Skip to main content
ARS Home » Northeast Area » Beltsville, Maryland (BARC) » Beltsville Agricultural Research Center » Environmental Microbial & Food Safety Laboratory » Research » Publications at this Location » Publication #281887
Title: Ensemble modeling with pedotransfer functions in the hydropedological context

Author
item Pachepsky, Yakov
item PAN, FENG - University Of Utah
item GUBER, ANDREY - Michigan State University
item YAKIREVICH, ALEXANDER - Ben Gurion University Of Negev
item JACQUES, DIEDERIK - Belgian Nuclear Research
item Gish, Timothy

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 5/29/2012
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Uncertainty of soil water content and/or soil water flux estimates with soil water models has recently become of a particular interest in various applications. This work provides examples of using pedotransfer functions (PTFs) to build ensembles of models to characterize the uncertainty of simulation results. Example I consists in comparison of ensemble predictions of soil water contents in field soil with ensembles built with laboratory water retention data and with PTF estimates. Data on soil water contents and pressure heads in layered loamy soil were collected during a year in Belgium. The ensemble of PTFs represented field water retention better than the laboratory data. PTF ensemble predictions were about two times more accurate compared with laboratory data-based ensembles. Using PTF ensembles appeared to be a viable method to downscale PTF-represented soil hydraulic properties for field-scale simulations purpose. Ensemble upscaling of the laboratory data resulted in less accurate simulations. Example 2 consists in evaluating PTFs as component of the data assimilation from soil water content sensors into soil water model. The dataset was the same as above. We evaluated the performance of the Ensemble Kalman Filter for the case when the ensemble members used different PTFs. Accounting for the temporal stability of water contents substantially decreased the estimated noise in data. Assimilating measurements from a single depth provided improvements at all other observation depths, i.e. point measurements were upscaled to the whole soil profile. Example 3 has had the objective of implementing and testing an information theory-based method for the sequential design of the observation network augmentation to discriminate between hydraulic models. The method was tested with the data from the tracer experiment at the USDA-ARS OPE3 integrated research site. Two conceptualizations of soil cover structure – layered and layered with lenses - were compared. A 3D flow and transport model was manually calibrated, and PTFs were used to build an ensemble of models which provided the information about locations for the future data collection and monitoring. We observed that using water retention PTFs to parameterize soil water models offers several advantages, including (a) elimination of the artificial source of uncertainty caused by the sampling parameters space under assumption of the absence of correlation between parameters, and (b) reducing the set of parameters to calibrate to the set of hydraulic conductivity parameters. Future research should include ensembles built with different conceptual models. Overall, using PTF-based model ensembles provides means of fusing pedologic information into hydrologic modeling and offers an efficient way to evaluate the uncertainty soil water modeling results.