Infiltration determines how much rainfall enters the soil versus running off the surface. Understanding this process and the models that predict it is central to managing water resources, designing drainage systems, and forecasting floods.

This section covers the stages of infiltration from start to steady-state, then walks through the major models used to estimate infiltration rates and their trade-offs.

Infiltration Process Stages

Stages of the infiltration process

Infiltration doesn't begin the instant rain hits the ground. Several things happen first, and the process evolves as the soil wets up.

1. Initial abstraction

Before any water infiltrates, some rainfall is "lost" to processes that never deliver water to the soil surface. Vegetation intercepts rain on leaves and stems. Small depressions on the ground fill up as puddles. Some water evaporates directly. Together, these reduce the volume of water actually available for infiltration. In hydrologic calculations, initial abstraction is often subtracted from total rainfall before computing infiltration or runoff.

2. Ponding

Once initial abstractions are satisfied, water reaches the bare soil surface. If rainfall intensity is less than the soil's infiltration capacity, all the water soaks in. Ponding begins when rainfall intensity exceeds the infiltration capacity. At that point, water accumulates on the surface. The time to ponding depends on both the rainfall rate and the soil's ability to absorb water, so a high-intensity storm on clay soil will pond much faster than a light drizzle on sand.

3. Steady-state infiltration

As water continues to enter the soil, the infiltration rate gradually decreases. Early on, strong capillary (matric suction) forces pull water into dry pore spaces, so the rate is high. Over time, those pores fill, suction gradients weaken, and gravity becomes the dominant driving force. Eventually the infiltration rate levels off at a roughly constant value. This steady-state rate is controlled by the soil's saturated hydraulic conductivity ( $K_s$ ), which reflects how easily water moves through fully saturated soil. Coarse sands might have $K_s$ values above 100 mm/hr, while dense clays can be below 1 mm/hr.

Infiltration Models

Stages of infiltration process, HESS - On the representation of water reservoir storage and operations in large-scale ...

Application of infiltration models

Green-Ampt model

The Green-Ampt model pictures a distinct "piston-like" wetting front moving downward through the soil. Above the front, the soil is saturated; below it, the soil remains at its initial moisture content. This sharp-front assumption makes the math tractable while still capturing the key physics.

The infiltration rate is:

$f(t) = K_s \left(1 + \frac{\psi \, \Delta\theta}{F(t)}\right)$

where:

$K_s$ = saturated hydraulic conductivity (mm/hr)
$\psi$ = wetting front suction head (mm), representing the capillary pull at the front
$\Delta\theta$ = change in volumetric moisture content (initial to saturated, dimensionless)
$F(t)$ = cumulative infiltration at time $t$ (mm)

Notice that as $F(t)$ grows (more water has infiltrated), the fraction $\frac{\psi \, \Delta\theta}{F(t)}$ shrinks, and $f(t)$ approaches $K_s$ . That matches the physical reality: early infiltration is fast because suction is strong, then it tapers to the steady-state rate.

The model works well for relatively homogeneous soils (uniform sands, loams) and is widely used in watershed models like SWMM and HEC-HMS.

Philip's equation

Philip's equation comes from an analytical solution to the Richards equation for one-dimensional vertical infiltration into a uniform soil:

$f(t) = \frac{1}{2} S \, t^{-1/2} + A$

where:

$S$ = sorptivity (mm/hr $^{1/2}$ ), which captures the capillary absorption component
$A$ = a parameter approximating the steady-state infiltration rate (mm/hr)
$t$ = time since infiltration began (hr)

At early times, the $S \, t^{-1/2}$ term dominates because dry soil pulls water in strongly through capillarity. As $t$ increases, that term fades and $f(t)$ approaches $A$ , reflecting gravity-driven flow. This two-term structure gives Philip's equation a clean physical interpretation while remaining simple to apply.

Compared to Green-Ampt, Philip's equation doesn't require the sharp wetting front assumption, so it can handle a broader range of soil moisture profiles. However, it's still limited to homogeneous, deep soil columns.

Empirical vs. physically-based models

These two categories differ in how they're built and what they can do.

Empirical models are fitted to observed infiltration data. They describe the pattern of infiltration decline over time without explicitly representing the underlying physics.

Common examples: Kostiakov ( $f = a \, t^{-b}$ ), Horton (exponential decay from an initial rate to a final rate), and Holtan (relates infiltration to available soil storage).
Advantages: simple to calibrate, require few inputs (basically time-series infiltration data and soil type).
Disadvantages: parameters are site-specific. A Horton equation calibrated on a silty loam in Iowa won't reliably predict infiltration on a laterite soil in West Africa. They also don't explain why infiltration behaves the way it does.

Physically-based models are derived from fundamental principles like Darcy's law and conservation of mass.

Examples: Green-Ampt, Philip's equation, and the full Richards equation.
Advantages: parameters have physical meaning (hydraulic conductivity, porosity, suction head), so they can be estimated from soil surveys or lab tests. They're more transferable across sites and conditions.
Disadvantages: they need detailed soil property data, and solving the Richards equation numerically can be computationally expensive, especially for layered or heterogeneous profiles.

Choosing a model: For quick engineering estimates with local calibration data, empirical models are practical. For ungauged basins or scenario analysis (e.g., "what happens if we change land use?"), physically-based models are more defensible because their parameters connect to measurable soil properties.

Limitations of infiltration models

No infiltration model perfectly represents field conditions. Here are the main limitations to keep in mind.

Green-Ampt model

The sharp wetting front assumption breaks down in layered soils. If a clay layer sits above a sand layer, the wetting front doesn't behave like a clean piston.
It ignores preferential flow through macropores, root channels, and cracks, which can deliver water to depth much faster than matrix flow alone.
In heterogeneous soils with spatially variable properties, the model tends to overestimate infiltration because it assumes uniform conductivity throughout.

Philip's equation

Assumes a semi-infinite, homogeneous soil column. Real soils have layers, and a shallow water table or restrictive horizon violates this assumption.
The two-term form is most accurate at relatively short times. Over long durations, the approximation can drift from the true Richards equation solution.
It doesn't handle ponded-surface boundary conditions well; the derivation assumes a constant supply of water at the surface, which may not match real storm patterns.

Empirical models (Kostiakov, Horton, Holtan)

Calibrated parameters are tied to the specific site, soil, and conditions of the original dataset. Extrapolating to different soils or climates is risky.
They don't account for soil physical properties explicitly, so they can't predict how infiltration would change if, say, soil structure degraded due to compaction.
Performance degrades for very long events or extreme conditions (major floods, prolonged droughts altering soil structure) that fall outside the calibration range.

2,589 studying →