Disaggregation of global ensemble rainfall forecasts for improved stormwater management
|Advisor:||Anctil, François; Pelletier, Geneviève|
|Abstract:||Dry detention ponds are commonly implemented to mitigate the impacts of urban runoff on receiving water bodies. They currently rely on static control through a fixed limitation of their maximum outflow rate. Real-Time Control (RTC) allows optimizing their performance by manipulation of an outlet valve. This thesis developed several enhanced RTC scenarios of a dry detention pond located at the outlet of a small (3.5 km2) urban catchment near Québec City, Canada. The catchment's runoff quantity and Total Suspended Solids' (TSS) concentration were simulated by the SWMM5 model with an improved wash-off formulation. The control procedures rely on rain gauge data, on measures of the pond's water height, and, in some of the RTC scenarios, on rainfall forecasts. Rainfall forecasts are indeed valuable to a wide variety of end users in the field of flood risk assessment and water management, as they allow some anticipation of the behaviour of the system under consideration. Ensemble rainfall forecasts thus provide an explicit and dynamic assessment of the uncertainty in the forecast. However, for hydrological forecasting, their low resolution currently limits their use to large watersheds. Therefore, this thesis explores rendering the Canadian Ensemble Prediction System's (EPS's) rainfall forecasts more appropriate for hydrological modeling of such a small urban catchment as the one studied here. To bridge this spatial gap, various implementations of the spatial statistical downscaling method proposed by Perica and Foufoula-Georgiou (1996b) were compared, bringing Environment Canada's (EC's) global Ensemble Rainfall Forecasts (ERFs) from a 100-km by 70-km resolution down to 6-km by 4-km, while increasing each pixel's rainfall variance and preserving its original mean. These ERFs were issued by the Canadian Global Ensemble Prediction System (GEPS) in its 2009 operational version. The statistical downscaling method of Skaugen (2002) was also applied to these ERFs, producing rainfall fields with a resolution of 10 km by 7 km. For comparison purposes, simpler methods were also implemented such as the bi-linear interpolation, which disaggregates global forecasts without modifying their variance. The downscaled meteorological products were evaluated, using different scores and diagrams, from both a meteorological and a hydrological view points. The rainfall forecasts were compared against nine days (presenting strong precipitation events) of observed values taken from Québec City's rain gauge database. Ensemble Hydrologic Forecasts (EHFs) with a time step of 3 and 24 hours were performed over a 3-month period for the original and disaggregated rainfall forecasts. This hydro-meteorological operational forecasting chain was conducted using hydrological models GR4J, a modified version of GR4J, and SWMM5. These models were implemented on four catchments ranging between 5 and 350 km2, and located in the Québec City region. The hydrological evaluation was based on the comparison of forecasted flows to the observed ones. Results obtained with the method of Skaugen (2002) were not as interesting as those based on the technique of Perica and Foufoula-Georgiou (1996b). This is due to the fact that with the method of Skaugen (2002), the final rainfall field corresponds to the average of ten downscaled fields, what tends to dampen the variance added through the disaggregation process. For the technique of Perica and Foufoula-Georgiou (1996b), the most important conclusions are: 1) the overall quality of the forecasts is preserved during the disaggregation procedure and 2) the disaggregated products using the variance-enhancing method are of similar quality than bi-linear interpolation products. However, variance and dispersion of the different members are, of course, much improved for the variance-enhanced products, compared to the bi-linear interpolation, which is a decisive advantage. These results were confirmed by the hydrological evaluation. The disaggregation technique of Perica and Foufoula-Georgiou (1996b) hence represents an interesting way of bridging the gap between the resolution of meteorological models and the high degree of spatial precision sometimes required (in the precipitation representation) by semi-distributed hydrological models and by models built on small watersheds. RTC strategies of the studied dry pond allowed for a substantial improvement of the performance compared to those with its current static control– the TSS removal efficiency increased from 46 to about 90% - while remaining safe and taking a mosquito-breeding risk constraint into account. However, the downscaled rainfall forecasts were not superior to the original ones (issued by the Canadian GEPS) in this context, as they led to the same performance for the RTC scenarios relying on rainfall forecasts. |
|Document Type:||Thèse de doctorat|
|Open Access Date:||20 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.