King et al. (2026). Soil Organic Carbon Improves Crop Yield and Yield Resilience

| Eunkyoung Choi

| Jesse Burkhardt

| Nathaniel D. Mueller

2,4

1

School of Food and Agriculture, University of Maine, Orono, Maine, USA |

2

Department of Ecosystem Science and Sustainability, Colorado State

University, Fort Collins, Colorado, USA |

3

Department of Agricultural and Resource Economics, Colorado State University, Fort Collins, Colorado,

USA |

4

Department of Soil and Crop Sciences, Colorado State University, Fort Collins, Colorado, USA

Correspondence:

Alison E. King (

alison.king@maine.edu

)

Received:

15 August 2025 |

Revised:

15 March 2026 |

Accepted:

23 March 2026

Keywords:

ABSTRACT

Increasing soil organic carbon (SOC) has been proposed as a strategy to promote crop yield resilience under extreme hydro-
climatic stress, particularly drought, due to its positive effect on soil available water- holding capacity. We analyze how SOC
mediates the relationship between US rainfed crop yields (1981–2020) and growing season hydroclimatic conditions—defined
by soil water supply and atmospheric water demand—over 67,000 county- years across three major crops. We show that higher
SOC is consistently associated with increased crop yields and yield resilience, evidenced through reduced interannual variabil-
ity. Contrary to prevailing expectations, the largest yield gains from SOC occur during moderate water supply conditions—not
drought. Under drought, water supply to crops may be more limited by water inputs to soil than by soil water- holding capac-
ity, constraining the benefit provided by SOC regeneration. Furthermore, because moderate conditions are more frequent than
drought or wet extremes, the largest production gains from SOC accumulate under moderate conditions. These findings indicate
that SOC regeneration can enhance drought resilience to some degree but cannot compensate for extreme water scarcity; the
services SOC provides to crops, including water storage, require water for their effective delivery.

1 | Introduction

Climate change is driving rising temperatures and altering

precipitation patterns, leading to more frequent departures

from historically average conditions (Lesk et al.

; Ridder

et al.

; Rosenzweig et al.

2002

). Agriculture, which is in-

herently exposed to climate variability, is under growing pres-

sure to adapt in order to sustain food security (Schmidhuber

and Tubiello

2007

; Wang et al.

). Increasing soil organic

carbon (SOC) is proposed as a strategy to promote crop yield re-

silience under a changing climate (Bradford et al.

; Chenu

et al.

; Paustian et al.

2016

; Wang et al.

), often with

the expectation that SOC will reduce crop yield losses due to

extreme weather (Bot and Benites

; Bradford et al.

Lal

). The regeneration of SOC also contributes to cli-

mate change mitigation through increased terrestrial C storage

(Amelung et al.

; Minasny et al.

), therefore offering a

promising “win- win” strategy for both adaptation and mitiga-

tion. While positive associations between SOC and mean crop

yield are well- documented (Ma et al.

; Oldfield et al.

),

it remains unclear the extent to which SOC enhances crop yield

resilience.

Soil organic carbon is expected to confer yield resilience by pro-

tecting crops specifically against drought- related yield losses

(Kane et al.

; Lal

). Mechanistically, SOC is thought to

reduce crop exposure to water stress during drought by increas-

ing soil plant- available water holding capacity (AWC) (Bagnall

et al.

; Hudson

1994

; Minasny and McBratney

2018

;

Rawls et al.

2003

). However, studies that quantify both SOC

and AWC cast doubt on this mechanism, finding that the ef-

fect of SOC on crop yield is not attributable to changes in AWC

2 of 13

Global Change Biology,

2026

(Baldwin- Kordick et al.

; Kane et al.

; Renwick

).

The disconnect between the lack of support for the AWC mech-

anism and evidence that SOC supports yields under drought

(Kane et al.

) underscores uncertainty about when and

how SOC enhances crop yields. SOC provides multiple services

that could support crop yield across diverse growing conditions,

including nitrogen (N) supply from associated organic mat-

ter and soil structural improvements (Booth et al.

2005

; King

et al.

). Understanding which growing conditions activate

these benefits will be critical to identifying the contexts in which

SOC benefits crops.

Statistical models of crop yield provide a valuable framework

for evaluating how hydroclimatic conditions, such as drought,

affect crop yield resilience. Models that incorporate both atmo-

spheric water demand and root- zone soil moisture have been

shown to outperform those relying solely on temperature or

precipitation in predicting US maize yields (Choi et al.

;

Rigden et al.

). While precipitation is often used as a

proxy for water supply (Li et al.

; Lobell et al.

Rosenzweig et al.

), soil moisture offers a more accurate

representation by accounting for runoff, drainage, and evapo-

transpiration (Basso and Ritchie

2014

). Atmospheric water de-

mand frequently coincides with high temperatures and plays

a critical role in modulating crop sensitivity to water avail-

ability (Butler and Huybers

2015

; Carter et al.

2018

; Lobell

et al.

2011

; Ortiz- Bobea et al.

; Urban et al.

2015

; Zhao

et al.

). Despite these insights, no studies thus far have

jointly considered water supply and water demand to investi-

gate the potential role of SOC in buffering crop yields against

hydroclimatic stress.

Here, we analyze how SOC influences US rainfed crop yields

and yield resilience (1981–2020) across the full spectrum of hy-

droclimatic conditions, represented by the interaction between

growing season atmospheric water demand (vapor pressure defi-

cit, VPD) and soil water supply (plant- available water). Using a

statistical regression model, we evaluate the joint impacts of

VPD and plant- available water on the yields of maize, soybean,

and spring wheat as a function of observed SOC (“true SOC”).

To quantify the potential benefits of SOC regeneration, we simu-

late two SOC regeneration scenarios (“regen. SOC”): a uniform

increase of 10 g SOC kg soil

−

1

and SOC regeneration informed

by the Intergovernmental Panel on Climate Change (IPCC) SOC

change factors (Ogle

). For each crop, we compare yield

under three categories of hydroclimatic conditions (drier, mod-

erate, wetter) with and without SOC regeneration.

2 | Materials and Methods

2.1 | Data

We used crop yield and weather data from Choi et al. (

).

Briefly, annual rainfed yields for maize, soybeans, and spring

wheat at county- level were downloaded from the USDA National

Agricultural Statistics Service (USDA NASS

). We used SOC

range of the cropped counties as a criterion for inclusion of each

crop. While sorghum was also reported in Choi et al. (

), the

SOC range of sorghum- cropped counties was too narrow to be

included in this study. We converted yields that were reported in

bushel per acre to kg/ha using crop conversion factors (USDA

ERS

1992

). As our analysis was focused on rainfed agriculture,

we limited our analysis to counties with non- irrigated agri-

cultural comprising at least 10% of harvested area (Butler and

Huybers

2013

; Choi et al.

; Ortiz- Bobea et al.

). Weather

data for statistical modeling was downloaded from the PRISM

Climate Group (Oregon State University

), which included

daily minimum and maximum vapor pressure deficit (VPD) and

daily minimum and maximum temperature. Weather data for

IPCC climate classification were downloaded from the NASA

Monthly Historical Water Balance for the CONUS archive, which

included potential evapotranspiration (PET) (NASA

2025

).

Daily root- zone soil water was obtained from the National

Climate Assessment- Land Data Assimilation Systems (NCA-

LDAS) Version 3 (Jasinski et al.

), and we used averaged soil

water (m

3

/m

3

) over the 0–40 cm layer. The gridded PET, VPD,

and soil moisture (SM) data were averaged to county- level after

accounting for grid cells with cropland, based on the Global

Food Security- Support Analysis Cropland Extent 2010 North

American 30 m data (Massey et al.

). County boundaries

were based on the map from the 2017 US Census Bureau.

Soil organic matter (SOM), sand, and clay values were extracted

from gSSURGO (Soil Survey Staff

2015

). The gSSURGO da-

tabase is organized as a series of polygons, termed map units,

which are composed of component units. Soils information is

reported at the level of component unit with variable soil depths.

We used the

mpspline2

package in R (O'Brien

) to spline soil

properties to a consistent 0–40 cm layer, and then averaged soil

properties to the map unit level based on the proportional area

of component units. We then converted soil data to raster for-

mat and applied the same cropland filter used for SM and VPD

data, described above, to calculate county- average soil data. Soil

organic matter was converted to soil organic carbon via multi-

plication by 0.58 (Pribyl

2010

). Counties with greater than 100 g

SOC kg soil

−

1

(17% OM) were deemed extreme (Kane et al.

)

and excluded from subsequent analyses.

2.2 | Plant- Available Water

We calculated plant- available water using the estimates of water

held at permanent wilting point and field capacity based on the

county- specific sand, clay, and SOM values derived from gS-

SURGO and pedotransfer functions (Saxton and Rawls

2006

).

Specifically, if volumetric water was less than water held at

permanent wilting point, we used 0 for plant- available water.

If volumetric water was between field capacity and permanent

wilting point, we calculated plant- available water as the differ-

ence between volumetric water and water held at permanent

wilting point. If volumetric water was greater than water held at

field capacity, we calculated plant- available water as the differ-

ence between water held at field capacity and permanent wilt-

ing point.

2.3 | Soil- Agnostic Water Supply

The effect of SOC on crop yield could manifest through an ef-

fect of SOC on exposure to different water contents throughout

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

3 of 13

Global Change Biology,

2026

the growing season (Bagnall et al.

). To the extent this

is the case, including raw water content values in our model

could prevent accurate estimation of SOC effects on yield. We

therefore aimed to calculate a water content index that would

disregard soil edaphic influence on water content and instead

reflect purely the influence of water input from precipitation

and water loss from evapotranspiration. Therefore, for each

crop, we identified the best- fitting linear model of LDAS SM

(0–40 cm) using a candidate set of predictors: vapor pressure

deficit (VPD), precipitation (

P

), and daily average tempera-

ture (

T

) on the current day and preceding day, soil water

the preceding day (SM

t

=

−

1

) and the interaction of VPD and

daily average temperature on both current and preceding day

(VPD ×

T

av

). Best- subset selection was performed on a ran-

dom 10% subsample of county- days using the

leaps

package

in R (Lumley

), with the best model identified by AIC

(Table

S1

). We extracted model intercept and coefficients for

each predictor variable and applied these recursively to each

day and county in the dataset. In this approach, we initial-

ized the calculation of the water supply index for each county

with a single starting SM value (0.3), but every subsequent

SM value was calculated using weather input specific to that

county and day as mediated by the global influence of weather

on SM derived from the entire dataset. The SM values for the

preceding day were recursively used as input to calculate SM

on the current day.

2.4 | County Clusters

Statistical models of panel data often control for time- invariant

factors by including fixed effects for spatial units. While these

fixed effects can be composed of a single county for some ap-

plications (Choi et al.

), in order to leverage variability

in SOC to estimate the effect of SOC on yield and yield sta-

bility, larger clusters are needed. At the same time, clusters

should not be so large as to introduce unmitigated variability

in yield within the clusters. Previous studies seeking to iden-

tify between- county spatial drivers have leveraged variability

within states (Kane et al.

). However, using states as the

spatial unit introduces unnecessary variation in ecological

conditions within the state, as well as uneven variation in the

range of ecological conditions between states. We therefore

developed a process to use size- constrained clustering in the

scclust

package in R (Savje et al.

) to delineate ecologically

relevant clusters.

First, we identified variables that influenced yield (average

growing season precipitation, temperature, and VPD; lati-

tude and longitude) and identified the change in each of these

variables that corresponded to a 15% distance along the total

yield range for each crop. We did this by regressing average

county- level yield of each crop against each of these five vari-

ables, fitting with either a linear or quadratic model based on

visual inspection of the relationships. We then determined the

change in predictor variable that was associated with the fixed

change in yield. For quadratic relationships, this range was

determined by taking the derivative of the slope at the 25th,

50th, and 75th percentile of the predictor data and then aver-

aging these slopes.

Then, we applied size- constrained clustering (

scclust

package)

across a range of cluster sizes. At each cluster size, we checked

the proportion of biophysical clustering variables (average

growing season precipitation, temperature, and VPD; latitude

and longitude) for which the range was smaller than the range

associated with 15% of the yield range, and defined “success”

as the range of the biophysical variable in the cluster not ex-

ceeding the range associated with 15% of the yield range. We

chose the cluster size (number of counties) that corresponded

to a 75% success of these tests. To evaluate how the choice of

threshold for crop yield range influenced model output, we

also evaluated the use of 10% and 20% of the crop yield range

in the clustering procedure for the SOC regeneration scenario

(Table

S2

).

2.5 | Statistical Regression Model

We adapted a statistical regression framework that examines the

relationship between crop yield and a matrix of daily SM and

VPD exposure (Choi et al.

; Rigden et al.

) by incorpo-

rating SOC into this framework. We used the state- specific full

growing seasons of Choi et al. (

) to delineate calendar days

used in the statistical model. Daily water supply (plant- available,

volumetric, or soil- agnostic), and VPD values were binned into

their 0–20th, 20th–80th, and 80th–100th percentiles (Table

).

We included interaction of SOC with each water supply- VPD bin

to analyze the response of rainfed yields

Y

in county cluster

i

and year

t

:

where

X

d,s,i,t

is the exposure time (% of growing season) to each

VPD

d

and water supply

s

bins. We excluded one of the nine

water supply- VPD bins to avoid perfect multicollinearity of the

explanatory variables.

𝛽

d,s

are the coefficients for % of grow-

ing season exposure to each water supply- VPD conditions, and

they are interpreted as the change in yield (kg/ha/%) relative

to the excluded bin. The interaction of SOC with water- supply-

VPD bin is described by

SOC

i

⋅

𝜃

d,s

.

Z

i,t

is the exposure time (%)

to low temperature with

𝛾

as its coefficient; for spring wheat,

the effect of exposure to low temperatures (

Z

i,t

) was excluded

(Choi et al.

).

f

i

are year- level fixed effects that account for

changes in technology or policy, and

𝛿

t

are the county cluster

fixed effects, described above, that account for spatial variabil-

ity. Performance of models for all crops and water supply scenar-

ios are shown in Table

S4

We used the

predict()

function in R (v 4.2.3) and SOC values with

either a uniform 10 g kg soil

−

1

increase or an increase in SOC

informed by IPCC SOC change factors (Ogle

) to calculate

projected changes in crop yield dynamics in the modeled SOC

regeneration scenario, based on the fitted model described in

Equation (

). To visualize how SOC moderates the effect of ex-

posure to hydroclimatic conditions on crop yield, we again used

the

predict()

function to generate crop yield at the 10th, 50th,

and 90th percentile of SOC by modifying the SOC variable in

the dataset.

(1)

Y

i,t

=

3

∑

d

=

1

3

∑

s

=

1

(

X

d,s,i,t

𝛽

d,s

+

X

d,s,i,t

SOC

i

𝜃

d,s

)

+

Z

i,t

𝛾

+

SOC

i

𝜓

+

f

i

+

𝛿

t

+ ∈

i,t

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

4 of 13

Global Change Biology,

2026

2.6 | Hydroclimatic Year Clustering

To classify counties and years based on hydroclimatic condi-

tions and modeled yield, we performed unsupervised cluster-

ing using a combination of water supply percentile bins and

VPD bins (described above), and yield. Yield was scaled to a

0–100 range to ensure equal weighing with the hydroclimatic

exposure bins, which were in the 0–100 range. To determine the

optimal number of clusters, we conducted a silhouette analysis

on a random subsample of the data (

n

= 10,000 for maize and

soybeans;

n

= 1000 for spring wheat). For each candidate num-

ber of clusters (

k

= 2–6), we computed k- means partitions using

the SimpleKMeans function from the RWeka package (Hornik

et al.

2009

; Witten and Frank

2005

) and calculated the average

silhouette width. Based on this analysis, we selected

k

= 3 for

all crops.

We then performed k- means clustering on the full dataset using

k

= 3 and assigned each county and year to one of three clusters.

After visual inspection of yield and VPD trends across clusters,

the clusters were labeled post hoc based on the average repre-

sentation of soil water supply bins within each group: clusters

with higher frequency in the lowest water percentile bins were

labeled as “drier,” those with intermediate frequency as “mod-

erate,” and those with higher representation in the highest soil

water bins as “wetter” (Table

S4

).

2.7 | Effects of SOC Regeneration on Yield

and Production

We explored two SOC regeneration scenarios: (1) “Regen IPCC”: in-

creases in SOC guided by IPCC SOC change factors and (2) “Regen

+ 10”: a uniform SOC increase of 10 g SOC kg soil

−

1

. For the IPCC

scenario, counties were assigned climate groups based on the clas-

sification scheme for default climate regions (Reddy et al.

).

Across all crops, counties were predominantly classified as warm

temperate moist (52%), followed by cool temperate moist (23%),

warm temperate dry (14%), cool temperate dry (11%), and tropical

dry (1%). Initial gSSURGO concentrations for SOC were multi-

plied by F

MG

factors for zero- tillage and F

I

factors for “high with-

out manure input”, representing a no- till cover cropped system

(Ogle

). Yield differences for each scenario were calculated as

[predicted crop yield under SOC regeneration]

−

[predicted crop

yield under true SOC]. Production differences were calculated as

the sum of [yield difference × hectares in crop production] across

all counties for a single year. Values outside the 1st–99th percen-

tile range within crop were excluded prior to analysis and visu-

alization to reduce the influence of extreme outliers. Differences

in crop yield between SOC scenarios or hydroclimatic year type

were evaluated within crop using paired Wilcoxon signed- rank

tests comparing each SOC regeneration scenario to the baseline

(true SOC), with Bonferroni correction applied for multiple test-

ing. Differences in yield response among hydroclimatic year types

were evaluated separately within crop and scenario using analysis

of variance followed by Tukey's HSD test.

3 | Results and Discussion

3.1 | Yields and Resilience Benefits From SOC

Regeneration Across All Years

Under projected SOC regeneration scenarios, crop yields and yield

resilience increased for all crops (Figure

). Average absolute yield

increases were greater under a 10 g SOC kg soil

−

1

regeneration

than IPCC- informed SOC regeneration, ranging from 384 kg ha

−

1

(maize) to 160 kg ha

−

1

(soybeans) to 94 kg ha

−

1

(spring wheat). Yield

FIGURE 1

| Changes in crop yield and yield resilience under two modeled soil organic carbon (SOC) regeneration scenarios, shown for all years.

(A) Yield and yield differences, (B) coefficient of variation (CV) and CV differences, (C) standard deviation (SD) and SD differences. True SOC = ob-
served SOC. Regen IPCC = regenerated SOC, increased from true SOC based on IPCC SOC change factors. Regen +10 = regenerated SOC, increased
from true SOC by 10 g kg soil

−

1

for all counties. Boxes (main plots) show the interquartile range with median lines; whiskers extend to the most ex-

treme values within 1× the interquartile range. Circles (inset plots) represent differences in yield between each SOC regeneration scenario and true
SOC. Prongs on circles represent 1st and 3rd interquartile range. The coefficient of variation (CV) and standard deviation (SD) are calculated within
counties over all years. Asterisks indicate difference between true SOC and nearest SOC regeneration scenario, ***

p

< 0.001 (Wilcoxon signed rank

test with Bonferroni correction). Water supply (WS) conditions delineated with plant- available water (0–40 cm).

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

5 of 13

Global Change Biology,

2026

increases under IPCC- informed SOC regeneration were smaller

(129 kg ha

−

1

for maize; 53 kg ha

−

1

for soybeans; 19 kg ha

−

1

for spring

wheat). These yield increases were associated with greater yield

resilience, as measured by decreases in the coefficient of variation

(CV) and standard deviation (SD) of yields (Figure

1B,C

). The re-

duction in SD with the 10 g kg

−

1

SOC regeneration was largest in

maize (from 1652 to 1619 kg ha

−

1

) and smallest for soybean (from

454 to 442 kg ha

−

1

). The reduction in CV with SOC regeneration

was largest for spring wheat (from 23.7% to 21.3%) and smallest for

soybean (19.5%–17.4%).

Yield improvements in the 10 g kg soil

−

1

regeneration scenario

were larger because average SOC change under the IPCC SOC

factors we selected was smaller. For maize, the mean SOC in-

crease was 3.0 g SOC kg soil

−

1

in the IPCC scenario (interquartile

range: 2.1–4.6). This magnitude of SOC increase was partly due

to the choice of SOC factors, which did not include the highest F

I

factors that represent manure input (Ogle

). Manure inputs

are infeasible over large areas but, when applied at high rates,

can increase SOC far above the 10 g SOC kg soil

−

1

change repre-

sented in the uniform regeneration scenario (Gulde et al.

2008

).

Thus, the IPCC regeneration scenario represents a moderate

SOC increase, in which SOC changes are modulated by climate.

Increases in SOC are lower in more arid areas, which reflects

the limitation to carbon inputs to soil that constrain accumula-

tion of SOC under regenerative management (King et al.

).

The uniform SOC increase scenario represents a higher, on av-

erage, SOC increase that is feasible under some conditions with

long- term regenerative management (King et al.

), or exog-

enous C inputs. This scenario allows for the examination of the

interaction of SOC regeneration with hydroclimatic conditions,

without a confounding influence of SOC change.

Regenerating SOC in croplands is promoted as a way to restore some

of the SOC lost from land conversion to agriculture (Sanderman

et al.

) and advance nature- based climate solutions (Amelung

et al.

; Minasny et al.

). While the climate mitigation poten-

tial of SOC- regenerating practices is relatively modest (McClelland

et al.

2025

), realizing this potential depends on farmers' willing-

ness to adopt regenerative practices. Soil carbon markets aim to

incentivize such adoption (Kreibich and Hermwille

), but

uptake has been limited, in part due to barriers to market entry.

Instead, farmers often cite on- farm co- benefits—rather than cash

payments—as motivations for implementing regenerative prac-

tices (Buck and Palumbo- Compton

). Although inherent soil

properties, such as texture, may also influence crop yield resil-

ience (Qiao et al.

), since texture is not usually considered a

management- responsive soil property, our analyses focus on SOC.

Our results suggest that investing in SOC can produce co- benefits

for farmers, indicating potential for SOC regeneration to “incentiv-

ize itself” via yield increases and yield stabilization.

Soil physicochemical properties, including SOC, can influ-

ence soil water status (Araya and Ghezzehei

; Bagnall

et al.

). We therefore developed and applied a soil- agnostic

water supply index to reveal solely the influence of climatic

drivers on soil water supply (Table

S1

). We replaced plant-

available soil water (shown throughout Main text) with the

soil- agnostic water supply or volumetric water in the statisti-

cal model. We found that key results emerging from analyses

are robust to different ways of conceptualizing water supply.

For Figure

, alternative water supply results (Figures

S1

and

S2

) reveal very similar effects of SOC on crop yield and yield

resilience.

3.2 | Yield Benefits of SOC Are Greatest in

Moderate and Wetter Years

We show that crop yield benefits from SOC regeneration were

greatest in moderate or wetter years, relative to drier years, for

maize and soybeans (Figure

2A,B

). This finding contrasts with

widely held expectations and prior works (Bot and Benites

Bradford et al.

; Kane et al.

; Lal

), which suggest

that SOC provides the greatest yield benefits under extreme

drought. This discrepancy may stem from methodological dif-

ferences in how climatic conditions are defined. Prior studies

typically isolate drought years, combining all non- drought years

into reference groups (Iizumi and Wagai

; Kane et al.

Mahmood et al.

). As a result, yield benefits from SOC may

appear strongest during drought simply because the comparison

group(s) includes both optimal and suboptimal years for SOC

effects (see Figure

). In contrast, we examine growing- season

exposure to the full spectrum of hydroclimatic conditions

(Figure

3A

), then use k- means clustering to identify three data-

informed hydroclimatic clusters across crops (Table

S2

), cor-

responding to wetter, drier, and moderate years. Furthermore,

because we define hydroclimatic exposure in terms of absolute,

rather than locally normalized, water supply and evaporative de-

mand, our hydroclimatic conditions are consistent within each

crop (Table

), which may also contribute to the identification

of extreme drought as a suboptimal condition for SOC- mediated

benefits to crops.

Crop yield benefits can also be conceptualized in terms of pro-

duction by multiplying yield benefits by total harvested area for

each year. We show that the frequency of hydroclimatic condi-

tions moderates the impact of these hydroclimatic conditions

on crop production. For instance, moderate years were approx-

imately 2.3–2.8 times more likely to occur than either drier or

wetter years (Table

S5

). Therefore, production benefits from

SOC were highest for moderate years for all crops (Figures

S5–

S7

). Our analyses reveal a predominant and overlooked pathway

through which SOC improves cropping system output—by sup-

porting yield under the moderate growing conditions that occur

most frequently.

3.3 | Protective Effects of SOC Across the Full

Spectrum of Hydroclimatic Conditions

When considering the full range of hydroclimatic conditions

outlined in our model (Figure

3A

), the ability of SOC to promote

yield was most pronounced under a specific condition: moder-

ate water supply and high atmospheric water demand (Figure

,

WS

medium

WD

high

panels). In Figure

, each panel displays rela-

tionships between growing season exposure to panel- specific

hydroclimatic growing conditions and crop yield. Each hydro-

climatic condition is defined by specific thresholds of plant-

available soil moisture and VPD (Table

). The lines within

each panel represent predicted yield at different levels of SOC.

Differences between the lines in each panel therefore signify the

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

6 of 13

Global Change Biology,

2026

interaction of SOC with increasing exposure to each hydrocli-

matic condition.

Our finding that the greatest benefits from SOC occur under

moderate water supply, specifically WS

medium

WD

high

, contra-

dicts expectations that SOC is most effective at protecting crops

from drought (i.e., low water supply and high atmospheric water

demand, WS

low

WD

high

). However, the result also corroborates

our finding that yield effects from a simulated SOC increase

were usually smaller in drier years than moderate or wetter

years (Figure

). The significant, but relatively limited, effect

of SOC in increasing crop yields due to exposure to drought

(WS

low

WD

high

) as compared to other growing conditions

(Figure

) likely reflects the inability of SOC to increase water

supply—or provide other benefits—when plant- available water

is consistently low.

To understand why SOC benefitted yield under moderate water

supply more so than drought, we looked at how often the soil

reached its maximum water storage capacity (i.e., field ca-

pacity, Figure

S10

). First, we identified extreme conditions:

years and counties in which exposure to either WS

low

WD

high

(drought) or WS

medium

WD

high

was in the 90th percentile or

greater. Then, we counted the number of days for which plant-

available water was at its maximum, as the influence of SOC on

available water storage is expected to be realized by increasing

water held at this maximum, which can then be drawn down

by transpiration. The relatively few days in extreme drought

years (WS

low

WD

high

) in which plant- available water reached

its maximum (mean = 0.8 days; Figure

S10

) indicates that water

supply was limited by inputs of water from precipitation, rather

than by the capacity of soil to store water. In contrast, years and

counties with extreme WS

medium

WD

high

conditions exhibited

more days in which plant- available water reached its maximum

(mean = 7.9 days, Figure

S10

). These are the conditions under

which SOC regeneration could augment plant- available water

storage, which aligns with our observation that the interaction

of SOC with WS

medium

WD

high

was greater than for drought. Put

succinctly, SOC cannot store more water if there is no water

to store.

The protective effect of SOC under crop exposure to these

medium water supply and high atmospheric water demand

(Figure

, WS

medium

WD

high

panels) may also relate to inter-

actions between water, heat stress, and nutrient availability.

First, exposure to high atmospheric water demand (WD

high

)

correlates with warmer air and soil temperatures, so crops

may experience heat stress on a portion of medium water sup-

ply and high atmospheric water demand (WS

medium

WD

high

)

days (Grossiord et al.

; Lobell et al.

2011

). Given the pres-

ence of soil water, SOC may increase the water transpired by a

growing crop, thus sustaining gas exchange and cooling leaves

(Li et al.

). Second, soils high in SOC under warm condi-

tions with moderate soil water represent ideal conditions for

mineralization of soil organic nitrogen (N) to plant- available

forms (Booth et al.

2005

; Guntiñas et al.

2012

). Nutrient sup-

ply can help crops withstand heat stress (Waraich et al.

2012

)

in addition to supporting productivity. Third, high SOC may

FIGURE 2

| Projected changes in crop yield under two modeled soil organic carbon (SOC) regeneration scenarios, shown for clusters of hydrocli-

matic years. All outcomes shown for maize (A), soybeans (B), and spring wheat (C). Regen IPCC = regenerated SOC, increased from true SOC based
on IPCC SOC change factors. Regen +10 = regenerated SOC, increased from true SOC by 10 g kg soil

−

1

for all counties. Types of hydroclimatic years

shown along x axis are identified based on k- means clustering of unique counties and years. Boxes (main plots) show the interquartile range with
median lines; whiskers extend to the most extreme values within 1× the interquartile range. Circles (inset plots) represent differences in yield be-
tween each SOC regeneration scenario and true SOC. Prongs on circles represent 1st and 3rd interquartile range. Letters beneath inset circles indicate
statistically distinct yield responses to SOC regeneration between weather conditions within a SOC regeneration scenario, as determined by Tukey's
Honest Significant Difference (HSD) test (

α

= 0.05) following a one- way ANOVA. Water supply (WS) conditions delineated with plant- available water

(0–40 cm). These analyses with different water supply indices (soil- agnostic and volumetric) are shown in Figures

and

S4

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

7 of 13

Global Change Biology,

2026

enhance capacity for crop nitrogen or water capture due to soil

structural improvements (King et al.

). As testing these

mechanisms directly is beyond the scope of this work, future

experimental or process- based model studies are needed to

validate the role of each mechanism.

3.4 | Spatial Variability in Crop Response to SOC

Regeneration

The spatial distribution of yield responses to SOC regeneration

(Figure

4

) reflects several factors, including the county- level

FIGURE 3

| Soil organic carbon (SOC) moderates the effect of exposure to hydroclimatic conditions on crop yield. (A) Growing season exposure

to different hydroclimatic conditions, averaged across all crops, where tile size is proportional to growing season exposure. X- axis is water supply
(WS): Plant- available soil water, 0–40 cm. Y- axis is water demand (WD): Vapor pressure deficit (VPD). The average percentage of growing season
exposure is noted within tile. Shades of orange correspond to low soil water, shades of green to medium soil water, and shades of purple to high soil
water. Darker shades of each color correspond to higher VPD. Percentile thresholds for each crop are provided in Table

. (B–D) Effect of increasing

exposure to hydroclimatic conditions on crop yields of maize (A), soybeans (B), and spring wheat (C) at three different levels of SOC (solid = 90th SOC
percentile; dashed = 50th SOC percentile; dotted = 10th SOC percentile). Asterisks indicate the

p

- value of the interaction effect between hydroclimat-

ic condition and SOC: ***

p

≤

0.001; **

p

≤

0.01; *

p

< 0.05. Each plot also includes a joint density plot of yield data by hydroclimatic condition. Model

equation in Section

. Water supply (WS) conditions delineated with plant- available water (0–40 cm). These analyses with different water supply

indices (soil- agnostic and volumetric) are shown in Figures

S8

and

S9

.

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

8 of 13

Global Change Biology,

2026

FIGURE 4

| Legend on next page.

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

9 of 13

Global Change Biology,

2026

hydroclimatic conditions (i.e., Figure

) and initial SOC con-

centrations (in the 10 g kg

−

1

SOC increase scenario), onto which

are layered the feasibility of SOC regeneration, as determined

by IPCC SOC change factors (in the IPCC- informed scenario).

Counties with lower initial SOC content exhibited greater in-

creases in crop yield and yield resilience relative to soils with

higher initial SOC content across all crops (Figures

S13–S17

),

suggestive of a nonlinear response to SOC regeneration. Current

estimates of the thresholds of SOC concentration at which crops

no longer respond to increases in SOC differ wildly between

studies and crops, from < 12 g kg

−

1

SOC (Vendig et al.

) to

> 40 g kg

−

1

SOC (Ma et al.

https:// gdg. sc. egov. usda. gov/

); our results converge in a more

moderate SOC response threshold, with a range of initial SOC

concentrations of ~20–30 g SOC kg soil

−

1

. The IPCC- informed

SOC regeneration scenario shows the intersection of these

conditions, for instance, in the western range of spring wheat

(Figure

4E

). This area is classified as “cool temperate – dry” in

the IPCC classification, which has lower SOC change factors

relative to other climate groups in this study. Since the IPCC

SOC change factors are multiplied by the initial SOC content,

soils with lower initial SOC increase by a lower absolute amount

compared to soils with a higher initial SOC content. The west-

ern range of spring wheat is also characterized by a low initial

SOC content, resulting in a minimal SOC increase (< 1 g SOC

kg soil

−

1

in some counties) that corresponds to the minimal pre-

dicted yield change.

3.5 | Implications for Climate Mitigation

and Adaptation

Our results imply that regenerating SOC can simultaneously pro-

mote crop yield and crop yield resilience, offering a “win- win”

for climate change adaptation and mitigation. The most extreme

hydroclimatic conditions, in terms of drought, are associated

with reduced benefits of SOC to crops, compared to moderate

conditions. Therefore, promoting SOC regeneration for drought

resilience is not inaccurate, but also does not convey the greatest

service that SOC provides or the greatest potential benefits of

SOC regeneration. Indeed, the utility of future SOC regeneration

relative to historical climates may be reduced relative to histor-

ical climates if drought conditions become more severe or more

prevalent. The benefits the SOC provides for crops—including

water storage and nitrogen supply—require soil water in order

to be effectively delivered.

Author Contributions
Eunkyoung Choi:

writing – review and editing, data curation, for-

mal analysis, visualization.

Nathaniel D. Mueller:

writing – review

and editing, conceptualization, formal analysis, visualization.

Alison

E. King:

writing – original draft, writing – review and editing, data

curation, conceptualization, formal analysis, visualization.

Jesse

Burkhardt:

writing – review and editing, formal analysis, visualization.

Acknowledgments

We thank Chris Pires for organizational support.

Funding

A.E.K. was supported by the Colorado State University Soil and Crop

Sciences Department and Hatch Proposal ME022510 through the Maine

Agricultural Forest and Experiment Station. N.D.M. was supported by

AI- LEAF (AI Institute for Land, Economy, Agriculture and Forestry),

the USDA National Institute of Food and Agriculture (NIFA), and the

National Science Foundation (NSF) National AI Research Institutes

Competitive Award no. 2023- 67021- 39829.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The data that support the findings of this study were derived from

the following resources available in the public domain: USDA NRCS

Geospatial Data Gateway—gSSURGO,

—

USDA National Agricultural Statistics Service,

https:// quick stats. nass.

usda. gov/

—PRISM Climate Group,

http:// prism. orego nstate. edu

—

National Climate Assessment- Land Data Assimilation System,

doi. org/ 10. 1175/ JHM- D- 17- 0234. 1

—Global Food Security-

Support

Analysis Cropland Extent 2010 N,

https:// www. earth data. nasa. gov/

data/ catal og/ lpclo ud- gfsad 30nac e- 001

-

NASA Monthly Historical

Water Balance Products for the CONUS

−

https:// data. nasa. gov/ datas

et/ month ly- histo rical - water - balan ce- produ cts- for- the- conus

. Code and

processed data used for analyses are archived at

https:// zenodo. org/

recor ds/ 18916052

.

References

Amelung, W., D. Bossio, W. De Vries, et al. 2020. “Towards a Global-

Scale Soil Climate Mitigation Strategy.”

Nature Communications

11, no.

1: 5427.

https:// doi. org/ 10. 1038/ s4146 7- 020- 18887 - 7

https:// doi. org/ 10. 1029/ 2018W R024357

Araya, S. N., and T. A. Ghezzehei. 2019. “Using Machine Learning

for Prediction of Saturated Hydraulic Conductivity and Its Sensitivity

to Soil Structural Perturbations.”

Water Resources Research

55, no. 7:

5715–5737.

.

Bagnall, D. K., S. Cappellazzi, C. L. S. Morgan, et al. 2022. “Carbon-

Sensitive Pedotransfer Functions for Plant Available Water.”

Soil

Science Society of America Journal

86: 612–629.

https:// doi. org/ 10. 1002/

saj2. 20395

.

Baldwin- Kordick, R., M. De, M. D. Lopez, et al. 2022. “Comprehensive

Impacts of Diversified Cropping on Soil Health and Sustainability.”

Agroecology and Sustainable Food Systems

46, no. 3: 331–363.

doi. org/ 10. 1080/ 21683 565. 2021. 2019167

.

Basso, B., and J. Ritchie. 2014. “Temperature and Drought Effects on

Maize Yield.”

Nature Climate Change

4, no. 4: 233.

https:// doi. org/ 10. 1080/ 03650 34021 4162

1038/ nclim ate2139

.

Booth, M. S., J. M. Stark, and E. Rastetter. 2005. “Controls on Nitrogen

Cycling in Terrestrial Ecosystems: A Synthetic Analysis of Literature

Data.”

Ecological Monographs

75, no. 2: 139–157.

Bot, A., and J. Benites. 2005.

The Importance of Soil Organic Matter

. FAO

Soils Bulletin.

https:// doi. org/ 10. 1038/ s4189 3- 019- 0431- y

FIGURE 4

| Map of average projected changes in crop yield under two modeled SOC regeneration scenarios. (A, C, E) IPCC- informed SOC re-

generation; (B, D, F) SOC regeneration using uniform 10 g SOC kg soil

−

1

. Model estimates of crop yield are calculated for every year and county, then

averaged within county. Water supply (WS) conditions delineated with plant- available water (0–40 cm). These analyses with different water supply
indices (soil- agnostic and volumetric) are shown in Figures

S11

and

S12

.

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

10 of 13

Global Change Biology,

2026

Bradford, M. A., C. J. Carey, L. Atwood, et al. 2019. “Soil Carbon Science

for Policy and Practice.”

Nature Sustainability

2, no. December: 1070–

1072.

.

Buck, H. J., and A. Palumbo-

Compton. 2022. “Soil Carbon

Sequestration as a Climate Strategy: What Do Farmers Think?”

Biogeochemistry

161, no. 1: 59–70.

https:// doi. org/ 10. 1007/ s1053 3-

022- 00948 - 2

.

Butler, E. E., and P. Huybers. 2013. “Adaptation of US Maize to

Temperature Variations.”

Nature Climate Change

3, no. 1: 68–72.

https:// doi. org/ 10. 1038/ nclim ate1585

Butler, E. E., and P. Huybers. 2015. “Variations in the Sensitivity of US

Maize Yield to Extreme Temperatures by Region and Growth Phase.”

Environmental Research Letters

10, no. 3: 034009.

1088/ 1748- 9326/ 10/3/ 034009

https:// doi. org/ 10. 1016/j. agrfo rmet.

Carter, E. K., J. Melkonian, S. Steinschneider, and S. J. Riha. 2018.

“Rainfed Maize Yield Response to Management and Climate

Covariability at Large Spatial Scales.”

Agricultural and Forest

Meteorology

256–257: 242–252.

2018. 02. 029

https:// doi. org/ 10. 1016/j. still. 2018. 04. 011

Chenu, C., D. A. Angers, P. Barré, D. Derrien, D. Arrouays, and J.

Balesdent. 2019. “Increasing Organic Stocks in Agricultural Soils:

Knowledge Gaps and Potential Innovations.”

Soil and Tillage Research

188: 41–52.

Choi, E., A. J. Rigden, N. Tangdamrongsub, M. F. Jasinski, and N. D.

Mueller. 2024. “US Crop Yield Losses From Hydroclimatic Hazards.”

Environmental Research Letters

19, no. 1: 014005.

https:// doi. org/ 10. 1111/ nph. 16485

1088/ 1748- 9326/ ad0c87

.

Grossiord, C., T. N. Buckley, L. A. Cernusak, et al. 2020. “Plant

Responses to Rising Vapor Pressure Deficit.”

New Phytologist

226, no. 6:

1550–1566.

https:// doi. org/ 10. 1016/j. ejsobi. 2011. 07. 015

Gulde, S., H. Chung, W. Amelung, C. Chang, and J. Six. 2008. “Soil

Carbon Saturation Controls Labile and Stable Carbon Pool Dynamics.”

Soil Science Society of America Journal

72, no. 3: 605–612.

Guntiñas, M. E., M. C. Leirós, C. Trasar- Cepeda, and F. Gil- Sotres.

2012. “Effects of Moisture and Temperature on Net Soil Nitrogen

Mineralization: A Laboratory Study.”

European Journal of Soil Biology

48: 73–80.

https:// doi. org/ 10. 1007/ s0018 0- 008- 0119- 7

Hornik, K., C. Buchta, and A. Zeileis. 2009. “Open- Source Machine

Learning: R Meets Weka.”

Computational Statistics

24, no. 2: 225–232.

Hudson, B. D. 1994. “Soil Organic Matter and Available Water Capacity.”

Journal of Soil and Water Conservation

49: 189–194.

Iizumi, T., and R. Wagai. 2019. “Leveraging Drought Risk Reduction

for Sustainable Food, Soil and Climate via Soil Organic Carbon

Sequestration.”

Scientific Reports

9, no. 1: 19744.

1038/ s4159 8- 019- 55835 - y

Jasinski, M. F., J. S. Borak, S. V. Kumar, et al. 2019. “NCA- LDAS:

Overview and Analysis of Hydrologic Trends for the National Climate

Assessment.”

Journal of Hydrometeorology

20, no. 8: 1595–1617.

doi. org/ 10. 1175/ JHM- D- 17- 0234. 1

https:// doi. org/ 10. 1088/ 1748- 9326/ abe492

Kane, D. A., M. A. Bradford, E. Fuller, E. E. Oldfield, and S. A. Wood.

2021. “Soil Organic Matter Protects US Maize Yields and Lowers Crop

Insurance Payouts Under Drought.”

Environmental Research Letters

16,

no. 4: 044018.

.

King, A. E., G. A. Ali, A. W. Gillespie, and C. Wagner- Riddle. 2020. “Soil

Organic Matter as Catalyst of Crop Resource Capture.”

Frontiers in

Environmental Science

8: 50.

https:// doi. org/ 10. 3389/ fenvs. 2020. 00050

.

King, A. E., J. P. Amsili, S. C. Córdova, et al. 2024. “Constraints on

Mineral- Associated and Particulate Organic Carbon Response to

Regenerative Management: Carbon Inputs and Saturation Deficit.”

Soil

and Tillage Research

238: 106008.

https:// doi. org/ 10. 1016/j. still. 2024.

106008

https:// doi. org/ 10. 1080/ 14693 062. 2021.

Kreibich, N., and L. Hermwille. 2021. “Caught in Between: Credibility

and Feasibility of the Voluntary Carbon Market Post- 2020.”

Climate

Policy

21, no. 7: 939–957.

1948384

https:// doi. org/ 10. 1002/ agj2. 20282

Lal, R. 2020. “Soil Organic Matter and Water Retention.”

Agronomy

Journal

112, no. 5: 3265–3277.

https:// doi. org/ 10. 1038/ s4301 7- 022- 00368 - 8

Lesk, C., W. Anderson, A. Rigden, et al. 2022. “Compound Heat and

Moisture Extreme Impacts on Global Crop Yields Under Climate

Change.”

Nature Reviews Earth & Environment

3, no. 12: 872–889.

.

Li, Y., K. Guan, B. Peng, T. E. Franz, B. Wardlow, and M. Pan. 2020.

“Quantifying Irrigation Cooling Benefits to Maize Yield in the US

Midwest.”

Global Change Biology

26, no. 5: 3065–3078.

https:// doi. org/

10. 1111/ gcb. 15002

.

Li, Y., K. Guan, G. D. Schnitkey, E. DeLucia, and B. Peng. 2019.

“Excessive Rainfall Leads to Maize Yield Loss of a Comparable

Magnitude to Extreme Drought in the United States.”

Global Change

Biology

25, no. 7: 2325–2337.

https:// doi. org/ 10. 1111/ gcb. 14628

Lobell, D. B., M. Bänziger, C. Magorokosho, and B. Vivek. 2011.

“Nonlinear Heat Effects on African Maize as Evidenced by Historical

Yield Trials.”

Nature Climate Change

1, no. 1: 42–45.

1038/ nclim ate1043

https:// doi. org/ 10. 1038/ s4301 6- 020- 00165 - w

Lobell, D. B., J. M. Deines, and S. D. Tommaso. 2020. “Changes in the

Drought Sensitivity of US Maize Yields.”

Nature Food

1, no. 11: 729–735.

.

Lumley, T. 2024. “leaps: Regression Subset Selection (p. 3.2) [Dataset].”

https:// doi. org/ 10. 32614/ CRAN. packa ge. leaps

.

Ma, Y., D. Woolf, M. Fan, L. Qiao, R. Li, and J. Lehmann. 2023. “Global

Crop Production Increase by Soil Organic Carbon.”

Nature Geoscience

16, no. 12: 1159–1165.

https:// doi. org/ 10. 1038/ s4156 1- 023- 01302 - 3

10. 1016/j. seh. 2023. 100048

Mahmood, S., M. R. Nunes, D. A. Kane, and Y. Lin. 2023. “Soil Health

Explains the Yield- Stabilizing Effects of Soil Organic Matter Under

Drought.”

Soil & Environmental Health

1, no. 4: 100048.

https:// doi. org/

https:// lpdaac. usgs. gov/ produ cts/ gfsad 30nac ev001/

Massey, R., T. Sankey, K. Yadav, R. Congalton, J. Tilton, and P.

Thenkabail. 2017. “NASA Making Earth System Data Records for Use

in Research Environments (MEaSUREs) Global Food Security- Support

Analysis Data (GFSAD) Cropland Extent 2010 North America 30 m

V001 [Dataset].”

.

McClelland, S. C., D. Bossio, D. R. Gordon, et al. 2025. “Managing for

Climate and Production Goals on Crop- Lands.”

Nature Climate Change

15: 642–649.

https:// doi. org/ 10. 1038/ s4155 8- 025- 02337 - 7

.

Minasny, B., B. P. Malone, A. B. McBratney, et al. 2017. “Soil Carbon

4 Per Mille.”

Geoderma

292: 59–86.

https:// doi. org/ 10. 1016/j. geode rma.

2017. 01. 002

https:// doi. org/ 10. 1111/ ejss. 12475

Minasny, B., and A. B. McBratney. 2018. “Limited Effect of Organic

Matter on Soil Available Water Capacity.”

European Journal of Soil

Science

69, no. 1: 39–47.

.

NASA. 2025. “Monthly Historical Water Balance Products for the

CONUS (Nos. C2674700048- LPCLOUD) [Dataset].”

https:// data. nasa.

gov/ datas et/ month ly- histo rical - water - balan ce- produ cts- for- the- conus

https:// doi. org/ 10. 32614/ CRAN. packa ge.

O'Brien, L. 2022. “mpspline2: Mass- Preserving Spline Functions for

Soil Data (p. 0.1.6) [Dataset].”

mpspl ine2

https:// doi. org/ 10. 5194/ soil- 5- 15- 2019

Ogle, S. M. 2019. “Chapter 5: Cropland (Volume 4: Agriculture, Forestry

and Other Land Use; 2019 Refinement to the 2006 IPCC Guidelines for

National Greenhouse Gas Inventories).”

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

11 of 13

Global Change Biology,

2026

Oldfield, E. E., M. A. Bradford, and S. A. Wood. 2019. “Global Meta-

Analysis of the Relationship Between Soil Organic Matter and Crop

Yields.”

SOIL

5: 15–32.

.

Oregon State University. 2021. “PRISM Climate Data [Dataset].”

prism. orego nstate. edu/

Ortiz- Bobea, A., H. Wang, C. M. Carrillo, and T. R. Ault. 2019.

“Unpacking the Climatic Drivers of US Agricultural Yields.”

Environmental Research Letters

14, no. 6: 064003.

1088/ 1748- 9326/ ab1e75

Paustian, K., J. Lehmann, S. Ogle, D. Reay, G. P. Robertson, and P.

Smith. 2016. “Climate- Smart Soils.”

Nature

532, no. 7597: 49–57.

doi. org/ 10. 1038/ natur e17174

Pribyl, D. W. 2010. “A Critical Review of the Conventional SOC to SOM

Conversion Factor.”

Geoderma

156, no. 3–4: 75–83.

1016/j. geode rma. 2010. 02. 003

.

Qiao, L., X. Wang, P. Smith, et al. 2022. “Soil Quality Both Increases

Crop Production and Improves Resilience to Climate Change.”

Nature

Climate Change

12, no. 6: 574–580.

https:// doi. org/ 10. 1038/ s4155 8- 022-

01376 - 8

.

Rawls, W. J., Y. A. Pachepsky, J. C. Ritchie, T. M. Sobecki, and H.

Bloodworth. 2003. “Effect of Soil Organic Carbon on Soil Water

Retention.”

Geoderma

116, no. 1–2: 61–76.

https:// doi. org/ 10. 1016/ S0016

- 7061(03) 00094 - 6

https:// doi. org/ 10. 1038/ s4161 2- 021- 00224 - 4

Reddy, S., L. Panichelli, R. M. Waterworth, et al. 2019. “Chapter 3:

Consistent Representation of Lands.” Volume 4: Agriculture, Foresty

and Other Land Use; Refinement to the 2006 IPCC Guidelines for

National Greenhouse Gas Inventories.
Renwick, L. L. R. 2021. “Long- Term Crop Rotation Diversification

Enhances Maize Drought Resistance Through Soil Organic Matter

Gains.”

Environmental Research Letters

16: 084067.

Ridder, N. N., A. M. Ukkola, A. J. Pitman, and S. E. Perkins- Kirkpatrick.

2022. “Increased Occurrence of High Impact Compound Events Under

Climate Change.”

NPJ Climate and Atmospheric Science

5, no. 1: 3.

.

Rigden, A. J., N. D. Mueller, N. M. Holbrook, N. Pillai, and P. Huybers.

2020. “Combined Influence of Soil Moisture and Atmospheric

Evaporative Demand Is Important for Accurately Predicting US Maize

Yields.”

Nature Food

1, no. 2: 127–133.

https:// doi. org/ 10. 1038/ s4301

6- 020- 0028- 7

https:// doi. org/ 10. 1016/ S0959 - 3780(02) 00008 - 0

Rosenzweig, C., F. N. Tubiello, R. Goldberg, E. Mills, and J. Bloomfield.

2002. “Increased Crop Damage in the US From Excess Precipitation

Under Climate Change.”

Global Environmental Change

12, no. 3: 197–

202.

https:// doi. org/ 10. 1073/ pnas. 18009

Sanderman, J., T. Hengl, and G. J. Fiske. 2017. “Soil Carbon Debt of

12,000 Years of Human Land Use.”

Proceedings of the National Academy

of Sciences

115, no. 7: 201800925.

25115

https:// doi. org/ 10. 32614/ CRAN. packa

Savje, F., M. Higgins, and J. Sekhon. 2017. “scclust: Size- Constrained

Clustering (p. 0.2.5) [Dataset].”

ge. scclust

.

Saxton, K., and W. Rawls. 2006. “Soil Water Characteristic Estimates

by Texture and Organic Matter for Hydrologic Solutions.”

Soil Science

Society of America Journal

70: 1569–1578.

https:// doi. org/ 10. 2136/ sssaj

2005. 0117

https:// doi. org/ 10. 1073/ pnas. 07019

Schmidhuber, J., and F. N. Tubiello. 2007. “Global Food Security

Under Climate Change.”

Proceedings of the National Academy of

Sciences

104, no. 50: 19703–19708.

76104

.

Soil Survey Staff. 2015. “Gridded Soil Survey Geographic (gSSURGO)

Database for the United States of America and the Territories,

Commonwealths, and Island Nations Served by the USDA- NRCS

[Dataset].”

https:// gdg. sc. egov. usda. gov/

https:// doi. org/ 10. 1088/ 1748- 9326/

Urban, D. W., J. Sheffield, and D. B. Lobell. 2015. “The Impacts of Future

Climate and Carbon Dioxide Changes on the Average and Variability

of US Maize Yields Under Two Emission Scenarios.”

Environmental

Research Letters

10, no. 4: 045003.

10/4/ 045003

https:// www. nass. usda.

USDA ERS. 1992. “Weights, Measures, and Conversion Factors for

Agricultural Commodities and Their Products.”
USDA NASS. 2021. “Quick Stats [Dataset].”

gov/ Quick_ Stats/

https:// doi. org/ 10. 1038/ s4189 3- 023-

Vendig, I., A. Guzman, G. De La Cerda, et al. 2023. “Quantifying Direct

Yield Benefits of Soil Carbon Increases From Cover Cropping.”

Nature

Sustainability

6, no. 9: 1125–1134.

01131 - 7

.

Wang, F., J. D. Harindintwali, K. Wei, et al. 2023. “Climate Change:

Strategies for Mitigation and Adaptation.”

Innovation Geoscience

1, no.

1: 100015.

https:// doi. org/ 10. 59717/ j. xinn- geo. 2023. 100015

https:// doi. org/ 10. 4067/ S0718 - 95162 01200 0200003

Waraich, E. A., R. Ahmad, A. Halim, and T. Aziz. 2012. “Alleviation

of Temperature Stress by Nutrient Management in Crop Plants: A

Review.”

Journal of Soil Science and Plant Nutrition

12, no. 2: 221–244.

.

Witten, I. H., and E. Frank. 2005.

Data Mining: Practical Machine

Learning Tools and Techniques

. 2nd ed. Morgan Kaufmann.

Zhao, C., B. Liu, S. Piao, et al. 2017. “Temperature Increase Reduces

Global Yields of Major Crops in Four Independent Estimates.”

Proceedings of the National Academy of Sciences of the United States

of America

114, no. 35: 9326–9331.

https:// doi. org/ 10. 1073/ pnas. 17017

62114

.

Supporting Information

Additional supporting information can be found online in the

Supporting Information section.

Figure S1:

Projected changes in crop

yield and yield stability under a modeled soil organic carbon (SOC) re-

generation scenario, shown for all years. (A) Yield and yield differences,

(B) Coefficient of variation (CV) and CV differences, (C) Standard devi-

ation (SD) and SD differences. Regen. = regenerated SOC, increased

from true SOC by 10 g kg soil

−

1

for all counties, with minimum values

determined by the lowest SOC observed across all counties for each

crop. The coefficient of variation (CV) and standard deviation (SD) are

calculated within counties over all years. Asterisks indicate difference

between SOC scenarios, ***

p

< 0.001 (Wilcoxon signed rank test). Water

supply (WS) conditions delineated with soil- agnostic water (0–40 cm).

Figure S2:

Projected changes in crop yield and yield stability under a

modeled soil organic carbon (SOC) regeneration scenario, shown for all

years. (A) Yield and yield differences, (B) Coefficient of variation (CV)

and CV differences, (C) Standard deviation (SD) and SD differences.

Regen. = regenerated SOC, increased from true SOC by 10 g kg soil

−

1

for

all counties, with minimum values determined by the lowest SOC ob-

served across all counties for each crop. The coefficient of variation

(CV) and standard deviation (SD) are calculated within counties over all

years. Asterisks indicate difference between SOC scenarios, ***

p

< 0.001

(Wilcoxon signed rank test). Water supply (WS) conditions delineated

with volumetric soil water (0–40 cm).

Figure S3:

Projected changes in

crop yield under a modeled soil organic carbon (SOC) regeneration sce-

nario, shown for clusters of hydroclimatic years. All outcomes shown

for maize (A), soybeans (B), and spring wheat (C). Types of hydrocli-

matic years shown along x axis are identified based k- means clustering

of unique counties and years. Boxplots represent predicted yield under

true SOC (lighter colors) or regenerated SOC (darker colors). Circles

(inset plots) represent differences in yield. Prongs on circles represent

1st and 3rd interquartile range. Letters beneath inset circles indicate

statistically distinct yield responses to SOC regeneration between

weather conditions, as determined by Tukey's Honest Significant

Difference (HSD) test (

α

= 0.05) following a one- way ANOVA. For visu-

alization, the 1st and 99th percentiles are removed from yield and pro-

duction values. Water supply (WS) conditions delineated with

13652486, 2026, 5, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/gcb.70909 by Hayley Peter-Contesse - Oregon State University , Wiley Online Library on [19/10/2026]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License

12 of 13

Global Change Biology,

2026

soil- agnostic water (0–40 cm).

Figure S4:

Projected changes in crop

yield under a modeled soil organic carbon (SOC) regeneration scenario,

shown for clusters of hydroclimatic years. All outcomes shown for

maize (A), soybeans (B), and spring wheat (C). Types of hydroclimatic

years shown along x axis are identified based k- means clustering of

unique counties and years. Boxplots represent predicted yield under

true SOC (lighter colors) or regenerated SOC (darker colors). Circles

(inset plots) represent differences in yield. Prongs on circles represent

1st and 3rd interquartile range. Letters beneath inset circles indicate

statistically distinct yield responses to SOC regeneration between

weather conditions, as determined by Tukey's Honest Significant

Difference (HSD) test (

α

= 0.05) following a one- way ANOVA. For visu-

alization, the 1st and 99th percentiles are removed from yield and pro-

duction values. Water supply (WS) conditions delineated with soil

volumetric water (0–40 cm).

Figure S5:

Projected changes in crop pro-

duction under a modeled soil organic carbon (SOC) regeneration sce-

nario, shown for clusters of hydroclimatic years. All outcomes shown

for maize (A), soybeans (B), and spring wheat (C). Types of hydrocli-

matic years shown along x axis are identified based k- means clustering

of unique counties and years. Boxplots represent predicted production

under true SOC (lighter colors) or regenerated SOC (darker colors).

Circles (inset plots) represent differences in production. Production dif-

ferences represent the sum of [yield difference × hectares in crop pro-

duction] across all counties for a single year. Prongs on circles represent

1st and 3rd interquartile range. Letters beneath inset circles indicate

statistically distinct yield responses to SOC regeneration between

weather conditions, as determined by Tukey's Honest Significant

Difference (HSD) test (

α

= 0.05) following a one- way ANOVA. For visu-

alization, the 1st and 99th percentiles are removed from yield and pro-

duction values. Water supply (WS) conditions delineated with

plant- available soil water (0–40 cm).

Figure S6:

Projected changes in

crop production under a modeled soil organic carbon (SOC) regenera-

tion scenario, shown for clusters of hydroclimatic years. All outcomes

shown for maize (A), soybeans (B), and spring wheat (C). Types of hy-

droclimatic years shown along x axis are identified based k- means clus-

tering of unique counties and years (Figure

2B

). Boxplots represent

predicted production under true SOC (lighter colors) or regenerated

SOC (darker colors). Circles (inset plots) represent differences in pro-

duction. Production differences represent the sum of [yield differ-

ence × hectares in crop production] across all counties for a single year.

Prongs on circles represent 1st and 3rd interquartile range. Letters be-

neath inset circles indicate statistically distinct yield responses to SOC

regeneration between weather conditions, as determined by Tukey's

Honest Significant Difference (HSD) test (

α

= 0.05) following a one- way

ANOVA. For visualization, the 1st and 99th percentiles are removed

from yield and production values. Water supply (WS) conditions delin-

eated with soil- agnostic water (0–40 cm).

Figure S7:

Projected changes

in crop production under a modeled soil organic carbon (SOC) regener-

ation scenario, shown for clusters of hydroclimatic years. All outcomes

shown for maize (A), soybeans (B), and spring wheat (C). Types of hy-

droclimatic years shown along x axis are identified based k- means clus-

tering of unique counties and years (Figure

2B

). Boxplots represent

predicted production under true SOC (lighter colors) or regenerated

SOC (darker colors). Circles (inset plots) represent differences in pro-

duction. Production differences represent the sum of [yield differ-

ence × hectares in crop production] across all counties for a single year.

Prongs on circles represent 1st and 3rd interquartile range. Letters be-

neath inset circles indicate statistically distinct yield responses to SOC

regeneration between weather conditions, as determined by Tukey's

Honest Significant Difference (HSD) test (

α

= 0.05) following a one- way

ANOVA. For visualization, the 1st and 99th percentiles are removed

from yield and production values. Water supply (WS) conditions delin-

eated with soil volumetric water (0–40 cm).

Figure S8:

Soil organic car-

bon (SOC) moderates the effect of exposure to hydroclimatic conditions

on crop yield. Effect of increasing exposure to hydroclimatic conditions

on crop yields of maize (A), soybeans (B), and spring wheat (C) at three

different levels of SOC (solid = 90th SOC percentile; dashed = 50th SOC

percentile; dotted = 10th SOC percentile). Asterisks indicate the

p

- value

of the interaction effect between hydroclimatic condition and SOC:

***

p

≤

0.001; **

p

≤

0.01; *

p

< 0.05. Each plot also includes a joint density

plot of yield data by hydroclimatic condition. Model equation in

Section

(Equation

). Water supply (WS) conditions delineated with

soil- agnostic water (0–40 cm).

Figure S9:

Soil organic carbon (SOC)

moderates the effect of exposure to hydroclimatic conditions on crop

yield. Effect of increasing exposure to hydroclimatic conditions on crop

yields of maize (A), soybeans (B), and spring wheat (C) at three different

levels of SOC (solid = 90th SOC percentile; dashed = 50th SOC percen-

tile; dotted = 10th SOC percentile). Asterisks indicate the

p

- value of the

interaction effect between hydroclimatic condition and SOC:

***

p

≤

0.001; **

p

≤

0.01; *

p

< 0.05. Each plot also includes a joint density

plot of yield data by hydroclimatic condition. Model equation in

Section

(Equation 4). Water supply (WS) conditions delineated with

volumetric soil water (0–40 cm).

Figure S10:

Number of days during the

maize growing season in which plant- available water reaches its maxi-

mum (calculated as [water held at field capacity]

−

[water held at perma-

nent wilting point]), for counties and years in which exposure to each

hydroclimatic condition (WS

low

WD

high

or WD

med

WD

high

) was in the

90th percentile or greater. The few days of maximum plant- available

water in years with extreme WS

low

WD

high

exposure (orange boxplot; rel-

ative to extreme WD

med

WD

high

exposure shown with green boxplot)

suggests that water supply from soil was limited by water inputs (precip-

itation that infiltrates to soil) rather than soil available water- holding

capacity.

Figure S11:

Map of projected changes in crop yield under a

modeled SOC regeneration scenario in which all counties increase SOC

concentration by 10 g SOC kg soil

−

1

. Changes in crop yield estimated for

maize (A), soybeans (B), spring wheat (C). Model estimates of crop yield

are calculated for every year and county, then averaged within county.

Water supply (WS) conditions delineated with soil- agnostic water

(0–40 cm).

Figure S12:

Map of projected changes in crop yield under a

modeled SOC regeneration scenario in which all counties increase SOC

concentration by 10 g SOC kg soil

−

1

. Changes in crop yield estimated for

maize (A), soybeans (B), spring wheat (C). Model estimates of crop yield

are calculated for every year and county, then averaged within county.

Water supply (WS) conditions delineated with volumetric water

(0–40 cm).

Figure S13:

Yield difference under soil organic carbon

(SOC) regeneration scenario, as a function of initial SOC concentration.

Lines represent smoothing from general additive models (GAMs) of sin-

gle predictor and response variables shown in plots, EDF = effective de-

grees of freedom. Water supply (WS) conditions delineated with

plant- available water (0–40 cm).

Figure S14:

Yield difference under soil

organic carbon (SOC) regeneration scenario, as a function of initial SOC

concentration. Lines represent smoothing from general additive models

(GAMs) of single predictor and response variables shown in plots,

EDF = effective degrees of freedom. Water supply (WS) conditions de-

lineated with soil- agnostic water (0–40 cm).

Figure S15:

Yield differ-

ence under soil organic carbon (SOC) regeneration scenario, as a

function of initial SOC concentration. Lines represent smoothing from

general additive models (GAMs) of single predictor and response vari-

ables shown in plots, EDF = effective degrees of freedom. Water supply

(WS) conditions delineated with volumetric soil water (0–40 cm).

Figure S16:

Coefficient of variation (CV) difference under soil organic

carbon (SOC) regeneration scenario, as a function of initial SOC con-

centration. Lines represent smoothing from general additive models

(GAMs) of single predictor and response variables shown in plots,

EDF = effective degrees of freedom. Water supply (WS) conditions de-

lineated with plant- available (0–40 cm).

Figure S17:

Standard deviation

(SD) difference under soil organic carbon (SOC) regeneration scenario,

as a function of initial SOC concentration. Lines represent smoothing

from general additive models (GAMs) of single predictor and response

variables shown in plots, EDF = effective degrees of freedom. Water sup-

ply (WS) conditions delineated with plant- available (0–40 cm).

Figure

S18:

Nine randomly selected county clusters for maize (A), soybeans

(B), and spring wheat (C). For details on approach to county clusters, see

Section