Journal list menu

Volume 86, Issue 3 p. 612-629
SOIL PHYSICS & HYDROLOGY
Open Access

Carbon-sensitive pedotransfer functions for plant available water

Dianna K. Bagnall

Corresponding Author

Dianna K. Bagnall

Soil Health Institute, Morrisville, NC, 27560 USA

Correspondence

Dianna K. Bagnall, Soil Health Institute, Morrisville, NC 27560, USA.

Email: [email protected]

Contribution: Conceptualization, Formal analysis, ​Investigation, Methodology, Writing - original draft

Search for more papers by this author
Cristine L. S. Morgan

Cristine L. S. Morgan

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Conceptualization, Funding acquisition, Supervision, Writing - review & editing

Search for more papers by this author
Michael Cope

Michael Cope

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Writing - review & editing

Search for more papers by this author
Gregory M. Bean

Gregory M. Bean

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Writing - review & editing

Search for more papers by this author
Shannon Cappellazzi

Shannon Cappellazzi

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Writing - review & editing

Search for more papers by this author
Kelsey Greub

Kelsey Greub

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Writing - review & editing

Search for more papers by this author
Daniel Liptzin

Daniel Liptzin

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Formal analysis

Search for more papers by this author
Charlotte L. Norris

Charlotte L. Norris

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Writing - review & editing

Search for more papers by this author
Elizabeth Rieke

Elizabeth Rieke

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Writing - review & editing

Search for more papers by this author
Paul Tracy

Paul Tracy

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Data curation

Search for more papers by this author
Ezra Aberle

Ezra Aberle

Carrington Research Extension Center, North Dakota State Univ., Carrington, ND, 58421 USA

Contribution: Data curation

Search for more papers by this author
Amanda Ashworth

Amanda Ashworth

Poultry Production and Product Safety Research Unit, USDA ARS, Fayetteville, AR, 72701 USA

Search for more papers by this author
Oscar Bañuelos Tavarez

Oscar Bañuelos Tavarez

Centro Internacional de Mejoramiento de Maiz y Trigo (CIMMYT) Tlaltizapán, Mor, 62770 Mexico

Contribution: Data curation

Search for more papers by this author
Andy Bary

Andy Bary

Dep. of Crop and Soil Sciences, Washington State Univ., Puyallup, WA, 98371 USA

Contribution: Data curation

Search for more papers by this author
R. Louis Baumhardt

R. Louis Baumhardt

Soil and Water Management Research, USDA-ARS, Bushland, TX, 79012 USA

Contribution: Data curation

Search for more papers by this author
Alberto Borbón Gracia

Alberto Borbón Gracia

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
Daniel Brainard

Daniel Brainard

Dep. of Horticulture, Michigan State Univ., East Lansing, MI, 48824 USA

Contribution: Data curation

Search for more papers by this author
Jameson Brennan

Jameson Brennan

West River Ag Center, South Dakota State Univ., Rapid City, SD, 57703 USA

Contribution: Data curation

Search for more papers by this author
Dolores Briones Reyes

Dolores Briones Reyes

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
Darren Bruhjell

Darren Bruhjell

Agriculture and Agri-Food Canada, Edmonton, AB, T5J 4C3 Canada

Contribution: Data curation

Search for more papers by this author
Cameron Carlyle

Cameron Carlyle

Food and Nutritional Science, Univ. of Alberta Dep. of Agriculture, Edmonton, AB, T6G 2P5 Canada

Contribution: Data curation

Search for more papers by this author
James Crawford

James Crawford

Extension Agricultural Engineering, Univ. of Missouri, Rock Port, MO, 64482 USA

Contribution: Data curation

Search for more papers by this author
Cody Creech

Cody Creech

Panhandle Research and Extension Center, Univ. of Nebraska-Lincoln, Scottsbluff, NE, 69361 USA

Contribution: Data curation

Search for more papers by this author
Steven Culman

Steven Culman

School of Environment & Natural Resources, Ohio State Univ., Wooster, OH, 44691 USA

Contribution: Data curation

Search for more papers by this author
William Deen

William Deen

Dep. of Plant Agriculture, Univ. of Guelph, Guelph, ON, N1G 2W1 Canada

Contribution: Data curation

Search for more papers by this author
Curtis Dell

Curtis Dell

Pasture Systems & Watershed Management Research Unit, USDA ARS, University Park, PA, 16802 USA

Contribution: Data curation

Search for more papers by this author
Justin Derner

Justin Derner

Rangeland Resources and Systems Research Unit, USDA ARS, Cheyenne, WY, 82009 USA

Contribution: Data curation

Search for more papers by this author
Thomas Ducey

Thomas Ducey

Coastal Plains Soil, Water, & Plant Research Center, USDA-ARS, Florence, SC, 29501 USA

Contribution: Data curation

Search for more papers by this author
Sjoerd Willem Duiker

Sjoerd Willem Duiker

Dep. of Plant Science, Pennsylvania State Univ., University Park, PA, 16802 USA

Contribution: Data curation

Search for more papers by this author
Miles Dyck

Miles Dyck

Dep. of Renewable Resources, Univ. of Alberta, Edmonton, AB, T6G 2H1 Canada

Contribution: Data curation

Search for more papers by this author
Benjamin Ellert

Benjamin Ellert

Lethbridge Research and Development Centre, Agriculture and Agri-Food Canada, Lethbridge, AB, T1J 4B1 Canada

Contribution: Data curation

Search for more papers by this author
Martin Entz

Martin Entz

Dep. of Plant Science, Univ. of Manitoba, Winnipeg, MB, R3T 2N2 Canada

Contribution: Data curation

Search for more papers by this author
Avelino Espinosa Solorio

Avelino Espinosa Solorio

Santiago de Querétaro, Sustentabilidad Agropecuaria Querétaro (SAQ), Qro, 76030 Mexico

Contribution: Data curation

Search for more papers by this author
Steven J. Fonte

Steven J. Fonte

Dep. of Soil and Crop Sciences, Colorado State Univ., Fort Collins, CO, 80523 USA

Contribution: Data curation

Search for more papers by this author
Simon Fonteyne

Simon Fonteyne

Centro Internacional de Mejoramiento de Maiz y Trigo (CIMMYT) Texcoco, Ver, 56237 Mexico

Contribution: Data curation

Search for more papers by this author
Ann-Marie Fortuna

Ann-Marie Fortuna

Agroclimate and Natural Resources Research, USDA-ARS, El Reno, OK, 73036 USA

Contribution: Data curation

Search for more papers by this author
Jamie Foster

Jamie Foster

Texas A&M AgriLife Research, Beeville, TX, 78102 USA

Contribution: Data curation

Search for more papers by this author
Lisa Fultz

Lisa Fultz

Environment & Soil Sciences, Louisiana State Univ. School of Plant, Baton Rouge, LA, 70803 USA

Contribution: Data curation

Search for more papers by this author
Audrey V. Gamble

Audrey V. Gamble

Dep. of Crop, Soil and Environmental Sciences, Auburn Univ., Auburn, AL, 36849 USA

Contribution: Data curation

Search for more papers by this author
Charles Geddes

Charles Geddes

Lethbridge Research and Development Centre, Agriculture and Agri-Food Canada, Lethbridge, AB, T1J 4B1 Canada

Contribution: Data curation

Search for more papers by this author
Deirdre Griffin-LaHue

Deirdre Griffin-LaHue

Dep. of Crop and Soil Sciences, Washington State Univ., Mount Vernon, WA, 98273 USA

Contribution: Data curation

Search for more papers by this author
John Grove

John Grove

Dep. of Plant and Soil Science, Research and Education Center, Univ. of Kentucky, Princeton, KY, 42445 USA

Contribution: Data curation

Search for more papers by this author
Stephen K. Hamilton

Stephen K. Hamilton

W.K. Kellogg Biological Station, Michigan State Univ., Hickory Corners, MI, 49060 USA

Contribution: Data curation

Search for more papers by this author
Xiying Hao

Xiying Hao

Lethbridge Research and Development Centre, Agriculture and Agri-Food Canada, Lethbridge, AB, T1J 4B1 Canada

Contribution: Data curation

Search for more papers by this author
Z. D. Hayden

Z. D. Hayden

Dep. of Horticulture, Michigan State Univ., East Lansing, MI, 48824 USA

Contribution: Data curation

Search for more papers by this author
Julie Howe

Julie Howe

Dep. of Soil and Crop Sciences, Texas A&M AgriLife Research, College Station, TX, 77843 USA

Contribution: Data curation, Writing - review & editing

Search for more papers by this author
James Ippolito

James Ippolito

Dep. of Soil and Crop Sciences, Colorado State Univ., Fort Collins, CO, 80523 USA

Contribution: Data curation

Search for more papers by this author
Gregg Johnson

Gregg Johnson

Southern Research and Outreach Center, Univ. of Minnesota, Waseca, MN, 56093 USA

Contribution: Data curation

Search for more papers by this author
Mark Kautz

Mark Kautz

USDA ARS, Tucson, AZ, 85719 USA

Contribution: Data curation

Search for more papers by this author
Newell Kitchen

Newell Kitchen

Cropping Systems and Water Quality Research Unit, USDA ARS USDA-ARS, Columbia, MO, 65211 USA

Contribution: Data curation

Search for more papers by this author
Sandeep Kumar

Sandeep Kumar

Dep. of Agronomy, Horticulture and Plant Science, South Dakota State Univ., Brookings, SD, 57007 USA

Contribution: Data curation

Search for more papers by this author
Kirsten Kurtz

Kirsten Kurtz

Dep. of Natural Resources Cornell Soil Health Laboratory, Dep. of Soil and Crop Sciences, Cornell Univ., Ithaca, NY, 14853 USA

Contribution: Data curation

Search for more papers by this author
Francis Larney

Francis Larney

Lethbridge Research and Development Centre, Agriculture and Agri-Food Canada, Lethbridge, AB, T1J 4B1 Canada

Contribution: Data curation

Search for more papers by this author
Katie Lewis

Katie Lewis

Dep. of Soil and Crop Sciences, Texas A&M Agrilife Research, Lubbock, TX, 79403 USA

Contribution: Data curation

Search for more papers by this author
Matt Liebman

Matt Liebman

Dep. of Agronomy, Iowa State Univ., Ames, IA, 50011 USA

Contribution: Data curation

Search for more papers by this author
Antonio Lopez Ramirez

Antonio Lopez Ramirez

Centro de Bachillerato Tecnológico Agropecuario, No. 305 - CBTA 305 Molcaxac, Pue, 75650 Mexico

Contribution: Data curation

Search for more papers by this author
Stephen Machado

Stephen Machado

Columbia Basin Agricultural Research Center, Oregon State Univ., Adams, OR, 97810 USA

Contribution: Data curation

Search for more papers by this author
Bijesh Maharjan

Bijesh Maharjan

Panhandle Research and Extension Center, Univ. of Nebraska-Lincoln, Scottsbluff, NE, 69361 USA

Contribution: Data curation

Search for more papers by this author
Miguel Angel Martinez Gamiño

Miguel Angel Martinez Gamiño

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
William May

William May

Agriculture and Agri-Food Canada Canada, Indian Head Research Farm, Indian Head, SK, S0G 2K0 Canada

Contribution: Data curation

Search for more papers by this author
Mitchel McClaran

Mitchel McClaran

School of Natural Resources & the Environment, Univ. of Arizona, Tucson, AZ, 85721 USA

Contribution: Data curation

Search for more papers by this author
Marshall McDaniel

Marshall McDaniel

Dep. of Agronomy, Iowa State Univ., Ames, IA, 50011 USA

Contribution: Data curation

Search for more papers by this author
Neville Millar

Neville Millar

W.K. Kellogg Biological Station, Michigan State Univ., Hickory Corners, MI, 49060 USA

Contribution: Data curation

Search for more papers by this author
Jeffrey P. Mitchell

Jeffrey P. Mitchell

Davis Dep. of Plant Sciences, Univ. of California, Davis, CA, 95616 USA

Contribution: Data curation

Search for more papers by this author
Philip A. Moore

Philip A. Moore

Poultry Production and Product Safety Research Unit, USDA ARS, Fayetteville, AR, 72701 USA

Contribution: Data curation

Search for more papers by this author
Amber Moore

Amber Moore

Dep. of Crop and Soil Science, Oregon State Univ., Corvallis, OR, 97331 USA

Contribution: Data curation

Search for more papers by this author
Manuel Mora Gutiérrez

Manuel Mora Gutiérrez

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
Kelly A. Nelson

Kelly A. Nelson

Greenley Research Center, Univ. of Missouri, Novelty, MO, 63460 USA

Contribution: Data curation

Search for more papers by this author
Emmanuel Omondi

Emmanuel Omondi

Dep. of Agricultural and Environmental Sciences, Tennessee State Univ., Nashville, TN, 37209 USA

Contribution: Data curation

Search for more papers by this author
Shannon Osborne

Shannon Osborne

North Central Agricultural Research Laboratory, USDA ARS, Brookings, SD, 57006 USA

Contribution: Data curation

Search for more papers by this author
Leodegario Osorio Alcalá

Leodegario Osorio Alcalá

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
Philip Owens

Philip Owens

Poultry Production and Product Safety Research Unit, USDA ARS, Fayetteville, AR, 72701 USA

Contribution: Data curation

Search for more papers by this author
Eugenia M. Pena-Yewtukhiw

Eugenia M. Pena-Yewtukhiw

Davis College of Agriculture, Natural Resources and Design, West Virginia Univ., Morgantown, WV, 26506 USA

Contribution: Data curation, Writing - review & editing

Search for more papers by this author
Hanna Poffenbarger

Hanna Poffenbarger

Dep. of Plant and Soil Science, Univ. of Kentucky, Lexington, KY, 40546 USA

Contribution: Data curation

Search for more papers by this author
Brenda Ponce Lira

Brenda Ponce Lira

Universidad Politécnica de Francisco I. Madero (UPFIM) San Juan Tepa, Hgo, 42660 Mexico

Contribution: Data curation

Search for more papers by this author
Jennifer Reeve

Jennifer Reeve

Soils and Climate Dep., Utah State Univ. Plants, Logan, UT, 84322 USA

Contribution: Data curation

Search for more papers by this author
Timothy Reinbott

Timothy Reinbott

Bradford Research Center College of Agriculture, Food, and Natural Resources, Univ. of Missouri, Columbia, MO, 65201 USA

Contribution: Data curation

Search for more papers by this author
Mark Reiter

Mark Reiter

Eastern Shore Agricultural Research and Extension Center, Virginia Tech, Painter, VA, 23420 USA

Contribution: Data curation

Search for more papers by this author
Edwin Ritchey

Edwin Ritchey

Dep. of Plant and Soil Science, Research and Education Center, Univ. of Kentucky, Princeton, KY, 42445 USA

Contribution: Data curation

Search for more papers by this author
Kraig L. Roozeboom

Kraig L. Roozeboom

Dep. of Agronomy, Kansas State Univ., Manhattan, KS, 66506 USA

Contribution: Data curation

Search for more papers by this author
Ichao Rui

Ichao Rui

Rodale Institute Farming Systems Trial, Kutztown, PA, 19530 USA

Contribution: Data curation

Search for more papers by this author
Amir Sadeghpour

Amir Sadeghpour

Crops, Soils, and Environmental Management Program, School of Agricultural Sciences, Southern Illinois Univ., Carbondale, IL, 62901 USA

Contribution: Data curation

Search for more papers by this author
Upendra M. Sainju

Upendra M. Sainju

Northern Plains Agricultural Research Laboratory, USDA ARS, Sidney, MT, 59270 USA

Contribution: Data curation

Search for more papers by this author
Gregg Sanford

Gregg Sanford

Dep. of Agronomy, Univ. of Wisconsin, Madison, WI, 53706 USA

Contribution: Data curation

Search for more papers by this author
William Schillinger

William Schillinger

Dryland Research Station, Washington State Univ., Lind, WA, 99341 USA

Contribution: Data curation

Search for more papers by this author
Robert R. Schindelbeck

Robert R. Schindelbeck

Dep. of Soil and Crop Sciences, Cornell Univ. Cornell Soil Health Laboratory, Ithaca, NY, 14853 USA

Contribution: Data curation

Search for more papers by this author
Meagan Schipanski

Meagan Schipanski

Dep. of Soil and Crop Sciences, Colorado State Univ., Fort Collins, CO, 80523 USA

Contribution: Data curation

Search for more papers by this author
Alan Schlegel

Alan Schlegel

SW Research - Extension Center, Kansas State Univ., Tribune, KS, 67879 USA

Contribution: Data curation

Search for more papers by this author
Kate Scow

Kate Scow

Davis Dep. of Land, Air, and Water Resources, Univ. of California, Davis, CA, 95616 USA

Contribution: Data curation

Search for more papers by this author
Lucretia Sherrod

Lucretia Sherrod

Center for Agricultural Resources Research, USDA ARS, Fort Collins, CO, 80526 USA

Contribution: Data curation

Search for more papers by this author
Sudeep Sidhu

Sudeep Sidhu

Univ. of Florida, Quincy, FL, 32351 USA

Contribution: Data curation

Search for more papers by this author
Ernesto Solís Moya

Ernesto Solís Moya

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
Mervin St. Luce

Mervin St. Luce

Agriculture and Agri-Food Canada Swift Current Research and Development Centre, Swift Current, SK, S9H 3×2 Canada

Contribution: Data curation

Search for more papers by this author
Jeffrey Strock

Jeffrey Strock

Dep. of Soil, Water and Climate, SW Research & Outreach Center, Univ. of Minnesota, Lamberton, MN, 56152 USA

Contribution: Data curation

Search for more papers by this author
Andrew Suyker

Andrew Suyker

Lincoln School of Natural Resources, Univ. of Nebraska, Lincoln, NE, 68583 USA

Contribution: Data curation

Search for more papers by this author
Virginia Sykes

Virginia Sykes

Dep. of Plant Sciences, Univ. of Tennessee, Knoxville, TN, 37996 USA

Contribution: Data curation

Search for more papers by this author
Haiying Tao

Haiying Tao

Dep. of Crop & Soil Sciences, Washington State Univ., Pullman, WA, 99164 USA

Contribution: Data curation

Search for more papers by this author
Alberto Trujillo Campos

Alberto Trujillo Campos

Instituto Nacional de Investigaciones Forestales, Agrícolas y Pecuarias (INIFAP), Ciudad de México, CDMX, 04010 Mexico

Contribution: Data curation

Search for more papers by this author
Laura L. Van Eerd

Laura L. Van Eerd

School of Environmental Sciences - Ridgetown Campus, Univ. of Guelph, Ridgetown, ON, N0P 2C0 Canada

Contribution: Data curation

Search for more papers by this author
Nele Verhulst

Nele Verhulst

International Maize and Wheat Improvement Center (CIMMYT), Carretera Mexico-Veracruz km 45, Texcoco, Mexico

Contribution: Data curation

Search for more papers by this author
Tony John Vyn

Tony John Vyn

Agronomy Dep., Purdue Univ., West Lafayette, IN, 47907 USA

Contribution: Data curation

Search for more papers by this author
Yutao Wang

Yutao Wang

Agriculture and Agri-Food Canada Harrow Research and Development Center, Harrow, ON, N0R 1G0 Canada

Contribution: Data curation

Search for more papers by this author
Dexter Watts

Dexter Watts

National Soil Dynamics Laboratory, USDA-NSDL, Auburn, AL, 36832 USA

Contribution: Data curation

Search for more papers by this author
David Wright

David Wright

Univ. of Florida, Quincy, FL, 32351 USA

Contribution: Data curation

Search for more papers by this author
Tiequan Zhang

Tiequan Zhang

Agriculture and Agri-Food Canada Harrow Research and Development Center, Harrow, ON, N0R 1G0 Canada

Contribution: Data curation

Search for more papers by this author
Charles Wayne Honeycutt

Charles Wayne Honeycutt

Soil Health Institute, Morrisville, NC, 27560 USA

Contribution: Data curation

Search for more papers by this author
First published: 23 February 2022
Citations: 20

Assigned to Associate Editor Ryan Stewart.

Abstract

Currently accepted pedotransfer functions show negligible effect of management-induced changes to soil organic carbon (SOC) on plant available water holding capacity (θAWHC), while some studies show the ability to substantially increase θAWHC through management. The Soil Health Institute's North America Project to Evaluate Soil Health Measurements measured water content at field capacity using intact soil cores across 124 long-term research sites that contained increases in SOC as a result of management treatments such as reduced tillage and cover cropping. Pedotransfer functions were created for volumetric water content at field capacity (θFC) and permanent wilting point (θPWP). New pedotransfer functions had predictions of θAWHC that were similarly accurate compared with Saxton and Rawls when tested on samples from the National Soil Characterization database. Further, the new pedotransfer functions showed substantial effects of soil calcareousness and SOC on θAWHC. For an increase in SOC of 10 g kg–1 (1%) in noncalcareous soils, an average increase in θAWHC of 3.0 mm 100 mm–1 soil (0.03 m3 m–3) on average across all soil texture classes was found. This SOC related increase in θAWHC is about double previous estimates. Calcareous soils had an increase in θAWHC of 1.2 mm 100 mm–1 soil associated with a 10 g kg–1 increase in SOC, across all soil texture classes. New equations can aid in quantifying benefits of soil management practices that increase SOC and can be used to model the effect of changes in management on drought resilience.

Abbreviations

  • NAPESHM
  • North American Project to Evaluate Soil Health Measurements
  • SOC
  • soil organic carbon
  • θAWHC
  • plant available water holding capacity
  • θFC
  • volumetric water content at field capacity
  • θPWP
  • volumetric water content at permanent wilting point.
  • 1 INTRODUCTION

    If plant available water holding capacity (θAWHC) increases in a meaningful way when soil organic carbon (SOC) increases, the outcome is that management practices that increase SOC in soils simultaneously change water retention by soils. This change in water retention has implications for hydrology, the energy balance, and crop production. Thus, a positive, causal, relationship between SOC and θAWHC has direct benefit through increased cropping system resilience to drought and provides an incentive for the adoption of practices that benefit society through climate change mitigation and adaptation (Lal, 2004; Lal, 2006; A. Williams et al., 2016; Yang et al., 2014). Soil science literature is inconclusive on whether this relationship exists to a meaningful degree in agricultural soils. Some studies have demonstrated substantial improvements in θAWHC as a result of increasing SOC (Ankenbaur & Loheide, 2017; Bouyoucos, 1939; Hudson, 1994; Maynard, 2000; Salter & Howarth, 1961) and others have not (Bauer & Black, 1992; Bell & Van Keulen, 1995; Feustal & Byers, 1936). A recent review concluded that the effect of SOC on θAWHC was limited; a 10 g kg−1 increase in SOC concentration resulted in a θAWHC increase of 1.2 mm 100 mm−1 soil (0.012 m3 m−3) across all soil textures (Minasny & McBratney, 2018). Meanwhile, those that promote management changes to improve soil health and functioning in agricultural landscapes, such as drought resilience, are limited to providing regional and anecdotal evidence that increased SOC improves θAWHC.

    While this debate is ongoing, researchers in disciplines like hydrology and land surface modeling, are relying on existing pedotransfer functions relating SOC to water characteristics. Pedotransfer functions allow practical estimation of field and laboratory soil measurements that are costly, time-consuming, and can be impractical to measure (Bouma, 1989) using proxy variables such as particle size, bulk density, and organic C to predict soil hydraulic properties of interest (Wösten et al., 2001). Soil scientists and engineers have a long history of estimating soil water characteristics that are difficult to measure (Brooks & Corey, 1964; Campell, 1974; Rawls et al., 1992; Van Genuchten, 1980) and early efforts demonstrated that soil particle size could predict soil water characteristics to provide adequate estimates for many decisions (Ahuja et al., 1985; Arya & Paris, 1981; Gupta & Larson, 1979; Saxton et, al.,1986; J. Williams et al., 1983). Early pedotransfer functions were accurate, but many had limited geographic use because they were based on regional data (Gijsman et al., 2002).

    Saxton and Rawls (2006) updated pedotransfer functions (Saxton, Rawls, Romberger & Papendick, 1986) for volumetric water content at field capacity (estimated by water retained at −33 kPa) and permanent wilting point (estimated by water retained at −1,500 kPa). The difference between volumetric water content held at field capacity (θFC) and wilting point (θPWP) estimates θAWHC. The Saxton and Rawls (2006) pedotransfer functions were created with approximately 2,000 soil samples from A horizons obtained from the Natural Resource Conservation Service (NRCS) National Soil Characterization database (Soil Survey Staff, 2004). Saxton and Rawls (2006) did not report the depth of A horizon used, but the depth is likely well represented by the horizons in the National Cooperative Soil Survey Characterization (NCSS) Microsoft Access database, in which A master horizons were log normally distributed and ranged in depth from 0 to 60 cm with a median depth of 13 cm. The pedotransfer functions were fit using continuous sand, clay, and soil organic matter content and two-way interaction terms as predictor variables in a multiple linear regression. To improve the fit, a second nonlinear function was added resulting in two combined, dependent equations for prediction of θFC and θPWP.

    The accuracy and continental distribution of the input data enabled these pedotransfer functions to be widely used (1,185 citations according to Scopus as of August 2021), while their simplicity resulted in their incorporation into many models. The high degree of use of these pedotransfer functions is illustrated by the number of agronomic, ecological, hydrological, land surface, and meteorological models listed in publications that cite Saxton and Rawls (2006) including the Soil Water Assessment Tool (SWAT), AquaCROP, Agricultural Policy/Environmental eXtender (APEX), Noah-MP land surface model, Soil-Plant-Air-Water (SPAW), Variable Infiltration Capacity (VIC), TOPMODEL, Agricultural Production System Simulator (APSIM), Annualized Agricultural Non-Point Source Pollution Model (AnnAGNPS), DeNitrification-DeComposition and DayCent. Because pedotransfer functions may be either hardcoded into models or used in the data preparation step to generate model input, this is not an exhaustive measure of the use of these functions.

    Pedotransfer functions may be as simple as a look-up tables or as complex as machine learning techniques, including artificial neural networks and bootstrapping (Moosavi & Sepaskhah, 2012), support vector machines (Twarakavi et al., 2009), classification and regression trees (Pachepsky et al., 2006), and random forests and boosted regression trees (Jorda et al., 2015). Machine learning can provide more accurate predictions (Jorda et al., 2015) but their complexity means that the mathematical structure of the pedotransfer function is not easily published. Pedotransfer functions built using machine learning require software development to support them (Zhang & Schaap, 2017) and reduce the ease of implementing the function in larger computer models (Schaap et al., 2004). A comparison of 11 pedotransfer functions with varying levels of complexity for water retention found no superior model (Schaap, Nemes & van Genuchten, 2004), indicating that simple models can be adequate.

    Most existing pedotransfer functions include SOC (or soil organic matter), but the effects of SOC on θAWHC are reported as negligible (e.g., Saxton & Rawls, 2006; Minasny & McBratney, 2018). There is a mismatch between soil science textbooks and pedotransfer functions on the effects of SOC on θAWHC. For example, a textbook by Brady and Weil (2002) states that “Recognizing the beneficial effects of organic matter on plant available water is essential to wise soil management” indicating that SOC increases θAWHC to a meaningful degree in terms of crop production, yet pedotransfer functions do not give such results. This discrepancy could be explained by two nonmutually exclusive elements of the underlying data. First, θFC measurements may lack the effects of soil structure. This is particularly relevant for soil samples that are dried and sieved prior to θFC measurement. Disturbed soils may show the direct effect of SOC on water retention, but do not capture the secondary effects of SOC that manifest through soil structure because they do not contain interpedal pores. It is preferable to use intact soil clods or cores because they are more likely to capture both direct effect of SOC on water retention and the secondary effects of SOC that manifest through soil structure (Dane & Hopmans, 2018; J. Williams et al., 1983). Soil organic C has a strong link to soil structure (Bronick & Lal, 2005), and soil structure is linked to pore size distribution, which affects soil water at the macro scale (Nimmo & Akstin, 1988; Pachepsky & Rawls, 2003; J. Williams et al., 1983). Second, many datasets may include measurements of intact soil structures for building pedotransfer functions but were collected to represent changes between soil pedons (at the m scale) for the purpose of mapping and inventory. Hence when considering a given texture class, changes in SOC in such databases are not primarily influenced by management, but rather the landscape position and climate from which the soils were collected.

    Core Ideas

    • New pedotransfer functions show organic C increases plant available water.
    • Noncalcareous soils show greater effects of organic C on plant available water.
    • Increase in plant available water from organic C is more than double previous estimates.
    • These pedotransfer functions can easily be used in hydrologic models.
    • A gap is bridged for modeling the effect of increased soil organic C on plant available water.

    An additional consideration in creating pedotransfer functions for θAWHC is the effect of calcareousness. Calcareous soils have been reported to require more frequent irrigation than noncalcareous soils to achieve the same crop yield, indicating that they may have lower θAWHC. Substantial amounts of calcium carbonate have been found to lower water retention in repacked soil samples (Stakman & Bishay, 1975) and to increase bulk density (Habel, 2014). It has been proposed that presence of calcium carbonate may alter water retention not only through changes in effective soil texture (e.g., carbonates the size of silicate clays), but also via alteration of soil structure and pores (Jackson & Eire, 1973), although analysis on intact samples would be needed to confirm such an effect. We found, in this data set, a significant effect of calcium carbonate on predictions of θAWHC.

    The Soil Health Institute's North America Project to Evaluate Soil Health Measurements (NAPESHM) provided a unique dataset to investigate whether increases in SOC as a result of management correspond to increases in θAWHC. The NAPESHM dataset contains a broad distribution of agricultural soils, primarily managed for row crops, across the major cropland regions of United States, Canada, and Mexico. Data were collected on replicated experimental units (either plots or fields that represent one replication of a treatment) under long-term treatments (10 yr or more) within research sites (n = 124). This approach is expected to capture variation in SOC and measured θAWHC that are both management induced (within pedons) and inherent (between pedons). Additionally, θFC (estimated by water retained at −33 kPa) was measured on intact cores, thus capturing the effect of SOC on the soil matrix and bulk soil (including soil structure).

    The goal of this work was to create a simple-to-implement pedotransfer function for θAWHC that is sensitive to changes in SOC. To meet our goal, we used simple linear regression to fit functions for θFC and θPWP using NAPESHM data. We assessed the accuracy of predicted θFC, θPWP, and θAWHC (by subtraction of predicted θFC and θPWP) to the accuracy of Saxton and Rawls (2006) on 1,797 soils from NCSS. The Saxton and Rawls (2006) approach was chosen as a comparison because of its ubiquity in hydrology and energy models, its representation of similar geographical extent, and its similar simplicity of modeling approach. We also considered how four levels of SOC affected predictions made by the new pedotransfer functions and compared these results to Saxton and Rawls (2006) pedotransfer functions and literature.

    2 MATERIALS AND METHODS

    Data used to develop new pedotransfer functions were drawn from the NAPESHM database, which captured a range in climate, management practices, and inherent soil properties using 124 long-term agricultural research sites across North America and were uniformly sampled in 2019 (Norris et al., 2020). Soil orders in the dataset included the Soil Taxonomy orders of Ultisol, Alfisol, Mollisol, Vertisol, Aridisol, Inceptisol, and Entisol. The replicated treatments included tillage, residue management, cover crop use and type, crop rotation, grazing, and nutrient type and rate (including organic amendments). At the time of sampling, most treatments in the NAPESHM database were continuous for 10 or more years.

    To develop new pedotransfer functions, we used measurements of particle size distribution, SOC, inorganic C, bulk density, and gravimetric water content measured at field capacity (−33 kPa) and permanent wilting point (−1,500 kPa). Soil particle size analysis, pH, and SOC were measured at The Ohio State Soil Water and Environmental Lab. Bulk density and all measures of water retention were done at the Cornell Soil Health Laboratory (Ithaca, NY). Particle size distribution, SOC, and θPWP were measured using a composite soil sample collected from 0-to-15-cm depth from four to six sampling locations within each experimental unit (Norris et al., 2020). The sieve and pipette method with three size classes (2,000–50, 50–2, and <2 μm) was used to measure soil particle size distribution (Gee & Bauder, 1986). Total C was measured by dry combustion (Nelson & Sommers, 1996) using an NC 2100 soil analyzer made by CE instruments (Lakewood, NJ). Soils were oven dried and ground. Soils with >7.2 pH (1:2, soil/water) (Thomas, 1996) were tested for effervescence with 10% HCl. Those that effervesced were analyzed for carbonates using the Chittick's volumetric calcimeter method (Dreimanis, 1962). For soils with carbonates, SOC was calculated by subtracting the value of inorganic C from total C obtained by dry combustion. For all other samples, total C obtained using dry combustion represented SOC.

    In addition to the composite soil sample, four 7.6-cm diam. soil cores were collected to a depth of 7.6 cm in each experimental unit and maintained intact using a plastic sleeve. Two of the four cores were kept intact for measuring bulk density and θFC. The remaining two cores were composited into one bag, sieved to remove coarse fragments >2 mm, and used to calculate soil bulk density. For any sample with <2% coarse fragments by weight (determined during preparation for particle size analysis), bulk density was calculated as the mean bulk density of all four cores—two intact and two composited. Ninety percent of the experimental units in this study had <2% coarse fragments by weight. For the remaining 10% with coarse fragments >2% by weight, bulk density was calculated as the mean of the two composited cores, following removal of coarse fragments and adjustments for the weight and volume of coarse fragments.

    Gravimetric water content at permanent wilting point was measured on pressure plates at 1,500 kPa (Reynolds & Topp, 2008) using repacked soil from the composite sample. Each sample used 15 g of soil that was dried (105 °C), ground (2-mm sieve), saturated, repacked, and equilibrated for 7 d on the pressure plate at 1,500 kPa. Two replications were measured per experimental unit and the mean was calculated to represent θPWP for each experimental unit. Gravimetric water content at field capacity was measured on both intact cores using a tension table at 33 kPa (Hao et al., 2008; Topp et al., 1993). The mean of the two intact cores represented the θFC for each experimental unit. For the intact cores used to calculate θFC, the cores were first saturated (4–7 d), equilibration times on the tension table were between 4 and 7 d, and reference samples were used (weight recorded daily) for quality control. Both θFC and θPWP were calculated by multiplying gravimetric water content by the mean bulk density of the experimental unit and assuming a density of water of 1.0 Mg m−3. Plant available water-holding capacity was calculated as the difference between θFC and θPWP.

    We removed extreme values from our analysis. Experimental units that were excluded were those with a mean bulk density >1.8 Mg m−3 and SOC concentrations >46.5 g kg−1. No samples had mean bulk density of <1.0 Mg m−3. Our threshold values for bulk density and SOC were the same as those used to develop the Saxton and Rawls (2006) pedotransfer functions. The final NAPESHM dataset used in this study included 1,731 samples for development of pedotransfer functions. These 1,731 samples represented 547 unique management treatments across 119 sites. Most sites (107) had been managed continuously for 10 or more years, although 12 sites had at least one treatment that had been in place for between 5 and 10 yr. Figure 1 shows the distribution of site locations across North America, with Hargreaves’ moisture deficit, to demonstrate the range in climatic conditions. Eleven of the 12 USDA soil texture classes were represented in the NAPESHM data used in this study (Table 1); sandy clay textures were not represented. The most abundant texture classes were silt loam and loam. Soils were predominantly sampled in the United States, with 1,335 experimental units from 91 sites. Fifteen sites (237 experimental units) were in Canada, and 13 sites (161 experimental units) were in Mexico. Soil organic C ranged from 2 to 44 g kg−1 with a median of 14 g kg−1. In total, 335 samples from 35 sites effervesced when treated with 10% HCl. Few calcareous soils had relatively high SOC; there were 63 experimental units with >20 g kg−1 SOC, including 12 >30 g kg−1 SOC, and 3 >40 g kg−1 SOC (Figure 2a). For noncalcareous soils, there were 322, 32, and 2 experimental units that were >20, 30, and 40 g kg−1 SOC, respectively. Most of the noncalcareous soils with greater SOC were between 10 and 40% clay content (Figure 2a) and two experimental units were >30 g kg−1 SOC and more than 40% clay. Calcareous soils were well distributed across the observed clay contents and largely followed the same trends as noncalcareous soils when θFC and θPWP were plotted against clay content (Figure 2b,c). There was greater variance in water content for θFC than for θPWP, which is consistent with theory and practice (Pachepsky & Rawls, 2003; Saxton & Rawls, 2006).

    Details are in the caption following the image
    Soils were sampled at 124 sites for the North American Project to Evaluate Soil Health Measurements. The maps shows Hargreaves’ climate moisture deficit for the 119 sites used in this study
    TABLE 1. Count of experimental units (n = 1,731) and sites (n = 119) by USDA soil texture class. A single site may have multiple soil texture classes, so the sum of sites shown is greater than the number of sites in the study. Data shown is from North American Project to Evaluate Soil Health Measurements
    USDA soil texture class Sites Experimental units
    Noncalcareous
    Clay 7 17
    Clay loam 20 148
    Loam 34 265
    Loamy sand 14 96
    Sand 5 31
    Sandy loam 26 193
    Sandy clay loam 10 46
    Silt 5 10
    Silt loam 40 480
    Silty clay 4 9
    Silty clay loam 14 85
    Calcareous
    Clay 6 60
    Clay loam 12 25
    Loam 18 112
    Sandy loam 4 20
    Sandy clay loam 1 2
    Silt loam 6 23
    Silty clay 6 63
    Details are in the caption following the image
    Organic C (%) and volumetric water content for permanent wilting point (θPWP) and field capacity (θFC), plotted against clay content. Data shown is from North American Project to Evaluate Soil Health Measurements

    Hargreaves’ moisture deficit used for the map in this study was generated with the ClimateNAv5.10 software package (available at http://tinyurl.com/ClimateNA) based on methodology described by Wang et al. (2016). Hargreaves’ moisture index has a value of zero for any month within a year that has greater precipitation than reference evapotranspiration. For all months in which the precipitation is less than reference evapotranspiration, the difference is summed to arrive at the annual moisture deficit (mm).

    To test the accuracy of the new pedotransfer functions, we obtained data from 1,797 soils from the National Cooperative Soil Survey Characterization (NCSS) database (http://ncsslabdatamart.sc.egov.usda.gov/; Soil Survey Staff, 1995). These data included sampling locations from 39 U.S. states. All 12 USDA soil texture classes were represented in the noncalcareous soils and 10 classes were represented within calcareous soils. Each horizon started at 0 cm and ended at or before 15-cm depth and had measurements of bulk density (volume measured at −33 kPa), gravimetric water content at both permanent wilting point (−1,500 kPa) and field capacity (at −33 kPa; clod method; Soil Survey Staff, 2014), and total C, calcium carbonate, sand, and clay content. We multiplied gravimetric water content by bulk density to obtain θFC and θPWP. Soil organic C was calculated by subtracting the quantity of inorganic C in carbonates from total C. The NCSS data were developed with standard laboratory procedures (Klute, 1986; USDA-SCS, 1982).

    We developed pedotransfer functions by initially fitting multiple linear regression models to the NAPESHM data using ordinary least squares for θFC and θPWP using clay, sand, SOC content (all units are in 10 g kg−1, which is equivalent to 1.0%), and all two-way interaction terms as predictor variables. To determine whether there was a significant difference in the predictions of the pedotransfer functions for soils that were and were not calcareous, we conducted a one-way ANOVA on the regression residuals using a categorical indicator for effervescence when treated with HCl as the only factor. For all statistical analyses, R statistical software (R Development Core Team, 2020) and α = .001 were used.

    The ANOVA p value for the effect of calcareousness on water content was significant for model predictions of θFC (p = .03) but not for θPWP (p = .45). We fit separate models for calcareous soils (those that effervesced when treated with HCl) and noncalcareous soils. We then used backwards stepwise selection for each model by applying the step function from the stats package in R (Hastie & Pregibon, 1992). For both calcareous and noncalcareous soils, stepwise selection for θFC models showed the lowest AIC for the full models (clay, sand, SOC content, and all two-way interaction terms). For calcareous soils, the θPWP model that had the lowest AIC included clay, sand, and SOC content, and the two-way interaction terms for both SOC by sand content and SOC by clay content. For calcareous soils, stepwise selection for the θPWP model found that the lowest AIC resulted from clay, sand, and SOC content, and the two-way interaction term between sand and clay content. For each linear model, plots of model residuals against model-fitted values and predictors were used to verify equal error variance and the Breusch–Pagan test against heteroskedasticity was also used. Plots of residuals vs. leverage (Cook's distance cutoff of 0.5) were used to check for influential values, and theoretical quantile-quantile plots were used to check normality of the residuals.

    To investigate whether θFC and θPWP were responding to effects of soil management that were not manifested in SOC changes, we plotted model residuals against aggregate stability measured with a Cornell rainfall simulator, the average Soil Tillage Intensity Rating (STIR, Karlen et al., 2008; U.S. Department of Agriculture, 2019) calculated for the past 5 yr, and categorical variables designating crop category (row crop, perennial, integrated row crop and perennial, nonfarmed, or woody perennial) and nutrient type (none, organic, synthetic, or synthetic and organic). We tested the strength of the relationship between model residuals and these variables (aggregate stability, STIR, crop category, nutrient type) using regression for continuous variables and ANOVA for categorical variables and found they did not explain substantial variance as shown by an R2 of .03 or less for each regression or ANOVA. We reported the root mean square error (RMSE) and adjusted R2 from regressions of predictions on measurements for θFC, θPWP, and θAWHC

    We predicted θFC and θPWP for 1,797 soils from the NCSS database using both the new pedotransfer functions and Saxton and Rawls (2006). We computed the deviation from measured values (predicted-measured) and RMSE for θFC, θPWP, and θAWHCAWHC calculated by subtraction) for both pedotransfer function predictions by USDA particle size class. We used a paired t test to determine significant differences from measured water content for both pedotransfer functions. Saxton and Rawls (2006) pedotransfer functions include an organic matter parameter, but we used a SOC parameter. Because the van Bemmelen factor (0.58) was used to convert SOC to organic matter to develop the Saxton and Rawls (2006) functions, we converted NCSS SOC values to organic matter by multiplying SOC by the reciprocal of the van Bemmelen factor.

    To investigate the effect of SOC on predicted θAWHC, we generated values that represent possible combinations of SOC, clay, and sand content. These combinations of possible SOC, clay, and sand content values were not measured soil samples – rather they were created to evaluate a wide range of sand, clay, and SOC content. Soil organic C values of 10, 20, 30, and 40 g kg−1 (1, 2, 3, and 4%) and clay and sand values from 5 to 95% in 5% increments were used and repeated so that each sand content was paired with every clay content and every level of SOC; resulting in 760 combinations in all. We used these combinations of SOC, clay, and sand content to generated predictions of θFC and θPWP using the new pedotransfer functions (both noncalcareous and calcareous) and Saxton and Rawls (2006) pedotransfer functions. Locally estimated scatterplot smoothing (LOESS) curves were fit to the predictions for each level of SOC for visual evaluation.

    3 RESULTS AND DISCUSSION

    The pedotransfer functions for volumetric water content at θPWP and θFC are given in Equations 1 and 2 for noncalcareous soils and in Equations 3 and 4 for calcareous, respectively. All units are in 10 g kg−1.
    θ PWP = 7.222 + 0.296 Clay 0.074 Sand 0.309 SOC ̲ + 0.022 Sand × SOC + 0.022 Clay × SOC \begin{eqnarray} \hspace*{0.28em}{\mathrm{\theta}}_{\mathrm{PWP}}& & =7.222+0.296\mathrm{Clay}-0.074\mathrm{Sand}-\underline{0.309\mathrm{SOC}}\nonumber\\ & & +\hspace*{0.28em}0.022\left(\mathrm{Sand}\ensuremath{\times{}}\mathrm{SOC}\right)+0.022\left(\mathrm{Clay}\ensuremath{\times{}}\mathrm{SOC}\right) \end{eqnarray} (1)
    θ FC = 37.217 0.140 Clay 0.304 Sand 0.222 SOC ̲ + 0.051 Sand × SOC + 0.085 Clay × SOC + 0.002 Clay × Sand \begin{eqnarray} {\mathrm{\theta}}_{\mathrm{FC}}& & =37.217-0.140\mathrm{Clay}-0.304\mathrm{Sand}-\underline{0.222\mathrm{SOC}}\nonumber\\ & & +\hspace*{0.28em}0.051\left(\mathrm{Sand}\ensuremath{\times{}}\mathrm{SOC}\right)+0.085\left(\mathrm{Clay}\ensuremath{\times{}}\mathrm{SOC}\right)\nonumber\\ & & +\hspace*{0.28em}0.002\left(\mathrm{Clay}\ensuremath{\times{}}\mathrm{Sand}\right) \end{eqnarray} (2)
    θ PWP , calc = 7.907 + 0.236 Clay 0.082 Sand + 0.441 SOC + 0.002 Clay × Sand \begin{eqnarray} {\mathrm{\theta}}_{\mathrm{PWP},\mathrm{calc}}& & =7.907+0.236\mathrm{Clay}-0.082\mathrm{Sand}+0.441\mathrm{SOC}\nonumber\\ & & +\hspace*{0.28em}0.002\left(\mathrm{Clay}\ensuremath{\times{}}\mathrm{Sand}\right) \end{eqnarray} (3)
    θ FC , calc = 33.351 + 0.020 Clay ̲ 0.446 Sand + 1.398 SOC ̲ + 0.052 Sand × SOC 0.077 Clay × SOC + 0.011 Clay × Sand \begin{eqnarray} {\mathrm{\theta}}_{\mathrm{FC},\mathrm{calc}}& & =33.351+\underline{0.020\mathrm{Clay}}-0.446\mathrm{Sand}+\underline{1.398\mathrm{SOC}}\nonumber\\ & & +\hspace*{0.28em}0.052\left(\mathrm{Sand}\ensuremath{\times{}}\mathrm{SOC}\right)-0.077\left(\mathrm{Clay}\ensuremath{\times{}}\mathrm{SOC}\right)\nonumber\\ & & +\hspace*{0.28em}0.011\left(\mathrm{Clay}\ensuremath{\times{}}\mathrm{Sand}\right) \end{eqnarray} (4)

    All terms were significant or were nonsignificant main effects that were retained as a component of a significant two-way interaction. The nonsignificant main effect coefficients are underlined in the equations above.

    3.1 Accuracy of new pedotransfer functions

    We used regressions between pedotransfer function predictions and the measured NAPESHM data that was used to create the models as one test of accuracy (Figure 3). Models for predicting θPWP had better accuracy than for θFC, indicated by R2 and RMSE. The models for noncalcareous soils performed better than models for calcareous soils. The new pedotransfer functions are near the lower end of RMSE values for water retention predictions reported by Schaap et al. (2004), who found that of the 11 modes they considered, mean RMSE was between 3.2 and 6.9 mm 100 mm−1. The error was slightly greater for θAWHC, ranging from 5.8 to 8.0 mm 100 mm−1. Model performance of the Saxton and Rawls (2006) equations were also similar with R2 and RMSE of 0.86 and 2.0 mm 100 mm−1 for θPWP and 0.63 and 5.0 mm 100 mm−1 for θFC, respectively. The lesser accuracy of the models for calcareous soils is likely from two sources. One being fewer experimental units used in the fit; the calcareous fit had 335, while the noncalcareous model had 1,396, and about 2,000 were used in Saxton and Rawls (2006). Secondly, the physical complexity of calcareous soils could also add to the poorer fit.

    Details are in the caption following the image
    (a and b) Measured vs. predicted volumetric water content at permanent wilting point (θPWP) (c and d) field capacity (θFC), and (e and f) plant available water holding capacity (θAWHC) for (left) noncalcareous and (right) calcareous soils. Measured values are from the from North American Project to Evaluate Soil Health Measurements and predicted values are from new pedotransfer functions. Predictions of θAWHC are calculated by subtraction of predicted θFC and θPWP. Regressions are significant (p value < .001). Blue, solid lines are regression fits and black, dashed lines are one-to-one. RMSE is root mean square error

    As a further test of accuracy and to provide context to evaluate the new pedotransfer functions, the new pedotransfer functions and Saxton and Rawls (2006) were used to calculate predictions of θPWP, θFC, and θAWHC (by subtraction of predicted θFC and θPWP) for 1,797 soils from the National NCSS database. Because the Saxton and Rawls (2006) pedotransfer functions were developed using NCSS data including θFC, it was expected that they would be more accurate in this test compared with the new pedotransfer functions using NAPESHM data including θFC. However, in general, new pedotransfer functions for noncalcareous soils performed similarly to Saxton and Rawls (2006) in regard to their RMSE (Table 2), although there were differences between particle size classes. Specifically, the new pedotransfer functions for noncalcareous soils had smaller RMSE values than Saxton and Rawls (2006) functions for θPWP, θFC, and θAWHC for coarse texture classes (sand, loamy sand, sandy loam, and silt loam). Saxton and Rawls (2006) functions had smaller RMSE values for θPWP, θFC, and θAWHC in finer textures (silty clay loam, sandy clay loam, clay loam, and all clays). For loams, the new pedotransfer functions had smaller RMSE values compared with Saxton and Rawls (2006) for θFC and θAWHC, but not for θPWP, while in silts, the reverse was true (Table 2). The greatest difference in RMSE magnitude for θAWHC of noncalcareous soils occurred in clay where the RMSE of the new pedotransfer functions was 3.3 mm 100 mm−1 greater than the Saxton and Rawls (2006) pedotransfer functions (Table 2). In calcareous soils, unlike noncalcareous soils, RMSE values for the new pedotransfer functions were greater in coarser textures; sand had an RMSE for θAWHC that was 7.4 mm 100 mm−1 greater compared with Saxton and Rawls (2006) but both pedotransfer functions had a relatively large RMSE for sand.

    TABLE 2. Root mean square error (RMSE) values for volumetric water content at permanent wilting point (θPWP), field capacity (θFC), and plant available water holding capacity (θAWHC) predicted using both Saxton and Rawls (2006) (S&R) and the new pedotransfer functions using 1,797 soil samples from the National Cooperative Soil Survey Characterization database. The RMSEs are grouped by USDA soil texture classes and the number of samples (n) are shown for each texture class
    θPWP θFC θAWH
    RMSE (mm 100 mm−1)
    USDA soil texture class n S&R New S&R New S&R New
    Noncalcareous
    Sand 39 3.4 2.3 10.2 7.8 7.9 6.9
    Loamy sand 45 3.9 2.7 10.6 6.0 7.3 5.0
    Sandy loam 169 7.3 6.5 13.6 9.0 8.3 5.8
    Silt loam 273 4.9 4.0 8.9 6.0 6.3 5.4
    Silt 4 3.6 2.1 5.6 6.1 5.4 6.6
    Loam 198 4.5 4.9 9.6 7.3 7.0 5.2
    Silty clay loam 201 2.8 4.7 4.2 5.2 4.7 4.8
    Sandy clay loam 39 4.1 6.8 6.9 9.5 5.3 5.5
    Clay loam 87 4.0 7.0 5.7 9.6 4.9 5.6
    Silty clay 71 4.3 8.0 5.5 10.9 7.0 7.5
    Sandy clay 10 3.1 8.4 4.8 13.6 4.9 7.0
    Clay 113 5.2 8.1 5.1 14.0 5.1 8.4
    Calcareous
    Sand 5 3.0 2.1 18.3 25.7 17.1 24.5
    Loamy sand 12 5.3 4.7 12.6 15.9 8.5 12.4
    Sandy loam 88 3.4 3.2 9.7 11.0 8.5 10.2
    Silt loam 77 7.8 6.3 11.6 10.0 10.4 10.6
    Loam 88 5.1 4.8 10.9 10.7 12.8 12.9
    Silty clay loam 62 4.8 5.5 6.4 5.8 4.8 4.9
    Sandy clay loam 51 2.9 3.2 4.4 4.3 3.4 3.6
    Clay loam 55 7.0 6.0 4.3 5.3 8.7 8.3
    Silty clay 34 5.6 6.8 6.6 6.8 3.9 3.9
    Clay 75 10.1 7.6 5.8 9.2 9.7 9.4

    Deviations from measured NCSS data for both Saxton and Rawls (2006) and new pedotransfer function predictions are shown in Figure 4, and asterisks indicate that paired t tests found significant differences between predicted and measured water content (α = .001). In noncalcareous soils, the new pedotransfer function predictions were significantly different from θPWP measurements in 11 texture classes, and Saxton and Rawls (2006) predictions were different for eight texture classes (Figure 4a). In calcareous soils, both the new pedotransfer function predictions and Saxton and Rawls (2006) were significantly different from θPWP measurements in four texture classes (Figure 4b). For θFC, and in noncalcareous soils, the new pedotransfer function predictions were significantly different from measurement in nine texture classes and Saxton and Rawls (2006) were significantly different in eight (Figure 4c). For calcareous soils, the new pedotransfer functions were significantly different from measured θFC in six texture classes and five for Saxton and Rawls (2006) (Figure 4d). For the new pedotransfer functions, deviations from NCSS measurements in θFC and θPWP had bias in the same direction, and the bias was lost in subtraction when calculating θAWHC. This resulted in improved performance for θAWHC (Figure 4e,f). The new pedotransfer functions predictions for noncalcareous soil were different from measurements in clay and silty clay loam textures, while Saxton and Rawls (2006) predictions were different from measurements in six texture classes (Figure 4e). For calcareous soils, both pedotransfer functions were similarly different from NCSS measurements in four to five texture classes (Figure 4f). For θAWHC predictions, there was only one soil texture in which new pedotransfer function predictions were different from NCSS measurements. For noncalcareous soils, the difference occurred in clay and in loamy sand for calcareous soil.

    Details are in the caption following the image
    Deviation from measured (measured minus predicted) water contents. Measured data is from soil samples from the National Cooperative Soil Survey Characterization database for permanent wilting point (θPWP), field capacity (θFC), and plant available water holding capacity (θAWHC). Predicated values were generated with Saxton and Rawls (2006) and the new pedotransfer functions

    For noncalcareous soils, the largest mean deviation from measured θAWHC was from Saxton and Rawls (2006) predictions in sandy loams and was 5.9 mm 100 mm−1 (Figure 4e). For calcareous soils, the largest mean deviation from measured θAWHC was from the new pedotransfer function in sand and was 19.8 mm 100 mm−1 (Figure 4f). Sand was underrepresented in the model and the next highest mean deviation (loamy sand) was half as large at 10.8 mm 100 mm−1. The comparison to NCSS data points out the new pedotransfer functions provided less accurate predictions of θPWP and θFC than Saxton and Rawls (2006), but predictions of θAWHC that were as or more accurate. The accuracy of the new pedotransfer functions is notable and encouraging, given that Saxton and Rawls (2006) functions were trained on NCSS data all using the same laboratory methodology for measurement (the clod method).

    3.2 Effect of soil organic C on predicted θAWHC

    The point of presenting the new set of pedotransfer functions is their greater response to increases in SOC relative to existing equations. The overlapping of all four gold LOESS fits in Figure 5a through 5f demonstrate that while organic matter content is statistically significant in Saxton and Rawls (2006) pedotransfer function there is little discernable change in θPWP, θFC, and θAWHC as SOC varies (<0.074% volumetric water content in response to an increase from 1 to 4% SOC content). Conversely, the new pedotransfer functions show changes in θPWP, θFC, and θAWHC with discernable changes in SOC values. The LOESS curves fit to predictions of water content show that changes in response to SOC were greatest at θFC and least for θPWP (Figure 5). Earlier knowledge described uniform effect of SOC on water retention (Jong, 1983; Riley,1981); however, our results agree with later findings by Minasny & McBratney (2018) that changes in water retention respond to SOC more for field capacity than for permanent wilting point. This effect of SOC was more pronounced for noncalcareous soils. Importantly, the effect of SOC on water retention was consistent across for all levels of SOC, that is, a 10 g kg−1 increase in SOC produced the same increase in volumetric water content regardless of the initial SOC content (Figure 5).

    Details are in the caption following the image
    LOESS fits to predicted volumetric water contents for permanent wilting point (θPWP), field capacity (θFC), and plant available water holding capacity (θAWHC) using simulated data for both new and Saxton and Rawls (2006) pedotransfer functions. Lines for Saxton and Rawls (2006) pedotransfer functions at the four levels of soil organic C overlap

    For noncalcareous soils with clay content not represented in the NAPESHM dataset (>60%) and SOC <20 g kg−1, the new pedotransfer functions predicted negative changes θAWHC values resulting from increased SOC. At these high clay contents, greater SOC causes θPWP to increase at a greater rate than θFC. This phenomenon was not expected, though similar outcomes have been reported in pedotransfer functions for saturated hydraulic conductivity (Ks). For example, Nemes et al. (2005) found negative correlations between Ks and organic matter increases in several pedotransfer functions and the range of soils that exhibited this negative correlation was dataset dependent. The pedotransfer function may be improved by adding observations representing these greater clay contents. At present, we recommend restricting clay contents to 60% when using Equations 1 and 2, which is the same range of clay content for the dataset used to create Saxton and Rawls (2006) as well as other datasets, for example, UNSODA (Nemes et al., 2001).

    Increases in SOC for calcareous soils with >35% clay content are also not represented well in the new pedotransfer functions, and there were only four calcareous experimental units that had 20 g kg−1 SOC or more. Because of this, we conclude that the usefulness of the calcareous functions are to illustrate that calcareous soils have different relationships with water holding capacity than noncalcareous soils. This difference is not well discussed in the soil science literature and needs more development. The new pedotransfer functions for calcareous soils (Equations 3 and 4) is not useful in soils with >35% clay, and we have limited Figure 3e and f accordingly.

    It is notable that Saxton and Rawls (2006) pedotransfer functions have similar predictions to the new pedotransfer functions in situations of high and low clay content and correspondingly low and high SOC. For example, the LOESS curves fit to predictions in Figure 5c are all similar for 10 g kg−1 SOC at 10% clay content and 40 g kg−1 SOC at 60% clay content. We speculate that this is because SOC (and its effects on θAWHC) in the NCSS data used to fit Saxton and Rawls (2006) is not management induced. Rather, it is driven largely by inherent properties so that, in essence, the effects of SOC are indistinguishable from the effects of clay content. For this reason, Saxton and Rawls (2006) pedotransfer functions are not able to capture the effects of improved soil management on θPWP, θFC, and θAWHC.

    To quantify the effect of SOC on the new predictions, we calculated summary statistics for predictions of θPWP, θFC, and θAWHC using the same simulated soil samples to which we fit LOESS curves. The summary statistics demonstrate that effects of SOC on predicted θPWP, θFC, and θAWHC are not constant across soil texture classes (Table 3). For noncalcareous soils, the effect of increasing SOC is most prominent in θAWHC, because of the combined effects at θPWP and θFC. This absolute effect is largest for fine textured soils. For example, changes in θAWHC due to increased SOC for fine textured soils is 30% greater relative to the mean of all textures, and coarse soils are 30% less relative to the mean of all textures. In calcareous soils, because SOC has no interaction term with texture for θPWP, the effect of increasing SOC is only in θFC. Coarse-textured calcareous soils have twice the water retention due to increased SOC compared with the mean of all textures (Table 3). Previous findings have shown that the magnitude of change resulting from increases in SOC was greatest in coarse-textured soil and least in fine-textured soils for θPWP, θFC, and θAWHC (Hudson, 1994; Minasny & McBratney, 2018). Our findings agree with past literature for the repacked cores used to make new θPWP pedotransfer functions, but our findings are not consistent with the literature for noncalcareous soils for either θFC or θAWHC. The differences between our study and previous work are likely caused by differences in procedures, especially (a) the experimental design of the NAPESHM dataset was developed to capture changes in SOC and water retention due to management (treatment) effects within sites as well as across sites; (b) previous studies combined calcareous and noncalcareous soils, but we separate them; and (c) differences in the θFC measurement methodology from some past studies.

    TABLE 3. Summary statistics for the effects of increase in soil organic C for predicted volumetric water content at permanent wilting point (θPWP), field capacity (θFC), and plant available water holding capacity (θAWHC) using the new pedotransfer functions on simulated data. For calcareous soils n = 270 and for noncalcareous soil n = 384. Particle size group, clay, sandy clay, silty clay, clay loam and silty clay loam are fine; loam, silt loam, silt, and sandy clay loam are medium; and sand, loamy sand, and sandy loam are coarse
    Change in water content (mm 100 mm−1) from 10 g kg−1 Increase in soil organic C
    Particle size group Min. Mean Max.
    Noncalcareous
    θPWP
    Coarse 0.9 1.5 1.9
    Medium –0.1 0.8 1.9
    Fine 0.5 1.3 1.9
    All –0.1 1.2 1.9
    θFC
    Coarse 2.8 4.2 5.2
    Medium 0.5 3.1 5.9
    Fine 2.6 5.1 7.0
    All 0.5 4.1 7.0
    θAWHC
    Coarse 1.9 2.7 3.4
    Medium 0.6 2.2 4.0
    Fine 2.1 3.8 5.0
    All 0.6 3.0 5.0
    Calcareous
    θPWP
    Coarse 0.5 0.5 0.5
    Medium 0.5 0.5 0.5
    Fine 0.5 0.5 0.5
    All 0.5 0.5 0.5
    θFC
    Coarse 3.0 4.1 5.6
    Medium –0.2 1.7 3.3
    Fine –0.9 0.2 1.2
    All –0.9 2.0 5.6
    θAWHC
    Coarse 2.5 3.6 5.2
    Medium –0.7 1.2 2.9
    Fine –1.3 -0.3 0.8
    All –1.1 1.5 5.2

    Increases in SOC can in some cases result in small reductions in predicted water content for the new pedotransfer functions. For noncalcareous soils, there was one simulated example of a silt (sand and clay at 5%) in which predicted θPWP decreased with more SOC (Table 3). This negative prediction was smaller than the RMSE (Equation 2) and is represented as the minimum value for noncalcareous θPWP predictions in medium textures soils in Table 3. For calcareous soils, there were 25 simulated examples that resulted in negative predicted changes in water content when SOC increased and these contributes to the negative values in Table 3 for calcareous soils. All were less than one-third of the RMSE for their respective equations. Thus, the new pedotransfer functions can predict small decreases in water content as the result of increased SOC. Given that the magnitude of these predictions is less than the RMSE of the equations, users may choose to interpret them as no change in water content.

    For noncalcareous soils across texture classes, the increase in θAWHC associated with a 10 g kg−1 increase in SOC ranged from 3.0 to 5.0 mm 100 mm−1 (Table 3). In calcareous soils, predicted changes in θAWHC from a 10 g kg−1 increase on SOC were smaller on average, but some larger changes were predicted. The range was 1.6–5.5 mm 100 mm−1. For comparison to these predicted changes, the mean change in θAWHC associated with a 10 g kg−1 increase in SOC reported by Minasny and McBratney (2018) was 1.2 mm 100 mm−1 across all textures. Thus, the mean effect of SOC on θAWHC reported by Minasny and McBratney (2018) is about the same as our finding for calcareous soil and about a third as large as our findings for noncalcareous soils. We attribute the greater increase in predicted θAWHC for the new pedotransfer functions to the fact that NAPESHM data used to create the function captures variance in SOC and measured θAWHC within pedons due to management practices. Another source of improvement may be the somewhat larger sample size for measures θFC; the intact clods used by Saxton and Rawls (2006) would have been approximately 8 × 6 × 6 cm so that they would fit snugly in the clod box (Soil Science Division Staff, 2017). The volume of a soil clod fitting in a clod box is approximately 72–78% of the volume of the core used in this investigation.

    This study demonstrates that, while there are cases in which the magnitude of change in θAWHC resulting from increased SOC will be negligible, other cases will show meaningful benefits of increasing SOC. To illustrate how these changes might be meaningful to crop production, a fine-textured soil with a 20 g kg−1 increase in SOC would increase θAWHC by 7.6 mm 100 mm−1 (Table 3). Extending this to 150-mm depth would result in 11.4 mm of additional plant available water. Assuming that plants took up this 11.4 mm of additional water and that it was recharged in five rainfall events throughout a growing season, the additional water available to plants over the growing season would be 57 mm (2.2 in, 570,000 L ha−1 yr−1). For reference as to whether this may be a meaningful increase, an estimate for the amount of water that corn needs for transpiration during the reproductive phase is about 5 mm per day. To determine whether increases in θAWHC are meaningful to stakeholders, considerations such as rainfall amount and timing and planting and harvesting of crops should inform situational analysis. This study enables such situational analyses by making available new pedotransfer functions for θPWP, θFC, and θAWHC that meaningfully respond to changes in SOC that result from soil management.

    4 CONCLUSIONS

    The newly developed pedotransfer functions for soil water retention have similar performance compared with previous models and is more sensitive to changes in SOC. The new functions provided robust estimates for 1,731 surface soils from the NCSS database. The magnitude of predicted increases in θPWP, θFC, and θAWHC in response to increased SOC was greater for noncalcareous soils, which showed a mean increase in θAWHC across all texture classes that was more than double that reported in earlier studies. The accuracy of new functions provides confidence that they are suitable to incorporate into models. Because they demonstrate a change in response to varying levels of SOC, they bridge the long-standing discrepancy between soil science textbooks and pedotransfer functions on the effects of SOC on θAWHC.

    While there was broad representation in soil types in the NAPESHM dataset used to develop the new pedotransfer functions, there were some conditions with limited data to train the model. For example, the functionality of the new pedotransfer function for noncalcareous soil is limited to no more than 60% clay. This study highlights the need for more mechanistic investigation and pedotransfer function development to understand water retention in calcareous soils.

    The newly developed pedotransfer functions show substantial effects of SOC on θAWHC and will enable future modeling to illuminate scenarios for which changes in soil management that result in increased SOC are likely to provide changes in water supplied to plants by the soil that are meaningful for stakeholders.

    ACKNOWLEDGMENTS

    The NAPESHM project is part of a broader effort titled, “Assessing and Expanding Soil Health for Production, Economic, and Environmental Benefits”. The project is funded by the Foundation for Food and Agricultural Research (grant ID 523926), General Mills, and The Samuel Roberts Noble Foundation. The content of this publication is solely the responsibility of the authors and does not necessarily represent the official views of the Foundation for Food and Agriculture Research. The application of NAPESHM data to develop pedotransfer functions for plant available water was supported by the United Soybean Board (project number 1920-172-0118). The authors acknowledge the following individuals and groups for their contribution to the long-term research sites: Melissa Bell, Nancy Creamer, Alan Franzluebbers, Tomas Moreno, Paul Mueller, Chris Reberg-Horton, Mike Zink, Matthew Helmers, Lisa Schulte-Moore, Matt Mortenson, Sean Vink, George Kapusta, Ronald Krausz, Karla Gage, Dr. Rachel Cook, Amanda Weidhuner, Orla Willoughby, Deanna L. Osmond, Bob Blevins, Donald C. Watts, Dr. Kenneth A. Barbarick, Robert S. Dungan, Joshua L. Heitman, James Custis, John Mason, April B. Leytem, Mark A. Liebig, Amy L. Shober, Michael L. Thompson, Bryan B. William, and Lamesa Cotton Growers. Foundation for Food and Agricultural Research (grant ID 523926), United Soybean Board (project no. 1920-172-0118), General Mills, and The Samuel Roberts Noble Foundation.

      AUTHOR CONTRIBUTIONS

      Dianna K. Bagnall: Conceptualization; Formal analysis; Investigation; Methodology; Writing-original draft. Cristine L. S. Morgan: Conceptualization; Funding acquisition; Supervision; Writing-review & editing, Michael Cope: Writing-review & editing. Gregory M. Bean: Writing-review & editing. Shannon Cappellazzi: Writing-review & editing. Kelsey Greub: Writing-review & editing. Daniel Liptzin: Formal analysis. Charlotte L. Norris: Writing-review & editing. Elizabeth Rieke: Writing-review & editing. Paul Tracy: Data curation. Ezra Aberle: Data curation. Oscar Bañuelos Tavarez: Data curation. Andy Bary: Data curation. R. Louis Baumhardt: Data curation. Alberto Borbón Gracia: Data curation. Daniel Brainard: Data curation. Jameson Brennan: Data curation. Dolores Briones Reyes: Data curation. Darren Bruhjell: Data curation. Cameron Carlyle: Data curation. James Crawford: Data curation. Cody Creech: Data curation. Steven Culman: Data curation. William Deen: Data curation. Curtis Dell: Data curation. Justin Derner: Data curation. Thomas Ducey: Data curation. Sjoerd Willem Duiker: Data curation. Miles Dyck: Data curation. Benjamin Ellert: Data curation. Martin Entz: Data curation. Avelino Espinosa Solorio: Data curation. Steven J. Fonte: Data curation. Simon Fonteyne: Data curation. Ann-Marie Fortuna: Data curation. Jamie Foster: Data curation. Lisa Fultz: Data curation. Audrey V. Gamble: Data curation. Charles Geddes: Data curation. Deirdre Griffin-LaHue: Data curation. John Grove: Data curation. Stephen K. Hamilton: Data curation. Xiying Hao: Data curation. Z. D. Hayden: Data curation. Julie Howe: Data curation; Writing-review & editing. James Ippolito: Data curation. Gregg Johnson: Data curation. Mark Kautz: Data curation. Newell Kitchen: Data curation. Sandeep Kumar: Data curation. Kirsten Kurtz: Data curation. Francis Larney: Data curation. Katie Lewis: Data curation. Matt Liebman: Data curation. Antonio Lopez Ramirez: Data curation. Stephen Machado: Data curation. Bijesh Maharjan: Data curation. Miguel Angel Martinez Gamiño: Data curation. William May: Data curation. Mitchel McClaran: Data curation. Marshall McDaniel: Data curation. Neville Millar: Data curation. Jeffrey P. Mitchell: Data curation. Philip A. Moore: Data curation. Amber Moore: Data curation. Manuel Mora Gutiérrez: Data curation. Kelly A. Nelson: Data curation. Emmanuel Omondi: Data curation. Shannon Osborne: Data curation. Leodegario Osorio Alcalá: Data curation. Philip Owens: Data curation. Eugenia M. Pena-Yewtukhiw: Data curation; Writing-review & editing. Hanna Poffenbarger: Data curation. Brenda Ponce Lira: Data curation. Jennifer Reeve: Data curation. Timothy Reinbott: Data curation. Mark Reiter: Data curation. Edwin Ritchey: Data curation. Kraig L. Roozeboom: Data curation. Ichao Rui: Data curation. Amir Sadeghpour: Data curation. Upendra M. Sainju: Data curation. Gregg Sanford: Data curation. William Schillinger: Data curation. Robert R. Schindelbeck: Data curation. Meagan Schipanski: Data curation. Alan Schlegel: Data curation. Kate Scow: Data curation. Lucretia Sherrod: Data curation. Sudeep Sidhu: Data curation. Ernesto Solís Moya: Data curation. Mervin St. Luce: Data curation. Jeffrey Strock: Data curation. Andrew Suyker: Data curation. Virginia Sykes: Data curation. Haiying Tao: Data curation. Alberto Trujillo Campos: Data curation. Laura L. Van Eerd: Data curation. Nele Verhulst: Data curation. Tony John Vyn: Data curation. Yutao Wang: Data curation. Dexter Watts: Data curation. David Wright: Data curation. Tiequan Zhang: Data curation. Charles Wayne Honeycutt: Data curation

      CONFLICT OF INTEREST

      The authors declare no conflict of interest.