FLOOD05577 - Laserfiche WebLink

Home Browse Search

L~ R.J. HoJ~~I<Jl./J""fJl"I<ifH_,JroI"I..:!J6/19\IlIJ 12~-/J6 app.e:LfS to ha\.e lillle. if any. effetton improving Ihe mood.s eft'kieno:y. It" Ihe effect of the friO:lion coeffi- denl is sruJied. there appears 10 be liule impact on Ihe model eftkiency with no definite pauem emerging. The gnph suggests that the spatial resollllion has a grealer etfa:t on Ihe mo.Jel"s predictive ability than d<-...esthe~.pi.:alcahbr-J.tionparameter. oJ_oJ_ bl.J,,,,l<Iill r<,sulTs Eight no.b.l points .....ere identified wilhin lhe dom:.tin 31 which individual hydruulic rcsuhs (II. II .lnJ \"l \\ ue analysed. Owing to Ihe manner in Ilhi..::h thex meshes were constructed. in the I- DEAS mesh generation ~ckage, only in-channel .lnJ nNal points on banks had identkal x. )' co-ordi. nates iorall se\"enmeshes.lhusdireo:t1yaffecling the .:hvke of points. This .....as. however. felt to be a neces- ~. r~uiumenl for rigorous inter-comparison of mesh m.;'ll~tion eITects. The location of the points is dem,:,nHr:ltN in Table ~. The resulls presented here :lfe iro:n the simulations of events 2 by analysis of ><:::l.I.o.r do....- f3te tresolving II and v) and water depth. \\"h.ilsl;.")s,ible auto-correlation effetts between mesh re>olucvns makes the interpretation of these results nther ;;('mplell. analysis of results showed an effi~.:e.j panern of systematic behaviour which ;;auld :-e u;..ed to formulate a first-order analysis. .J.':'.1. S,'.:;.!ar flow raT~ aT id~lItifit!d inlt!mal poinlS These r.e;;ults are shown in Fig. 6(a)-(d). The most iJentifu.,epattern in all the graphs is tlw: hydrograph di:eo:tioR_ \lith the \.elocity increasing to the pcilk then &.:r~3.)ing \-matioD with changing spatial resolution i. of II:ore interest. The differences in the calculated \.e:OOI:' ':311 be quite dramatic and arc demonstrated in Fij'. 6l:'llthe result from point 4). Primarily, there i. m ill.:rease in the velocity as the element decreases in siu and later the velocity levels-off. However. bciow a spatial resolution of 54 m2 there is a dramatic decrea><:: in the \.elocityrate prior to a rapid increase ior the hj,l.hest re.olution mesh. The scalar flow rale ranges r-r.;m 0.367 to OA2~ m S-I, a difference of 1~.53,(. This panern was also identified for the >ame loc~tion in event 3 with a 14.S4'A> variation in the scalar dOIL- rate at the peak flow. Similar patterns may be idenlified for the other poinu a1dlou!h there is no consistency in the trends from mesh to mesh. Trends ~. however identifiable from event to event. su!gesting that it is somewhat de~ndent on the geographical location of the nodal point within the domain. This may be because of a number of factors including the complCll feedback proceS.\il:s oper-...ting in the domain as well as the effect of spatial discretization on the governing equations. ./../.1. W(lfer,!ep,1I (If idelltified iluenUlI poims The results are shown in Fig. 7(a)-(d). These indi. cate very similar pallerns and interestingly. not an inverse panem. to those identified in the velocity rate lhrough the hydrograph direction. If the node is located in an 'active' region of the mesh. such as Ihe outer section of a meander bend where water tends to move out of bank. the difference in the simulaled variables will be greater than in an 'inactive' area. such as the outer floodplain. where the effect of mesh resolution may nOl have such a great eITect. The resullS from this analysis have not produced clearly identifiable trends. in contrast to the analysis of the bulk Row. However, it has been demonstrated that the spatial resolution has a dramatic effect on internal hydraulic predictions. The geographicalloca- tion within the domain is also an important factor, although it is nOl yet possible to identify a location and characterise the trend. For example. by having a nodal point on the apex of the meander bend. water depth will increase with spatial resolution. although the solution will be a reRection of calculations made at surrounding nodes and thus imply spatial feedback. It is tncrefore not possible to dissect these results much further. though the analysis enables a cautionary note to be made thai the choice of element size is extremely imponant and will directly affect the internal hydrau- licprcdictions. 5. Discussion The results obtained in this paper have indicated the importance of spatial resolution to the predictions oblained from numerical simulations. This is an important result as in a classical sense all the meshes used in this analysis fulfil the tr:lditional criteria of Row length physics typically used 10 condition the choice of mesh resolution. Yet within this range of physically acceptable solution significant variation N.J. Hardy ~I ,,/./ J"u...w.if H.,Jrololl..1l6 /J999j J2~-/J6 in model results was nOled. The results have also indicated that; Spatial resolution directly alfecls bulk Row char. acteristics. For the me~hes studied, as the element size decreases. bulk Row increases up to a point of the penultimate mesh. The bulk Row tharacteris- tics for the highest resolution mesh. however. decrease. Spalial resolution directly aITects inundationexlent although it may be an eITect of the loss of topo- graphic information. Spatial resolution has a grcater effect than thetypi. cal calibration parameter. friction. in altering the hydraulic simulations. This indicates that initial model set-up needs 10 be carefuUy considered and the transfer of parameter values should not occur. 1bc spatial resolution has a dramatic effecl on the internal results. Identification of systematic trends is not feasible o\\ing 10 the complex nature of the system; however, the effect of the s~tial resolution should always be considered. Understanding the effects of mesh resolution in the development of a high resolution space/time model is clearly vital. Moreo\'er. one of the advanlages of the linile clement technique is that the concentration of elements in a specific area can be increased if this region is believed to be sensitive. This needs to be reconsidered. as the same is true if an adaptive mesh- ing technique is applied. \lhere the topographic gradi. ent dctennines the concentration of elemenlS. If an area has a hiJh concenlr.ltion of elemenlS (whether it is a subjective decision by the mesh user or has been created by the mesh generation procedure). then the simulaled h~.draulics Ch. ., and v) may be different in that area from what they would be if an equally weighted element size mesh had been created. This is demonstrated in the 15% variation of the scalar flow rate for a variation in mesh resolution. Although most hypotheses assume that the higher the spatial resolution the closer the simulated hydraulics are to the troe solulion, for field simulations there is currently no means of telling how close to Ihe troe solution the mesh actually is. It therefore appears that II complex feedback process operates within the modelling system driven by the spatial resolution of the mesh and has not pre\'iously been identified when applying distributed models to natural environments. '" Against this background the three central questions we r.used in relation 10 evidente of consislen.:y with process equlltions, mesh resolution guiddines and new inter-comparison methods are thus ~rnaps ambi- tious. Their cenlrality in terms of model infereno:c has. howeVer, been demonstrated by the results from this investigation. particularly given their prominence in a relatively small sample of the whole model parum<:ter space. We can. however. recommend that any future modelling projects. whether for this or other environ- mental problems, should construct at least four meshes of different spatial resolutions to ascertain the envelope of response to spatial resolulion. This would enable the construclion of boundaries for the mesh development. prior to more complex calibration processes. The transfer of such infonnation. in the conSltUction ofa numerical algorithm relating spatial resolution to reach size in a more general scnse. is not possible according to this initial study. However. further studies of this nature may provide an improved insight enabling a clearer definition of mesh/spatial boundaries to be achieved. Acknowledgements Richard Hardy grote:fully acknowledges the suppon of an NERC postgraduale scholarship. The TELE- MAC-2D code was kindly provided by Electricih~ de France. References Abbott. M_B.. BaohursI. J.C.. OltIp. J.A~ O.COIIIICll....E..Il....mU>- XII, J.. 1986. An i/ltlOliur:ti"" 10 !he E~ HydrolOJica! Sytem-Sy5li:me HydlOlojique Europecn. 'SHE" I: His!oty alldphiIO$Op/lyof.p/lysica!lyb:iKd.di<rribuledm<><klling sySlem.J. Hydtol. 81.4S-S9 Baird, L. Gee. D_M.. Andcno:l. M_O.. 1992. UflJaugN cO/oilmen! mo:w.kllillJ 2. Utiliu.Uon of hydraulic modcla for .~id;Won CalCna 19.33-42. Bale.. P.o.. Anderson, M.O.. Baird. L. W~bng. D,E.. Simm. D.. 1992. Modt!llingltoodplain lIow with. !wo dime,,,ional fin;t" .1~mcllllCh.,"". Eanh Sun. Proc. and Lolldfoml. 17. S75- n. Bal".. ".0.. Andcl'S<'ln. M.O.. HcI\'OllCL J.-M.. 1993, All in;!;;>! comp;lriJ.on of IWO 2-di,"",,"ionoIl fin;I" .Jcl1\lOJll cll<k. fOf ri.ernoodoimuL:llion.Proc.lnoI.Ci..EngfS..Walcr.Marilim" and Energy 112.238_248.