Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations
The coastal zone of Bangladesh comprises several polders, which are low-lying tracts of land surrounded by embankments to protect against tidal floods and saline water intrusion. They also enhance freshwater availability and aid in improving land productivity. These polders are equipped with sluice...
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Abstract |
| Language: | Inglés |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10568/168948 |
| _version_ | 1855538758551601152 |
|---|---|
| author | Behera, Abhijit Sena, Dipaka Ranjan Hasib, Md. R. Matheswaran, Karthikeyan Jampani, Mahesh Mizan, Syed Adil Islam, Md. J. Alam, R. Mondal, M. K. Sikka, Alok |
| author_browse | Alam, R. Behera, Abhijit Hasib, Md. R. Islam, Md. J. Jampani, Mahesh Matheswaran, Karthikeyan Mizan, Syed Adil Mondal, M. K. Sena, Dipaka Ranjan Sikka, Alok |
| author_facet | Behera, Abhijit Sena, Dipaka Ranjan Hasib, Md. R. Matheswaran, Karthikeyan Jampani, Mahesh Mizan, Syed Adil Islam, Md. J. Alam, R. Mondal, M. K. Sikka, Alok |
| author_sort | Behera, Abhijit |
| collection | Repository of Agricultural Research Outputs (CGSpace) |
| description | The coastal zone of Bangladesh comprises several polders, which are low-lying tracts of land surrounded by embankments to protect against tidal floods and saline water intrusion. They also enhance freshwater availability and aid in improving land productivity. These polders are equipped with sluice gates for water to drain out and intake into the polders. Each sluice has its own catchment area, defined by the elevation and connectivity with canal systems that carry fresh or saline water from surrounding rivers or streams. The sluice gates operation is influenced by in-polder water management for crop cultivation, diurnal tidal dynamics, and the seasonal variations of saline and fresh water in the peripheral river networks. During the dry season, limited flows in the lower Ganges River allow seawater to push inland, causing saltwater intrusion in the peripheral rivers until the rainy season. Community-coordinated sluice gate operations can improve water management, facilitating timely drainage and irrigation, which is essential for high-yielding rice and subsequent dry-season crops. To address these challenges, a multi-variate LSTM (Long Short-Term Memory) model was employed to forecast salinity levels in rivers near 29 sluice gates in a polder near Khulna City in southwest Bangladesh. Utilizing salinity data from July 2011 to December 2022, the models were trained (2011-18) and validated (2018-20) with covariates of discharge, water level, and an upstream reference station. A hierarchical variable additive approach was used to sequentially estimate salinity from upstream to downstream. The NSE was over 0.90 and PBIAS under 5% for all sluice gate locations, confirming accuracy in reconstructing the time series. For forecast testing, the 2020-22 dataset also showed significant confirmation with NSE values over 0.90 and PBIAS under 10%. With readily available input data, the developed salinity forecast model can effectively capture annual and seasonal salinity fluctuations along all sluice gate locations. These forecasting capabilities can potentially identify critical seasonal windows for sluice gate operations, giving the farmers in the polder a 30-day lead time for freshwater intake for irrigation and starting agricultural operations in the aman season. |
| format | Abstract |
| id | CGSpace168948 |
| institution | CGIAR Consortium |
| language | Inglés |
| publishDate | 2024 |
| publishDateRange | 2024 |
| publishDateSort | 2024 |
| record_format | dspace |
| spelling | CGSpace1689482025-10-14T15:09:09Z Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations Behera, Abhijit Sena, Dipaka Ranjan Hasib, Md. R. Matheswaran, Karthikeyan Jampani, Mahesh Mizan, Syed Adil Islam, Md. J. Alam, R. Mondal, M. K. Sikka, Alok saltwater intrusion salinity polders coastal areas machine learning modelling sluices The coastal zone of Bangladesh comprises several polders, which are low-lying tracts of land surrounded by embankments to protect against tidal floods and saline water intrusion. They also enhance freshwater availability and aid in improving land productivity. These polders are equipped with sluice gates for water to drain out and intake into the polders. Each sluice has its own catchment area, defined by the elevation and connectivity with canal systems that carry fresh or saline water from surrounding rivers or streams. The sluice gates operation is influenced by in-polder water management for crop cultivation, diurnal tidal dynamics, and the seasonal variations of saline and fresh water in the peripheral river networks. During the dry season, limited flows in the lower Ganges River allow seawater to push inland, causing saltwater intrusion in the peripheral rivers until the rainy season. Community-coordinated sluice gate operations can improve water management, facilitating timely drainage and irrigation, which is essential for high-yielding rice and subsequent dry-season crops. To address these challenges, a multi-variate LSTM (Long Short-Term Memory) model was employed to forecast salinity levels in rivers near 29 sluice gates in a polder near Khulna City in southwest Bangladesh. Utilizing salinity data from July 2011 to December 2022, the models were trained (2011-18) and validated (2018-20) with covariates of discharge, water level, and an upstream reference station. A hierarchical variable additive approach was used to sequentially estimate salinity from upstream to downstream. The NSE was over 0.90 and PBIAS under 5% for all sluice gate locations, confirming accuracy in reconstructing the time series. For forecast testing, the 2020-22 dataset also showed significant confirmation with NSE values over 0.90 and PBIAS under 10%. With readily available input data, the developed salinity forecast model can effectively capture annual and seasonal salinity fluctuations along all sluice gate locations. These forecasting capabilities can potentially identify critical seasonal windows for sluice gate operations, giving the farmers in the polder a 30-day lead time for freshwater intake for irrigation and starting agricultural operations in the aman season. 2024-12-11 2025-01-14T07:01:41Z 2025-01-14T07:01:41Z Abstract https://hdl.handle.net/10568/168948 en Open Access Behera, Abhijit; Sena, Dipaka Ranjan; Hasib, Md. R.; Matheswaran, Karthikeyan; Jampani, Mahesh; Mizan, Syed Adil; Islam, Md. J.; Alam, R.; Mondal, M. K.; Sikka, Alok Kumar. 2024. Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations [Abstract only]. Paper presented at the American Geophysical Union Annual Meeting 2024 (AGU24) on What’s Next for Science, Washington, DC, USA, 9-13 December 2024. 1p. |
| spellingShingle | saltwater intrusion salinity polders coastal areas machine learning modelling sluices Behera, Abhijit Sena, Dipaka Ranjan Hasib, Md. R. Matheswaran, Karthikeyan Jampani, Mahesh Mizan, Syed Adil Islam, Md. J. Alam, R. Mondal, M. K. Sikka, Alok Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations |
| title | Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations |
| title_full | Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations |
| title_fullStr | Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations |
| title_full_unstemmed | Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations |
| title_short | Addressing salinity intrusion in the polders of coastal Bangladesh: predictive machine-learning modeling for strategic sluice gate operations |
| title_sort | addressing salinity intrusion in the polders of coastal bangladesh predictive machine learning modeling for strategic sluice gate operations |
| topic | saltwater intrusion salinity polders coastal areas machine learning modelling sluices |
| url | https://hdl.handle.net/10568/168948 |
| work_keys_str_mv | AT beheraabhijit addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT senadipakaranjan addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT hasibmdr addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT matheswarankarthikeyan addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT jampanimahesh addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT mizansyedadil addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT islammdj addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT alamr addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT mondalmk addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations AT sikkaalok addressingsalinityintrusioninthepoldersofcoastalbangladeshpredictivemachinelearningmodelingforstrategicsluicegateoperations |