Comparing TSS Loads from a Catchment Model, Two Rating Curves and a Ratio Method without True Values: The Challenge

Due to lack of adequate observed suspended sediment concentration (SSC) data, sediment-rating curves and ratio methods have been commonly used to calculate suspended sediment loads. However, measurements of suspended sediment load concentration and flow and the subsequent use of rating curves or ratio methods to calculate suspended sediment loads alone would not answer many spatial and temporal scale related questions with regard to management decisions to improve water quality. These questions are better addressed using catchment scale water quality models. The SedNet (Sediment River Network) model is a sediment generation and transport model to estimate long-term mean annual end-of-valley and in-stream sediment loads. Due to lack of appropriate data, very little work has been done to test the accuracy of outputs SedNet in Queensland catchments. The objective of this study is to compare mean annual total suspended sediment (TSS) load data estimated from two rating curve methods, and the Beale ratio method with outputs of SedNet at six sites in the Burnett catchment in Queensland, Australia. SedNet consistently estimated much higher sediment loads than the other three methods. The power function method consistently predicted the lowest total load. The ratio (highest/lowest) in predicted TSS loads for the six sites varied from 16 to 36. Hence there is poor agreement across all four methods on the annual TSS load both at a site and across several sites. Possible reasons for the discrepancy between the estimates are discussed. This study has shown the challenge in choosing a relatively more accurate method of load estimation in the absence of actual suspended sediment load values against which estimates from the different methods are to be compared. Hence, it is suggested that a similar study be conducted on a site with a very high frequency data set, so that the comparison can be presented as deviation from a known true value and the results can be used to truly quantify error in the methods.