Physics-guided deep learning for rainfall-runoff modeling by considering extreme events and monotonic relationships

Xie, Kang; Liu, Pan; Zhang, Jianyun; Han, Dongyang; Wang, Guoqing

doi:10.1016/j.jhydrol.2021.127043

Cited by 87 publications

(56 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Many studies in hydrology have attempted to improve predictions of extremes such as floods (Mosavi et al, 2018) using multiple techniques that include using different criteria for model selection (Coulibaly et al, 2001), conditional density estimation networks (Cannon, 2012), training exclusively on extreme events such as historical high‐flow data (Fleming et al, 2015), adjustment of ML prediction bias to improve performance on the tails of the distribution (Belitz & Stackelberg, 2021), and using KGML models, for example by training an ML model on simulation data containing extremes that might not exist in the observation data (Read et al, 2017; Xie et al, 2021). However, in some cases KGML models may perform worse than traditional DL models; for example, Frame et al (2021) found that an LSTM constrained to conserve mass was not able to predict peak flows as well as the base LSTM, although both models had lower errors than the process‐based model used for comparison.…”

Section: Opportunities For Advancement Of Water Quality MLmentioning

confidence: 99%

Can machine learning accelerate process understanding and decision‐relevant predictions of river water quality?

Varadharajan

Appling

Arora

et al. 2022

Hydrological Processes

View full text Add to dashboard Cite

The global decline of water quality in rivers and streams has resulted in a pressing need to design new watershed management strategies. Water quality can be affected by multiple stressors including population growth, land use change, global warming, and extreme events, with repercussions on human and ecosystem health. A scientific understanding of factors affecting riverine water quality and predictions at local to regional scales, and at sub‐daily to decadal timescales are needed for optimal management of watersheds and river basins. Here, we discuss how machine learning (ML) can enable development of more accurate, computationally tractable, and scalable models for analysis and predictions of river water quality. We review relevant state‐of‐the art applications of ML for water quality models and discuss opportunities to improve the use of ML with emerging computational and mathematical methods for model selection, hyperparameter optimization, incorporating process knowledge into ML models, improving explainablity, uncertainty quantification, and model‐data integration. We then present considerations for using ML to address water quality problems given their scale and complexity, available data and computational resources, and stakeholder needs. When combined with decades of process understanding, interdisciplinary advances in knowledge‐guided ML, information theory, data integration, and analytics can help address fundamental science questions and enable decision‐relevant predictions of riverine water quality.

show abstract

Section: Opportunities For Advancement Of Water Quality MLmentioning

confidence: 99%

Can machine learning accelerate process understanding and decision‐relevant predictions of river water quality?

Varadharajan

Appling

Arora

et al. 2022

Hydrological Processes

View full text Add to dashboard Cite

show abstract

“…Driven by the increasingly powerful performance of computers and big data, statistical and non-inferential deep learning methods enable machines to have the same ability to analyze and learn as human beings (Kadow et al, 2020;Karpatne et al, 2018;Sit et al, 2020). Recent case studies have revealed that deep learning networks have succeeded in geoscience fields (Karpatne et al, 2018;Xie et al, 2021). It has been widely used for spatial missing data (Kadow et al, 2020), spatial downscaling (Jiang et al, 2021;Nearing et al, 2021), rainfall simulation improvement (Liu et al, 2020), and spatial phenomena prediction (Pan et al, 2019).…”

Section: Deep Residual Networkmentioning

confidence: 99%

Global soil moisture storage capacity at 0.5° resolution for geoscientific modelling

Xie

Liu

Xia

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Abstract. Soil moisture storage capacity (SMSC) links the atmosphere and terrestrial ecosystems, which is required as spatial parameters for geoscientific models. However, there are currently no available common datasets of the SMSC on a global scale, especially for hydrological models since conventional evapotranspiration-derived estimates cannot represent the extra storage capacity for the lateral flow and runoff generation. Here, we produce a dataset of the SMSC parameter for global hydrological models. Joint parameter calibration of three commonly used monthly water balance models provides the labels for a deep residual network. The global SMSC is constructed based on the deep residual network at 0.5° resolution by integrating 15 types of meteorological forcings, underlying surface properties, and runoff data. SMSC products are validated with the spatial distribution against root zone depth datasets and validated in the simulation efficiency on global grids and typical catchments from different climatic regions. We provide the global SMSC parameter dataset as a benchmark for geoscientific modelling by users.

show abstract

“…En la actualidad, existen varios modelos hidrológicos de lluvia-escorrentía que se diferencian por los detalles en los procesos, es decir, la conceptualización de los procesos, los datos requeridos de entrada y la resolución espacial-temporal (Nonki et al, 2021). Generalmente se debe elegir entre diversos modelos: continuos, agrupados, distribuidos, deterministas o estocásticos, dependiendo de los datos de entradas que requieren y se tengan para ejecutarlos Sin embargo, los modelos lluvia-escorrentía tienen un mayor uso, debido principalmente a los pocos datos de entrada que requieren, lo cual los hace útiles en zonas con escasez de información (Tegegne et al, 2017) Los modelos lluvia-escorrentía son un campo clave de las investigaciones hidrológicas, ya q que proporcionan información fundamental para la gestión del recurso hídrico (Tajiki et al, 2020;Xie et al, 2020;Xie, Liu, Zhang, Han, et al, 2021), principalmente en cuanto a desastres por sequías (Kang & Sridhar, 2017;Pan et al, 2020) e inundaciones (Gao et al, 2019;Hostache et al, 2018;X. Zhang et al, 2018).…”

Section: Los Modelos Hidrológicos Empleadosunclassified

Oferta hídrica superficial del humedal el Gallinazo, ubicado en Aguachica-Colombia

Escorcia

Herrera

Galviz

2022

rev.investig.agrar.ambient.

View full text Add to dashboard Cite

Contextualización: Hoy en día los humedales se encuentran en deterioro a causa principalmente de las actividades antrópicas que provocan procesos de sustitución y degradación por el cambio en los usos de los suelos causando conflicto de uso así como consecuencias en el suministro de los recursos. En especial, el recurso hídrico el cual presenta diversas alteraciones lo que limita los servicios ecosistémicos que los humedales prestan. Vacío de conocimiento: Información inexistente sobre la disponibilidad del recurso hídrico en la zona del humedal que sirve de base para la gobernanza del agua eficaz, eficiente e incluyente en la zona, permitiendo contribuir en la formulación de políticas públicas en la gestión integral del recurso hídrico maximizando así el bienestar social y económico en el municipio, y se hace necesario la determinación de esta con el fin de conservar los ecosistemas periurbanos de importancia regional. Propósito: Este trabajo se estimó la oferta hídrica superficial neta disponible mediante el método lluvia-escorrentía, en el humedal El Gallinazo, Aguachica (Cesar – Colombia) como aporte a la gestión integral del recurso. Metodología: Se basó en el establecimiento de la cobertura vegetal por medio del sistema Corine Land Cover adaptado para Colombia, el grupo hidrológico mediante la textura del suelo y con esto se determinó a través de tablas el número de curvas. La investigación incluyó el cálculo de la precipitación media anual empleando el método de interpolación de Kriging con ayuda del software ArcGIS y datos meteorológicos extraídos de estaciones del IDEAM, para luego determinar la precipitación efectiva de la zona. Además, se utilizó el método lluvia-escorrentía en función al número de curvas (SCS-CN) y la resolución 865 de 2004 para establecer la oferta hídrica neta disponible del humedal El Gallinazo. Resultados y conclusiones: La zona posee dos tipos de coberturas vegetales, tres grupos hidrológicos de suelo y un CN promedio de 59,29 demostrando un potencial de escorrentía moderadamente alto y una capacidad de infiltración moderada; la oferta hídrica anual disponible fue de 42573,6 m3. El comportamiento de la disponibilidad del agua depende de las actividades antrópica dentro y fuera del humedal así como del régimen de precipitación.

show abstract

Physics-guided deep learning for rainfall-runoff modeling by considering extreme events and monotonic relationships

Cited by 87 publications

References 49 publications

Can machine learning accelerate process understanding and decision‐relevant predictions of river water quality?

Can machine learning accelerate process understanding and decision‐relevant predictions of river water quality?

Global soil moisture storage capacity at 0.5° resolution for geoscientific modelling

Oferta hídrica superficial del humedal el Gallinazo, ubicado en Aguachica-Colombia

Contact Info

Product

Resources

About