The major inputs to the RRL are as follows:

• Rainfall – a continuous time series of rainfall data that represents the rainfall across the catchment
• Potential evapotranspiration – a continuous time series of potential evapotranspiration or evapotranspiration data that represents the evapotranspiration across the catchment.
• Flow gaugings – daily runoff values for the gauging station that is to be modelled. This data are used for model calibration and checking.
• Catchment area – this is used to convert inputs and outputs between flow and depth of runoff.
  

Note: Various time series formats are supported within the TIME environment. Some of these include:

• Comma separated value time series format (.cdt). Note that this supersedes the files that used to have the too generic .csv extension. If you used some .csv files prior to June 2004, you just need to change the extension to .cdt
• IQQM daily time series format (.iqqm)
• AWBM daily time series format (.awb). Note that existing datasets may not always have the dates included with the data: you are strongly encouraged to add them.
• Time series format from the Francis Chiew and Tom McMahon dataset of Australian catchments (.dat)
• QDNR SILO daily time series format (.silo5)
• SWAT daily time series (.pcp)
• Tarsier daily time series format (.tts)

File Formats

More information about file formats supported by this application can be found in the Toolkit Data Files guide.

There are five basic options for production of outputs from RRL:

• Time Series Graphs
• Scatter plots (daily and monthly)
• Difference of flows
• Flow duration curves
• Export

Note: All graphs may also be plotted with a log axis.

The accuracy of rainfall runoff models is largely dependent upon the representativeness of the input data. It is strongly recommend that a very thorough review of input data is undertaken prior to using the streamflows generated by any of the models. Some tools are included in the RRL to help users to check the basic statistics of input data to detect inconsistencies, but other specific verifications may have to be performed prior to using the RRL on a case-by-case basis. For instance a rainfall input may be have been collated using Theissen weighting of several gauges, some possibly outside of the catchment, and may overestimate the number of rain days.

• Automated calibration is not a substitute for a sound knowledge of each model:
  1. Automated calibration scheme may result in a good mathematical fit with parameter values which are not realistic given the structure of the conceptual model
  2. It is advised to set reasonable allowed ranges for parameters prior to calibration; this will improve the average calibration time, and increase the chances of getting a good fit and realistic result. These reasonable parameter boundaries depend on the characteristics of the catchment being modelled and the relations they bear with the model parameters
• Care should be taken in the calibration to ensure that the model is calibrated during a period that represents sufficient climatic variability to improve the robustness of the model
• It is important to select the objective functions appropriately for the task at hand. For example a standard Nash-Sutcliffe coefficient of efficiency is unlikely to be appropriate if the main goal of the modelling is to reproduce low flows particularly well.
• Global search algorithms are very powerful for calibration of non-linear models, but also very demanding in terms of computation. Users are advised to get at least a basic understanding of the algorithms involved to parameterise those algorithms themselves for an efficient search in the parameter space.