How can we predict the probability distribution of peak outflows from a detention dam without running thousands of Monte Carlo simulations? This work answers that question with a closed-form, theoretically derived probability distribution (TDD) of peak outflows from in-line detention dams, validated across the full Generalized Extreme Value (GEV) family. The approach was presented at the EWaS5 International Conference (Naples, 2022) and published in Environmental Sciences Proceedings.
Quick links: Framework · GEV validation · Applications · MATLAB code · How to cite · FAQ
Why an analytical distribution of detention dam outflows matters
Detention dams are among the most effective hydraulic structures for flood mitigation. Yet most flood frequency analyses still treat reservoirs as a black box, or rely on computationally expensive Monte Carlo simulations to quantify their downstream effect. As climate change, soil sealing and the proliferation of small in-line basins reshape catchment dynamics, the lack of an analytical, distribution-agnostic framework for outflow probability becomes a real bottleneck for risk-based design.
A theoretically derived probability distribution (TDD) closes that gap. Once the inflow flood-peak distribution is known, the outflow distribution can be computed analytically — no simulation loop required, no proprietary solver. The full mathematical derivation appears in Manfreda, Miglino & Albertini (2021), Hydrology and Earth System Sciences; this paper extends the validation to the entire GEV family.
The TDD framework: three regimes of detention dam outflow
The peak outflow probability density splits naturally into three physical regimes, each tied to the discharge level relative to the design control discharge Qc of the low-level opening:
1. Undisturbed flow regime (Qp,out < Qc)
The outflow distribution coincides with the inflow distribution — small floods pass through the bottom outlet undisturbed and the dam has no measurable mitigation effect.
2. Reservoir filling regime (Qp,out = Qc)
A probability mass concentrates exactly at Qc: as long as the storage volume Wmax is filling, the outflow is clamped to the bottom-outlet design discharge. This is where most of the flood-mitigation benefit lies for moderate events.
3. Spillway activation regime (Qp,out > Qc)
Once the water level reaches the crest spillway elevation hs, the outflow follows a transformed distribution — TDD3 — derived from the inverse of the inflow–outflow functional relationship. This is the regime that controls residual risk for the largest design events.
The framework is distribution-agnostic: any inflow probability distribution used in flood frequency analysis can be plugged in (Gumbel, Fréchet, Weibull, log-normal, Pearson III, etc.), making the method directly transferable to most case studies.
Validation across the GEV family: Gumbel, Fréchet, Weibull
We tested the TDD with a Generalized Extreme Value (GEV) distribution for the inflow peaks, varying the shape parameter ξ to span the three asymptotic sub-families:
- Gumbel (ξ = 0) — light-tailed, EV Type I
- Fréchet (ξ = 0.5) — heavy-tailed, EV Type II
- Weibull (ξ = −0.5) — bounded, EV Type III
Across all three families, the analytical TDD reproduces the empirical pdf obtained via numerical hydraulic simulations almost perfectly — both in the bulk and in the tails, exactly where flood-risk design is most sensitive. The agreement confirms that the closed-form solution is a robust replacement for simulation-based outflow analysis on small detention basins with impulsive hydrological response.
Practical applications for flood-risk engineers and modellers
- Sizing of small detention dams — analytical design of spillway geometry and storage volume against a target downstream return period, without simulation loops.
- Flood-risk reassessment of existing infrastructure — quantify residual risk downstream of legacy dams and identify retrofit priorities.
- Optimization workflows at catchment scale — the TDD has been combined with geomorphic flood mapping to optimize detention-basin networks across entire catchments.
- Non-stationary climate scenarios — because any inflow distribution is admissible, the framework absorbs climate-adjusted or trend-corrected flood distributions out of the box.
- Probabilistic assessment of dam break or overtopping risk — the explicit probability mass at Qc and the spillway-regime tail support direct return-period reasoning.
Open MATLAB code & reproducibility
The MATLAB implementation of the peak-outflow TDD is freely available — open code, fully reproducible:
👉 Download: Peak outflows of a detention basin on MATLAB File Exchange
👉 HydroTool page on this site — usage notes, parameters, examples
Related research and the broader TDD programme
This work belongs to a wider research line on theoretically derived probability distributions in hydrology, developed within the HydroLAB group:
- Manfreda, Miglino & Albertini (2021) — the foundational HESS paper introducing the closed-form inflow–outflow relationship.
- Manfreda, Link & Pizarro (2018) — TDD of bridge-pier scour.
- Gioia, Iacobellis, Manfreda & Fiorentino (2008) — runoff thresholds in derived flood frequency distributions.
- Stochastic Hydrology research area on this site.
The work was developed within the MATCAS project — Hydraulic risk mitigation in coastal basins with in-line expansion tanks — funded by the Italian Ministry of Environment (CUP E68D20000010001). See the dedicated MATCAS project page.
How to cite this work
Manfreda, S.; Miglino, D.; Albertini, C. The Theoretical Probability Distribution of Peak Outflows of Small Detention Dams. Environmental Sciences Proceedings 2022, 21(1), 90.
https://doi.org/10.3390/environsciproc2022021090
FAQ
What is a theoretically derived probability distribution (TDD) of detention dam outflows?
A TDD is a closed-form probability distribution of peak outflows from a detention dam, derived analytically from the inflow flood-peak distribution and the dam’s hydraulic equations — without Monte Carlo simulation. It splits the outflow regime into three phases: undisturbed flow, reservoir filling, and crest spillway activation.
Which inflow probability distributions are compatible with this method?
The framework is distribution-agnostic. Any flood-peak probability distribution used in flood frequency analysis can be plugged in, including the full GEV family (Gumbel, Fréchet, Weibull), log-normal, Pearson III and others.
Where can I download the code to compute the peak-outflow TDD?
The MATLAB implementation is freely available on MATLAB Central File Exchange under the title Peak outflows of a detention basin.
Is the method valid for large reservoirs as well?
The formulation is intended for small in-line detention basins on river basins with an impulsive hydrological response. For large reservoirs with complex operating rules, the rectangular-hydrograph hypothesis becomes restrictive and a more detailed routing model is recommended.
How does the TDD compare with Monte Carlo simulation?
Across the three GEV sub-families tested (Gumbel, Fréchet, Weibull), the analytical TDD matches Monte Carlo / numerical hydraulic simulations very closely in both the bulk and the tail of the outflow distribution, while running orders of magnitude faster and providing closed-form sensitivity to the design parameters.
