Data conditioning and climate sensitivity analysis of a probablistic rainfall-runoff model

Loading...
Thumbnail Image
Files
Nowbakht_Parvaneh_MEngSc2020.pdf(9.31 MB)
Full Text E-thesis
Date
2020-04-02
Authors
Nowbakht, Parvaneh
Journal Title
Journal ISSN
Volume Title
Publisher
University College Cork
Published Version
Research Projects
Organizational Units
Journal Issue
Abstract
The Munster Blackwater catchment, in the South West of Ireland, was regularly subject to flooding, prior to flood allevation works. The towns of Mallow and Fermoy within the catchment suffered many disturbances for their inhabitants with sometimes severe economic losses. A good knowledge of rainfall-runoff processes is important in order to understand the causes of flooding to be able to develop new infrastructure to manage flooding. The first part of this project focuses on the rainfall and river flow data collection from different sources: the 15-minute time step precipitation data from the OPW, the 15-minute river level/river flow from the OPW and the EPA and the precipitation data from MÉRA (Met Éireann ReAnalysis- Climate ReAnalysis). MÉRA is a very high resolution climate reanalysis dataset which was used to calculate the monthly and annual rainfall in a specific year, for example for 2010 for selected locations (the nearest point to each rain gauge). Initial analysis of the measured OPW data shows significant numbers of missing values and outliers for the precipitation data. A method was developed to cluster the rain gauges with similar precipitation patterns based on the amount of precipitation of the nearest points to these rain gauges from MÉRA. Then a gap filling method was applied in each cluster to fill the missing values of each rain gauge with its cluster members. Other methods were also examined to obtain quality controlled data. The second part of this project applies a conceptual hydrological model, PDM (Probability Distributed Model) developed by Moore (Moore, 2007) to the Munster Blackwater catchment. The model considers each point of a catchment as a single storage unit with a specific storage capacity (depth) that can be described by a Pareto distribution. PDM is suitable for a variety of catchments, and has minimal data and computational requirements. The input is 15-minute precipitation data from different rain gauges and 15-minute river level/river flow data from river stations along the river. The calibration was applied on three subcatchments of the Munster Blackwater catchment. The validation was applied for years between 2010 to 2017. The calibrations and validations indicate that the PDM model can explain most of the variability of observed flows in the different subcatchments over a period of years, especially when a high standard of data quality is available, for example in 2015. Then validation of the model for flood events was examined. Validation was applied for the highest flood event in each year during 2010 to 2017. The accuracy of the model runs are different for each subcatchment with the best accuracy of 93% in the Dromcummer subcatchment and the accuracies in Mallow Rail BR and Killavullen being 80 % and 78% respectively. The model estimates the peak and low flow very well in Dromcummer. The computed flow is underestimated in Mallow and overestimated in Killavullen. The third part of the project is to use the PDM model in a precipitation and river flow sensitivity analysis. This was achieved by increasing the precipitation amounts in the datasets by 10, 15, 20, 25 and 30% to examine how the peak flows and low flows respond. It was found that the peak flows increase by amounts similar to the precipitation increases. The low flows increase at a much lower rate than the precipitation increases. It is known that in a scenario of climate change for a warming world that the precipitation increases by a maximum of 7% per degree C increase in accordance with the Clausius-Clapeyron equation. However as a warming world also increases evaporation and will likely impact the soil moisture status, it is considered that flood flows might increase at a rate less than the precipitation increases. This can be examined by increasing the value of potential evaporation by 10, 15, 20, 25 and 30% .These conditions were not included in this and it is ecommended that further research be done in this area for Ireland.
Description
Keywords
Climate , Hydrology , Machine learning , PDM model , Data conditioning , Rainfall-runoff models
Citation
Nowbakht, P. 2020. Data conditioning and climate sensitivity analysis of a probablistic rainfall-runoff model. MRes Thesis, University College Cork.