Explore the vast range of titles available.

This study identifies major climate extreme indices as drivers of stream flow discharge variability using machine learning and the SHaply Additive Explanation. The homogenized and bias-corrected downscaled rainfall and temperatures were used to compute fifteen climate extreme indices using RClimdex. The data set was partitioned into 70% and 30% for Auto Machine Learning (AutoML) training and testing of the 38 machine learning models. The coefficient of determinant ( R 2 ), mean square error (MSE), and root squared mean error (RSME) were used to evaluate the models. The variability and trend of the stream flow discharge was assessed using Mann Kendall and Sen’s slope. The findings revealed that there is high variability with an insignificant negative trend ( Z : – 0.90, P  > 0.05) in inter-annual discharge. In addition, among the 38 ML, the extra trees regression proved to perform better with a coefficient of determinant ( R 2  > 0.9) and minimal MSE (less than 7 m 3 ) and RMSE (less than 4 m 3 ) for the training and testing. In addition, the result revealed that the maximum temperature (TMAX), total precipitation (PRCPTOT), and monthly maximum consecutive 5-day precipitation (RX5DAY) are the major climate indices that influence the stream flow discharge variability of the Kaduna River catchment area. From the findings, it was concluded that climate indices have a greater impact on water resources than average daily precipitation and temperature. As such, it is recommended that the combination of machine learning and SHaply Additive Explanation value makes it a great tool for exploring the impact of climate on water resources for policy making and sustainable water resource development and management.

Arctic rivers drain ~15% of the global land surface and significantly influence local communities and economies, freshwater and marine ecosystems, and global climate. However, trusted and public knowledge of pan-Arctic rivers is inadequate, especially for small rivers and across Eurasia, inhibiting understanding of the Arctic response to climate change. Here, we calculate daily streamflow in 486,493 pan-Arctic river reaches from 1984-2018 by assimilating 9.18 million river discharge estimates made from 155,710 satellite images into hydrologic model simulations. We reveal larger and more heterogenous total water export (3-17% greater) and water export acceleration (factor of 1.2-3.3 larger) than previously reported, with substantial differences across basins, ecoregions, stream orders, human regulation, and permafrost regimes. We also find significant changes in the spring freshet and summer stream intermittency. Ultimately, our results represent an updated, publicly available, and more accurate daily understanding of Arctic rivers uniquely enabled by recent advances in hydrologic modeling and remote sensing. The authors combine satellite data with hydrologic models to investigate recent changes in pan-Arctic river discharge magnitude, trends, and seasonality for nearly half a million rivers. They reveal that these rivers likely exported 3-17% more water to the global ocean than previously thought from 1984-2018.

Groundwater discharge generates streamflow and influences stream thermal regimes. However, the water quality and thermal buffering capacity of groundwater depends on the aquifer source-depth. Here, we pair multi-year air and stream temperature signals to categorize 1729 sites across the continental United States as having major dam influence, shallow or deep groundwater signatures, or lack of pronounced groundwater (atmospheric) signatures. Approximately 40% of non-dam stream sites have substantial groundwater contributions as indicated by characteristic paired air and stream temperature signal metrics. Streams with shallow groundwater signatures account for half of all groundwater signature sites and show reduced baseflow and a higher proportion of warming trends compared to sites with deep groundwater signatures. These findings align with theory that shallow groundwater is more vulnerable to temperature increase and depletion. Streams with atmospheric signatures tend to drain watersheds with low slope and greater human disturbance, indicating reduced stream-groundwater connectivity in populated valley settings. Groundwater discharge generates streamflow and influences stream thermal regimes. Classifying more than 1700 streams across the US by using an empirically-based approach the study shows that the vulnerability of streams to stressors depends on the aquifer source-depth of groundwater discharge

Groundwater is the world’s largest freshwater resource and is critically important for irrigation, and hence for global food security 1 – 3 . Already, unsustainable groundwater pumping exceeds recharge from precipitation and rivers 4 , leading to substantial drops in the levels of groundwater and losses of groundwater from its storage, especially in intensively irrigated regions 5 – 7 . When groundwater levels drop, discharges from groundwater to streams decline, reverse in direction or even stop completely, thereby decreasing streamflow, with potentially devastating effects on aquatic ecosystems. Here we link declines in the levels of groundwater that result from groundwater pumping to decreases in streamflow globally, and estimate where and when environmentally critical streamflows—which are required to maintain healthy ecosystems—will no longer be sustained. We estimate that, by 2050, environmental flow limits will be reached for approximately 42 to 79 per cent of the watersheds in which there is groundwater pumping worldwide, and that this will generally occur before substantial losses in groundwater storage are experienced. Only a small decline in groundwater level is needed to affect streamflow, making our estimates uncertain for streams near a transition to reversed groundwater discharge. However, for many areas, groundwater pumping rates are high and environmental flow limits are known to be severely exceeded. Compared to surface-water use, the effects of groundwater pumping are markedly delayed. Our results thus reveal the current and future environmental legacy of groundwater use. Estimates for when critical environmental streamflow limits will be reached—with potentially devastating economic and environmental effects—are obtained using a global model that links groundwater pumping with the groundwater flow to rivers.

Groundwater resources are vital to ecosystems and livelihoods. Excessive groundwater withdrawals can cause groundwater levels to decline 1 – 10 , resulting in seawater intrusion 11 , land subsidence 12 , 13 , streamflow depletion 14 – 16 and wells running dry 17 . However, the global pace and prevalence of local groundwater declines are poorly constrained, because in situ groundwater levels have not been synthesized at the global scale. Here we analyse in situ groundwater-level trends for 170,000 monitoring wells and 1,693 aquifer systems in countries that encompass approximately 75% of global groundwater withdrawals 18 . We show that rapid groundwater-level declines (>0.5 m year −1 ) are widespread in the twenty-first century, especially in dry regions with extensive croplands. Critically, we also show that groundwater-level declines have accelerated over the past four decades in 30% of the world’s regional aquifers. This widespread acceleration in groundwater-level deepening highlights an urgent need for more effective measures to address groundwater depletion. Our analysis also reveals specific cases in which depletion trends have reversed following policy changes, managed aquifer recharge and surface-water diversions, demonstrating the potential for depleted aquifer systems to recover. Analysis of about 170,000 monitoring wells and 1,693 aquifer systems worldwide shows that extensive and often accelerating groundwater declines are widespread in the twenty-first century, but that groundwater levels are recovering in some cases.

The behaviors and skills of models in many geosciences (e.g., hydrology and ecosystem sciences) strongly depend on spatially-varying parameters that need calibration. A well-calibrated model can reasonably propagate information from observations to unobserved variables via model physics, but traditional calibration is highly inefficient and results in non-unique solutions. Here we propose a novel differentiable parameter learning (dPL) framework that efficiently learns a global mapping between inputs (and optionally responses) and parameters. Crucially, dPL exhibits beneficial scaling curves not previously demonstrated to geoscientists: as training data increases, dPL achieves better performance, more physical coherence, and better generalizability (across space and uncalibrated variables), all with orders-of-magnitude lower computational cost. We demonstrate examples that learned from soil moisture and streamflow, where dPL drastically outperformed existing evolutionary and regionalization methods, or required only ~12.5% of the training data to achieve similar performance. The generic scheme promotes the integration of deep learning and process-based models, without mandating reimplementation. Much effort is invested in calibrating model parameters for accurate outputs, but established methods can be inefficient and generic. By learning from big dataset, a new differentiable framework for model parameterization outperforms state-of-the-art methods, produce more physically-coherent results, using a fraction of the training data, computational power, and time. The method promotes a deep integration of machine learning with process-based geoscientific models.

Human-induced climate change has increased the frequency and intensity of heavy precipitation 1 . Due to the complexity of runoff generation and the streamflow process, the historical impact of human-induced climate change on river flooding remains uncertain. Here, we address the question of whether anthropogenic climate change has altered the probability of the extreme river flood events for the period 1951–2010 based on simulated river discharge derived from large ensemble climate experiments with and without human-induced climate change. The results indicate that human-induced climate change altered the probabilities of 20 of the 52 analyzed flood events. Fourteen of these 20 flood events, which occurred mainly in Asia and South America, were very likely to have been enhanced by human-induced climate change due to an increase in heavy precipitation. Conversely, two flood events in North/South America and two flood events in Asia and two flood events in Europe were suppressed by human-induced climate change, perhaps as a result of lower snowfall. Human-induced climate change has enhanced flooding more prominently in recent years, providing important insights into potential adaptation strategies for river flooding.

Intermittent streams that regularly dry up constitute over half of the world's river network. They exhibit biogeochemical processes distinct from those of continuously flowing perennial rivers. In perennial rivers, discharge is often perceived as predominantly driving water chemistry, as demonstrated by the widespread use of concentration–discharge (CQ) relationships. Does discharge similarly drive water chemistry in intermittent streams? Given its extended periods of no flow, we hypothesized that stream chemistry depends less on discharge alone but more on the granularity of dry‐wet transitions, including their direction (drying or rewetting), history (antecedent conditions), and intermittency. We tested this hypothesis by analyzing three decades of streamflow and solute chemistry data from an intermittent stream (N04D) in the Konza Prairie Biological Station, a Long‐Term Ecological Research site in Kansas, USA. Results showed that concentrations are generally higher at no flow compared to flow conditions and depend on dry‐wet transitions. Geogenic solutes were predominantly chemostatic (relatively constant C without Q dependence), contrasting primarily dilution patterns (decreasing C with Q) in perennial rivers. Biogenic solutes did not exhibit pronounced discharge‐dependent patterns commonly observed in perennial rivers at decadal scale; at monthly scale, however, they exhibited a transition from highly variable CQ patterns at low flows to consistent flushing patterns (increasing C with Q) at flows higher than 2.5–5 mm/day. These observations support our hypothesis of weaker chemistry dependence on discharge in intermittent streams. We further hypothesize that the emerging discharge thresholds signal a tipping point at which intermittent streams switch from a dry state governed by intermittency‐driven biogeochemistry to a wet, discharge‐driven state resembling perennial rivers. The hypothesis calls for intensive data collection at dry‐wet transitions to develop theories and models for intermittent streams that have become increasingly prevalent globally.

When evaluating the performance of quantitative models, dimensioned errors often are characterized by sums-of-squares measures such as the mean squared error (MSE) or its square root, the root mean squared error (RMSE). In terms of quantifying average error, however, absolute-value-based measures such as the mean absolute error (MAE) are more interpretable than MSE or RMSE. Part of that historical preference for sums-of-squares measures is that they are mathematically amenable to decomposition and one can then form ratios, such as those based on separating MSE into its systematic and unsystematic components. Here, we develop and illustrate a decomposition of MAE into three useful submeasures: (1) bias error, (2) proportionality error, and (3) unsystematic error. This three-part decomposition of MAE is preferable to comparable decompositions of MSE because it provides more straightforward information on the nature of the model-error distribution. We illustrate the properties of our new three-part decomposition using a long-term reconstruction of streamflow for the Upper Colorado River.

High Mountain Asia hosts the largest glacier concentration outside the polar regions. These glaciers are important contributors to streamflow in one of the most populated areas of the world. Past studies have used methods that can provide only regionally averaged glacier mass balances to assess the glacier contribution to rivers and sea level rise. Here we compute the mass balance for about 92% of the glacierized area of High Mountain Asia using time series of digital elevation models derived from satellite stereo-imagery. We calculate a total mass change of −16.3 ± 3.5 Gt yr −1 (−0.18 ± 0.04 m w.e. yr −1 ) between 2000 and 2016, which is less negative than most previous estimates. Region-wide mass balances vary from −4.0 ± 1.5 Gt yr −1 (−0.62 ± 0.23 m w.e. yr −1 ) in Nyainqentanglha to +1.4 ± 0.8 Gt yr −1  (+0.14 ± 0.08 m w.e. yr −1 ) in Kunlun, with large intra-regional variability of individual glacier mass balances (standard deviation within a region ∼0.20 m w.e. yr −1 ). Specifically, our results shed light on the Nyainqentanglha and Pamir glacier mass changes, for which contradictory estimates exist in the literature. They provide crucial information for the calibration of the models used for projecting glacier response to climatic change, as these models do not capture the pattern, magnitude and intra-regional variability of glacier changes at present. Publisher Correction (18 June 2018) Glacier mass balances in High Mountain Asia are uncertain. Satellite stereo-imagery allows a spatially resolved estimate for about 92% of the glacierized area and yields a region-wide average of about 16 Gt yr −1 for 2000 to 2016.