The study, published in Hydrology and Earth System Sciences, relies on daily streamflow records from 1982 to 2020 at USGS gauges compiled in the GAGES-II dataset. After filtering for data completeness—requiring at least eight years of observations per decade with 95 percent of days recorded—the team settled on 3,198 gauges. To define drought objectively, they converted each day’s flow into a percentile using the Weibull plotting position, computed within a 30-day window centered on that calendar day across all years. That trick removes the seasonal cycle, so drought becomes any departure from what a river normally does at that time of year. Anything below the 20th percentile was flagged as drought, a threshold chosen to balance a usable definition against enough events to train a statistical model. Zero-flow days, common in arid and intermittent streams, were handled with a combined threshold-level and continuous-dry-period method that ranks consecutive dry days rather than treating them as ties.

For each gauge, the researchers built a random forest classifier fed with 130 candidate predictors: precipitation, soil moisture at multiple depths, potential evapotranspiration, snow water equivalent, temperature, the Standardized Precipitation Evapotranspiration Index, large-scale climate teleconnections such as El Niño–Southern Oscillation, the Pacific Decadal Oscillation, the Atlantic Multidecadal Oscillation, and the Pacific–North American pattern, plus even sunspot counts. Crucially, each variable entered the model in several forms—raw values, percentile-transformed values, and rolling averages over 30, 90, and 365 days—so the forests could learn which timescales of antecedent conditions matter most. Random forests were chosen for their resistance to overfitting, their tolerance of correlated predictors, and their built-in variable importance scores, which formed the analytical heart of the study. Models were trained on 1985–2015 and tested on the bookend periods 1982–1984 and 2016–2020, and scored with Cohen’s Kappa, a metric robust to the severe class imbalance created when drought occupies only 20 percent of the record.

The single most influential predictor nationwide turned out to be soil moisture at 40–100 centimeters depth, followed by soil moisture at 10–40 centimeters, then a trio of precipitation variables smoothed over 90 and 365 days and the 90-day SPEI. Nothing else came close. But the real story lay in the geography. Potential evapotranspiration dominated in the South, Southwest, and Southeast; snow water equivalent mattered most in the Southwest, Northern Rocky Mountains, Northwest, and West; and the teleconnection indices, while weak at individual gauges, carried outsized weight across the western half of the country. Deep soil moisture was especially important in humid regions, which the authors attribute to hydrologic memory: deficits stored deep in the soil profile sustain low flows long after a passing rainstorm has improved surface conditions, resisting rapid drought recovery.

To transform thousands of gauge-level importance scores into a coherent national picture, the team applied principal component analysis, compressing 130 variables into 27 components that together explained 95 percent of the variance. The first three components told most of the story. The first aligned teleconnections, temperature, and evaporative demand—a cluster centered on the West, Southwest, and Northern Rockies, and notably the same regions where the at-site models performed worst, largely because water management alters flows in ways the data cannot capture. The second component was a moisture axis pitting long-term soil moisture against short-term precipitation, cleanly dividing the humid East from the arid West. The third distinguished snow-driven basins in the mountains and northern latitudes from baseflow-driven, soil-moisture-dominated systems in the South and Southeast, drawing the map along the line between seasonal snowpack storage and subsurface storage.

The regressions linking these components to static basin characteristics revealed why certain regions respond to certain drivers. Basins where teleconnections and temperature loomed large tended to be warmer, rainier, forested or cultivated, and equipped with permeable soils but limited buffering capacity—meaning they lack the soil or snow storage to insulate streamflow from large-scale climate swings. Basins scoring high on the snow-and-soil-moisture component were characterized by high elevation, a large fraction of precipitation falling as snow, and cold freeze timing, consistent with evidence that earlier, slower snowmelt reduces streamflow production. In effect, the study shows that a basin’s physiography—its soils, topography, land cover, and snow regime—determines which meteorological levers can push it into drought, and how hard those levers must be pulled.

The second half of the study tackles a long-standing gap: roughly one-fifth of the nation’s land area drains through watersheds too small, too high, or too remote to appear in the gauge network, and the gaps are not randomly distributed. The team’s solution is a novel dynamic regionalization scheme built on donor gauges. For any hypothetical ungauged location, principal component scores are estimated from its physical characteristics using regressions trained on the 1,900 gauges whose models passed a Kappa threshold of 0.4. Each candidate donor is then scored by how closely its component profile matches the target, weighted by the variance each component explains. The 19 most similar donors—regardless of whether they sit across the country or across the county—are enlisted, their random forest models run on the target’s climate data, and their drought probability predictions averaged and thresholded at 0.35 to yield a final daily drought forecast.

The results were striking: donor-based predictions achieved a mean Kappa of 0.40 across all sites, only 0.02 below the at-site models trained locally on that same gauge’s own data. Even more intriguingly, the ensemble sometimes outperformed the local models, particularly where those models were weak—likely because averaging 19 similar gauges suppresses the noise that a single model overfits. Two worked examples illustrate the method’s flexibility. For Brodhead Creek in Pennsylvania, the selected donors averaged just 219 kilometers away and agreed tightly in their predictions. For the Gunnison River in Colorado, the nearest characteristic donors sat an average of 2,891 kilometers away and disagreed more, yet the ensemble still matched the local model’s accuracy almost exactly. Similarity, not proximity, is what counts.

Beyond its practical value for water managers, the framework offers a partial answer to the black-box criticism that haunts machine learning in high-stakes applications. Because the workflow is compartmentalized—drought definition, random forest training, dimensionality reduction, donor regression, ensemble prediction—a modeler can inspect and adjust each stage, swapping donor sets and observing how predictions change to gauge each donor’s influence. The authors note that tools such as partial dependence analysis or SHAP values could further disentangle the overlapping drivers, especially in the West, where multiple drought typologies stack on top of one another, and they caution that a season-specific modeling approach could better separate snowmelt deficits in spring from evaporative-demand-driven drought in late summer.

The timing of such a tool could hardly be better. Under continued warming, rising evaporative demand and shrinking snowpack are expected to amplify the importance of temperature and energy-balance drivers in the West and Northern Rockies, while shifting precipitation seasonality could redraw the driver map in the East. The drought typology established here provides a baseline against which those future shifts can be measured—and, because the donor method needs no local streamflow record, it extends early-warning capability to the small headwater basins and high-elevation catchments that supply so much of the nation’s water but that gauges have long overlooked.

Subject of Research: Machine learning prediction of streamflow drought and its meteorological drivers across the conterminous United States

Article Title: Predicting streamflow drought in the conterminous United States using machine learning and a donor-gage approach, 1982–2020

Article References: Heldmyer, A., Sando, R., Simeone, C., Wieczorek, M., Hamshaw, S., Goodling, P., McShane, R., Diaz, J., Watkins, D., Pulver, B., Shastry, A., Hafen, K., & Hammond, J. (2026). Predicting streamflow drought in the conterminous United States using machine learning and a donor-gage approach, 1982–2020. Hydrology and Earth System Sciences, 30(18), 5925-5945. https://doi.org/10.5194/hess-30-5925-2026

Image Credits: AI Generated

DOI: 10.5194/hess-30-5925-2026

Keywords: streamflow drought, machine learning, random forest, hydrology, donor gage, ungauged basins, snow water equivalent, soil moisture, teleconnections, USGS, drought prediction, principal component analysis

The framework rests on a deceptively simple question. When rain falls on a landscape, how predictably does that rainfall translate into a rise in streamflow? To answer it at global scale, the researchers compiled daily streamflow and climate time series for thousands of monitored catchments and identified thousands of individual rainfall–runoff events. For each catchment, they plotted how much rain fell during an event against how much stormflow appeared in the stream, and they characterized the shape of that relationship. Catchments were then assigned to one of three functional types. Simple catchments show a near-linear relationship, where a given amount of rain produces a proportional and reliable amount of runoff. Intermediate catchments display segmented or threshold-like behavior, responding one way until a saturation point is crossed and another way beyond it. Complex catchments produce scattered, highly variable responses, where the same rainfall can yield wildly different runoff outcomes depending on the state of the landscape.

The technical distinction matters because these three categories imply fundamentally different modeling requirements. A simple catchment can be represented with a straightforward linear transfer function between rainfall and runoff, requiring few parameters and little ancillary data. An intermediate catchment demands at least a two-regime model that captures a breakpoint—for instance, the transition examined in a French catchment where the runoff response steepened sharply beyond roughly 110 millimeters of event rainfall. A complex catchment, by contrast, defies any single tidy equation; its behavior is shaped by shifting storage thresholds, variable flow pathways and antecedent conditions, and it calls for richer model structures and greater caution when parameters are transferred from one basin to another. The classification thus functions as a diagnostic tool: it tells a modeler, before any calibration begins, how elaborate a model needs to be for a given place—neither simpler than necessary nor more complex than required.

When the team extrapolated their trained classification model to ungauged territory, using a machine learning approach built on the XGBoost tree-boosting algorithm, the global picture that emerged upended a long-standing assumption. Hydrologists have often hoped that catchment behavior could be generalized with relatively simple rules. Instead, the analysis found that catchments draining 87 percent of the evaluated ungauged land area fall into the complex category during the dormant season, and that the number of complex catchments climbs by a further 63 percent when the growing season is considered. Overall, complex catchments were found to drain some 121 million square kilometers of ungauged, rain-dominated land globally in the dormant season alone. In other words, the hardest-to-predict behavior is not an exception confined to a few quirky basins—it is the planetary norm, and it is concentrated precisely in the regions where streamflow is least observed.

The geography of complexity is far from uniform, and the seasonal dimension of the analysis reveals patterns that would be invisible in a static snapshot. Africa and much of Asia, with the notable exception of Japan, remain predominantly complex year-round, suggesting that hydrologic prediction in these regions will require persistent investment in sophisticated model structures and in new observational infrastructure. Europe undergoes a dramatic seasonal intensification: by the growing season, 94 percent of European catchments are classified as complex, with widespread shifts toward complexity documented across France, Germany, Denmark, Ireland and the southeastern United Kingdom. In North America, a pronounced swing toward complexity dominates the eastern United States as the growing season progresses, while the Pacific Northwest of the United States and the neighboring Canadian province of British Columbia stand out as rare strongholds where clusters of simple catchments persist. Meanwhile, southeastern South America and southeastern Australia display the opposite trajectory, with catchments transitioning from complex in the dormant season to intermediate in the growing season—demonstrating that a catchment’s functional type is not a fixed attribute but a shifting property of the coupled land–water system.

Beneath the regional patterns, the researchers searched for the environmental factors that govern where complexity takes hold. Climate emerged as the dominant control. Rainfall persistence indices, which measure the fraction of time in each season with meaningful rainfall, showed the strongest separation between classes: simple catchments tend to sit in persistently wet climates, complex catchments in more variable ones, with intermediate systems arrayed between. Overall water availability, measured through aridity and effective precipitation, reinforced the same divide between wet, predictable landscapes and drier, erratic ones. On top of this climatic template, physiographic and land-cover variables fine-tune the picture. Simple catchments tend to be smaller and steeper; complex catchments have greater depth to bedrock. Most provocatively, the analysis found that catchments with greater urban and agricultural land cover are consistently associated with greater functional complexity—a signal that human modification of landscapes may be actively reshaping how water moves through them, with implications for flood forecasting in developing regions.

The seasonal transitions themselves carry scientific weight. Among gauged catchments, the most common pathway between the two seasons was a shift from intermediate to complex behavior, with 402 European catchments and 314 in the United States making that jump, likely reflecting intensified runoff generation as vegetative growth, evapotranspiration and soil moisture dynamics reshape subsurface storage. Yet the opposite trend appeared in Latin America, the Caribbean and Oceania, where hundreds of catchments simplified from complex to intermediate. This heterogeneity underscores that the same broad climatic forces can push different landscapes in different functional directions, and it cautions against assuming that a model calibrated in one season will hold in another. For the field of large-sample hydrology, which seeks universal patterns across thousands of basins, the message is that functional diversity must be treated as dynamic, not static.

The practical implications extend well beyond academic taxonomy. Because gauges are expensive to install and maintain, water agencies face hard choices about where to deploy limited monitoring resources. The new framework provides an evidence-based blueprint for prioritizing streamflow gauging in under-represented complex catchments—exactly the places where a single well-placed gauge would yield the greatest scientific return, since complex basins both resist prediction and dominate the ungauged land area. The findings also speak directly to the boom in machine learning approaches to rainfall–runoff modeling. As deep learning models proliferate, knowing a catchment’s functional class offers a principled way to judge whether a model’s internal behavior is realistic, and whether a model trained in data-rich Europe can be credibly transferred to complex, ungauged African basins.

To make the work usable, the team has released the underlying dataset of streamflow, climate and catchment attributes through Figshare, published the trained classification code on GitHub, and built interactive global maps showing season-specific functional classes for the 77,585 ungauged catchments analyzed in the study. They have gone further, compiling a user-friendly web application that classifies any catchment from an uploaded boundary polygon, returning season-specific labels within moments. In an era when climate change is eroding the stationary assumptions on which decades of water management were built, a global, dynamic map of where catchments behave predictably—and where they do not—may prove one of the most consequential tools hydrology has produced for confronting the world’s unmeasured waters.

The study builds on two decades of community effort. The International Association of Hydrological Sciences’ decade on Predictions in Ungauged Basins, launched in 2003, framed the ungauged-stream problem as a central scientific challenge, and the new classification offers a functional rather than purely physical answer to it. Where earlier regionalization efforts grouped catchments by climate, geology or landscape descriptors, this framework groups them by observed behavior—how rainfall is actually converted into runoff across many individual events.

That behavioral emphasis connects the work to a broader shift in hydrology away from assuming linearity. Field studies of subsurface stormflow have long documented fill-and-spill threshold dynamics, in which hillslopes release water abruptly once a storage limit is exceeded, and meta-analyses of small forested catchments confirm that such nonlinearities are widespread. The three-class scheme effectively quantifies how often those threshold and scatter-dominated behaviors dominate at continental scale.

The open release of data and code also matters scientifically. Because the event catalogue, catchment attributes and trained model are publicly available, other researchers can test whether functional classes align with independent evidence such as water transit times, tracer-based flow-path diagnostics or satellite-extended gauge records, and can refine the seasonal boundaries as climate records lengthen.

Subject of Research: Global classification of hydrologic functional diversity in gauged and ungauged catchments

Article Title: A global classification of hydrologic functional diversity in gauged and ungauged catchments

Article References: Ameli, A. A., Sharif, H., & McDonnell, J. J. (2026). A global classification of hydrologic functional diversity in gauged and ungauged catchments. Nature Water. https://doi.org/10.1038/s44221-026-00699-6

Image Credits: AI Generated

DOI: 10.1038/s44221-026-00699-6

Keywords: hydrology, catchments, runoff, rainfall, ungauged basins, streamflow, machine learning, seasonal variability, land cover, water security, flood prediction, Nature Water