A deep learning approach for hydrological time-series prediction: A case study of Gilgit river basin

Hussain, Dostdar; Hussain, Tahir; Khan, Arif Raza; Naqvi, Syed Ali Asad; Jamıl, Akhtar

doi:10.1007/s12145-020-00477-2

Cited by 101 publications

(45 citation statements)

References 40 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Today, artificial intelligence (AI)-based techniques are being used more and more for problem solving in hydrology. A number of researchers predicted the hydrological parameter by developing artificial neural network (ANN)-based models [47,48]. For many years, ANN-based models have been used for rainfall estimation, runoff estimation, reservoir inflow prediction, suspended sediment prediction, reservoir level estimation, and reservoir operation [39,[49][50][51][52].…”

Section: Artificial Neural Network Model (Ann)mentioning

confidence: 99%

Simulating the Impact of Climate Change with Different Reservoir Operating Strategies on Sedimentation of the Mangla Reservoir, Northern Pakistan

Khan

Stamm

Haider

2020

Water

View full text Add to dashboard Cite

Reservoir sedimentation reduces the gross storage capacity of dams and also negatively impacts turbine functioning, posing a danger to turbine inlets. When the sediment delta approaches the dam, further concerns arise regarding sediments passing through turbine intakes, blades abrasion due to increased silt/sand concentration, choking of outlets, and dam safety. Thus, slowing down the delta advance rate is a worthy goal from a dam manager’s viewpoint. These problems can be solved through a flexible reservoir operation strategy that prioritize sediment deposition further away from the dam face. As a case study, the Mangla Reservoir in Pakistan is selected to elaborate the operational strategy. The methodology rests upon usage of a 1D sediment transport model to quantify the impact of different reservoir operating strategies on sedimentation. Further, in order to assess the long-term effect of a changing climate, a global climate model under representative concentration pathways scenarios 4.5 and 8.5 for the 21st century is used. The reduction of uncertainty in the suspended sediments concentration is achieved by employing an artificial neural networking technique. Moreover, a sensitivity analysis focused on estimating the impact of various parameters on sediment transport modelling was conducted. The results show that a gradual increase in the reservoir minimum operating level slows down the delta movement rate and the bed level close to the dam. However, it may compromise the downstream irrigation demand during periods of high water demand. The findings may help the reservoir managers to improve the reservoir operation rules and ultimately support the objective of a sustainable reservoir use for the societal benefit.

show abstract

Section: Artificial Neural Network Model (Ann)mentioning

confidence: 99%

Simulating the Impact of Climate Change with Different Reservoir Operating Strategies on Sedimentation of the Mangla Reservoir, Northern Pakistan

Khan

Stamm

Haider

2020

Water

View full text Add to dashboard Cite

show abstract

“…The use of deep learning (DL) has gained traction in geophysical disciplines (Bergen et al, 2019;Gagne II et al, 2019;Ham et al, 2019), including hydrology, providing alternative or complementary approaches to supplement traditional processbased modelling (Hussain et al, 2020;Kratzert et al, 2018Kratzert et al, , 2019aKratzert et al, , 2019bMarçais and de Dreuzy, 2017;Shen, 2018;Van et al, 2020). While substantial progress has been made towards distributed process-based hydrological models, input and target data are becoming available at increasingly higher spatiotemporal resolution leading to greater computational requirements and human labour (Marsh et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

“…Alternatively, LSTM models with convolutional rather than fully connected (or 'dense') layers have also been used to encode spatiotemporal information for applications including precipitation nowcasting (Shi et al, 2015). In hydrology, CNN (and particularly combined CNN-LSTM) models have seen fewer applications to date as compared to the LSTM approach, with recent work developing 1D CNNs for rainfallrunoff modelling (Hussain et al, 2020;Van et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

Evaluation and interpretation of convolutional-recurrent networks for regional hydrological modelling

Anderson

Radić

2021

Preprint

View full text Add to dashboard Cite

Abstract. Deep learning has emerged as a useful tool across geoscience disciplines; however, there remain outstanding questions regarding the suitability of unexplored model architectures and how to interpret model learning for regional scale hydrological modelling. Here we use a convolutional-recurrent network, a deep learning approach for learning both spatial and temporal patterns, to predict streamflow at 226 stream gauges across the region of southwestern Canada. The model is forced by gridded climate reanalysis data and trained to predict observed daily streamflow between 1979 and 2015. To interpret the model learning of both spatial and temporal patterns, we introduce two experiments with evaluation metrics to track the model's response to perturbations in the input data. The model performs well in simulating the daily streamflow over the testing period, with a median Nash-Sutcliffe Efficiency (NSE) of 0.68 and 35 % of stations having NSE > 0.8. When predicting streamflow, the model is most sensitive to perturbations in the input data prescribed near and within the basins being predicted, demonstrating that the model is automatically learning to focus on physically realistic areas. When uniformly perturbing input temperature timeseries to obtain relatively warmer and colder input data, the modelled freshet timing and intensity changes in accordance with the transition timing from below- to above-freezing temperatures. The results demonstrate the suitability of a convolutional-recurrent network architecture for spatiotemporal hydrological modelling, making progress towards interpretable deep learning hydrological models.

show abstract

“…Further, to combine different models to identify potential landslide areas through Landslide Inventory and Landslide Susceptibility Mapping for China Pakistan Economic Corridor (CPEC)'s main route (Karakorum Highway) [19] developing LSM contributed to manage risk in a sustainable way [7,8]. More recently Artificial Intelligence techniques have been used to identify potential landslides [9][10][11][12][13]. Similarly, combining different approaches that include interferometric synthetic aperture radar (InSAR) images for monitoring and assessment of landslides have led to high accuracy in identification of potential landslides [14][15][16][17][18].…”

Section: Introductionmentioning

confidence: 99%

Landslide Inventory and Landslide Susceptibility Mapping for China Pakistan Economic Corridor (CPEC)’s main route (Karakorum Highway)

Abbas¹,

Khan

Hussain³

et al. 2021

Journal of Applied and Emerging Sciences

Self Cite

View full text Add to dashboard Cite

The landslide inventory helps to develop landslide susceptibility maps which further assist to minimize economic and human losses as well as towards hazard management. This work investigates landslide hazard mapping in the rugged mountain terrain vis-a-vis highly economically significant route between China and Pakistan i.e. Karakuram Highway (KKH). KKH is passing through the Karakorum mountainous region where landslides events occur frequently and prone to serious risk to local travelers, tourists, and to trading caravans. In this work, landslide inventory was developed (302 landslides) along KKH by visual interpretation of Sentinel and google images. Field survey was also carried to validate landslide datasets. Traditional knowledge-based model i.e. Analytic Hierarchy Process (AHP) and data-based models that include Frequency Ratio (FR) and weight of evidence were applied and compared to develop landslide susceptibility maps (LSM). The landslide dataset was divided into modelling/training (70%) and testing/validation (30%) datasets. LSMs are validated by Area Under Curve (AUC) criterion. The results show that weight of evidence, AHP and FR have success rate curves of 61%, 72% and 84%, respectively. In addition, most highly accurate models are validated for their prediction power using testing landslide datasets. The results for prediction capacity for weight of evidence, AHP and FR are 72%, 58%, and 64%, respectively. Further, landslide susceptibility index (LSI) maps are classified into susceptibility zones. The validation and prediction results show that FR model is the most reliable and accurate model for our study area. Our results will be helpful to minimize landslide hazard losses along KKH, ultimately assisting in successful implementation of CPEC idea between China and Pakistan.

show abstract

A deep learning approach for hydrological time-series prediction: A case study of Gilgit river basin

Cited by 101 publications

References 40 publications

Simulating the Impact of Climate Change with Different Reservoir Operating Strategies on Sedimentation of the Mangla Reservoir, Northern Pakistan

Simulating the Impact of Climate Change with Different Reservoir Operating Strategies on Sedimentation of the Mangla Reservoir, Northern Pakistan

Evaluation and interpretation of convolutional-recurrent networks for regional hydrological modelling

Landslide Inventory and Landslide Susceptibility Mapping for China Pakistan Economic Corridor (CPEC)’s main route (Karakorum Highway)

Contact Info

Product

Resources

About