From back-of-envelope Glover calculations to a custom MODFLOW model, the available methods span four orders of magnitude in effort. Picking the right tool first is usually more important than running the wrong one carefully.
Closed-form formulas (Glover, Jenkins, Hantush, Hunt). One well, one stream, simple aquifer. Hours of effort. Pages 02–03 here.
Strength: transparent, fast, defensible for screening.
Limit: only as good as the assumptions.
Adds finite stream width, partial penetration, leaky/wedge aquifers, two-layer systems, meandering, finite extent (Butler & Tsou, Chen, Huang, Yeh, Hunt 2003/2009).
Strength: handle one or two relaxed assumptions.
Limit: still single-well / single-stream.
Analytical core + geometric apportionment across a stream network (Zipper et al.). Page 04 here. Scales to hundreds of wells in seconds.
Strength: basin-scale screening; many scenarios.
Limit: heterogeneity, disconnection.
MODFLOW + SFR/LAK packages, IWFM, ParFlow, HydroGeoSphere. Heterogeneous aquifers, transient stress, all the streams and ET in one place.
Strength: handles real complexity.
Limit: data hungry, slow, parameter uncertainty.
Differential gaging, seepage meters, temperature methods, isotopes, pumping tests interpreted for depletion (Lough & Hunt 2006). Page 07.
Strength: measure reality, not model it.
Limit: point-in-time, expensive to spatialize.
Calibrate analytical or numerical with monitoring data; machine learning on baseflow records; data assimilation for transient calibration.
Strength: tightens uncertainty; useful for adaptive management.
Limit: hard to extrapolate; needs records.
Describe your problem at left. The bar chart on the right gives a heuristic "fit score" for each method family. The recommendation is suggestive — not prescriptive. Use it to start a method-selection conversation; verify with the GWSW-MST framework (Hammett et al. 2022).
| Method | Effort | Data | Many wells/streams | Heterogeneity | Disconnection | Defensibility | Typical use |
|---|---|---|---|---|---|---|---|
| Analytical (Glover, Hunt) | Hours | 3–5 params | One well, one stream | Uniform only | Not handled | High for screening | First-pass; permit screening |
| Semi-analytical (Butler, Chen, Huang) | Days | 5–8 params | One/two wells | One relaxed assumption | Limited | Medium-high | When a specific assumption matters |
| ADF (Zipper et al.) | Weeks | Param fields + well locs | Hundreds of wells | Partial via zones | Not directly | Medium-high for screening | GSP-scale rapid analysis |
| Numerical (MODFLOW/IWFM) | Months to year+ | Full hydrostrat, BC, pumping | Unlimited | Full | With UZF/SFR packages | High when calibrated | Regulatory / 5-yr update |
| Field measurements | Days–years (campaign) | Field measurements | Point/reach by reach | Direct observation | Direct observation | Highest where measured | Calibration; ground-truth |
| Hybrid / data-driven | Varies | Long monitoring records | Basin scale | Implicit | Implicit | Medium; extrapolation risk | Adaptive management |
An open, structured framework that asks practitioners to characterize their problem along five axes — system type, conceptual model, study goal, data resources, and study scale — and returns a shortlist of appropriate methods.
Built specifically for GW–SW interaction problems, drawing on the catalog from Rassam & Werner (2008), Rathfelder (2016), Huang et al. (2018), and field-method reviews.
The matching algorithm is published open-source — see Hammett et al. (2022) for the decision tables.
A workflow that has held up well across California basins: