EGU2020-17646, updated on 12 Jun 2020
https://doi.org/10.5194/egusphere-egu2020-17646
EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Spatial and temporal patterns in seasonal forecast skill based on river flow persistence in Irish catchments

Daire Quinn1, Conor Murphy1, Robert L. Wilby2, Tom Matthews2, Ciaran Broderick3, Saeed Golian1, Seán Donegan1, and Shaun Harrigan4
Daire Quinn et al.
  • 1Irish Climate Analysis and Research UnitS (ICARUS), Department of Geography, Maynooth University, Co. Kildare, Ireland
  • 2Department of Geography and Environment, Loughborough University, Loughborough, UK.
  • 3Met Éireann, Glasnevin Hill, Dublin 9, Ireland.
  • 4Forecast Department, European Centre for Medium-Range Weather Forecasts (ECMWF), Shinfield Road, Reading, RG2 9AX, UK.

In this study we assess the seasonal hydrological forecast skill of river flow persistence across a sample of 46 catchments representative of Ireland’s diverse range of hydrogeological conditions. This statistical approach is straightforward to implement as it uses a river’s most recently observed flow anomaly (calculated over a predictor period of a given duration) as its forecasted flow anomaly (for a given horizon). In our hindcast experiment, persistence skill is evaluated against a streamflow climatology benchmark and by assessing the correlations between predicted and observed anomalies. Using the most skilful predictor period of 1-week, we find that the majority of persistence forecasts outperform the benchmark between April and September at the 1-month forecast horizon. However, this narrows to solely the summer months when using 2- and 3-month horizons.  Skill declines with increasing durations of the predictor and/ or forecast horizon period as a catchment is given more time to “forget” initial anomalous streamflow conditions and/or to be impacted by “new” anomalies. High rainfall events, for example, tend to disrupt the persistence of flows and greater forecast skill is thus found in the relatively drier months.

The degree of persistence skill is also strongly conditional on the “memory” inherent to each catchment (i.e. their storage capacity), as indicated by physical catchment descriptors such as the Base Flow Index (correlation ρ with skill = 0.86). Persistence skill is greatest in lowland regions characterised by permeable lithologies, well drained soils and lower annual average rainfall totals. Physical descriptors can thus be used to anticipate the likely performance of river flow persistence as a forecasting tool in rivers outside the catchment sample. Through multiple linear regression analysis, we identified the combination of predictors that produced the best-performing model (adjusted R2= 0.89) and used it to predict the persistence forecast skill level expected in 215 catchments across the country at different horizons and seasons. Highlighting exactly when and where persistence provides higher predictive skill than the reference climatology forecast, we show the value of statistical flow persistence methods as a tougher-to-beat benchmark in the development of more sophisticated seasonal river flow forecasting methods at the catchment-scale. This research also underscores the scope for development of dynamical hydrological forecasting approaches in the wetter, poorly drained catchments underlain by impermeable lithologies, found mainly in the north-western and south-western regions of Ireland.

How to cite: Quinn, D., Murphy, C., Wilby, R. L., Matthews, T., Broderick, C., Golian, S., Donegan, S., and Harrigan, S.: Spatial and temporal patterns in seasonal forecast skill based on river flow persistence in Irish catchments, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-17646, https://doi.org/10.5194/egusphere-egu2020-17646, 2020

Displays

Display file