Benchmarking data-driven rainfall–runoff models in Great Britain: a comparison of long short-term memory (LSTM)-based models with four lumped conceptual models

Hydrology and earth system sciences, Journal Year: 2021, Volume and Issue: 25(10), P. 5517 - 5534

Published: Oct. 21, 2021

Abstract. Long short-term memory (LSTM) models are recurrent neural networks from the field of deep learning (DL) which have shown promise for time series modelling, especially in conditions when data abundant. Previous studies demonstrated applicability LSTM-based rainfall–runoff modelling; however, LSTMs not been tested on catchments Great Britain (GB). Moreover, opportunities exist to use spatial and seasonal patterns model performances improve our understanding hydrological processes examine advantages disadvantages simulation. By training two LSTM architectures across a large sample 669 GB, we demonstrate that Entity Aware (EA LSTM) simulate discharge with median Nash–Sutcliffe efficiency (NSE) scores 0.88 0.86 respectively. We find outperform suite benchmark conceptual models, suggesting an opportunity additional refine models. In summary, show largest performance improvements north-east Scotland south-east England. The England remained difficult model, part due inability configured this study learn groundwater processes, human abstractions complex percolation properties hydro-meteorological variables typically employed modelling.

Language: Английский

---

Toward catchment hydro‐biogeochemical theories

Wiley Interdisciplinary Reviews Water, Journal Year: 2020, Volume and Issue: 8(1)

Published: Dec. 3, 2020

Abstract Headwater catchments are the fundamental units that connect land to ocean. Hydrological flow and biogeochemical processes intricately coupled, yet their respective sciences have progressed without much integration. Reaction kinetic theories prescribe rate dependence on environmental variables (e.g., temperature water content) advanced substantially, mostly in well‐mixed reactors, columns, warming experiments considering characteristics of hydrological at catchment scale. These shown significant divergence from observations natural systems. On other hand, theories, including transit time theory, substantially not been incorporated into understanding reactions Here we advocate for development integrated hydro‐biogeochemical across gradients climate, vegetation, geology conditions. The lack such presents barriers mechanisms forecasting future Critical Zone under human‐ climate‐induced perturbations. Although integration has started co‐located measurements well way, tremendous challenges remain. In particular, even this era “big data,” still limited by data will need (1) intensify beyond river channels characterize vertical connectivity broadly shallow deep subsurface; (2) expand older dating scales reflected stable isotopes; (3) combine use reactive solutes, nonreactive tracers, (4) augment environments undergoing rapid changes. To develop it is essential engage models all stages model‐informed collection strategies maximize usage; adopt a “simple but simplistic,” or fit‐for‐purpose approach include process‐based models; blend data‐driven framework “theory‐guided science.” Within hypothesis testing, model‐data fusion can advance mechanistically link catchments' internal structures external drivers functioning. It only field hydro‐biogeochemistry, also enable hind‐ fore‐casting serve society large. Broadly, education cultivate thinkers intersections traditional disciplines with hollistic approaches interacting complex earth This article categorized under: Engineering Water > Methods

Language: Английский

---

An overview of the hydrology of non‐perennial rivers and streams

Wiley Interdisciplinary Reviews Water, Journal Year: 2021, Volume and Issue: 8(2)

Published: Feb. 5, 2021

Abstract Non‐perennial rivers and streams are ubiquitous on our planet. Although several metrics have been used to statistically group or compare streamflow characteristics, there is currently no widely definition of how many days over what reach length surface flow must cease in order classify a river as non‐perennial. At the same time, breadth climate geographic settings for non‐perennial leads diversity their regimes, such often quickly they go dry. These rich expanding body literature addressing ecologic geomorphic features, but said be ignored by hydrologists. Yet much we do know about hydrology terms generation processes, water losses, variability flow. We also that while prevalent arid regions, occur across all types experience diverse set natural anthropogenic controls streamflow. Furthermore, measuring modeling these presents distinct challenges, research directions, which still require further attention. Therefore, present an overview current understanding, methodologic knowledge gaps, directions hydrologic understanding rivers; critical topics light both growing global scarcity ever‐changing laws policies dictate whether environmental protection receive. This article categorized under: Science Water >

Language: Английский

---

Zero or not? Causes and consequences of zero‐flow stream gage readings

Wiley Interdisciplinary Reviews Water, Journal Year: 2020, Volume and Issue: 7(3)

Published: April 13, 2020

Abstract Streamflow observations can be used to understand, predict, and contextualize hydrologic, ecological, biogeochemical processes conditions in streams. Stream gages are point measurements along rivers where streamflow is measured, often infer upstream watershed‐scale processes. When stream read zero, this may indicate that the has dried at location; however, zero‐flow readings also caused by a wide range of other factors. Our ability identify whether or not gage reading indicates dry fluvial system far reaching environmental implications. Incorrect identification interpretation data user lead inaccurate and/or predictions from models analyses. Here, we describe several causes readings: frozen surface water, flow reversals, instrument error, natural human‐driven source losses bypass flow. For these examples, discuss implications interpretations. We highlight additional methods for determining presence, including direct observations, statistical methods, hydrologic models, which applied interpret reach‐ dynamics. Such efforts necessary improve our understand predict activation, cessation, connectivity across river networks. Developing integrated understanding possible meanings zero‐flows will only attain greater importance more variable changing climate. This article categorized under: Science Water > Methods Hydrological Processes Life Conservation, Management, Awareness

Language: Английский

---

Temporal dynamics of tree xylem water isotopes: in situ monitoring and modeling

Biogeosciences, Journal Year: 2021, Volume and Issue: 18(15), P. 4603 - 4627

Published: Aug. 12, 2021

Abstract. We developed a setup for fully automated, high-frequency in situ monitoring system of the stable water isotope deuterium and 18O soil tree xylem. The was tested 12 weeks within an isotopic labeling experiment during large artificial sprinkling including three mature European beech (Fagus sylvatica) trees. Our allowed one measurement every 12–20 min, enabling us to obtain about seven measurements per day each our 15 probes While induced abrupt step pulse signature, it took 7 10 d until signatures at trees' stem bases reached their peak label concentrations 14 8 m height leveled off around same values. During experiment, we observed effects several rain events dry periods on xylem signatures, which fluctuated between measured upper lower horizons. In order explain observations, combined already existing root uptake (RWU) model with newly approach simulate propagation from tips base further up along stem. key proper simulation short-term dynamics isotopes accounting sap flow velocities path length distribution modeling framework identify parameter values that relate depth, horizontal wilting point. insights gained this study can help improve representation trees ecohydrological models prediction transit time age transpiration fluxes.

Language: Английский

---

Intraseasonal Drainage Network Dynamics in a Headwater Catchment of the Italian Alps

Water Resources Research, Journal Year: 2020, Volume and Issue: 56(4)

Published: March 23, 2020

Abstract In the majority of existing studies, streams are conceived as static objects that occupy predefined regions landscape. However, empirical observations suggest stream networks systematically and ubiquitously featured by significant expansion/retraction dynamics produced hydrologic climatic variability. This contribution presents novel data about active drainage network a 5 km 2 headwater catchment in Italian Alps. The has been extensively monitored with biweekly temporal resolution during field campaign conducted from July to November 2018. Our results reveal that, spite wet climate typical study area, more than 70% observed river is temporary, presence disconnected reaches periods. Available have used develop set simple statistical models were able properly reconstruct length function antecedent precipitation. rainfall timing intensity represent major controls on length, while evapotranspiration minor effect intraseasonal changes density. also indicate multiple expansion retraction cycles simultaneously operate at different time scales, response distinct hydrological processes. Furthermore, we found spatial patterns unchanneled lengths related underlying heterogeneity geological attributes. offers insights physical mechanisms driving low‐order alpine catchments.

Language: Английский

---

Intermittent rivers and ephemeral streams: Perspectives for critical zone science and research on socio‐ecosystems

Wiley Interdisciplinary Reviews Water, Journal Year: 2021, Volume and Issue: 8(4)

Published: May 4, 2021

Abstract Intermittent rivers and ephemeral streams (IRES) are now recognized to support specific freshwater biodiversity ecosystem services represent approximately half of the global river network, a fraction that is likely increase in context changes. Despite large research efforts on IRES during past few decades, there need for developing systemic approach considers their hydrological, hydrogeological, hydraulic, ecological, biogeochemical properties processes, as well interactions with human societies. Thus, we assert interdisciplinary promoted by critical zone sciences socio‐ecology relevant. These approaches rely infrastructure—Critical Zone Observatories (CZO) Long‐Term Socio‐Ecological Research (LTSER) platforms—that representative diversity (e.g., among climates or types geology. We illustrate this within French CZO LTSER, including socio‐ecosystems, detail IRES. networks also specialized long‐term observations required detect measure responses climate forcings despite delay buffering effects ecosystems. The LTSER platforms development innovative techniques data analysis methods can improve characterization IRES, particular monitoring flow regimes, groundwater‐surface water flow, biogeochemistry rewetting. provide scientific methodological perspectives which its associated infrastructure would relevant original insights help fill knowledge gaps about This article categorized under: Water Life > Stresses Pressures Ecosystems Science Hydrological Processes Conservation, Management, Awareness

Language: Английский

---

Aqua temporaria incognita

Hydrological Processes, Journal Year: 2020, Volume and Issue: 34(26), P. 5704 - 5711

Published: Nov. 18, 2020

It has been 12 years since Bishop et al. (2008) wrote the Invited Commentary "Aqua Incognita: unknown headwaters".They highlighted that "In most regions, overwhelming majority of stream length lies beyond frontiers any systematic documentation and would have to be represented as a blank space on assessment map.This means for streams support aquatic life, understanding is lacking water quality, habitat, biota, specific discharge, or even how many kilometers such are there.This so vast it deserves name help us at least remember there.We propose calling 'Aqua Incognita'" (Bishop al., 2008; p. 1239).We continue agree with this statement need understand headwater better.In commentary, we want draw attention particular type less frequently examined: flow intermittently, i.e., Aqua Temporaria Incognita.Question 3 23 unsolved problems in hydrology (Blöschl 2019) focuses ephemeral dryland streams.We argue focus needs broadening temporary because they ubiquitous all climates.Headwater feed larger perennial particularly sensitive climate change other human influences (Jaeger 2014;Reynolds 2015;Pumo 2016).Their effective management protection, therefore, requires an both natural artificial causes intermittence.Temporary among hydrologically variable systems (Wohl, 2017).They include intermittent seasonally, only response rainfall snowmelt events, episodic contain flowing during

Language: Английский

---

A 2D-SWEs framework for efficient catchment-scale simulations: Hydrodynamic scaling properties of river networks and implications for non-uniform grids generation

Journal of Hydrology, Journal Year: 2021, Volume and Issue: 599, P. 126306 - 126306

Published: April 21, 2021

Language: Английский

---

Simulating Fully‐Integrated Hydrological Dynamics in Complex Alpine Headwaters: Potential and Challenges

Water Resources Research, Journal Year: 2022, Volume and Issue: 58(4)

Published: March 11, 2022

Abstract Highly simplified approaches continue to underpin hydrological climate change impact assessments across the Earth's mountainous regions. Fully‐integrated surface‐subsurface models may hold far greater potential represent distinctive regimes of steep, geologically‐complex headwater catchments. However, their utility has not yet been tested a wide range settings. Here, an integrated model two adjacent calcareous Alpine headwaters that accounts for two‐dimensional surface flow, three‐dimensional (3D) variably‐saturated groundwater and evapotranspiration is presented. An energy balance‐based representation snow dynamics contributed model's high‐resolution forcing data, sophisticated 3D geological helped define parameterize its subsurface structure. In first known attempt calibrate catchment‐scale region automatically, numerous uncertain parameters were estimated. The salient features regime could ultimately be satisfactorily reproduced – over 11‐month evaluation period, Nash‐Sutcliffe efficiency simulated streamflow at main gauging station was 0.76. Spatio‐temporal visualization data responses further confirmed broad coherence. Presumably due unresolved local heterogeneity, closely replicating somewhat contrasting level signals observed near one another proved more elusive. Finally, we assessed impacts various simplifications assumptions are commonly employed in physically‐based modeling including use spatially uniform forcings, vertically limited domain, global products on key outputs, finding strongly affected performance many cases. Although certain outstanding challenges must overcome if uptake mountain regions around world increase, our work demonstrates feasibility benefits application such complex systems.

Language: Английский

---