Hydrological Datasets Research Articles

Multi-hazard susceptibility mapping based on Convolutional Neural Networks

Multi-hazard susceptibility prediction is an important component of disasters risk management plan. An effective multi-hazard risk mitigation strategy includes assessing individual hazards as well as their interactions. However, with the rapid development of artificial intelligence technology, multi-hazard susceptibility prediction techniques based on machine learning has encountered a huge bottleneck. In order to effectively solve this problem, this study proposes a multi-hazard susceptibility mapping framework using the classical deep learning algorithm of Convolutional Neural Networks (CNN). First, we use historical flash flood, debris flow and landslide locations based on Google Earth images, extensive field surveys, topography, hydrology, and environmental data sets to train and validate the proposed CNN method. Next, the proposed CNN method is assessed in comparison to conventional logistic regression and k-nearest neighbor methods using several objective criteria, i.e., coefficient of determination, overall accuracy, mean absolute error and the root mean square error. Experimental results show that the CNN method outperforms the conventional machine learning algorithms in predicting probability of flash floods, debris flows and landslides. Finally, the susceptibility maps of the three hazards based on CNN are combined to create a multi-hazard susceptibility map. It can be observed from the map that 62.43% of the study area are prone to hazards, while 37.57% of the study area are harmless. In hazard-prone areas, 16.14%, 4.94% and 30.66% of the study area are susceptible to flash floods, debris flows and landslides, respectively. In terms of concurrent hazards, 0.28%, 7.11% and 3.13% of the study area are susceptible to the joint occurrence of flash floods and debris flow, debris flow and landslides, and flash floods and landslides, respectively, whereas, 0.18% of the study area is subject to all the three hazards. The results of this study can benefit engineers, disaster managers and local government officials involved in sustainable land management and disaster risk mitigation.

Quantifying relations between altered hydrology and fish community responses for streams in Minnesota

BackgroundAltered hydrology is a stressor on aquatic life, but quantitative relations between specific aspects of streamflow alteration and biological responses have not been developed on a statewide scale in Minnesota. Best subsets regression analysis was used to develop linear regression models that quantify relations among five categories of hydrologic metrics (i.e., duration, frequency, magnitude, rate-of-change, and timing) computed from streamgage records and six categories of biological metrics (i.e., composition, habitat, life history, reproductive, tolerance, trophic) computed from fish-community samples, as well as fish-based indices of biotic integrity (FIBI) scores and FIBI scores normalized to an impairment threshold of the corresponding stream class (FIBI_BCG4). Relations between hydrology and fish community responses were examined using three hydrologic datasets that represented periods of record, long-term changes, and short-term changes to flow regimes in streams of Minnesota.ResultsRegression models demonstrated significant relations between hydrologic explanatory metrics and fish-based biological response metrics, and the five regression models with the strongest linear relations explained over 70% of the variability in the biological metric using three hydrologic metrics as explanatory variables. Tolerance-based biological metrics demonstrated the strongest linear relations to hydrologic metrics. The most commonly used hydrologic metrics were related to bankfull flows and aspects of flow variability.ConclusionsFinal regression models represent paired streamgage records and biological samples throughout the State of Minnesota and encompass differences in stream orders, hydrologic landscape units, and watershed sizes. Presented methods can support evaluations of stream fish communities and facilitate targeted efforts to improve the health of fish communities. Methods also can be applied to locations outside of Minnesota with continuous streamgage data and fish-community samples.

Drought recorded by Ba∕Ca in coastal benthic foraminifera

Abstract. Increasing occurrences of extreme weather events, such as the 2018 drought over northern Europe, are a concerning issue under global climate change. High-resolution archives of natural hydroclimate proxies, such as rapidly accumulating sediments containing biogenic carbonates, offer the potential to investigate the frequency and mechanisms of such events in the past. Droughts alter the barium (Ba) concentration of near-continent seawater through the reduction in Ba input from terrestrial runoff, which in turn may be recorded as changes in the chemical composition (Ba/Ca) of foraminiferal calcium carbonates accumulating in sediments. However, so far the use of Ba/Ca as a discharge indicator has been restricted to planktonic foraminifera, despite the high relative abundance of benthic species in coastal, shallow-water sites. Moreover, benthic foraminiferal Ba/Ca has mainly been used in open-ocean records as a proxy for paleo-productivity. Here we report on a new geochemical data set measured from living (CTG-labeled) benthic foraminiferal species to investigate the capability of benthic Ba/Ca to record changes in river runoff over a gradient of contrasting hydroclimatic conditions. Individual foraminifera (Bulimina marginata, Nonionellina labradorica) were analyzed by laser-ablation ICP-MS over a seasonal and spatial gradient within Gullmar Fjord, Swedish west coast, during 2018–2019. The results are compared to an extensive meteorological and hydrological data set, as well as sediment and pore-water geochemistry. Benthic foraminiferal Ba/Ca correlates significantly to riverine runoff; however, the signals contain both spatial trends with distance to Ba source and species-specific influences such as micro-habitat preferences. We deduce that shallow-infaunal foraminifera are especially suitable as proxy for terrestrial Ba input and discuss the potential influence of water-column and pore-water Ba cycling. While distance to Ba source, water depth, pore-water geochemistry, and species-specific effects need to be considered in interpreting the data, our results demonstrate confidence in the use of Ba/Ca of benthic foraminifera from near-continent records as a proxy for past riverine discharge and to identify periods of drought.

Groundwater potential characterisation over the Voltaian basin using geophysical, geological, hydrological and topographical datasets

The delineation of prospective zones of groundwater resources has been carried out over the Voltaian basin as a viable alternative source of potable water resource to complement increasing demand for domestic and industrial consumption, owing to the danger posed to surficial water resources within the basin by various anthropogenic activities. In achieving this objective, the groundwater potential of the area was assessed by generating a groundwater potential map (GPM) derived by integrating nine (9) evidential layers obtained from geological, geophysical, hydrological, and topographical datasets in geospatial setting using geographic information systems (GIS). The integration of these nine evidential layers was preceded by the assignment of a weight to each evidential layer based on the fuzzy analytical hierarchy process (FAHP) multi-criteria decision making (MCDM) technique. The assignment of weights to these 9 layers using the FAHP approach was done by three decision-makers (DMs) and dwelled on their relative relevance to groundwater resource delineation within the Voltaian basin. The three DMs employed in weighting these layers have expertise in applying geophysical, geological, hydrological and topographical datasets in groundwater prospecting. The weights obtained for the various evidential layers were subjected to normalisation and subsequently consolidated to produce the GPM using a fuzzy gamma operator value of 0.9 in GIS. The efficiency of the GPM was tested using the Nash-Sutclliffe efficiency (NSE) test and the Index of Agreement test. In all, 0.095%, 10.57%, 44.879%, 40.973%, and 3.974% of the study area were delineated as very high, good, moderate, poor, and very poor groundwater zones, respectively by the GPM. The NSE test yielded a value of 0.9996, and the Index of Agreement produced a value of 0.9999. In this regard, results obtained show that FAHP-based model produced is efficient and subsequently good for groundwater prediction over the Voltaian basin.

An automated machine learning methodology for the improved prediction of reference evapotranspiration using minimal input parameters

AbstractAn accurate and minimal meteorological parameter‐based reference evapotranspiration (ETo) computational model is highly essential to design a cost‐effective and automated solution for precise water allocation in various agricultural and hydrological applications. Automated machine learning (AutoML), a substantially progressive machine learning (ML) paradigm, expedites the data science applications in these domains by automating and alleviating the tedious tasks of developing ML pipelines, particularly in non‐linear hydrological process modelling. This paper proposes an AutoML solution for daily EToprediction in a limited input parameter scenario, for the first time in EToestimation research. Two AutoML frameworks, AutoGluon‐Tabular (AGT) and H2O AutoML were implemented using daily meteorological data of a humid tropical climatic region of Kerala, India and evaluated its performance with the radiation‐based empirical methods and traditional ML methods. The developed AutoML models imitated the input parameter combinations of radiation‐based empirical models such as Priestley‐Taylor, Makkink, Turc and Ritchie, with the Penman‐Monteith EToserving as the target for modelling process. An improvement of mean absolute error from 0.119 to 0.078 mm/day, mean squared error from 0.035 to 0.017 mm/day, root mean squared error from 0.186 to 0.129 mm/day, root mean squared logarithmic error from 0.078 to 0.026 mm/day, and coefficient of determination (R2) from 0.971 to 0.985 observed in the AutoML model, AGT relative to the empirical model, Turc describes its enhanced performance capability. For demonstrating the superiority of AutoML to traditional ML techniques, another set of AutoML models were also developed based on California Irrigation Management Information System datasets and evaluated them with existing study results in literature. The overall result analysis demonstrated the superiority of AGT model in EToprediction in all the weather stations considered in this study. The results of the study confirm that AutoML techniques can be applied to any hydrological dataset and provide an effective automated solution for hydrological process modelling.

Modeling stage\u2013discharge\u2013sediment using support vector machine and artificial neural network coupled with wavelet transform

Many real water issues involve rivers’ sediment load or the load that rivers can bring without degrading the fluvial ecosystem. Therefore, the assessment of sediments carried by a river is also crucial in the planning and designing of various water resource projects. In the current study, five different data-driven techniques, namely artificial neural network (ANN), wavelet-based artificial neural network (WANN), support vector machine (SVM), wavelet-based support vector machine (WSVM), and multiple-linear regression (MLR) techniques, were employed for time-series modeling of daily suspended sediment concentration (SSC). Hydrological datasets containing the daily stage (h), discharge (Q), and SSC for 10 years (2004–2013) from June to October at Adityapur and Ghatshila station of Subernrekha river basin, Jharkhand, India, were considered for analysis. The Gamma test was used to determine the input variables in the first step. Various combinations were made by lagging the maximum three-day time step for predicting current-day SSC. The outcomes of ANN, SVM, WAAN, WSVM, and MLR models were evaluated with the actual values of SSC based on statistical metrics. Pearson correlation coefficient (PCC), root-mean-square error (RMSE), Nash–Sutcliffe efficiency (NSE), and Wilmot index (WI) as well as visual inspection of time variation, scatter plots, and Taylor diagrams. Our results stated that the WSVM model discovered the best trustworthy models among all existing models. PCC, RMSE, NSE, and WI values were 0.844 and 0.781, 0.096 g/l and 0.057 g/l, 0.711 and 0.591, 0.907 and 0.878, respectively, throughout the training and testing processes at the Adityapur site. Also, at the Ghatshila location, it was the most accurate model. During the training and testing stages, PCC, RMSE, NSE, and WI values were 0.928 and 0.751, 0.117 g/l and 0.095 g/l, 0.861 and 0.541, 0.962 and 0.859, respectively. Our findings showed that the WANN model was the second-best model during the testing phase for both sites. Hence, the WSVM technique can model SSC at this location and other similar (i.e., geomorphology and flow regime type) rivers.

Identification of Artificial Levees in the Contiguous United States

AbstractArtificial levees are anthropogenic structures designed to hydrologically disconnect rivers from floodplains. The extent of artificial levees in the contiguous United States (CONUS) is unknown. To better estimate the distribution of artificial levees, we tested several different geomorphic, land cover, and spatial variables developed from the National Elevation Dataset, the National Land Cover database, and the National Hydrology Dataset HR Plus. We used known levee locations from the National Levee Database as training data. We tested machine learning and general logistic models’ ability to detect artificial levees in a 100‐year hydrogeomorphic floodplain of seven geographically diverse 8‐digit HUC basins. Random forest models outperformed other models in predicting the location of levees using variables representing geomorphic attributes, land cover, and distance from streams ranging in size between stream order one through six. To demonstrate the ability of our approach to detect unknown levees, we conducted a leave‐one‐out cross‐validation in the lower Mississippi Basin using approximately 1,100 artificial levees. This approach detected known levees constituting 94% of the total levee length in the basin. Scaling up to the CONUS, we applied a high performing (overall accuracy of 97%) random forest model using land cover and stream order variables. We detected 182,213 km of potential levees, mostly along streams of order 2–6 in the Mississippi and Missouri River Basins, indicating that the national levee database contains 20.4% of levee length. Potential levees and those documented in the national levee database modify 2% of the total length of streams in the contiguous United States.

Quantifying lake–aquifer water exchange: the case of Lake Urmia, Iran

ABSTRACT This study investigated the lake–aquifer hydraulic interactions in Lake Urmia (LU) as the second largest hypersaline lake in the world. Due to the scarcity of hydrogeological data required for modelling, a method based on Darcy’s Law and lake water budget was used to quantify the lake–aquifer interaction. Long-term ground- and satellite-based hydrological datasets over the time frame 2001–2019 were used. Results indicate that the groundwater flux between LU and the aquifers controls 18.74 ± 1.67% of the lake’s water storage. While 10 out of 14 adjacent aquifers recharge LU at a rate of less than 180 m3/m.month, one phreatic aquifer recharges the LU up to 1400 m3/m.month. Two aquifers are recharged from the seawater of LU at a rate of 110–640 m3/m.month. The results of this study lead to a holistic understanding of the hydraulic interaction of LU–aquifers, which is essential for the sustainable management of both settings.

Generalized ordinal patterns allowing for ties and their applications in hydrology

When using ordinal patterns, which describe the ordinal structure within a data vector, the problem of ties appears permanently. So far, model classes were used which do not allow for ties; randomization has been another attempt to overcome this problem. Often, time periods with constant values even have been counted as times of monotone increase. However, ties can contain valuable information which is disregarded by all of these approaches. To overcome this, a new approach is proposed: it explicitly allows for ties and, hence, considers more patterns than before. Ties are no longer seen as nuisance, but the information they carry is taken into account explicitly. Limit theorems in the new framework are provided, both, for a single time series and for the dependence between two time series. The methods are applied to hydrological data sets. In hydrology, it is common to distinguish five flood classes (plus ‘absence of flood’). Considering data vectors of these classes at a certain gauge in a river basin, one will usually encounter several ties. Co-monotonic behavior between the data sets of two gauges (increasing, constant, decreasing) can be detected by the method as well as spatial patterns. Thus, it helps to analyze the strength of dependence between different gauges in an intuitive way. This knowledge can be used to assess risk and to plan future construction projects.

Dynamics of self-regulatory processes in a lowland river due to seasonal changes in certain hydro-ecological and water quality factors

This study establishes an evaluable system of relationships among the hydromorphological and water quality factors of a temperate, lowland, regulated watercourse. The aim is to investigate the effects of the environmental factors that cause temporal changes in planktonic algal communities. Measured, real-time and modelled water quality and hydrological data sets were compared during the analysis. It is observed that the same factor can have both positive and negative effects in different periods. Different physiological and morphological characteristics determine environmental tolerance. Chlorophytes, which utilize the red light living in the upper layer, closely correlate with water mirror width (top width). Diatoms need dissolved and suspended silicon and are therefore related to the factors that cause sediment mixing and increased salinity (EC). Cyanobacteria and dinophytes are also able to utilize organic matter (mixotrophy and heterotrophic photosynthesis, respectively). They multiply easily when the organic nutrients that are abundant. In the absence of inorganic nitrogen compounds, cyanobacteria are preferred because of their ability to directly or indirectly take up the N2 molecule. Hydraulic changes have a significant impact on water quality and life processes, including biodiversity.

A multi-isotope and modelling approach for constraining hydro-connectivity in the East African Rift System, southern Ethiopia

During the last African Humid Period (AHP; 15–5 ka), many lakes in the East African Rift System (EARS) experienced pronounced lake-level variations that dramatically transformed the hydrological landscape. Currently dry, saline or marshy-wetland terminal lakes became vast waterbodies, interconnected via overflow sills resulting in the formation of a several thousand-kilometre-long chain of lakes in the EARS. A quantitative, process-based understanding of these hydrological systems can advance our interpretation of past hydroclimate variability from proxy records. Here, we provide a critical modern hydrological dataset for the data-sparse Lake Chew Bahir basin in southern Ethiopia. Driven by modern data, an isotope-enabled hydro-balance model was developed to assess how increases in rainfall modulate δ18O and 87Sr/86Sr variability. Considering a terminal Lake Chew Bahir scenario, humid conditions resulted in higher lake δ18O (∼+14‰) due to increased evaporation and longer water residence times. At the same time 87Sr/86Sr decreased from 0.7064 to 0.7061 due to an increased riverine Sr flux characterised by lower, unradiogenic 87Sr/86Sr ratios. In a modelling scenario where Lake Chew Bahir became a flow-through system with interconnectivity between lakes Abaya, Chamo, Chew Bahir and Turkana, higher lake δ18O (∼+12‰) relative to present was found, but δ18O was lower than in the terminal lake scenario. The lake water 87Sr/86Sr ratios (<0.7061) were also slightly lower. A moderate concomitant change in rainfall input δ18O of −1‰ in step with hydrological reorganisation resulted in the lowest lake δ18O (∼+5‰). Modelled δ18O values were similar to the δ18O range of endogenic carbonates from sedimentary cores from Lake Chew Bahir at the onset of the AHP, supporting the validity of our model, and suggesting that evaporation and the lake water residence time strongly influence lake water δ18O. However, the reported 87Sr/86Sr of fossil carbonates from Lake Chew Bahir during the AHP (0.7065–0.7060) could not be reproduced by our modelled scenarios without adjusting the surface-water-to-groundwater ratio, highlighting the potential role of groundwater as a water source in semi-arid regions. These results demonstrate the insights that can be gained from applying a process-based approach using O and Sr isotope hydro-balance modelling to aid interpretation of past hydro-balance and lake interconnectivity from lacustrine sedimentary records.

Hydrology and biogeochemistry datasets from Sleepers River Research Watershed, Danville, Vermont, USA

AbstractThe Sleepers River Research Watershed (SRRW) in Danville, Vermont, USA, was established in 1957 by the U.S Department of Agriculture, Agricultural Research Service (ARS). The 111‐km2 area was selected as a representative mixed‐land‐use upland landscape in the northeastern USA, with the objective to study fundamental hydrological processes. The U.S. Geological Survey (USGS) currently operates SRRW and is the fourth federal agency to lead it. SRRW has hosted historic research on snowmelt modelling and streamflow generation processes, and USGS has augmented this hydrological foundation with biogeochemical cycling research. This data note describes five freely available data releases in stable repositories, extending to the late 2010s, from the indicated start date: aqueous chemistry and isotopes (1991); precipitation, water temperature, and streamflow (1991); groundwater levels (1991); ground frost depth (1983); and snow depth and water equivalent (1960). USGS research at SRRW is expected to continue, and we plan to update these releases every 1–3 years.

Investigation of subsurface connectivity and subsurface stormflow in low mountain ranges—Findings from the two research catchments Obere Brachtpe and Bohlmicke (Germany)

AbstractThe two small research catchments Obere Brachtpe (2.6 km2; 50.989986, 7.752013) and Bohlmicke (1 km2, 51.079319, 7.892988) are located in the Rhenish Massif, a low mountain range in Germany. Land use in both catchments is dominated by pasture land, spruce stands and mixed forests. Mean annual temperature is 9.1°C, and mean annual total precipitation is 1250 mm, with 15%–20% of the annual precipitation falling as snow. The geology is characterized by sandy silty clay shale from the Lower and Middle Devonian. Loamy Cambisols derived from periglacial slope deposits, complemented by Leptosols and Stagnosols, are the most prominent soils in the catchments. Long‐term hydrological datasets of precipitation, throughfall, discharge, groundwater levels and soil moisture (at different soil depths) in a high temporal and spatial resolution are available for further scientific analysis. Both catchments were monitored within the time period 1999 and 2009, in order to understand how the antecedent soil moisture, stratified soils (periglacial cover beds) and topography (slope form) impacted the subsurface connectivity, and the subsurface stormflow generation ‐ a dominant runoff generation process in humid mountainous catchments. Detailed physically based investigations on runoff processes were carried out, and the obtained results helped to better understand subsurface stormflow generation and subsurface connectivity dynamics. The process knowledge gained, which was presented at several conferences, as well as publications, was the basis for the discussion of open questions within the scientific network ‘Subsurface Stormflow ‐ A well‐recognized, but still challenging process in Catchment Hydrology’ (2016–2021), and the research unit ‘Fast and invisible: conquering subsurface stormflow through an interdisciplinary multisite approach’ (2022–2025), both financed by the German Research Foundation (DFG).

A hydrological simulation dataset of the Upper Colorado River Basin from 1983 to 2019

This article presents a hydrological reconstruction of the Upper Colorado River Basin with an hourly temporal resolution, and 1-km spatial resolution from October 1982 to September 2019. The validated dataset includes a suite of hydrologic variables including streamflow, water table depth, snow water equivalent (SWE) and evapotranspiration (ET) simulated by an integrated hydrological model, ParFlow-CLM. The dataset was validated over the period with a combination of point observations and remotely sensed products. These datasets provide a long-term, natural-flow, simulation for one of the most over-allocated basins in the world.

Statistical modelling for a new family of generalized distributions with real data applications.

The modern trend in distribution theory is to propose hybrid generators and generalized families using existing algebraic generators along with some trigonometric functions to offer unique, more flexible, more efficient, and highly productive G-distributions to deal with new data sets emerging in different fields of applied research. This article aims to originate an odd sine generator of distributions and construct a new G-family called "The Odd Lomax Trigonometric Generalized Family of Distributions". The new densities, useful functions, and significant characteristics are thoroughly determined. Several specific models are also presented, along with graphical analysis and detailed description. A new distribution, "The Lomax cosecant Weibull" (LocscW), is studied in detail. The versatility, robustness, and competency of the LocscW model are confirmed by applications on hydrological and survival data sets. The skewness and kurtosis present in this model are explained using modern graphical methods, while the estimation and statistical inference are explored using many estimation approaches.

The gompertz exponential pareto distribution with the properties and applications to bladder cancer and hydrological datasets

Many existing distributions in literatures does not have the modeling fits capacity to adequately describe the real-life phenomena. The Exponential Pareto (EP) distribution has further gained some generalizations among several authors using different generator techniques with an aim to obtain a new distribution with greater flexibility. This article proposes Gompertz Exponential Pareto (GEP) distribution using the Gompertz generator. Findings from the study revealed some lifetime distributions as special cases and mathematical properties of the distribution investigated including the mean, variance, coefficient of variation, quantile, moment, moment generating function and, order statistics. The distribution can be positively or negatively skewed. It is unimodal with failure rates whose shapes could be reversed J bathtub, constant, decreasing and, increasing and the parameters were estimated using maximum likelihood estimation approach. Applications to two real-life datasets revealed the ability of GEP distribution to provide more flexibilities and better fit to the dataset compared to some previously proposed distributions for the data. The results also revealed that GEP had the superior performance over other generalizations of EP distribution existing in literatures and the performance has further strengthened the usefulness of the Gompertz-generator technique.

Modelling and assessing the impact of illegal water abstractions by upstream farmers on reservoir performance

This study proposes a reservoir yield analysis that incorporates the realities of upstream illegal human activities relating to water abstraction. The study assesses the impact of such unlawful human activities on reservoir storage and yields quantitatively. A reservoir operation water balance model was simulated and coupled with upstream irrigation users’ propensity to unauthorized water abstraction and set to co-evolve for the entire simulation period. The model was developed using four-state drivers (hydrological state, users’ compliance, management competence and reservoir performance). The impact of human behaviour (users’ and management) was assessed using 9 plausible human behaviour scenarios. The model was applied to a system of 5 reservoirs using the 90-year historical hydrologic dataset. The trajectories of the storage, yield-demand and storage-yield ratios were analyzed under different human behaviour scenarios. Both storage and yield were found to substantially decrease as users’ compliance and management competence deteriorated for the same reservoir hydrological state. Depending on the scenario, the annual yield (%) was observed to reduce from 100 to 80 or 50 or even 30 of the annual demand due to changing behaviour. Also, most of the years in which the yield differs significantly from one scenario to the other are years with shallow storage due to drought. A yield difference of about 23% was recorded between the scenarios without and with the highest unauthorized abstractions. The study, therefore, revealed how human behaviour can significantly affect reservoir storage and yield performances. This highlighted the need to be incorporating the impact of unlawful human activities into yield analysis models to quantitatively assess the impact of human behaviour on reservoir performance.

Disentangling the effect of climatic and hydrological predictor variables on benthic macroinvertebrate distributions from predictive models

Lotic freshwater macroinvertebrate species distribution models (SDMs) have been shown to improve when hydrological variables are included. However, most studies to date only include data describing climate or stream flow-related surrogates. We assessed the relative influence of climatic and hydrological predictor variables on the modelled distribution of macroinvertebrates, expecting model performance to improve when hydrological variables are included. We calibrated five SDMs using combinations of bioclimatic (bC), hydrological (H) and hydroclimatic (hC) predictor datasets and compared model performance as well as variance partition of all combinations. We investigated the difference in trait composition of communities that responded better to either bC or H configurations. The dataset bC had the most influence in terms of proportional variance, however model performance was increased with the addition of hC or H. Trait composition demonstrated distinct patterns between associated model configurations, where species that prefer intermediate to slow-flowing current conditions in regions further downstream performed better with bC–H. Including hydrological variables in SDMs contributes to improved performance, it is however, species-specific and future studies would benefit from hydrology-related variables to link environmental conditions and diverse communities. Consequently, SDMs that include climatic and hydrological variables could more accurately guide sustainable river ecosystem management.

CCAM: China Catchment Attributes and Meteorology dataset

Abstract. The absence of a compiled large-scale catchment characteristics dataset is a key obstacle limiting the development of large-sample hydrology research in China. We introduce the first large-scale catchment attribute dataset in China. We compiled diverse data sources, including soil, land cover, climate, topography, and geology, to develop the dataset. The dataset also includes catchment-scale 31-year meteorological time series from 1990 to 2020 for each basin. Potential evapotranspiration time series based on Penman's equation are derived for each basin. The 4911 catchments included in the dataset cover all of China. We introduced several new indicators that describe the catchment geography and the underlying surface differently from previously proposed datasets. The resulting dataset has a total of 125 catchment attributes and includes a separate HydroMLYR (hydrology dataset for machine learning in the Yellow River Basin) dataset containing standardized weekly averaged streamflow for 102 basins in the Yellow River Basin. The standardized streamflow data should be able to support machine learning hydrology research in the Yellow River Basin. The dataset is freely available at https://doi.org/10.5281/zenodo.5729444 (Zhen et al., 2021). In addition, the accompanying code used to generate the dataset is freely available at https://github.com/haozhen315/CCAM-China-Catchment-Attributes-and-Meteorology-dataset (last access: 26 November 2021) and supports the generation of catchment characteristics for any custom basin boundaries. Compiled data for the 4911 basins covering all of China and the open-source code should be able to support the study of any selected basins rather than being limited to only a few basins.

Explore Spatio‐Temporal Learning of Large Sample Hydrology Using Graph Neural Networks

AbstractStreamflow forecasting over gauged and ungauged basins play a vital role in water resources planning, especially under the changing climate. Increased availability of large sample hydrology data sets, together with recent advances in deep learning techniques, has presented new opportunities to explore temporal and spatial patterns in hydrological signatures for improving streamflow forecasting. The purpose of this study is to adapt and benchmark several state‐of‐the‐art graph neural network (GNN) architectures, including ChebNet, Graph Convolutional Network (GCN), and GraphWaveNet, for end‐to‐end graph learning. We explicitly represent river basins as nodes in a graph, learn the spatiotemporal nodal dependencies, and then use the learned relations to predict streamflow simultaneously across all nodes in the graph. The efficacy of the developed GNN models is investigated using the Catchment Attributes and MEteorology for Large‐sample Studies (CAMELS) data set under two settings, fixed graph topology (transductive learning), and variable graph topology (inductive learning), with the latter applicable to prediction in ungauged basins (PUB). Results indicate that GNNs are generally robust and computationally efficient, achieving similar or better performance than a baseline model trained using the long short‐term memory (LSTM) network. Further analyses are conducted to interpret the graph learning process at the edge and node levels and to investigate the effect of different model configurations. We conclude that graph learning constitutes a viable machine learning‐based method for aggregating spatiotemporal information from a multitude of sources for streamflow forecasting