Water retention curve

Due to the hysteretic effect of water filling and draining the pores, different wetting and drying curves may be distinguished.

As θ decreases, binding of the water becomes stronger, and at small potentials (more negative, approaching wilting point) water is strongly bound in the smallest of pores, at contact points between grains and as films bound by adsorptive forces around particles.

Sandy soils will involve mainly capillary binding, and will therefore release most of the water at higher potentials, while clayey soils, with adhesive and osmotic binding, will release water at lower (more negative) potentials.

[2] It was measured and made for six soils varying in texture from sand to clay.

The data came from experiments made on soil columns 48 inch tall, where a constant water level maintained about 2 inches above the bottom through periodic addition of water from a side tube.

[4][5] The accuracy of the estimated parameters will depend on the quality of the acquired dataset (

Structural overestimation or underestimation can occur when water retention curves are fitted with non-linear least squares.

In these cases, the representation of water retention curves can be improved in terms of accuracy and uncertainty by applying Gaussian Process regression to the residuals obtained after non-linear least-squares.

This is mostly due to the correlation between the data points, which is accounted for with Gaussian Process regression through the kernel function.