Four short, hands-on modules built from the course notes. Each one lets you play with the physics — drag gauges, break hydrogen bonds, propagate errors — then check your understanding with quick checkpoints and a scored module quiz.
Evaristo Critical Zone Hydrology Lab · University of Georgia
Scores are for self-assessment only and are not recorded. Progress resets if you reload the page.
Hydrosystems are absurdly complex — so hydrologists cheat. We draw an imaginary box, count what crosses its boundary, and let one theorem generate every governing equation we need.
Hydrosystem processes transform the space and time distribution of water — in watersheds, pipes, aquifers, even trees. What they all share is the same physical laws, and the consistent way to develop those laws is the control volume approach.
A system is whatever we choose to analyze. Its boundary is the control surface (CS) — the surface enclosing the control volume (CV). The CS may follow real physical boundaries (a pipe wall, a watershed divide), but parts of it can be purely hypothetical surfaces that fluid flows through. That freedom is the whole trick: put the imaginary surfaces where the accounting is easiest.
To apply laws written for discrete lumps of mass to fluid streaming through a control volume, we split every property into two flavors. Extensive properties scale with the total mass in the system: mass \(m\), momentum \(m\mathbf{v}\), and energy \(E\). Intensive properties are per-unit-mass and independent of how much fluid you have.
So the intensive partner of mass is \(\beta = 1\) (mass per unit mass), of momentum it’s velocity \(\mathbf{v}\), and of energy it’s energy per unit mass \(e\).
The volume rate of flow past an area is the dot product of velocity and the area vector, \(Q = \mathbf{V}\cdot\mathbf{A}\) (Eq. 1.2), with \(\mathbf{A}\) pointing outward from the control volume. That sign convention does real work: outflows count positive, inflows negative, so summing the dot product over the whole control surface gives the net outflow:
Multiply by density and you’re counting mass; multiply by \(\beta\rho\) and you’re counting any extensive property riding along with the flow (Eqs. 1.4–1.6). Try it on a pipe that changes size:
Watch a blob of fluid (the system) for a moment \(\Delta t\): some of what it carries is still inside the control volume, some has escaped, some new fluid has arrived. Taking the limit \(\Delta t \to 0\) (Eqs. 1.7–1.11) gives the general control volume equation, better known as the Reynolds transport theorem. It says the total rate of change of an extensive property of a flow equals the rate of change stored in the CV plus the net outflow through the CS.
Click each term to unpack it, then switch the property \(B\) and watch the same theorem become three different laws of physics:
Set \(B=m\), \(\beta=1\), and use conservation of mass (\(dm/dt = 0\)). For constant-density unsteady flow, calling the stored volume \(S\) and writing net outflow as outflow minus inflow, \(Q(t) - I(t)\), the Reynolds transport theorem collapses into the storage equation used all over hydrology:
Mixed formats — multiple choice, numeric entry, true/false, and one predict-then-check tied to the widgets above. Answer everything, then submit. Explanations appear after grading.
Every measurement, no matter how carefully made, differs from the “true” value. Good hydrologists don’t pretend otherwise — they quantify the doubt and carry it honestly through every calculation.
Before the error theory, one deceptively simple tool. The residence time \(T_R\) of a reservoir is the average time a “parcel” of water spends inside it. For a reservoir in steady state, divide average storage by average throughflow:
Storage acts like a shock absorber: the longer the residence time, the less variable and the more persistent the outflow becomes relative to the inflow. That’s why streams fed by big groundwater reservoirs run steadier than flashy, storage-poor catchments — and why we build water-supply reservoirs at all.
A measurement is written \(x_{meas} = x \pm \delta x\) (Eq. 1.28), where \(\delta x\) is the error — the bounds within which we’re reasonably confident the true value lies. Errors come in two very different flavors:
Systematic error (bias) pushes measurements consistently high or low. It hides well and can be serious: virtually all point precipitation methods undermeasure — globally by about 12%. There is no general formula to fix bias; you have to hunt it down by analyzing the method itself.
Random error is the unpredictable scatter, equally likely high or low. Its likelihood is inversely reflected in the precision of the measurement — and unlike bias, it obeys beautiful statistics.
Random errors follow the normal distribution: repeat a measurement many times and the histogram of results forms the bell curve, centered on the mean \(\mu_x\) with spread measured by the error standard deviation \(\sigma_x\). About 68% of measurements land within \(\pm 1\sigma_x\); the multiplier \(k(p)\) tells you how many standard deviations correspond to any confidence level \(p\):
You rarely get to repeat a field measurement hundreds of times, so agencies publish generally accepted precisions. The USGS rates individual discharge measurements at 68% confidence, and average annual streamflow at 95% confidence:
| Quality | Individual measurement ε% (p = 0.68) | Annual average ε% (p = 0.95) |
|---|---|---|
| “Excellent” | 2 | 5 |
| “Good” | 5 | 10 |
| “Fair” | 8 | 15 |
| “Poor” | > 8 | > 15 |
Use the widget in section 3 to turn any rating into a full confidence statement. For repeated measurements, the error standard deviation of the mean shrinks with sample size: \(\sigma_{\mu x} = \sigma_x / N^{1/2}\) (Eq. 1B1.1). For counted events (floods, rainy days), uncertainty comes free: \(\sigma_n = \mu_n^{1/2}\) (Eq. 1B1.10).
Any quantity computed from measurements inherits their uncertainty. For \(y = f(x_1, x_2, \ldots, x_m)\) with independent, random, normally distributed errors:
The classic application: estimating evapotranspiration as \(\mu_{ET} = \mu_P - \mu_Q\). Because ET is a difference, its error standard deviation is the quadrature sum \(\sigma_{\mu ET} = [\sigma_{\mu P}^2 + \sigma_{\mu Q}^2]^{1/2}\) (Eq. 1.40) — geometrically, the hypotenuse of a right triangle whose legs are the two measurement errors. Explore the Big Nemaha example from the notes:
Answer everything, then submit. Keep a calculator handy — a couple of questions want numbers, exactly like the exam will.
Rain gauges measure points; hydrology needs areas. Turning a handful of gauge readings into one regional average is a choice of weights — and different choices give different answers.
The spatial average of rainfall over a region with \(G\) gauges is estimated as a weighted average of the measured values:
The arithmetic mean method is the laziest defensible choice: \(w_g = 1/G\) for every gauge (Eq. 2). The Thiessen method weights each gauge by the fraction of the region closest to it. The isohyetal method goes further: interpolate a whole rainfall surface \(\hat{p}(x,y) = F_1(p_1, p_2, \ldots, p_G)\) (Eq. 3) over the region, then average the surface. Because the interpolated values define a surface — drawn with contours of equal precipitation called isohyets — interpolation methods are also called surface-fitting methods.
scatteredInterpolant, smooth, great for irregular gauge networks), linear (griddata, triangulates and interpolates within each triangle — fine when gauges are evenly spaced and rainfall varies gradually), and inverse distance weighting (IDW). The lab below is the same idea, live.
Drag the gauges. Edit their readings. Switch methods and watch both the map and the regional estimate \(\hat{P}\) respond. In IDW mode, click anywhere on the map to probe the interpolated value and see the weights at that spot.
IDW estimates a grid point \(u\) from every gauge \(g\), weighting by inverse distance raised to an exponent \(c\):
The exponent \(c\) is chosen by the analyst — typical values are \(c=1\) (inverse distance) or \(c=2\) (inverse distance squared). Small \(c\): distant gauges keep real influence, surfaces are smooth and “regional.” Large \(c\): the nearest gauge dominates, and the surface develops bullseyes. As \(c \to \infty\), every point simply takes its nearest gauge’s value — IDW quietly turns into Thiessen. In the X1 MATLAB function, \(c\) is called \(p\), the power parameter.
IDW shines when gauges are sparse or rainfall varies over short distances — but it needs a reasonable gauge density to avoid unrealistic estimates far from any gauge.
Weighted averages, distances, and method judgment calls. Answer everything, then submit.
Water is a deeply weird liquid — and every one of its anomalies traces back to a bent molecule with sticky, charged ends. This module connects the molecular story to the properties hydrologists use daily.
Two hydrogens share electrons with one oxygen through very strong covalent bonds — but they attach on one side, about 105° apart. The lopsided geometry gives the molecule a positive end (the hydrogen side) and a negative end, like a tiny magnet. These polar molecules attract each other end-to-end, forming hydrogen bonds — only about one-twentieth the strength of a covalent bond, yet absent in most other liquids, and the root cause of nearly every strange thing water does.
Compare water to the hydrides of its chemical cousins — H2S, H2Se, H2Te. All are Group VIa hydrides; all except water are nearly symmetrical, nonpolar molecules. Without strong intermolecular forces, melting and boiling temperatures should simply rise with molecular weight. They do — except, strikingly, for H2O. Toggle the hydrogen bonds off to see the world we’d live in without them:
Below 0°C, hydrogen bonds lock molecules into a hexagonal crystal lattice — open and roomy, which is why ice is only 91.7% as dense as liquid water at 0°C. Melting breaks ~15% of the hydrogen bonds and the lattice partially collapses, so the liquid actually packs tighter than the solid. Freezing releases the latent heat of freezing/melting, \(\lambda_f = 3.34\times 10^{5}\ \mathrm{J/kg}\); the same energy must be absorbed to melt each kilogram of ice.
Liquid water keeps the strangeness going: warm it from 0°C and density increases until the maximum at 3.98°C, then decreases like a normal liquid. Slide the temperature:
Pure water can also be supercooled to as low as −41°C if no ice or impurities are present — freezing needs a template. Ice particles or clay minerals (whose crystal structure resembles ice) act as growth nuclei that trigger freezing at 0°C. Supercooling is common in clouds and matters for how raindrops and snowflakes form; fast-flowing river reaches supercool by 0.01–0.1°C while forming ice.
Density, viscosity, surface tension, heat capacity, latent heat of vaporization — all vary with temperature, some steeply. The curves below use the empirical equations from the notes (Heggen 1983). Watch viscosity in particular: water at 0°C is nearly twice as viscous as at 25°C, which matters for everything from settling velocities to groundwater flow.
Two properties deserve their own headlines. Surface tension (\(\sigma = 0.0756\) N/m at 0°C, higher than almost any other liquid) plus attraction to mineral surfaces produces capillarity — the reason unsaturated soils hold and move water the way they do; the contact angle on most silicate minerals is essentially 0°. Heat capacity (\(c_p = 4{,}216\) J/kg·K at 0°C) is enormous because added heat goes into breaking hydrogen bonds instead of speeding molecules up — it lets warm-blooded organisms regulate temperature and makes oceans the planet’s thermostat.
Near a boundary, hydrogen bonds make water “stick” (the no-slip condition), and friction propagates into the flow as viscosity. Slow, thin flows slide in parallel layers (laminar); faster, deeper flows break into irregular eddies (turbulent), which add eddy viscosity far exceeding molecular viscosity. Which regime you’re in is decided by the Reynolds number — a ratio of inertia to viscous friction:
Consequences: pore flows are almost always laminar — Darcy’s law only applies then. Virtually all stream flows, even small ones, are turbulent. Overland flow can be either, or in between. Note the temperature slider: colder water is more viscous, so the same flow can change regime with the seasons.
A tiny fraction of water molecules dissociate into H+1 and OH−1 ions. Acidity is measured as \( \mathrm{pH} \equiv -\log_{10}([\mathrm{H}^{+1}]) \) (Eq. B.1): pure water sits at 7.00, cloud water equilibrated with atmospheric CO2 at about 5.7, and natural rain at 4.5–5.6 depending on location. pH controls water’s propensity to dissolve many elements.
Isotopes are heavier or lighter versions of the same element. About 99.8% of water is ordinary 1H216O; the rare heavy species — deuterium (2H, D) and 18O — are fractionated during phase changes: lighter isotopes evaporate more readily, heavier ones condense more readily. Composition is expressed relative to standard mean ocean water (SMOW):
So precipitation is depleted (\(\delta < 0\)), increasingly so toward the poles as rainout strips the heavy isotopes; evaporating lakes and oceans are relatively enriched (\(\delta > 0\)). Uptake by plants and transpiration do not fractionate — all water reaching the stomata evaporates. Radioactive tritium (3H, half-life 12.5 yr) once dated groundwater thanks to bomb-test fallout, a clock that has since wound down. These signatures let hydrologists trace where the water in a stream, glacier, or aquifer actually came from.
From molecular structure to Reynolds numbers. Answer everything, then submit.