A Transdisciplinary Review of Deep Learning Research and Its Relevance for Water Resources Scientists

Water Resources Research, Journal Year: 2018, Volume and Issue: 54(11), P. 8558 - 8593

Published: Aug. 30, 2018

Deep learning (DL), a new-generation of artificial neural network research, has transformed industries, daily lives and various scientific disciplines in recent years. DL represents significant progress the ability networks to automatically engineer problem-relevant features capture highly complex data distributions. I argue that can help address several major new old challenges facing research water sciences such as inter-disciplinarity, discoverability, hydrologic scaling, equifinality, needs for parameter regionalization. This review paper is intended provide resources scientists hydrologists particular with simple technical overview, trans-disciplinary update, source inspiration about relevance water. The reveals physical geoscientific have utilized challenges, improve efficiency, gain insights. especially suited information extraction from image-like sequential data. Techniques experiences presented other are high research. Meanwhile, less noticed may also serve exploratory tool. A area termed 'AI neuroscience,' where interpret decision process deep derive insights, been born. budding sub-discipline demonstrated methods including correlation-based analysis, inversion network-extracted features, reduced-order approximations by interpretable models, attribution decisions inputs. Moreover, use condition neurons mimic problem-specific fundamental organizing units, thus revealing emergent behaviors these units. Vast opportunities exist propel advances sciences.

Language: Английский

---

Global extent of rivers and streams

Science, Journal Year: 2018, Volume and Issue: 361(6402), P. 585 - 588

Published: June 28, 2018

Expanding the role of rivers The surfaces and streams are interfaces for a host chemical exchanges with atmosphere biosphere. For instance, carbon dioxide outgassing from is estimated to be equivalent one-fifth combined emissions fossil fuel combustion cement production. Allen Pavelsky used satellite imagery estimate surface area (see Perspective by Palmer Ruhi). stunning map that they generated results in an upward revision, about one-third, total on Earth. Science , this issue p. 585 ; see also 546

Language: Английский

---

An overview of current applications, challenges, and future trends in distributed process-based models in hydrology

Journal of Hydrology, Journal Year: 2016, Volume and Issue: 537, P. 45 - 60

Published: March 22, 2016

Language: Английский

---

Hillslope Hydrology in Global Change Research and Earth System Modeling

Water Resources Research, Journal Year: 2019, Volume and Issue: 55(2), P. 1737 - 1772

Published: Feb. 1, 2019

Abstract Earth System Models (ESMs) are essential tools for understanding and predicting global change, but they cannot explicitly resolve hillslope‐scale terrain structures that fundamentally organize water, energy, biogeochemical stores fluxes at subgrid scales. Here we bring together hydrologists, Critical Zone scientists, ESM developers, to explore how hillslope may modulate grid‐level fluxes. In contrast the one‐dimensional (1‐D), 2‐ 3‐m deep, free‐draining soil hydrology in most land models, hypothesize 3‐D, lateral ridge‐to‐valley flow through shallow deep paths insolation contrasts between sunny shady slopes top two globally quantifiable organizers of water energy (and vegetation) within an grid cell. We these processes likely impact predictions where when) and/or limiting. further that, if implemented will increase simulated continental storage residence time, buffering terrestrial ecosystems against seasonal interannual droughts. efficient ways capture mechanisms ESMs identify critical knowledge gaps preventing us from scaling up processes. One such gap is our extremely limited subsurface, stored (supporting released stream baseflow aquatic ecosystems). conclude with a set organizing hypotheses call syntheses activities model experiments assess on change predictions.

Language: Английский

---

Perspectives on the Future of Land Surface Models and the Challenges of Representing Complex Terrestrial Systems

Journal of Advances in Modeling Earth Systems, Journal Year: 2020, Volume and Issue: 12(4)

Published: March 11, 2020

Abstract Land surface models (LSMs) are a vital tool for understanding, projecting, and predicting the dynamics of land its role within Earth system, under global change. Driven by need to address set key questions, LSMs have grown in complexity from simplified representations biophysics encompass broad interrelated processes spanning disciplines biophysics, biogeochemistry, hydrology, ecosystem ecology, community human management, societal impacts. This vast scope complexity, while warranted problems designed solve, has led enormous challenges understanding attributing differences between LSM predictions. Meanwhile, wide range spatial scales that govern heterogeneity, spectrum timescales dynamics, create tractably representing LSMs. We identify three “grand challenges” development use LSMs, based around these issues: managing process parametric across asked changing world. In this review, we discuss progress been made, as well promising directions forward, each challenges.

Language: Английский

---

Connections between groundwater flow and transpiration partitioning

Science, Journal Year: 2016, Volume and Issue: 353(6297), P. 377 - 380

Published: July 22, 2016

Groundwater flow drives partitioning Soil evaporation and plant transpiration together contribute a substantial proportion of terrestrial freshwater fluxes. Land surface models are used to understand the these fluxes on continental scale; however, model outputs often inconsistent with stable isotope observations. Maxwell Condon incorporated dynamic groundwater into an integrated hydrologic simulation for entire United States. The showed that water table depth lateral strongly affect partitioning, thus explaining inconsistencies between observations models. Science , this issue p. 377

Language: Английский

---

The evolution of process-based hydrologic models: historical challenges and the collective quest for physical realism

Hydrology and earth system sciences, Journal Year: 2017, Volume and Issue: 21(7), P. 3427 - 3440

Published: July 11, 2017

Abstract. The diversity in hydrologic models has historically led to great controversy on the correct approach process-based modeling, with debates centered adequacy of process parameterizations, data limitations and uncertainty, computational constraints model analysis. In this paper, we revisit key modeling challenges requirements (1) define suitable equations, (2) adequate parameters, (3) cope computing power. We outline historical challenges, provide examples advances that address these outstanding research needs. illustrate how have been made by groups using different type complexity, argue for need more effectively use our approaches order advance collective quest physically realistic models.

Language: Английский

---

Evapotranspiration depletes groundwater under warming over the contiguous United States

Nature Communications, Journal Year: 2020, Volume and Issue: 11(1)

Published: Feb. 13, 2020

Abstract A warmer climate increases evaporative demand. However, response to warming depends on water availability. Existing earth system models represent soil moisture but simplify groundwater connections, a primary control moisture. Here we apply an integrated surface-groundwater hydrologic model evaluate the sensitivity of shallow across majority US. We show that as shifts balance between supply and demand, storage can buffer plant stress; only where connections are present, not indefinitely. As persists, be depleted lost. Similarly, in arid western US does result significant changes because this area is already largely limited. The direct demonstrates strong early effect low moderate may have evapotranspiration.

Language: Английский

---

Our Skill in Modeling Mountain Rain and Snow is Bypassing the Skill of Our Observational Networks

Bulletin of the American Meteorological Society, Journal Year: 2019, Volume and Issue: 100(12), P. 2473 - 2490

Published: July 30, 2019

Abstract In mountain terrain, well-configured high-resolution atmospheric models are able to simulate total annual rain and snowfall better than spatial estimates derived from in situ observational networks of precipitation gauges, significantly radar or satellite-derived estimates. This conclusion is primarily based on comparisons with streamflow snow basins across the western United States Iceland, Europe, Asia. Even though they outperform gridded datasets gauge networks, still disagree each other average often more their representation individual storms. Research address these difficulties must make use a wide range observations (snow, streamflow, ecology, radar, satellite) bring together scientists different disciplines communities.

Language: Английский

---

Prolongation of SMAP to Spatiotemporally Seamless Coverage of Continental U.S. Using a Deep Learning Neural Network

Geophysical Research Letters, Journal Year: 2017, Volume and Issue: 44(21)

Published: Oct. 16, 2017

The Soil Moisture Active Passive (SMAP) mission has delivered valuable sensing of surface soil moisture since 2015. However, it a short time span and irregular revisit schedule. Utilizing state-of-the-art time-series deep learning neural network, Long Short-Term Memory (LSTM), we created system that predicts SMAP level-3 data with atmospheric forcing, model-simulated moisture, static physiographic attributes as inputs. removes most the bias model simulations improves predicted climatology, achieving small test root-mean-squared error (<0.035) high correlation coefficient >0.87 for over 75\% Continental United States, including forested Southeast. As first application LSTM in hydrology, show proposed network avoids overfitting is robust both temporal spatial extrapolation tests. generalizes well across regions distinct climates physiography. With fidelity to SMAP, shows great potential hindcasting, assimilation, weather forecasting.

Language: Английский

---