16-Civ-B4 Engineering Hydrology · May 2013
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Paper format. National Examinations, May 2013 — 98-Civ-B4 Engineering Hydrology. Three hours, closed book, one candidate-prepared two-sided 8½″ × 11″ aid sheet, and one approved Casio or Sharp calculator whose model must be declared. Seven problems are printed; page-1 Note 4 states that any five (5) questions constitute a complete paper and that only the first five answers appearing in the workbook are marked. Each problem carries twenty (20) points, so the examinable total is 5 × 20 = 100 points. The page-7 marking scheme breaks each problem into its sub-parts. All seven problems are solved here, because this set is a study resource rather than a timed sitting; the sub-part mark values shown below are the printed ones.
Reference texts. V. T. Chow, D. R. Maidment and L. W. Mays, Applied Hydrology (hydrologic cycle, unit hydrographs, routing, frequency analysis, infiltration); L. W. Mays, Water Resources Engineering, 3rd ed. (design application, rainfall–runoff, groundwater); W. Viessman and G. L. Lewis, Introduction to Hydrology, 5th ed. (measurement, areal precipitation, snowmelt energy budget); V. T. Chow, Open-Channel Hydraulics (1959) (flood-wave propagation, gradually varied unsteady flow); C. W. Fetter, Applied Hydrogeology, 4th ed. (Darcy’s law, hydraulic conductivity, transmissivity). Canadian practice references: Environment and Climate Change Canada / Water Survey of Canada HYDAT archive and the ECCC Engineering Climate Datasets (IDF curves and the IDF_CC climate-adjustment tool), and the Transportation Association of Canada Guide to Bridge Hydraulics, 2nd ed.
Check — conventions used throughout this paper. A hydrologic year is taken as 365 days = 31 536 000 s unless a question says otherwise. Water density is 1000 kg/m3 and the latent heat of fusion of ice is 334 kJ/kg. Where the printed data are internally inconsistent — and Question 7(i) is such a case — the inconsistency is demonstrated arithmetically, the governing conservation requirement is stated, and the corrected reading actually used is declared at the point of use, as page-1 Note 1 invites (“the candidate is urged to submit… a clear statement of any assumptions made”).
Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.
The task described is model transfer, or regionalisation: a model whose parameters were fitted to one gauged basin must be used to predict the runoff hydrograph of another basin, typically one that is ungauged. The essential recognition is that a calibrated parameter set is not a property of nature but a property of the pairing of a model structure, a data record and an objective function; it cannot simply be carried across a divide. The defensible procedure has six stages.
Stage 1 — separate what is physical from what is fitted. Go through the parameter list and classify each entry. Quantities that can be measured or mapped independently in the new basin — drainage area, main-channel length and slope, soil group and hydrologic condition, land use and imperviousness, channel roughness — are re-derived from the new basin’s own data and are not transferred. Quantities that are purely conceptual — a linear-reservoir constant, a recession coefficient, a routing weighting factor, an initial-abstraction ratio — are the ones that must be transferred, and they are the ones that require the argument in the following stages.
Stage 2 — re-establish the model on the donor basin and confirm it is well identified. Split the historical record into a calibration period and an independent verification period, and confirm that the model calibrated on the first reproduces the second. Use more than one objective function, because each is blind to something: the Nash–Sutcliffe efficiency and the root-mean-square error weight the peaks, the volume error checks the water balance, the peak-timing error checks the response time, and the log-transformed criterion checks the low-flow behaviour. Run a sensitivity analysis to identify which parameters actually control the response; a parameter to which the model is insensitive is not identified by the data, so transferring its fitted value carries no information and it should be set from physical reasoning instead.
Stage 3 — establish hydrologic similarity between the donor and the target basin. Transfer is only defensible if the two basins belong to the same population. Compare drainage area (an order-of-magnitude difference is disqualifying), relief and main-channel slope, climate regime and the seasonality of runoff, the fraction of flow that is snowmelt-derived, soil and surficial geology, land use, lake and wetland storage, and the degree of regulation. In practice this is done by placing both basins in a hydrologic region — in Canada, using the regional flood-frequency and physiographic groupings established for the province — and confirming that they fall in the same one. If similarity fails on a first-order control such as lake storage or the snowmelt fraction, the correct conclusion is that this model should not be transferred at all.
Stage 4 — scale the transferable parameters by physiographic regression. Rather than copying values, express them as functions of measurable basin characteristics. The classic form is the Snyder synthetic unit hydrograph, in which the basin lag is regressed on channel geometry,
$$t_p = C_t \left( L\,L_{ca} \right)^{0.3}, \qquad Q_p = \frac{C_p\,A}{t_p}$$where L is the main-channel length, Lca the length to the centroid, and Ct, Cp are regional coefficients fitted on gauged basins in the region — including the donor. The same idea underlies the SCS dimensionless unit hydrograph with its lag–time-of-concentration relation, and the regional regressions used for time of concentration and for recession constants. This stage is what converts a single fitted value into a transferable relation: the coefficients travel, the parameters are recomputed from the target basin’s own geometry. Where several gauged donors exist in the region, fit the regression on all of them rather than on one, and prefer the physically-nearest or the most similar donor where a single choice is required.
Stage 5 — apply the model to the target basin and constrain it with whatever independent evidence exists. Even an “ungauged” basin usually offers something: a few spot discharge measurements, a high-water mark from a known storm, a nearby regional flood-frequency estimate for the 2-year or 100-year flow, a downstream gauge whose record contains the target basin as a sub-area, or a documented flood of record. Check that the transferred model reproduces these. A drainage-area-ratio scaling of a nearby gauged flow, Qtarget = Qdonor(Atarget/Adonor)n with n typically 0.6–1.0, provides a cheap independent order-of-magnitude test of the routed peak.
Stage 6 — quantify and report the uncertainty, and state the validity range. Propagate the parameter uncertainty from the donor calibration through to the target prediction, most simply by re-running with the parameter sets that lay within the acceptable range of the objective function, and report a band rather than a single hydrograph. State explicitly the range of conditions over which the transfer is claimed to be valid: the model has been calibrated on events of a certain magnitude and cannot be extrapolated to a design flood several times larger without a stated caveat, and it is invalid if the target basin’s land use changes materially. Finally, recommend that a gauge be installed on the target basin if the decision it supports warrants it — a transferred model is a provisional instrument, and the honest recommendation is often to replace it with data.
The conceptual diagram printed with the question arranges routing methods along an axis from basic hydrologic methods to complex hydraulic methods, and along a second axis from uniform steady flow to non-uniform unsteady flow. It makes the point that the choice of method is a choice about which flow conditions must be represented. The answer below selects one method — Muskingum — applied to one routing type — channel (river) routing, the hydrologic-method end of that diagram, and describes the process of applying it.
Step 1 — establish the physical premise. Treat the reach as a single storage element between a defined upstream section and a defined downstream section, with no significant lateral inflow, no diversion and no overbank spill outside the reach. Storage in the reach has the two components drawn in Figure 5.1: the prism, proportional to the outflow, and the wedge, proportional to the difference between inflow and outflow. Summing them gives the Muskingum storage relation,
$$S = K\,O + K\,X\,(I - O) = K\left[X\,I + (1-X)\,O\right]$$with K a time constant approximately equal to the travel time of the flood wave through the reach and X the wedge weighting, 0 ≤ X ≤ 0.5.
Step 2 — assemble the data. Obtain the inflow hydrograph I(t) at the upstream section, tabulated at a uniform interval, and the initial outflow O1 at the start of the event, normally taken equal to the initial inflow so the reach begins in steady state. For calibration, obtain at least one historical event for which both inflow and outflow were gauged.
Step 3 — determine K and X. From the historical event, compute the accumulated storage by continuity, Sj = Sj−1 + [(Ī − Ō)Δt]j, then plot S against the weighted flow [XI + (1−X)O] for a series of trial values of X — 0.1, 0.2, 0.3, 0.4. Each trial gives a loop; the value of X that makes the loop narrowest and closest to a single straight line is adopted, and K is the slope of that line, with units of time. Where no gauged pair exists, take K as the reach length divided by the flood-wave celerity (about 5/3 of the mean velocity for a wide channel) and X from experience — 0.2 to 0.3 for a natural river reach, near 0.5 for a steep reach that translates the wave with almost no attenuation, and 0 for a reach behaving as a level pool.
Step 4 — select the routing interval and compute the coefficients. Choose Δt to define the rising limb adequately — a common rule is no more than one-fifth of the time to peak — and check the stability requirement 2KX ≤ Δt ≤ 2K(1 − X). If the reach is long enough that K greatly exceeds the acceptable Δt, subdivide it into n sub-reaches each with K/n. Then evaluate the three routing coefficients given in Question 4(i) and confirm that C0 + C1 + C2 = 1 before proceeding — this single check catches most arithmetic errors.
Step 5 — march the routing forward. Tabulate time, I, the three products C0I2, C1I1 and C2O1, and the resulting O. At each interval the newly computed O2 becomes O1 for the next line. Continue until the outflow has receded to its initial value. Where the river has been subdivided, the outflow column of one sub-reach becomes the inflow column of the next, and tributary hydrographs are added at the appropriate junction before routing continues.
Step 6 — verify and interpret. Three checks are mandatory. The volumes under the inflow and outflow hydrographs must be equal to within the rounding of the table. The outflow peak must occur later than the inflow peak and be lower than it, and the outflow hydrograph must cross the inflow hydrograph exactly at the outflow peak — because at the instant of maximum outflow dS/dt = 0, which requires I = O. And there must be no negative outflow ordinates; if there are, the time step violated the stability range and must be re-selected. The routed peak and its arrival time are then the design outputs: the peak sizes the downstream conveyance or dike, and the arrival time sets the warning lead time.
Two remarks connect this back to the printed diagram. First, the same continuity framework with X = 0 and a storage–outflow relation taken from the reservoir stage–area curve and the spillway rating becomes level-pool (modified Puls) reservoir routing — the reservoir case is the degenerate Muskingum case in which the water surface is horizontal and storage is a single-valued function of outflow. Second, the Muskingum method sits at the “basic”, uniform-steady end of the diagram: it cannot represent backwater, reverse flow, an ice jam or a downstream control, and where any of those govern, the analysis must move to the hydraulic-method end and solve the Saint-Venant equations as described in Question 4(ii). The Muskingum–Cunge variant is the bridge between the two, deriving K and X from channel hydraulics rather than from a gauged pair, which allows the method to be applied to an ungauged reach with a defensible physical basis.
