Evaluation Of Simhyd, Sacramento And Gr4j Rainfall Runoff Models In Two Contrasting Great Barrier Reef Catchments

The eWater modelling framework has been used to predict streamflow and pollutant loads across the Great Barrier Reef (GBR) catchments, Queensland as part of the Reef Plan Paddock to Reef program. The Simhyd rainfall-runoff model was initially chosen to simulate streamflow across the GBR due to its extensive use for catchment hydrology studies across Australia, particularly in Queensland. The aim of this study was to determine whether an improved calibration could be achieved with alternative models, and secondly to identify the most suitable model and objective functions for high flow event simulations. Source Catchments was coupled to the Parameter Estimation Software Tool (PEST), to compare the performance of Simhyd to two widely-used rainfall-runoff models available in the Source framework, namely Sacramento and GR4J. Two contrasting catchments were selected with three nested gauging stations in each, with total drainage areas of approximately 10,000km(2). The first is in the upper Burdekin Dry Tropics Region and the second in the Wet Tropics region, both are located in the northern GBR.Statistical evaluation of the three rainfall-run models found that in Burdekin dry tropics catchment, all three models had a coefficient of determination (r(2)) greater than 0.75 and daily Nash Sutcliffe Coefficient (NSE) greater than 0.55 with the exception of a Simhyd result at one gauge. Overall, the Sacramento model achieved the highest r(2) and NSE values. Similarly, all three models have simulated total flow volume ratios (bias) within 20% with the exception being Simhyd at one gauge. Sacramento outperformed the other two models when comparing high flow volume ratios (bias). When simulating middle and low range flows, the GR4J model simulated volume ratios greater than 20% for all sites. In the Wet Tropics catchment, the Sacramento model achieved the highest r(2) and NSE compared to Simhyd and GR4J, although the latter two models achieved acceptable r(2) and NSE values according to the criteria applied in this study. All three models simulated acceptable volume ratios of total flow and high flow in this drier Burdekin catchment.The Sacramento model also achieved the lowest bias (volume ratio) in the simulation when comparing flow statistics for the three biggest flow events on record in both catchments. Visual assessment of the hydrographs also revealed that Sacramento model simulated the timing and magnitude of peak flow and the shape of recession curves of the large flow events better than the other two models.The preliminary results from the calibration suggest that the Sacramento model may provide an improved calibration performance over Simhyd and GR4J. The use of Source Catchments coupled to PEST using multiple objective functions also improved calibration statistics over Source Catchment's in-built calibration tool. Future work will look at assessing other objective functions and alternative rainfall inputs to refine high flow calibrations.