This is the second in a series of three blog posts written by students who have recently completed First Class undergraduate theses.
My thesis set out to establish a benchmark for seasonal streamflow forecasting in Irish catchments. A benchmark method of flow forecasting is a technique which is used to compare the relative potential of other methods to. This is essential in order to establish which methods provide the most accurate long-term flow forecasts in Irish catchments.
An integral component of researching my thesis was the identification of techniques that can be used for this purpose. I identified flow persistence as a flow forecasting method which could potentially be used as a benchmark. This involves using this month’s river flow to predict next month’s flow (i.e. the zero-order forecast).
I then aimed to investigate the relative potential of the Standardised Precipitation Index (SPI) for flow forecasting in Irish catchments. SPI values represent the number of standard deviations by which the observed precipitation anomaly deviates from the long-term mean. The SPI value of various precipitation accumulation periods can be used to predict the following month’s streamflow.
In order to establish the predictive skill of flow persistence methods and the relative potential of SPI for long-term flow forecasting in Ireland, I chose four representative catchments from across the island. I selected two groundwater dominated catchments, the Slaney and the Fergus, and two run-off dominated catchments, the Feale and the Owenea.
I found that the predictive skill of flow persistence methods differed depending on the type of catchment. Flow persistence offered skill in Irish groundwater dominated catchments where I calculated significant correlations between their mean monthly flows and these same flows lagged by one month. However, this predictive power of persistence diminished with longer lead times. For example, I found the correlations between the mean monthly flows and these same flows lagged by 3 months to be insignificant.
I found that flow persistence offered little skill in the run-off dominated catchments that I studied. This is because the skill of seasonal hydrological forecasting only comes from the hydrogeological memory of preceding conditions in groundwater dominated catchments. I identified June, July and August as the months with greatest predictive potential and winter as the least predictable season.
Using SPI, I found that the groundwater dominated catchments corresponded better with longer SPI accumulation periods than the run-off dominated catchments (e.g. the Slaney’s mean monthly flows corresponded best with the precipitation from the preceding 2 months while the Feale corresponded better with the precipitation from the preceding one-month period). This is because of the relatively slow hydrological response to rainfall of groundwater dominated catchments.
I then benchmarked the lagged SPI against flow persistence methods and found that its predictive potential was lower in comparison. The lagged SPI did produce slightly better results than persistence did from May to September in the run-off dominated River Feale catchment but these correlations were still very low and not much higher than the respective correlations calculated using persistence. While the lag-correlations using mean monthly flows and SPI were much stronger for the groundwater dominated catchments, flow persistence consistently yielded much more accurate forecasts than SPI in them. Thus, flow persistence was established as a benchmark for seasonal flow forecasting in Irish catchments. It can therefore be used for the assessment of more complex methods to determine how well they perform relative to persistence methods.
The predictive skill of flow persistence methods shows that knowledge of the initial hydrological conditions of a catchment demonstrates potential for long-term flow forecasting. Such seasonal flow forecasting can be used by water resources management and in early warning systems for floods and droughts. Such forecasts will increase Ireland’s resilience and reduce its vulnerability to these hydroclimatic extremes which are increasing in frequency and magnitude as a result of climate change.
A key lesson that I learned which I would like to pass on to future thesis-writers, particularly those working with quantitative data, is to calculate your results as early on as possible because they may lead to discoveries which will change the direction of your thesis. My thesis initially set out to explore the potential for forecasting drought occurrence in Irish catchments using SPI. When this produced relatively low correlations, my thesis supervisor suggested I check if flow persistence produced higher correlations than SPI for both droughts and floods. This gave a new direction to my thesis which involved reviewing more literature and carrying many more calculations. This requires adequate time.
I would also suggest future thesis-writers to avail of the fountain of knowledge and support from your thesis supervisor and other members of the Geography Department where necessary. My thesis supervisor, Dr. Conor Murphy, provided invaluable direction and guidance which was hugely instrumental to the success of my thesis. The first semester thesis preparation module also proved hugely beneficial in understanding how to properly approach my research and structure my thesis.
A final lesson I learned about my own learning process is that I can gain a much deeper understanding of an issue through primary research as opposed to simply consulting literature that already exists on the topic. The research process gave me the opportunity to engage with the subject in a way that enhanced my learning. I found that learning is even more enjoyable when carrying out primary research on the topic and finding my own results. The level of engagement and satisfaction I derived from carrying out this primary research has highlighted research as a potential area that I would like to pursue in the future.