Heat from wildfires can produce a two-layer system composed of extremely dry soil covered by a layer of ash, which when subjected to rainfall, may produce extreme floods. To understand the soil physics controlling runoff for these initial conditions, we used a small, portable disk infiltrometer to measure two hydraulic properties: (1) near-saturated hydraulic conductivity, K_f and (2) sorptivity, S(θ_i), as a function of initial soil moisture content, θ_i, ranging from extremely dry conditions (θ_i < 0.02 cm³ cm⁻³) to near saturation. In the field and in the laboratory replicate measurements were made of ash, reference soils, soils unaffected by fire, and fire-affected soils. Each has a different degrees of water repellency that influences K_f and S(θ_i).

Values of K_f ranged from 4.5 × 10⁻³ to 53 × 10⁻³ cm s⁻¹ for ash; from 0.93 × 10⁻³ to 130 × 10⁻³ cm s⁻¹ for reference soils; and from 0.86 × 10⁻³ to 3.0 × 10⁻³ cm s⁻¹, for soil unaffected by fire, which had the lowest values of K_f. Measurements indicated that S(θ_i) could be represented by an empirical non-linear function of θ_i with a sorptivity maximum of 0.18–0.20 cm s^−0.5, between 0.03 and 0.08 cm³ cm⁻³. This functional form differs from the monotonically decreasing non-linear functions often used to represent S(θ_i) for rainfall–runoff modeling. The sorptivity maximum may represent the combined effects of gravity, capillarity, and adsorption in a transitional domain corresponding to extremely dry soil, and moreover, it may explain the observed non-linear behavior, and the critical soil-moisture threshold of water repellent soils. Laboratory measurements of K_f and S(θ_i) are the first for ash and fire-affected soil, but additional measurements are needed of these hydraulic properties for in situ fire-affected soils. They provide insight into water repellency behavior and infiltration under extremely dry conditions. Most importantly, they indicate how existing rainfall–runoff models can be modified to accommodate a possible two-layer system in extremely dry conditions. These modified models can be used to predict floods from burned watersheds under these initial conditions.