep with river drying.

Why should rivers so often refuse to follow the atmosphere? The answer lies in what hydrologists call drought propagation — the chain of storage, routing and feedback processes that stand between a precipitation deficit and a flow deficit. Rain that falls on a catchment does not run straight into the channel. It infiltrates soils, recharges groundwater, fills lakes and wetlands, is intercepted by canopies and transpired back to the air by vegetation. Each of these stores acts as a buffer, and each buffer introduces nonlinearity. A modest atmospheric drying signal may be absorbed entirely by soil moisture storage before it ever reaches the stream, attenuating the hydrological response. In snow-dominated basins, warming can shift precipitation from snow to rain and alter the timing of melt, reshaping the seasonal signature of drought in ways that a simple climate index cannot anticipate. In some catchments, storage release can delay the emergence of a flow deficit by months or even years; in others, threshold behavior in groundwater or permafrost can amplify a modest climatic nudge into a dramatic collapse in low flows.

The team’s quantitative treatment of this propagation is one of the study’s most distinctive contributions. Using a basin-level response metric computed from paired atmospheric and hydrological drought trends, they classified each catchment’s hydrological response to changes in atmospheric drought frequency and magnitude as directionally decoupled, weakly responsive, approximately proportional or amplified. The results were strikingly asymmetric. When atmospheric drought frequency or magnitude increased, the hydrological response was dominated by weak or directionally decoupled behavior — together accounting for roughly 62 to 63 percent of catchments. In other words, in most places where the atmosphere grew more drought-prone, rivers did not follow proportionally. But when atmospheric drought conditions improved, the story flipped: decreases in either atmospheric drought characteristic produced amplified hydrological responses in about 55 percent of catchments. Drying signals get dampened on the way down to the river; wetting signals get magnified. This asymmetry, the authors argue, is a fingerprint of the nonlinear storage and release dynamics that govern how catchments hold and surrender water.

The implications of that asymmetry cut in uncomfortable directions for water managers. If worsening atmospheric drought is routinely attenuated before it reaches streamflow, then river-flow droughts may emerge later, weaker or in different places than atmospheric drought monitoring alone would suggest — a deceptive cushion that could encourage complacency until storage buffers are exhausted. Conversely, the amplification of improving conditions means that when the atmosphere does ease, rivers can rebound faster and more vigorously than the climate signal implies. Neither behavior is captured by drought indices computed purely from precipitation and evaporative demand, which remain the backbone of most operational drought early-warning systems. The study’s authors emphasize that assessing water-security risks under climate warming requires accounting explicitly for hydrological propagation rather than assuming a one-to-one transmission of atmospheric drying into water supply.

The findings also speak to a long-running scientific debate. Earlier work has oscillated between claims of intensifying global drought driven by rising evaporative demand and counterarguments that such indices overstate drying, in part because potential evaporation formulations can exaggerate temperature-driven increases in atmospheric thirst. Research published in recent years has documented warming-accelerated drought severity at the global scale, while other studies have highlighted disconnections between atmospheric drying trends and continental runoff, and the role of vegetation responses to elevated carbon dioxide in moderating evapotranspiration. The new analysis does not resolve that debate by picking a side on the atmosphere — it confirms that atmospheric drying is widespread — but it reframes the question. The more consequential uncertainty, the study suggests, lies not in the climate index but in the catchment: in soils, geology, snow, glaciers, vegetation and, increasingly, human engineering.

Human interventions loom large in that last category. Reservoirs regulate and redistribute flow, irrigation returns water to the atmosphere and depletes groundwater, deforestation and afforestation alter evapotranspiration and runoff partitioning, and water abstraction directly removes flow from channels. The study’s global sample inevitably includes catchments shaped by all of these forces, and the authors note that such landscape changes and human interventions can reinforce or counteract climate-driven trends in low flows. Documented examples abound in the recent literature: streamflow declines across China attributed to landscape change and human intervention, irrigation-driven groundwater depletion reducing downstream discharge in the Ganges-Brahmaputra system, and forest expansion and reservoir construction reshaping century-long discharge trends in the Mediterranean basin. Disentangling the climatic component of streamflow drought from the managed component remains one of the field’s hardest attribution problems, and the new global dataset — with metadata for all catchments made openly available through Zenodo — provides a foundation for that effort.

Methodologically, the study is careful about the uncertainties that plague large-sample hydrology. Streamflow records were quality-controlled and standardized, SPEI and SSI were computed in Python using distribution fitting with established statistical packages, and trend tests were applied with corrections for autocorrelation. Atmospheric drought results were cross-checked against four independent combinations of precipitation and evaporation products, and the analysis was repeated across both large and small non-nested catchments to guard against spatial-scale artifacts. Regional aggregation using IPCC reference regions confirmed that the divergence between atmospheric and hydrological drought trends is not an artifact of any single mapping choice. Where data coverage was insufficient for robust trend estimation, the authors said so rather than extrapolating.

What emerges is a picture of the global water cycle that is messier, more contingent and more place-specific than headline drought maps imply. The atmosphere is drying almost everywhere, but the rivers are writing their own regional stories, governed by the storage architectures and human histories of individual basins. For the roughly two-thirds of catchments where atmospheric intensification has so far failed to translate into proportional flow deficits, the message is not reassurance but a warning about buffers: soil moisture, snowpack and groundwater reservoirs are finite, and as warming deepens, the attenuation they provide may weaken. For the places where hydrological drought has already outpaced its atmospheric driver, the study offers an explanation — and a reminder that the water crisis of a warming century will be decided not just in the sky, but in the ground beneath our feet.

Subject of Research: Global observational assessment of how atmospheric drought propagates into hydrological drought in rivers

Article Title: Global hydrological drought diverges from atmospheric drying

Article References: Guo, Y., Xiong, J., Yang, Y., Liang, S., Guo, L., & Yang, D. (2026). Global hydrological drought diverges from atmospheric drying. Nature Water. https://doi.org/10.1038/s44221-026-00724-8

Image Credits: AI Generated

DOI: 10.1038/s44221-026-00724-8

Keywords: hydrological drought, atmospheric evaporative demand, streamflow, drought propagation, climate change, SPEI, SSI, catchments, water security, Nature Water, hydrology, river flow