An investigation of the predictive accuracy of salinity forecast using the source IMS for the Murray-Darling river

Mccullagh, Harry James (2016) An investigation of the predictive accuracy of salinity forecast using the source IMS for the Murray-Darling river. [USQ Project]

Text (Main Project)

Download (3MB) | Preview


The Murray Darling Basin (MDB) is Australia’s largest and most important river system. Today, the Murray Darling Basin Authority (MDBA) manages and operates the river system through the oversight of key components such as water storage, quality, markets, trade, sharing and salinity. In order to provide defensible operational decisions and enable effective planning, the MDBA has developed a model of the Lower Murray Darling River using the Source Integrated Modelling System (IMS).

A key functionality of the model is the ability to forecast salinity. The forecasting of salinity enables justification of key water sharing and management decisions in relation to their effects on future salinity levels. In order to predict salinity, the current method is driven by three key inputs being salinity concentration (mg/L), flow (ML) and inflow salt load (Tonnes). Currently, salinity and flow are forecast using trend or average functions while inflow salt load is forecast using the average of the most recent month extrapolated forward.

This research project worked to determine the current accuracy of salinity predictions within a new Source model and investigated methods used to estimate and forecast additional salt loads between the reaches. The project worked to improve the model prediction through investigating a variety of data smoothing methods in order to determine whether monthly averaging is the best representation of including the salt inflow loads within the current model. The project then worked to refine the existing forecast method using two approaches: one being trend extrapolation, and the second being application of an Artificial Neural Network (ANN).

The results of the data smoothing analysis indicate that monthly averaging is the best representation of additional salt inflow used within the model. The results of the forecast analysis indicate that rather than using the average of the most recent month for forecasting, trend methods may provide a more effective option. Finally, the research found that the developed neural network was unable to recognize patterns present in the salt inflow data enabling an effective forecast. However, the research highlighted that the application of artificial neural networks are well suited to the prediction of water resource variables such as salinity and would make an excellent option for future research.

Statistics for USQ ePrint 31440
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Additional Information: Bachelor of Engineering (Honours) Major Civil Engineering project
Faculty/School / Institute/Centre: Historic - Faculty of Health, Engineering and Sciences - School of Civil Engineering and Surveying (1 Jul 2013 - 31 Dec 2021)
Supervisors: Alam, Jahangir; Bethune, Mathew; Korn, Alistair
Date Deposited: 20 Jul 2017 23:59
Last Modified: 20 Jul 2017 23:59
Uncontrolled Keywords: predictive accuracy of salinity; Murray-Darling river; water sharing and management; data smoothing analysis; Artificial Neural Network
Fields of Research (2008): 09 Engineering > 0905 Civil Engineering > 090509 Water Resources Engineering
Fields of Research (2020): 40 ENGINEERING > 4005 Civil engineering > 400513 Water resources engineering

Actions (login required)

View Item Archive Repository Staff Only