Snyder vs Clark Unit Hydrograph: When to Use Each

Choosing a synthetic unit hydrograph transform usually comes down to a single question — do you have regional coefficients or do you have GIS terrain — and the Snyder and Clark methods sit on opposite answers to it. This decision guide contrasts the two so you can pick deliberately rather than by habit. It is a focused comparison under the Unit Hydrograph Methods guide, part of the wider Rainfall-Runoff Modeling & Hydrologic Simulation coverage on this site. Both methods produce a unit hydrograph that you then convolve with excess rainfall exactly as the parent guide describes; the difference is entirely in how the ordinates are generated and what data that generation demands.

Snyder is empirical and lumped: two calibrated regional coefficients fix the peak and lag, and the rest of the shape is inferred. Clark is conceptual and distributed: rainfall is translated along basin isochrones and then routed through a linear reservoir. The right choice depends on data availability, basin character, and the modelling platform you are targeting.

The Snyder Synthetic Unit Hydrograph

Snyder’s method was developed from gauged Appalachian basins and defines the UH through relationships driven by two regional coefficients. The basin lag is:

tp = Ct * (L * Lca)^0.3

where L is the main-channel length to the divide, Lca is the length along the channel to the point nearest the basin centroid, and Ct is a regional coefficient (commonly 1.8 to 2.2, higher in flatter regions). The peak discharge per unit runoff is:

qp = C * Cp * A / tp

with Cp a peaking coefficient (roughly 0.4 to 0.8) and C a unit constant (640 English, 2.75 SI). Snyder also gives the standard duration tr = tp/5.5 and the time base, plus empirical widths at 50 and 75 percent of peak to sketch the shape between the defined points. Its defining strength and weakness are the same: Ct and Cp are regional. Transferred from a calibrated gauged basin to a nearby ungauged one they give a defensible hydrograph with almost no data; used far outside their region of derivation they are guesses. Snyder needs no terrain grid at all, which is precisely why it survives as an ungauged-basin workhorse.

The Clark Time-Area Unit Hydrograph

Clark’s method builds the hydrograph from basin geometry in two conceptual stages. First, translation: a time-area histogram describes how much basin area contributes to the outlet at each travel time, delineated by isochrones (lines of equal travel time to the outlet). Rainfall is routed along this histogram, moving each area’s contribution to the outlet with pure delay. Second, attenuation: the translated hydrograph passes through a single linear reservoir whose storage S relates to outflow O by S = R·O, where R is the storage coefficient with units of time. The linear-reservoir routing is a simple recursion:

O_i = Ca * I_i + Cb * O_(i-1),   Ca = dt / (R + 0.5*dt),   Cb = 1 - Ca

Clark therefore has two parameters: the time of concentration Tc (the longest travel time, which sets the width of the time-area histogram) and the storage coefficient R (which controls attenuation). The time-area histogram is ideally derived from a GIS flow-length or travel-time grid, the same kind of terrain analysis used in watershed delineation and catchment synchronization. Because the shape comes from real basin geometry rather than an empirical template, Clark represents elongated, branching, or storage-dominated basins more faithfully than a lumped template can. The cost is the GIS work needed to build the time-area curve.

Head-to-Head Comparison

Decision Guidance

Use Snyder when you are working an ungauged basin inside a region where Ct and Cp have been established by a gauged-basin study or a state hydrology manual, and you have neither the time nor the terrain data to build a time-area histogram. It is the fastest defensible answer for small-to-medium basins whose response is dominated by lag rather than by internal storage or unusual shape, and it is the natural companion to the single-parameter SCS dimensionless unit hydrograph when you want a second empirical opinion on the peak.

Use Clark when basin geometry genuinely matters — elongated basins, strong internal storage, or wetlands and reservoirs that attenuate the flood wave — and you have or can build a GIS time-area curve. Because R explicitly represents storage, Clark also transfers between basins more physically than Snyder’s regional coefficients, and it underpins the gridded ModClark method used with radar rainfall on large basins. If you only have Tc and R but no GIS histogram, Clark still runs on a synthetic time-area curve, but at that point the practical gap between Clark and a lumped method narrows.

Both transforms are configured as subbasin transform methods in HEC-HMS, so the choice is not constrained by tooling; automating those runs is covered under HEC-HMS Python automation. Whichever you choose, calibrate the parameters against observed storms rather than trusting first-guess coefficients, using the objective functions described in model calibration & objective functions.

A quick way to frame the choice: if the dominant uncertainty in your basin is timing — how long water takes to reach the outlet — and you have a regional lag study, Snyder is efficient. If the dominant uncertainty is shape and attenuation — how a flood wave spreads and flattens as it drains through storage — and you have terrain data, Clark rewards the extra effort. On a basin that is both gauged and geometrically ordinary, the two methods often converge to similar peaks once calibrated, and Snyder wins on simplicity.

Estimating the Parameters in Practice

For Snyder, measure L and Lca directly from the channel network — L is the length along the main channel from the outlet to the basin divide, and Lca is the length along that same channel to the point nearest the basin centroid. Both come straight from a stream-network layer. The coefficients Ct and Cp are the hard part: pull them from a regional regression study, a state drainage manual, or by back-calculating tp and qp from a gauged storm on a neighbouring basin of similar physiography. A single well-observed event on a nearby gauge is worth more than any tabulated default.

For Clark, Tc is the travel time of the hydraulically most distant point, estimated by the segmental velocity method or an empirical Tc equation, and the time-area histogram is built by binning basin area against travel time from a flow-length grid. The storage coefficient R is rarely known a priori; the usual practice is to fix the regional ratio R/(Tc+R) — often between 0.5 and 0.7 — and solve for R once Tc is set, then refine R by calibration. Because R and Tc trade off against each other, calibrating both freely against one storm overfits; anchoring the ratio keeps the fit physically meaningful.

Gotchas

• Snyder coefficients are not universal. Ct and Cp derived in one physiographic region can be badly wrong in another. Never import a textbook value without checking it against a regional study; a wrong Ct shifts the entire hydrograph in time.
• Clark Tc and R are correlated. The ratio R/(Tc+R) tends to be roughly constant within a region, so calibrating Tc and R independently against a single storm is ill-posed. Constrain the ratio from regional data and calibrate one parameter.
• Time-area orientation errors. A time-area histogram built from a flow-length grid with an inverted or mis-registered isochrone set will translate rainfall to the wrong travel times, distorting the Clark shape even when Tc and R are correct.
• Duration consistency still applies. Both methods produce a UH of a specific duration. As with any unit hydrograph, the excess-rainfall pulse duration must match it or be reconciled through the S-curve before convolution.
• Do not double-count storage. If you apply Clark attenuation and then route the outflow through a separate reservoir model, you may attenuate the same storage twice. Keep the linear-reservoir R consistent with any downstream routing.

Related Topics

• Unit Hydrograph Methods — parent guide covering LTI theory, convolution, and the S-curve both methods feed into
• Deriving SCS Dimensionless Unit Hydrographs in Python — the single-parameter empirical alternative to contrast against Snyder and Clark
• HEC-HMS Python Automation — where Snyder and Clark transforms are configured and batch-run
• Model Calibration & Objective Functions — calibrating Ct, Cp, Tc, and R against observed storms