Water quality sampling provides an overview of the underlying trends within a catchment. In many catchments rainfall events contribute large portions to the annual sediment and nutrient load, therefore it is crucial to characterise these events. While most catchments in Australia rely on monthly sampling, increasingly automatic samplers are used for event-based sampling. This generally involves sampling at equal intervals in time once the sampler has been triggered. Such data is systematic in nature and requires the use of model-based statistics to be analysed correctly. Random sampling allows the use of easier to use design-based statistics. In this work we propose the use of stratified random sampling for sampling events using auto-samplers. Our approach is to divide the event into strata which represent key features such as the rising and falling limb. Within each we randomly sample. One problem is that we have to estimate the length of the strata, i.e. rising limb, from historical data which will be in most cases an approximation. We show how the discharge data from an event can be used to retrospectively re-stratify the samples. Such an approach allows unbiased estimates of loads and mean concentrations (with a known precision) for events and the strata within each, for example the rising and falling limb. Using this approach we can also estimate how many samples are needed to characterise an event with a known precision. We illustrate the approach with an example from the Muttama creek, a tributary of the Murrumbidgee river.
|Title of host publication||Proceedings of the 34th Hydrology & Water Resources Symposium|
|Publication status||Published - 2012|
|Event||34th Hydrology and Water Resources Symposium - Sydney, Australia|
Duration: 26 Jun 2012 → …
|Conference||34th Hydrology and Water Resources Symposium|
|Period||26/06/12 → …|