16-Civ-B4 Engineering Hydrology · December 2014
Nivaar worked solution (AI-drafted; not reviewed by a licensed engineer)
Paper format. National Exams, December 2014 — 98-Civ-B4 Engineering Hydrology. Three hours; closed book with one candidate-prepared double-sided 8½″ × 11″ aid sheet; Casio or Sharp approved calculator. Seven problems are printed, each worth 20 marks; any five constitute a complete paper and only the first five answers in the work book are marked, for a maximum of 100 marks. All seven problems are solved below, because the full set is the more useful study resource.
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, reservoir operation); W. Viessman and G. L. Lewis, Introduction to Hydrology, 5th ed. (measurement, areal precipitation, energy budget); V. T. Chow, Open-Channel Hydraulics (1959) (flood-wave propagation and gradually varied unsteady flow); C. W. Fetter, Applied Hydrogeology, 4th ed. (Darcy’s law, hydraulic conductivity, recharge). 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); the ISO 1100 / WMO Manual on Stream Gauging series as adopted by the Water Survey of Canada; the Canadian Dam Association Dam Safety Guidelines (inflow design flood and dam-break consequence classification); and the Transportation Association of Canada Guide to Bridge Hydraulics, 2nd ed.
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.
1. The independent variables retained. Lumped (hydrologic) routing treats the reservoir as a single storage element and solves continuity alone,
$$\frac{dS}{dt}=I(t)-O(t)$$so flow is a function of time only. Distributed (hydraulic) routing solves the full one-dimensional St Venant equations — continuity plus momentum — so flow is a function of both distance and time, and the water surface is computed at every section rather than assumed. The distinction is exactly that between an ordinary and a partial differential equation.
2. The treatment of the water surface. Lumped reservoir routing assumes a level pool: the water surface is horizontal, so storage is a unique function of outflow, $S = f(O)$, and the storage-indication (modified Puls) method closes the problem with the outlet rating curve. Distributed routing computes a sloping, dynamic water surface and therefore represents the wedge storage that develops during the passage of a flood wave. On a long, narrow lake the level-pool assumption can fail badly, because the inflow end genuinely stands higher than the outlet during a rapid rise.
3. The physics that can be represented. Because it omits momentum, lumped routing cannot represent backwater from a downstream control, wind set-up, seiching, tidal influence, or a flow reversal — it will always produce an outflow peak that lies on the recession of the inflow hydrograph, which is the correct signature of a genuine level pool but wrong when a downstream control governs. Distributed routing captures all of these, and is required for dam-break analysis, for lakes with a downstream backwater control, and wherever the flood wave is steep.
Relative accuracy, and how to improve each. For a genuine reservoir — wide, deep, with a well-defined outlet structure and a flood wave slow relative to the basin’s response — lumped level-pool routing is highly accurate, and the added complexity of a hydraulic model buys little. Its accuracy degrades as the water body becomes long and narrow, as the inflow becomes steeper, and as downstream backwater becomes significant. Distributed routing is intrinsically more accurate but is only as good as its inputs: it demands surveyed cross-sections, calibrated roughness coefficients and a stable numerical scheme, and a poorly surveyed hydraulic model can easily be worse than a well-calibrated lumped one.
Lumped routing is improved by refining the stage–storage and stage–discharge curves from a bathymetric survey and a proper outlet rating, by shortening the routing time step (a common rule is $\Delta t \le$ one fifth of the time to peak of the inflow hydrograph), and by subdividing a long lake into several level-pool cells in series. Distributed routing is improved by denser cross-section spacing, by calibrating Manning’s n against observed flood profiles and high-water marks, by checking Courant-number stability, and by moving from a simplified (kinematic or diffusion) formulation to the full dynamic wave where backwater matters. In both cases the single largest improvement is verification against a measured flood.
Part (a) — Adequate delineation of the watershed boundaries (4 marks)
Delineation fixes the contributing area, and area is the multiplier on nearly every quantity the model produces — volume, peak discharge, and yield. An error here propagates through the whole analysis and cannot be recovered by calibration; it is simply absorbed into distorted parameter values, which then fail on the next event. On large watersheds the delineation is not trivial: the divide must be traced from a DEM of sufficient resolution, and the automatic result must be checked against topographic maps and aerial imagery. Three traps recur. Flat terrain and wetlands give ambiguous divides where small DEM errors move the boundary by kilometres. Non-contributing areas — internally drained prairie potholes, a widespread feature across the Canadian Prairies — must be excluded from the effective drainage area, or the model will over-predict every flood. And artificial diversions, storm sewers, irrigation canals and inter-basin transfers can route water across a topographic divide, so the hydrologic boundary is not the topographic one. Delineation also fixes the subbasin structure and hence the model’s spatial resolution, and it must be done consistently with the outlet chosen for calibration, which is normally a stream gauge.
Part (b) — Obtaining appropriate and sufficient hydrologic and geographical data (4 marks)
A hydrologic model is a data-transformation device: its output can be no better than its inputs, and the single most common cause of a failed flood study is inadequate data rather than an inadequate model. Four categories are needed. Meteorological: precipitation at sufficient gauge density and temporal resolution to resolve the storm — hourly or finer for flood work — plus temperature for snowmelt, and increasingly radar or gridded reanalysis products to capture spatial structure that a sparse gauge network misses. Streamflow: gauged hydrographs at the outlet and, ideally, at interior points, without which the model cannot be calibrated or verified at all. Geographical: a DEM, soils mapping, land cover, and channel geometry. Water management: reservoir operating rules, diversions and withdrawals.
“Appropriate” carries as much weight as “sufficient”. Data must be at a resolution matched to the processes being modelled, must span a period long enough to include events of the magnitude of interest, and must be quality-controlled for gaps, station moves and rating shifts. Crucially the record must be split: one period for calibration and an independent period for verification. A model calibrated and tested on the same data proves nothing. Where data are sparse — the normal condition in northern Canada — regionalisation from hydrologically similar gauged basins is the accepted substitute, and the resulting uncertainty must be stated rather than hidden.
Part (c) — Spatial and temporal scaling of model parameters and model selection (4 marks)
Parameters measured or defined at one scale do not transfer unchanged to another, and ignoring this is one of the deepest errors in hydrologic modelling. A hydraulic conductivity measured on a 100 mm soil core does not describe a 10 km2 subbasin, in which macropores, preferential pathways and heterogeneity dominate; an effective parameter at the model’s grid scale is required, and it is generally obtained by calibration rather than measurement. Likewise a runoff coefficient calibrated on a 1 km2 urban catchment cannot be applied to a 500 km2 basin, where channel routing and partial-area contribution change the response entirely.
Temporal scaling is the same problem in time. The time step must resolve the process: a daily step cannot represent an urban flood peak that rises in 40 minutes, while an hourly step over a 50-year continuous simulation is unnecessarily expensive. The step must also satisfy numerical stability — the Courant condition in hydraulic routing, and the requirement in Muskingum routing that $2KX \le \Delta t \le 2K(1-X)$.
These considerations drive model selection. The appropriate model is the simplest one whose spatial and temporal resolution matches both the decision being made and the data available. A lumped event model suffices to size a culvert on a small basin; a semi-distributed continuous model is needed for reservoir yield; a fully distributed hydraulic model is warranted for floodplain mapping or dam-break analysis. Choosing a model finer than the data can support produces an over-parameterised system in which many parameter sets fit the record equally well — the equifinality problem — giving false confidence and unreliable prediction.
