A multi-collaborative technique for assessing the reservoir sedimentation with futuristic capacity prediction using ANN model

• Published in: Environmental earth sciences 2024-07, Vol.83 (14), p.439, Article 439
• Main Authors: Mishra, Kartikeya, Tiwari, H. L.
• Format: Article
• Language: English
• Subjects: Algorithms; Artificial neural networks; Back propagation networks; Banded structure; Biogeosciences; Collaboration; Deposition; Earth and Environmental Science; Earth Sciences; Environmental Science and Engineering; Floodplains; Foothills; Geochemistry; Geographic information systems; Geographical information systems; Geology; Hydrology/Water Resources; Information systems; Infrared analysis; Multilayer perceptrons; Multilayers; Neural networks; Original Article; Pattern analysis; Performance prediction; Remote sensing; Reservoir capacity; Reservoir sedimentation; Reservoirs; Satellite data; Satellites; Sedimentation; Sedimentation & deposition; Siltation; Spacecraft recovery; Terrestrial Pollution
• ISSN: 1866-6280; 1866-6299
• DOI: 10.1007/s12665-024-11757-1

This study illustrates the comprehensive investigation to assess the sedimentation, deposition pattern, and futuristic active capacity of the reservoir in a minimal period. A multi-collaborative methodology was developed using Geographic information system (GIS) & Artificial Neural Networks (ANN). The sedimentation analysis was carried out on a GIS environment using satellite rasters. The satellite data captures the live water region with a combination of visible and Near-Infrared (NIR) bands. The ANN model was developed using feed-forward backpropagation algorithm to forecast the revised water spread. A Multi-Layer Perceptron [2-1(10)-1(1)-1] ANN structure best captures the trend of water-spread reduction with the coefficient of determination (R 2 )1,1,1 & 0.977 for training, testing, validation, and overall performance respectively. The designed approach provides a performance comparison of GIS and ANN methods for the prediction of reservoir capacity. Also, the observations of sedimentation analysis were superimposed on the Borland & Miller graph to portray the pattern of deposition. The research framework was applied to the Kerwan reservoir located in the capital of central India. This analysis reveals that the useful capacity of the reservoir had reduced from 22.67 to 15.13 Mm 3 in 46 years (1976–2022) and the depositing pattern was shifting towards Type-II (Flood Plain-Foot Hill) which was designed as Type-III like Hilly reservoir. From the Neural Network fittings, it was concluded that Kerwan would be suppressed to 59.95% in 2030 and reduced up to 49.49% & 40.84% for 2050 & 2070 respectively, if siltation carried on.