Firstly, a word of caution: baseflow separation methods give an estimate of the baseflow based on the assumptions built into them. As such, the results can have large uncertainties. Generally baseflow separation methods are applied at a daily timescale, though some can be applied at other scales.

You haven't said anything about the catchment you are studying (e.g. size, climate) so it is difficult to give specific advice. There are models that can be applied at a 5 minute resolution - however, at that resolution you would be resolving the flow peaks, so you will need to use a cascade to capture the rise to peak. This means a fairly complex model, which will result in identification issues (discrete time applications to high resolution data can be a problem). This means that there can be more than one structure that could fit the data, leading to uncertainty in the result.

Another issue is whether the components of the model actually represent the baseflow, or merely separate flow into quick and slow flow components. Generally speaking, models will give you separation into slow flow - interpretation of that as baseflow is common, but generally not very reliable as there can be a fast response from an aquifer, implying that part of the quick flow component may be due to baseflow. All depends on the specific conditions for the catchment, and hinges on expert knowledge (without additional information).

All that said, given a 5 minute resolution, and assuming that you have high resolution climate data also, you could use something like the Captain toolbox (a Matlab toolbox available from http://captaintoolbox.co.uk/) which can be used in discrete or continuous time mode - the continuous time mode may be better in this case. Alternatively, the Hydromad package (http://hydromad.catchment.org/) has a suite of models, some of which have been applied at very high resolution (down to the minute scale).

Maybe you could try separating the baseflow component at different temporal resolutions and see how the results compare. I recommend caution using any baseflow separation technique on any catchment, and would be very cautious using baseflow estimates at a 5 minute resolution however unless you have an extremely small catchment (time to peak

Thanks for these links, Barry. I am new to the world of hydrological modelling so this is very helpful. My catchment is predominantly underlain by impermeable bedrock - do you think this would help reduce the uncertainty in the models (eliminating the possibility of an aquifer)? Also, what is this "cascade" method you mention?

G'day Anna,

There would be an aquifer above the bedrock, so wouldn't eliminate the possibility of an aquifer. Since you are interested in baseflow, you are looking for a component of flow that has come through a subsurface flow path. What is the size of your catchment? What data do you have (aside from 5min streamflow)?

The cascade is a combination several stores in series. The simplest form is the Nash cascade, which has n identical exponentially decaying stores (an exponentially decaying store has the discharge equal to a constant times the storage, making a linear first order ODE, and is hence often called a linear store).

Hi Barry,

I am studying 4 watersheds with areas of 332ha, 481ha, 780ha, and 1279ha. In addition to streamflow, I have continuous stage, water temperature, turbidity and precipitation data.

G'day Anna,

At that size, 5 minute resolution should probably be okay. Does depend on topography, soils etc, so can only be sure by looking at the data. Have you done a cross correlation analysis on the rainfall and streamflow data?

For example:

Article Evaluation of approaches for estimation of rainfall and the ...