Artificial neural networks for estimating soil hydraulic parameters from dynamic flow experiments

Schmitz, Gerd; Puhlmann, Heike; Dröge, Werner; Lennartz, Franz

doi:10.1111/j.1351-0754.2004.00631.x

Cited by 14 publications

(7 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The flow experiment is numerically simulated based on the hydraulic functions defined in Eqs. 1 to 6, and using our own numerical code RIAN (Schmitz et al, 2005). An optimization algorithm (in our case, the Annealing-Simplex approach of Pan and Wu, 1998) estimates the unknown soil parameters through minimization of the difference between observed and simulated flow variables (residuals) involving an iterative solution of the Richards equation (Richards, 1931).…”

Section: Multistep Outflow Experiments and Inverse Parameter Optimizamentioning

confidence: 99%

See 1 more Smart Citation

Pedotransfer functions for water retention and unsaturated hydraulic conductivity of forest soils

Puhlmann

Wilpert²

2012

Z. Pflanzenernähr. Bodenk.

View full text Add to dashboard Cite

The hydraulic properties of soils, i.e., their ability to store and conduct water, mainly govern the availability of soil water for plants. Information on the hydraulic properties is needed, e.g., for the quantification of drought risk at a given site. Furthermore, knowledge of the water transport is the precondition for the estimation of element fluxes in the soil, e.g., when predicting element leaching from the root zone to the groundwater. For forest soils, only few systematic investigations of their hydraulic properties exist. Within the 2nd forest-soil survey of Germany, soil samples were taken along a regular 8 km 8 km grid in the forests of the State of BadenWürttemberg and the hydraulic properties were estimated in the laboratory by multistep outflow experiments. Besides the soil-hydraulic measurements, numerous additional soil chemical and physical analyses were carried out and comprehensive profile descriptions were compiled and integrated in a hydraulic database. Based on this database, multiple-linear-regression techniques were used to develop pedotransfer functions for the water-retention curve and the unsaturated-hydraulic-conductivity curve using the parametric models of Mualem/van-Genuchten. Our work fills a gap since to our knowledge, no pedotransfer functions for the unsaturated hydraulic conductivity for forest soils exist so far. The predictive accuracy of the established pedotransfer functions, both for the water-retention curve and the hydraulic-conductivity curve, is in the range of (and in some cases better than) other published pedotransfer functions that were mostly derived for agricultural soils.

show abstract

Section: Multistep Outflow Experiments and Inverse Parameter Optimizamentioning

confidence: 99%

“…During the experiment, the water volume and the pressure head in the sample were recorded. The Mualem/van Genuchten (MvG) parameters were estimated for all of the MSO experiments using the software RIAN (Schmitz et al, 2005).…”

Section: Data Materialsmentioning

confidence: 99%

Pedotransfer functions for water retention and unsaturated hydraulic conductivity of forest soils

Puhlmann

Wilpert²

2012

Z. Pflanzenernähr. Bodenk.

View full text Add to dashboard Cite

show abstract

“…The Annealing‐Simplex algorithm used for parameter estimation provided unique results in many cases (Schmitz et al , 2005). Possible problems with non‐unique solutions of the inverse routine were handled by 20 repeated optimization runs with varying initial values of the soil hydraulic parameters and varying settings for the Annealing‐Simplex algorithm (e.g.…”

Section: Resultsmentioning

confidence: 99%

“…This resulted in the following weighting factors: g ψ = g V = 1/2 for the soil samples drained in the MSO experiments applying pressure regime BC1 or BC2 (loams and clays), g ψ = 1/3, g V = 2/3 for pressure regime BC3 (loamy sands) and g ψ = 1/4, g V = 3/4 for BC4 (pure sands). The Annealing-Simplex algorithm used for parameter estimation provided unique results in many cases (Schmitz et al ., 2005). Possible problems with non-unique solutions of the inverse routine were handled by 20 repeated optimization runs with varying initial values of the soil hydraulic parameters and varying settings for the Annealing-Simplex algorithm (e.g.…”

Section: Estimated Soil Hydraulic Parametersmentioning

confidence: 99%

Multistep outflow experiments to derive a soil hydraulic database for forest soils

Puhlmann¹,

Wilpert²,

Lukes³

et al. 2009

European J Soil Science

Self Cite

View full text Add to dashboard Cite

Forest soils are under-represented throughout the existing soil hydraulic databases. Classical measurement methods for the hydraulic soil characteristics are very time-consuming and are not suited for routine application in the course of soil survey. We are presently engaged in establishing a database of the hydraulic properties of forest soils using multistep outflow experiments together with an inverse optimization of the Mualem/van Genuchten parameters from the tension and outflow measurements. The lower boundary in the multistep outflow experiments was modified such that, firstly, the experiment can be conducted as fast as possible and hence can be incorporated in the framework of routine soil monitoring. Secondly, the experimental design should be identical for a large texture range and therefore allow for a statistical comparison of the hydraulic properties of a broad variety of soils. Thirdly, the experiments must be designed such that they yield sufficient measurements for a unique and stable inverse parameter estimation. The inverse parameter estimates of the presently available 141 multistep outflow experiments were compared with predictions of two pedotransfer functions. For most soil samples, the PTF predictions differed highly from the MSO estimates. Reasons for this are, among others, insufficient hydraulic measurements included in the datasets used for establishing the PTFs and, specifically, insufficient information on water retention/conductivity in the range of the air entry value. Here, the MSO experiments can provide much more detailed information on the water retention and conductivity curves because they yield measurements over a continuous pressure range and are conducted such that the pressure range of the air entry value is covered by sufficient tension and outflow measurements. The presented research aims to: (i) introduce the measurement of soil hydraulic properties in an ongoing soil monitoring campaign, (ii) complement existing soil hydraulic databases with data from forest soils, and (iii) enhance the reliability of soil hydraulic databases through a standardization of the measuring approaches.

show abstract

“…This is especially true for surface irrigation where the nonlinearity associated with surge flow, and especially with the surge tip, is extreme. An interesting new approach (introduced by Schmitz et al, 2005; Schütze et al, 2005) is to replace the full solution of the subsurface component with a trained neural network. By that, the computational effort is dramatically reduced but the advantages associated with the physical modeling of subsurface flow are essentially maintained.…”

Section: Overview Of Computational Tools and Applicationsmentioning

confidence: 99%

Modeling Coupled Surface–Subsurface Flow Processes: A Review

Furman

2008

Vadose Zone Journal

174

112

View full text Add to dashboard Cite

A: PDE, partial diff erential equation.S S : V Z M Surface and subsurface fl ow systems are inherently unifi ed systems that are o en broken into sec ons for logical (e.g., me scales) and technical (e.g., analy cal and computa onal solvability) reasons. While the basic physical laws are common to surface and subsurface systems, spa al and temporal dimensions as well as the con nuum approach used for the subsurface lead to diff erent formula ons of the governing par al diff eren al equa ons. While in most applica ons such decoupling of the systems works well and allows a very accurate and effi cient descrip on of the individual system by trea ng the adjacent system as a boundary condi on, in the case of water fl ow over a porous medium, it does not. Therefore coupled models are in increasing use in this fi eld, led mostly by watershed and surface irriga on modelers.The governing equa ons of each component of the coupled system and the coupling physics and mathema cs are reviewed fi rst. Three diff erent coupling schemes are iden fi ed, namely the uncoupled (with the degenerated uncoupled scheme being a special case of the uncoupled), the itera vely coupled, and the fully coupled. Next, the diff erent applica ons of the diff erent coupling schemes, sorted by fi eld of applica on, are reviewed. Finally, some research gaps are discussed, led by the need to include ver cal momentum transfer and to expand the use of fully coupled models toward surface irriga on applica ons.F . 1. Rela ons between diff erent components of the hydrologic cycle. Black lines indicate vapor processes, blue lines indicate non-vapor processes. The solid, bold lines are the focus of this discussion.

show abstract

Artificial neural networks for estimating soil hydraulic parameters from dynamic flow experiments

Cited by 14 publications

References 34 publications

Pedotransfer functions for water retention and unsaturated hydraulic conductivity of forest soils

Pedotransfer functions for water retention and unsaturated hydraulic conductivity of forest soils

Multistep outflow experiments to derive a soil hydraulic database for forest soils

Modeling Coupled Surface–Subsurface Flow Processes: A Review

Contact Info

Product

Resources

About