FLOOD05577 - Laserfiche WebLink

Home Browse Search

128 RJ H"rJ)' tf ,,1./JUlmuJI "I HyJro',,!.,' :!J6 (/99\11 /2./-136 Fi~,!. Th< dilf.",,,,,,, in 'p.mal ,...,Iulion be,ween m..h I and me.... 7 for:ln ""'3 ",,,,,,..nlinl 3 near ch:mnel "'I'on of \he dom3in. flowing: down the middl~. ~~ Fig. I. From this t~mplat~ sev~n ~paratc ~s~s were constructed. ranging: from 1169 10 II 890 el~m~nts (a\'erage clemenl size 71.54-24.11 m~). using the I-DEAS mesh g~neraling package. A statistical summary of the meshes is presented in Table I while the actual difference in spatial resolution can be seen in Figs. I and 2. Simple topography was prescribed for the domain. as it is the Question of resolution that is of interest here and not the effector topography. The cross-section of me domain was divided into three sections; above and below an absolule .\' value of 200 m, a gradient of 0.01 was specified; between + 200 m and - 200 m a gradient of 0.001. A longitudinal downstream gradi- ent of 0.005 was impo~d. The dimensions of the channel also altered as the resolution increased. because of the means employed to <kfine the wpography. As the number of nodes varied across the channel. from 4 to 7. the cross- sectional area of the channel varies. Therefore, prior to any difference that may be generaled in the solution of the equations. the actual channel volume is differ- ent. However. no scaling corrections were made as one of the first effects ofspatial resolution is the filter- ing of infonnation. Funher. if different resolution meshes were applied to nalUral reach input hydro- graphs and stage data. these would not be scaled fL>\' the representation of the channel. Manning friction coefficients were prescribed of 0.025 fOf the channel and 0.06 for the floodplain. The boundary conditions prescribed were an impo~d flow rate at the upstream end of the reach \\'hil~ at th~ downstream boundary all variables wue allowed to vary fr~ely. A synthetic downstream boundary condition was considered an unnecessary constraint on model behaviour as the results obtained would primarily reflect the boundary conditions assumed. Furth~r it was felt that this would have l~d to additional complications in the interpretation of results. T~st simulations showed that any enhance- menl of wat~r surface slopes only occurred within an ~xtrem~ly localised region. less than 50 m from the downstream boundary. and thai the upstream water levels were relatively insensitive to this simpli. fyingassumption. As we are here concerned with overbank flow. simulations commenced at bank-full discharge. Initial conditions therefore consisted of steady stale flow with a near bank-full water depth of 1.75 m within the channel. while on the floodplain no water was prescribed. The simulations were then run fOf 20 000 s allowing all perturbations caused by the start-up procedure to propagate out of the domain. until a true steady state exists. Identical numerical techniques were implemented fOf all seven meshes. Although TELEMAC-2D is an implicit code and therefore stability may be maintained fOf higher Cour- ant numbers. the results reported in this article always involve a Courant number less than I (Cr..... = 0.7). This guarantees that it is the effect of resolution which is being analysed and not the quality of the simulations. Three separate events were applied to the domain. 40 R.J. H"rJ_~ tf <1/./ }"..m<l/ "f HyJro/"IC-" !16/I999J 124-1J6 ", 35 3<l ~ 25 . t 20 ~ . i 15 o ~_.~ ---,_1 -, .----_l --. -..-.-., -)04<0106 =..:~1<~2 T....(_OfthI o o 10000 20000 30000 40000 50000 60000 70000 Fl,.J. The o.npu'hydrogmpluforevenl2. therefore essential that the flood wat~r level can be accurately simulated forriskasscssment. Relative sensitivity of spatial resolution versus cu/ibrcuion: The effect of spatial resolution and a fieldrepresentalivecalibrationcoefficient(thefric- tion coefficienl) arc considered in relation to a measure of model predictive ability, to identify which has more influence on the model. Currently. model sensitivity to friction has been relatively well explored; however. the same cannot be said with regard 10 mesh resolution. In particular. we wish to detennine whether the calibration for an event is stationary between different mesh resolu- tions or whether some feedback occurs. [n-dumai" results: These consist of the actualu. v and h values calculated by TELEMAC-2D at specific x. y co-ordinates during the simulation. This analysis examines the eff~ct of a change in the spatial resolution on the gOl/erning equations. Primarily, attention was directed towards the nodes located on the channel banks, as it was believed that this was one of the most sensitive locations in the domain when considering floodplain model- ling. owing to the momentum exchange mechan- ism operating between the main channel and floodplain flows. '0 5 The magnitudes of the ev~nt5 were calculated from the dimensions of the channel and floodplain. The research design was based on simulalions of (i) a high magnitude low freQuency event where the whole domain flooded with appreciable water stand- ing on the floodplain: (ii) an intenncdiate event where up to 50% of the domain flooded and; (iii) a low magnilude high frequency event where overbank flow occurred. although large areas of the domain were not inundated. The model results were analysed in four differ- ent contexts enabling funher insight into the hydraulic processes operafing within a floodplain, in order to examine the strengths of the arguments commonly cited as advantages of increasinl spatial resolution. OutPUJ hydrograph (to/al flow J: This type of model output is typically used to validate model predic- tive ability in field applications as this is usually the only data available. Inundation e.l/ent: One of the growing advantages of the application of high resolution two-dimen- sional hydraulic models is the prediction of inun- dation extent for flood proteclion schemes. insurance sUr\leys and similar applications. It is