Simulation of water temperature at the watershed scale with HYDROTEL and long short-term memory networks (LSTMs)
This project aims to implement a new thermal simulation module in the PHYSITEL-HYDROTEL hydrological platform, as well as an LSTM deep learning model, to simulate the temperature of rivers in southern Quebec for a future edition of the Hydroclimatic Atlas of Southern Québec.
Project details
Principal(s) investigator(s)

Context
Water temperature is a key indicator for riparian ecosystems and human activities, which are particularly vulnerable during dry spells. These hydrothermal conditions are likely to become critical due to global warming and increased water withdrawals resulting from socio-economic growth. Since the PHYSITEL-HYDROTEL platform is already used to model flow trends for the Hydroclimatic Atlas of Southern Québec, the addition of a water temperature simulation module is a natural next step.
Objective(s)
To develop the PHYSITEL-HYDROTEL platform’s capacity for mechanistic modelling of river temperature dynamics
To simulate water temperature on a daily basis for all sections of watercourses in the Hydroclimatic Atlas of Southern Québec
To offer complementary deep learning-based water temperature modelling compatible with the data used in the production of the Hydroclimatic Atlas
Methodology
Construction of a database of existing water temperature histories (e.g. RivTemp, BQMA) and data collected by remote sensing (i.e. thermal infrared remote sensing, land surface temperature) for southern Quebec to support the development of thermal models
Integration of the CEQUEAU water temperature mechanistic modelling module into the PHYSITEL-HYDROTEL platform
Development of a computational intelligence module for water temperature by means of deep learning based on long short-term memory (LSTM) networks, using the data used by the Hydroclimatic Atlas platform
Expected results
A database of historical water temperature data for southern Quebec
A version of the PHYSITEL-HYDROTEL platform with the ability to simulate water temperature using a mechanistic modelling module, compatible with the Hydroclimatic Atlas of Southern Québec
A water temperature model based on LSTM deep learning, compatible with the data used for the Hydroclimatic Atlas
Funding

Other participants
Yegane Khoskhalam, INRS
Stéphane Savary, INRS
Sébastien Tremblay, INRS
Louiné Célicourt, INRS
Related projects
712800