Mahmood et al. (2023). Soil health explains the yield-stabilizing effects of soil organic matter under drought

). However, it is unclear how effectively

they can capture and predict the key outcomes of healthy soil, such as
agricultural productivity and carbon sequestration. For example, estab-
lishing the causality between soil health and crop yield continues to be a
complex challenge (

Crookston et al., 2022

;

Multiple popular approaches for assessing soil health are based on the

idea of variance partitioning (

;

). For a given soil health indicator (

e.g.

, soil organic

matter or aggregate stability), its variance can be explained by two types
of factors: inherent site-speci

ﬁ

c drivers and dynamic drivers. The former

include texture, climate, taxonomy, and other edaphic factors that are not
easily altered by management practices and do not change drastically
over time (

Wiesmeier et al., 2019

). In contrast, dynamic drivers consist of

land use and management practices that can quickly change soil health
over time. These soil health assessment methods aim to discern the
variance in the given soil health indicator that can be attributed to
inherent properties, typically using a peer group approach. After ac-
counting for the variance due to inherent properties, soil health status is
re

ﬂ

ected in the variance associated with dynamic drivers. For example,

soils are categorized into three peer groups based on texture, as in the
Comprehensive Assessment of Soil Health (CASH;

Moebius-Clune et al.,

2016

). The variability of soil health indicators,

e.g.

, soil organic carbon

(SOC), among these texture groups re

ﬂ

ects the effects of texture on soil

health. Then, the within-group variability of the soil health indicators is
used to quantify soil health conditions. In this instance, high SOC values
correspond to healthier soils in each texture group of CASH.

The Soil Health Assessment Protocol and Evaluation (SHAPE) is

another example of the peer group approach. However, SHAPE is more
robust as compared to CASH, taking into account not only the variation in
soil texture but also climate (local precipitation and temperature) and
soil taxonomy across the continental U.S. (

). In addi-

tion, SHAPE was developed using a large, continental-scale database,
making it more representative of U.S. edaphoclimatic conditions. Using
SOC as an example, the SHAPE framework categorizes soils into

ﬁ

ve soil

texture classes and

ﬁ

ve soil suborder classes (

) based on the

inherent potential of each soil to accumulate organic carbon. This results
in the creation of 25 peer groups, each representing a unique combina-
tion of texture and suborder classes (

C). Using a Bayesian modeling

approach, it also adjusts the scoring function for variations in annual
temperature and precipitation (

). Similar to CASH, it

produces scores in the range of 0

–

100% that correspond to the quantile of

SOC values within each peer group.

Methods like CASH and SHAPE rely on several assumptions to

quantify soil health: (1) the peer group approach effectively accounts for
the variance associated with inherent properties, (2) an increase in a soil

health indicator within a peer group corresponds to an increase in soil
health, and (3) the increase in soil health leads to an increase in
ecosystem functions (such as improved nutrient cycling and retention).
From the evolution of CASH to SHAPE, these methods have increasingly
incorporated inherent drivers to more comprehensively account for their
impacts on soil health indicators (1st assumption). While the 2nd
assumption is intrinsic to these methods, they enable us to evaluate the
3rd assumption when we compile a dataset with matching records of soil
health and ecosystem functions. The outcome of these evaluations will
ultimately inform us about the effectiveness of these methods in assessing
soil health.

Soil organic matter (SOM) content is a critical indicator of soil health

and is fundamental to various ecosystem processes. High SOM concen-
trations have been shown to bene

ﬁ

t crop yields (

). For instance, using county-level

corn yield data from the U.S.,

showed that soils with

higher SOM were associated with lower mean yield losses under drought
conditions. Their results indicate that SOM-rich soils help to stabilize
corn yield under drought conditions. However, it is unclear whether
these yield bene

ﬁ

ts are driven by soil health or the inherent site-speci

ﬁ

c

drivers. We obtained predominantly rainfed county-level corn yields in
the U.S. from 2000 to 2016 to

ﬁ

ll this research gap. Building upon

Kane

et al. (2021)

, we used linear mixed effects models to assess the effects of

SOM-derived soil health scores (SHS) on corn yields under different
drought conditions. We hypothesized that counties with higher SHS,
associated with higher SOM levels in the root zone (0

–

30 cm), would

have greater yield stability during drought. In other words, soils with
high SHS are associated with high yield under severe drought conditions.
To evaluate whether inherent factors (

i.e.

, soil texture, suborder, and

climate conditions) were responsible for the positive effects of SOM on
crop yield, we followed the peer group approach of SHAPE to categorize
soils into different texture and suborder groups (

C) as well as

compared the effects of SOM on yield between peer groups. If the effects
of the SOM are independent of peer groups, then a similar linear rela-
tionship between SOM and yield would be observed both within and
between peer groups, as illustrated in scenario A (

A). In contrast, if

the effects of SOM are driven by differences in inherent factors between
peer groups, then the SOM-yield relationship between peer groups would
drastically differ from those within peer groups, as illustrated in scenario
B (

National Environmental Public

B).

2. Materials and methods

2.1. Data collection

County-level soil attributes, corn yields, and Standardized Precipita-

tion Evapotranspiration Index (SPEI) data were obtained from

Kane et al.

(2021)

. These data were all derived from publicly available sources,

including the USDA's corn-frequency raster, the Gridded Soil Survey
Geographic (gSSURGO) database, the National Agricultural Statistics
Service's corn yield data (

USDA, 2018

), and the Centers for Disease

Control and Prevention SPEI database (

Health Tracking Network, 2018

). Details of data processing can be found

in

.

2.2. Soil health scoring

We adopted the SHAPE framework to derive SHS from soil organic

carbon (SOC) concentrations (

). The SHAPE framework

scores soil health based on scoring curves developed from a

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

2

comprehensive national database and Bayesian model-derived condi-
tional cumulative distribution functions. It considers all possible com-
binations of

ﬁ

ve soil suborders and

ﬁ

ve texture classes as peer groups

(

) and accounts for variations in temperature and precipitation

(

i.e.

, climate variables). The SHS indicates the quantile of a SOM value in

relation to other soils with similar inherent and climate conditions,

i.e.

,

the SHAPE approach accounts for the inherent factors driving soil health
before accounting for the dynamic (management) impact. It is worth
noting that a soil from the Midwestern U.S. can have the same SHS as a
soil from the Southeastern U.S. This implies that both soils are positioned
similarly in their SOM concentrations when compared to soils with
similar inherent and climate conditions. However, their SOM concen-
trations differ because of regional differences in soil-forming factors.
Thus, the SHAPE approach controls for the regional variability in SOM
values by accounting for their inherent and climate drivers. The mean
SOM concentration decreases as the suborder class number increases
from S1 to S5, while it increases from texture class T1 (coarse) to T5
(

ﬁ

ne) (

). These groupings were chosen to consolidate the large

number of suborders (

n

¼

75) and texture classes (

n

¼

12) available in

Soil Taxonomy and to amplify the differences in SOM among groups and
the variance explained by them (

).

County-level values of SOM (%), sand, silt, and clay were acquired

from the gSSURGO database following

https://github.com/paparker/SHAPE

. Brie

ﬂ

y, the

aqp

package in R (

Beaudette et al., 2023

) was used to convert the gSSURGO

characterization data for each series to the 0

–

30 cm soil depth, which

corresponds to a typical root depth of corn (

Feldman, 1994

). Then, we

estimated the mean value of selected soil properties for each map unit by
weighing the coverages of the included soil series and converted the data
to a raster format. These raster layers were then masked to remove areas
that did not grow corn consistently and used to calculate county-level
mean SOM (%, corresponding to g SOM 100 g

1

soil) values weighted

by the relative proportions of map units. The SOC (g 100 g

1

soil) was

calculated by dividing the SOM by the van Bemmelen factor (1.724)
(

Nelson and Sommers, 1996

). Soil suborder data were obtained in a

similar fashion, where the suborder of each map unit was determined as
that of the major component with the largest spatial coverage in the
given map unit. County-level suborders were determined as the suborder
with the largest spatial coverage. The mean annual temperature (MAT)
and the mean annual precipitation (MAP) at the county level were ob-
tained from the NOAA Climate Divisional Database (nClimDiv) (

NOAA,

2018

). The SHSs were calculated using a customized R script available at

. The SHSs vary between 0 and 1

(0

–

100%), where the higher value represents healthier soil within the

peer group. Even though the SHS is derived from only SOM, the SHAPE
approach enables us to answer our primary question,

i.e.

, whether the

bene

ﬁ

ts of SOM on yield were driven by soil health or inherent factors.

2.3. Yield and yield de

ﬁ

cit modeling

First, we created four different drought categories based on the mean

and standard deviation of the SPEI values. The SPEI is a drought index
where the differences between cumulative monthly precipitation and
potential evapotranspiration are determined from long-term climate data
(

National Environmental Public Health Tracking Network, 2018

). It is a

robust measurement of drought severity through time and space across a
wide range of climates (

Vicente-Serrano et al., 2010

). As drought effects

are expected to be more distinct on corn growth and yield during summer
(

Kane et al., 2021

), we considered SPEI values for summer months (

i.e.

,

May to August). The drought categories were as follows: very severe
drought, greater than or equal to two standard deviations (SPEI

1.02); severe drought, between one and two standard deviations

(

1.02

<

SPEI

0.46); moderate drought, between one standard

Fig. 1.

Schematic diagrams comparing the yield-stabilizing effects of soil organic matter (SOM) across peer groups and how the SHAPE tool works to score soil health.

T2 and T3 refer to two texture groups, while S3 and S4 refer to two soil suborder groups (

). Since soils from T2 have coarser texture than those from T3, a T2

soil with 3 g 100 g

1

SOM receives a higher soil heath score (SHS) than a T3 soil with the same SOM concentration. As soils from S3 are more SOM-rich than those

from S4, a S3 soil with 3 g 100 g

1

SOM receives a lower SHS than a S4 soil with the same SOM concentration.

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

3

deviation and the mean (

0.46

<

SPEI

0.10); and normal condition,

greater than or equal to the mean (SPEI

0.10). We standardized all

relevant data following

Gelman (2008)

by subtracting the mean from

each observation and dividing the resulting value by two standard de-
viations. We built linear mixed models of yield (Mg ha

1

) using SHS, soil

clay content (%), soil pH, and available water capacity (AWC; %) as in-
dependent variables. We added clay content, pH, and AWC to soils in
addition to SHS in the model to account for other potential drivers of
yield. We calculated the variance in

ﬂ

ation factors (VIF) to identify the

variables with high collinearities with others (

Table S2

). Due to its high

VIF (

>

5), cation exchange capacity was excluded from the models. We

also added a random effect of states following

to

capture the regional variability of agricultural management and envi-
ronment in the model results. We acknowledge that the state is only a
coarse indicator of this regional variability. Then, a linear mixed model
was constructed for each drought category. The

lme4

package in R (

Bates

et al., 2015

) was used to

ﬁ

t and analyze linear mixed effect models, and

the

lmerTest

package (

Kuznetsova et al., 2017

) was used to obtain the

p-

values of model coef

ﬁ

cients. The model was coded in

lme4

as follows:

Yield ~ SHS * clay * pH * AWC

þ

(1|State)

We estimated the marginal and conditional coef

ﬁ

cient of determi-

nation (R

2

) of the models with the

MuMIn

package (

Barton, 2020

). We

observed the effects of SHS by plotting SHS against yields, adapted for
each drought category. We also estimated the yields for each drought
category for a one-unit increase in SHS by dividing the standardized
coef

ﬁ

cients from the yield models by two standard deviations, as it

converted the standardized data back to their raw values (

Gelman,

2008

).

We calculated yield de

ﬁ

cits across all counties for very severe, severe,

and moderate drought categories by subtracting the mean yield under
normal conditions in the county from the mean yield under each drought
category. This approach helps to control for the temporal autocorrelation
in yields across different years in the same county. Yield de

ﬁ

cit is

expressed as negative values, and large absolute values indicate a
stronger decline in yield under drought. To evaluate the effect of SHS on
the yield de

ﬁ

cits, we built similar linear mixed models with yield de

ﬁ

cit

(Mg ha

1

) as the response using SHS, soil clay content (%), soil pH, and

available water capacity (AWC; %) as independent variables after stan-
dardization of data following

Gelman (2008)

. Similar to yield models, we

also added a random effect of states following

to

capture the regional variability in the yield de

ﬁ

cit model results. The

model was coded in

lme4

as follows:

Yield de

ﬁ

cit ~ SHS * clay * pH * AWC

þ

(1|State)

To evaluate whether the effects of SHS on yield (and yield de

ﬁ

cit)

differed between different soil texture classes, we built additional linear
models of yields (and yield de

ﬁ

cits) at each drought class with inde-

pendent variables including the SHS, soil texture class, and their inter-
action. The models were coded in R as follows:

Yield ~ SHS * Texture class

Yield de

ﬁ

cit ~ SHS * Texture class

A signi

ﬁ

cant interaction effect (

α

¼

0.05) would indicate that the

regression coef

ﬁ

cients of SHS differed signi

ﬁ

cantly across different

texture classes. The signi

ﬁ

cant interaction effect was also followed by

pairwise comparisons of the regression coef

ﬁ

cients between texture

classes using the

emmeans

package in R (

Lenth et al., 2022

). We repeated

this analysis across soil suborder classes. Similarly, the models were
coded in R as follows:

Yield ~ SHS * Suborder class

Yield de

ﬁ

cit ~ SHS * Suborder class

We omitted the very severe drought class from the suborder models

due to their small sample size (less than 100;

). The relatively

small sample size also prevented us from conducting similar analyses in
each combination of texture and suborder classes. The county-wise
extent of mean SHS and yield de

ﬁ

cits under different drought cate-

gories are shown in

Fig. S1

3. Results and discussions

3.1. Effects of SOM and soil health on corn yield

We found that soil health scores (SHS) signi

ﬁ

cantly affected corn

yield in four drought categories (

A). During the most severe

drought events, an increase of 0.5 (out of 1) in SHS was associated with a
1.15

0.18 Mg ha

1

(mean

standard error unless otherwise noted)

increase in corn yield (

p

<

0.001). Smaller, yet statistically signi

ﬁ

cant

effects were observed with the decrease in drought intensity (

Table S3A

;

all

p

<

0.001). For instance, an increment of 0.5 in SHS only resulted in a

yield increase of 0.40

0.03 Mg ha

1

under normal conditions. These

results expand upon the prior research conducted by

and align with previous studies that documented positive correlations
between SOM and yield under drought conditions in the U.S. (

Williams

et al., 2016

;

Zhou et al., 2021

). Our results also demonstrate that SHS

becomes more effective in predicting corn yield under increasing drought
severity, highlighting a yield-stabilizing role of SOM.

Compared to SOM concentration, SHS had a slightly less pronounced

effect on yields, as it had a smaller standardized coef

ﬁ

cient and a slope

less steep than SOM in each drought category (

Table S3A

). The SOM

being slightly better in predicting yields than SHS could be due to the
added effect of inherent soil and climate drivers with SOM, indicating the
signi

ﬁ

cance of regional variability. The SHS models also had slightly

lower marginal R

2

values than the SOM models but higher conditional R

2

values. For instance, under very severe drought, the

ﬁ

xed effects of the

SOM model explained 21.5% of the variability in yield (

i.e.

, marginal R

2

¼

0.215), while those of SHS explained 18.0% of the variability. Under

severe drought, the marginal R

2

was higher in the SOM model than in the

SHS model (0.239 vs. 0.188, respectively). After accounting for the
inherent drivers of SOM, SHS models still retained most of the variation
in yields. These results suggest that soil health is the main factor in
explaining the yield stabilizing effect of SOM. Other factors, including
inherent ones, likely explained the rest of the yield stabilizing effect of
SOM. Since the difference between conditional R

2

and marginal R

2

cor-

responds to the effects of the random factor (

i.e.

, state), it is worth noting

that the state explained more yield variability than SOM (or SHS). The
state re

ﬂ

ects regional variability in inherent soil-forming factors and

management practices, such as fertilization application, highlighting
their roles in regulating corn yield at the continental scale (

Kane et al.,

2021

).

The SHS also signi

ﬁ

cantly affected yield de

ﬁ

cits under severe

droughts (

B;

Table S3B

). With increasing SHS, the yield de

ﬁ

cit

decreased in all drought categories (

p

<

0.001, and

p

<

0.05). During a

very severe drought, an increase of 0.5 in SHS resulted in a 0.5

0.15

Mg ha

1

decrease in yield de

ﬁ

cit. Smaller changes were observed as the

drought severity decreased. For example, under moderate drought, a 0.5
increase in SHS only decreased the yield de

ﬁ

cit by 0.12

0.05 Mg ha

1

.

The impact of SHS on yield de

ﬁ

cits was consistent with their effects on

yields, indicative of the robustness of these yield stabilizing effects.
Interestingly, yield de

ﬁ

cits under very severe drought exhibited sub-

stantial variability in instances of low SHSs, with some of these less
healthy soils registering near-zero yield de

ﬁ

cits. In contrast, soils with

high SHS displayed a decrease in yield de

ﬁ

cit values approaching zero.

Although these patterns could have been in

ﬂ

uenced by the uneven dis-

tribution of samples across the full range of SHS, these results could also
suggest that corn yield was more stable against intensive drought when
SOM levels were high.

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

4

Soil pH, clay content, and AWC all had meaningful effects on corn

yield, as their main effects were signi

ﬁ

cant in most of the models studied

(

Table S3A

). These factors have been well-documented to in

ﬂ

uence corn

productivity. For example, most corn cultivars prefer a soil pH range of
5

–

6.2 (

Heiniger, 2018

), and high acidity reduces corn yield during

drought (

Caires et al., 2008

). Soil AWC plays a critical role in buffering

plant water stress for rainfed corn (

Benjamin et al., 2015

;

), while clay content is a key factor in

ﬂ

uencing SOM

and AWC (

Georgiou et al., 2022

Minasny and McBratney, 2018

). Despite

their signi

ﬁ

cant main effects, many of the interaction effects between

these factors and the SHS (or SOM) were not statistically signi

ﬁ

cant. Only

one three-way interaction between SHS, clay, and AWC was consistently
signi

ﬁ

cant in all models. These results suggest that the effect of SHS on

yield was largely independent of clay, pH, and AWC.

3.2. Effects of soil texture and suborder on yield and yield de

ﬁ

cit prediction

curves

To evaluate whether soil texture in

ﬂ

uenced the effects of SHS on

yield, we built a linear model using SHS, texture classes, and their
interaction as the main factors. Under very severe drought, SHS was a
signi

ﬁ

cant predictor of yield (

p

<

0.001), and the full model explained

13.5% of the variability in yield (

Table S4

;

Fig. 3

). The interaction effect

between SHS and texture classes on yield was not statistically signi

ﬁ

cant

(

Table S4A

). When yield was plotted against SHS, the slopes of different

texture classes were indistinguishable from each other. Similarly, the

interaction between SOM and texture classes did not affect corn yield,
and the slopes of SOM models were largely comparable across different
texture classes (

Fig. S2

;

Table S4

). We then repeated this analysis for

county-level yield de

ﬁ

cit and found the same patterns where SHS was a

signi

ﬁ

cant predictor of yield de

ﬁ

cit (

p

<

0.001;

Fig. 3

). However, the

interaction between SHS and texture class did not in

ﬂ

uence yield de

ﬁ

cit

(

Table S4B

). Together, these results suggest that the effect of soil health

on yield is independent of soil texture under very severe drought. In other
words, inherent variability in soil texture cannot explain the yield ben-
e

ﬁ

ts of SOM during extreme drought events (

Kane et al., 2021

;

Williams

et al., 2016

). Our study thus indicates that SOM can contribute to corn

yield stability under extreme conditions across a wide range of soil
textures.

We also observed that the intercepts of the SHS models differed

among texture classes under very severe drought. The intercept here
represents the estimated yield when the SHS is zero. Therefore, the dif-
ferences in the intercepts re

ﬂ

ect the effect of texture on corn yield when

the SOM has been taken into account. Texture class T2 had a lower
intercept than the other two classes, as it contained more coarse-textured
soils and, therefore, had a lower capacity to retain SOM. Although the
soils in texture class T4 contained more clay particles than those in T3, T4
had a lower intercept than T3. Previous studies have reported a bimodal
relationship between clay content and yield (

Katerji and Mastrorilli,

2009

Zipper et al., 2015

). Clay-rich soils with a low level of SOM are

particularly vulnerable to compaction, which subsequently limits root
growth and crop productivity (

Parent et al., 2008

).

Fig. 2.

(A) Yield stability increases with soil health scores (SHS) throughout county years and different texture peer groups, with effects becoming more pronounced

with increasing drought severity. (B) Yield de

ﬁ

cit decreases with increasing SHS with the effects becoming stronger with increasing drought severity. Trend lines

represent predicted yields (or yield de

ﬁ

cits) based on the marginal effect. The

p

values indicate the signi

ﬁ

cance of the effect of SHS. The standardized coef

ﬁ

cients (or

slopes) represent the yield (or yield de

ﬁ

cit) prediction in response to the marginal effect of SHS. Explained variation indicates marginal R

2

of the model.

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

5

Under less severe drought and normal conditions, soil texture regu-

lated the effects of SHS on yield, as indicated by a signi

ﬁ

cant interaction

effect between texture and SHS (

Table S4A

). Pairwise comparisons

further show that the regression slope of the T3 texture class was lower
than those of the other classes in each drought class, while there was no
difference in the slope between T2 and T4 (

Table S5A

Figs. S2-5

). The T3

is made of soils in the texture classes of silt and loamy silt, many of which
are from Midwestern U.S., including Illinois, Wisconsin, Indiana, Ohio,
and Kentucky. Noticeably, all T3 soils had SHS less than 0.5. Since SHAPE
was constructed using data from both natural and agricultural soils, our
results suggest that SOM is depleted in agricultural T3 soils compared to
their counterparts in natural systems. Given the prevalence of T3 soils in
the U.S., further examination is warranted to explain why these SOM
depletions were predominantly observed in T3 soils. We speculate that
this pattern re

ﬂ

ects the degradation of surface soil and SOM in the Corn

Belt, which has been well documented (

). Nonetheless, the effects of SHS on yield remained positive and

statistically signi

ﬁ

cant in each texture class (all

p

<

0.001;

Table S4A

),

regardless of drought categories, indicative of the bene

ﬁ

cial effects of

SOM on corn yield.

Under severe drought, the effect of SHS on yield differed between soil

suborder classes, as indicated by a signi

ﬁ

cant interaction between sub-

order and SHS (

Table S6A

). The slope of the yield model was lower in

suborder class S5 than in S4, while other pairwise comparisons of slopes
were not statistically signi

ﬁ

cant (

Fig. 4

Table S7A

). Similar results were

also found in the SOM model of yield, such that the regression slope of S5
was lower than that of S4 (

Tables S6A and S7A

). However, the interaction

between the suborder and SHS was not detected in the model of yield
de

ﬁ

cit. The S5 consisted of soils with the lowest mean SOM of all sub-

order classes, implying that these soils had the lowest capacity to retain
the SOM. Most of the S5 soils in this assessment are Udults (111 out of
115 soils) from states including Alabama, Kentucky, North Carolina,
Tennessee, and Virginia. As Udults tend to be heavily weathered, corn
productivity is likely limited by other soil properties, such as base satu-
ration and nutrient supply capacity. Their low SOM level also helps to
explain why S5 had the lowest regression intercept of the three suborder
classes (

Tables S7A

). Nevertheless, SHS consistently positively affected

yield in each suborder class (all

p

<

0.001;

Table S6A

). Overall, our re-

sults point to a moderate effect of suborder on the yield bene

ﬁ

ts of soil

health under severe drought.

The suborder also regulated the effect of soil health on yield under

moderate and normal conditions. In both drought categories, the regres-
sion slope of S5 was lower than those of S3 and S4 (

In fact, both slopes of S5 were indistinguishable from 0 under moderate
drought and normal conditions (

Fig. S8

). Consistent with those under

severe drought, these results further illustrate that SOM has a relatively
small impact on corn yield in Udults at regional scales. We did not conduct
similar analyses under very severe drought due to the limited sample size.
These results suggest that the role of suborders in regulating the soil
health-yield relationships became stronger under less severe drought.

Our calculation of SHS was in

ﬂ

uenced by the conversion coef

ﬁ

cient

between SOC and SOM. While SHAPE was initially formulated using
SOC, SSURGO provided data in terms of SOM.

Pribyl (2010)

suggested

that a factor of 2 (

i.e.

, SOM

¼

2

SOC) could be more accurate in most

soils. Using a conversion factor of 2, the SHS values decreased compared
to those derived from the van Bemmelen factor (

Fig. S9

); however, our

main

ﬁ

ndings hold, as SHS remains more effective in predicting corn

yield with increasing drought severity. These results indicate that our

ﬁ

ndings are robust against the conversion coef

ﬁ

cient between SOC and

SOM.

3.3. Implications for soil management

Overall, the results support our hypothesis that healthy soils stabilize

corn yield during droughts, and the effect is particularly strong when the
drought is severe. Our work builds upon past research (

Kane et al., 2021

;

Williams et al., 2016

;

Zhou et al., 2021

) and identi

ﬁ

es soil health, rather

than inherent soil properties, as the main driver explaining the
yield-stabilizing effects of SOM during drought. Thus, building SOM and
soil health has the potential to protect corn yield in a variety of soils with
distinct textures and taxonomy. Although the bene

ﬁ

ts of soil health

appear to be weaker in certain regions,

e.g.

, those with Udults, our work

supports the notion that the bene

ﬁ

ts of restoring SOM can be scaled down

to the farm scale. Since conservation practices, such as cover cropping
and crop rotations, have been shown to be effective in restoring SOM
(

Joshi et al., 2023

Vendig et al., 2023

), they could potentially increase

yield stability, especially when extreme weather events are becoming
more frequent. Our work also provides evidence supporting the current
efforts of large-scale soil health initiatives that seek to promote agricul-
tural resilience by increasing SOM (

Lehmann et al., 2020

).

Fig. 3.

The effects of SHS on yields (A) and on yield de

ﬁ

cits (B) by soil texture classes under very severe droughts. The

p

values indicate the signi

ﬁ

cance of the effect of

SHS. Explained variations indicate the R

2

of the full model, including the SHS, texture classes, and their interaction.

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

6

Our

ﬁ

ndings also directly link SHS to yield, an important ecosystem

function of soil, af

ﬁ

rming the SHAPE's approach to assessing soil health.

Following the peer-group approach, SHAPE offers exciting opportunities
to quantify the role of soil health in regulating other soil functions. We
also recognize that SHS remains a somewhat ambiguous value that acts as
a catch-all for potential mechanisms. For example, multiple physico-
chemical mechanisms have been proposed to explain the bene

ﬁ

ts of SOM

on yield, including enhanced water holding capacity, better aggregation,
and improved nutrient cycling and retention (

;

). Our results cannot be used to discern the

importance of these mechanisms directly. Instead, we primarily focused
on comparing the effectiveness of inherent vs. dynamic factors in regu-
lating the relationships between SOM and corn yield. Nonetheless, the
physiochemical mechanisms underpinning the yield bene

ﬁ

ts of SOM

warrant further research.

Even though SOM is a major soil health indicator, this assessment

showed that SOM is not the most important predictor of corn yield across
the U.S. Our models have shown that the random factor (

i.e.

, state) could

explain similar, if not greater, variation in yield compared to the com-
bined effects of SOM, pH, clay, and AWC (

Table S3

), highlighting the

importance of regional variation in soil characteristics and management
in regulating yield. Yield was also extremely sensitive to drought (

),

and SHS or SOM alone only explained 8

–

15% of the variation in yield in

each drought category (

;

Table S3

). These results are consistent with

other modeling studies of corn yield in the U.S. (

Li et al., 2019

;

Xu et al.,

2021

). Our unique contribution is to identify that soil health, rather than

inherent properties, is responsible for the yield-stabilizing effects of SOM.

Within this assessment, we took into consideration and discussed

several aspects of soil health, including inherent and dynamic soil health
indicators (

e.g.

, clay and silt content, AWC, pH, SOM, SOC), with the po-

tential to impact crop yield responses to drought across the U.S. Never-
theless, only SOC was scored and interpreted using the SHAPE approach
(

https://doi.org/10.5281/zenodo.10047650

). Incorporating other soil health indicators would have

advanced the current understanding of the correlation and causality be-
tween crop yield and soil health. Yet, the current approach is adequate to
answer our primary question,

i.e.

, whether the yield-stabilizing bene

ﬁ

ts of

SOM could be attributed to soil health or inherent drivers. In addition, we
lack large-scale data on other soil health indicators to conduct similar
county-level analyses. The SOM remains the most widely adopted soil
health indicator, given its close associations with many soil physical,

chemical, and biological processes (

Karlen et al., 2008

). Other indicators

beyond SOC (or SOM) are currently being included in SHAPE (personal
communication; MR Nunes). It is also worth noting that our SOM and
other soil characteristics are derived from a mapping product and are
expected to carry relatively high uncertainty (

Libohova et al., 2019

). The

utilization of high-resolution soil characteristics and yield maps would
have improved the statistical power of our analyses. Additionally, we
lacked reliable county-level fertilization data, a crucial factor contributing
to the variability in yield. Future large-scale fertilization data can improve
the understanding of soil health and crop yield. Finally, we studied rainfed
corn, a drought-sensitive crop, for which the results for other crops in
different agroecological settings might not show similar drought resil-
ience with improved soil health as our

ﬁ

ndings showed.

4. Conclusions

Our results demonstrate that the soil health score explained a ma-

jority of the effects of SOM on county-level corn yields in the U.S. The
effects of the soil health score on yield were largely consistent across
different soil textures and soil suborder groups. Thus, our results
demonstrate that soil health, rather than inherent properties, is largely
responsible for the yield-stabilizing effects of SOM. Therefore, it is
potentially feasible to increase the resilience of corn yield against
drought by adopting agronomic practices that build SOM and soil health.
The bene

ﬁ

ts of soil health are largely independent of texture and soil

suborder, so building up soil health has the potential to improve crop
productivity across a wide range of geographical locations and site
characteristics.

Data availability

Data required for reproducing the statistical analyses and

ﬁ

gures are

available at

org/10.1016/j.seh.2023.100048

Declaration of Competing Interest

The authors declare that they have no known competing

ﬁ

nancial

interests or personal relationships that could have appeared to in

ﬂ

uence

the work reported in this paper.

Fig. 4.

The effects of SHS on yields (A) and on yield de

ﬁ

cits (B) by soil suborder classes under severe droughts. The

p

values indicate the signi

ﬁ

cance of the effect of

SHS. Explained variations indicate the R

2

of the full model, including the SHS, texture classes, and their interaction.

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

7

Acknowledgments

We thank the members of the Soil Health Laboratory for their com-

ments. This work was supported by a U.S. Department of Agriculture
Hatch grant FLA-SWS-006103 to YL. Comments from the Editors, Drs. Ma
and Hayat, and anonymous reviewers greatly improved an earlier version
of this manuscript.

Appendix A. Supplementary data

Supplementary data to this article can be found online at

https://doi.

References

Andrews, S.S., Karlen, D.L., Cambardella, C.A., 2004. The soil management assessment

framework. Soil Sci. Soc. Am. J. 68 (6), 1945

–

1962.

https://doi.org/10.2136/

sssaj2004.1945

https://cran.r-project.org/web/pa

Barton, K., 2020. Package

“

MuMIn

”

(1.43.17). CRAN.

ckages/MuMIn/MuMIn.pdf

.

Bates, D., M

€

achler, M., Bolker, B.M., Walker, S.C., 2015. Fitting linear mixed-effects

models using lme4. J. Stat. Software 67 (1).

https://doi.org/10.18637/JSS.V067.I01

.

Beaudette, D.E., Roudier, P., O'Geen, A.T., 2023. Package

“

aqp

”

.

https://cran.r-project

.org/web/packages/aqp/aqp.pdf

.

Benjamin, J.G., Nielsen, D.C., Vigil, M.F., Mikha, M.M., Calderon, F., 2015. Cumulative

de

ﬁ

cit irrigation effects on corn biomass and grain yield under two tillage systems.

Agric. Water Manag. 159, 107

–

114.

https://doi.org/10.1016/j.agwat.2015.05.025

.

Caires, E.F., Garbuio, F.J., Churka, S., Barth, G., Corr

^

ea, J.C.L., 2008. Effects of soil acidity

amelioration by surface liming on no-till corn, soybean, and wheat root growth and
yield. Eur. J. Agron. 28 (1), 57

–

64.

https://doi.org/10.1016/j.eja.2007.05.002

.

Chang, T., Feng, G., Paul, V., Adeli, A., Brooks, J.P., 2022. Soil health assessment

methods: progress, applications and comparison. Adv. Agron. 172, 129

–

210.

https://

doi.org/10.1016/bs.agron.2021.10.002

.

Crookston, B., Yost, M., Bowman, M., Veum, K., 2022. Relationships of on-farm soil

health scores with corn and soybean yield in the midwestern United States. Soil Sci.
Soc. Am. J. 86 (1), 91

–

105.

https://doi.org/10.1002/saj2.20355

.

de Araujo Ru

ﬁ

no, C., Fernandes-Vieira, J., Martín-Gil, J., de Souza Abreu Júnior, J.,

Tavares, L.C., Fernandes-Correa, M., Martín-Ramos, P., 2018. Water stress in

ﬂ

uence

on the vegetative period yield components of different maize genotypes. Agronomy 8
(8), 1

–

9.

https://doi.org/10.3390/agronomy8080151

Feldman, L., 1994. The maize root. The Maize Handbook 29

–

37.

10.1007/978-1-4612-2694-9_4

.

Gelman, A., 2008. Scaling regression inputs by dividing by two standard deviations. Stat.

Med. 27 (15), 2865

–

2873.

https://doi.org/10.1002/SIM.3107

.

Georgiou, K., Jackson, R.B., Vindu

skov

a, O., Abramoff, R.Z., Ahlstr

€

om, A., Feng, W.,

Harden, J.W., Pellegrini, A.F.A., Polley, H.W., Soong, J.L., Riley, W.J., Torn, M.S.,
2022. Global stocks and capacity of mineral-associated soil organic carbon. Nat.
Commun. 13 (1), 1

–

12.

https://doi.org/10.1038/s41467-022-31540-9

.

Haney, R.L., Haney, E.B., Smith, D.R., Harmel, R.D., White, M.J., 2018. The soil health

tool

—

theory and initial broad-scale application. Appl. Soil Ecol. 125 (February),

162

–

168.

https://doi.org/10.1016/j.apsoil.2017.07.035

https://corn.ces.ncsu.edu/corn-production-infor

Heiniger, R.W., 2018. How Soil pH Affects Wheat and Corn Yields. NC State Extension.

Corn Production Information 4

–

7.

mation/how-soil-ph-affects-wheat-and-corn-yields/

https://doi.org/10.1002/agj2.21340

Joshi, D.R., Sieverding, H.L., Xu, H., Kwon, H., Wang, M., Clay, S.A., Johnson, J.M.,

Thapa, R., Westhoff, S., Clay, D.E., 2023. A global meta-analysis of cover crop
response on soil carbon storage within a corn production system. Agronomy Journal,
May 1543

–

1556.

.

Kane, D.A., Bradford, M.A., Fuller, E., Old

ﬁ

eld, E.E., Wood, S.A., 2021. Soil organic

matter protects US maize yields and lowers crop insurance payouts under drought.
Environ. Res. Lett. 16 (4).

https://doi.org/10.1088/1748-9326/abe492

Karlen, D.L., Andrews, S.S., Wienhold, B.J., Zobeck, T.M., 2008. Soil quality assessment:

past, present and future. J. Integr. Biosci. 6 (1), 3

–

14

https://doi.org/10.1016/J.STILL.2019.104365

Karlen, D.L., Veum, K.S., Sudduth, K.A., Obrycki, J.F., Nunes, M.R., 2019. Soil Health

Assessment: Past Accomplishments, current Activities, and future opportunities. Soil
and Tillage Research 195.

https://doi.org/10.1016/J.EJA.2008.07.009

Katerji, N., Mastrorilli, M., 2009. The effect of soil texture on the water use ef

ﬁ

ciency of

irrigated crops: results of a multi-year experiment carried out in the Mediterranean
region. Eur. J. Agron. 30 (2), 95

–

100.

https://doi.org/10.1098/rstb.2007.2178

Kibblewhite, M.G., Ritz, K., Swift, M.J., 2008. Soil health in agricultural systems. Phil.

Trans. Biol. Sci. 363 (1492), 685

–

701.

.

Kuznetsova, A., Brockhoff, P.B., Christensen, R.H.B., 2017. lmerTest package: tests in

linear mixed effects models. J. Stat. Software 82 (13), 1

–

26.

https://doi.org/10.1002/fes3.96

10.18637/JSS.V082.I13

.

Lal, R., 2016. Soil health and carbon management. Food Energy Secur. 5 (4), 212

–

222.

.

Lehmann, J., Bossio, D.A., K

€

ogel-Knabner, I., Rillig, M.C., 2020. The concept and future

prospects of soil health. Nature Reviews Earth

&

Environment 2020 1:10 1 (10),

544

–

553.

https://doi.org/10.1038/s43017-020-0080-8

https://cran.r-project.org/web/

Lenth, R.V., Buerkner, P., Gin

e-V

azquez, I., Herve, M., Jung, M., Love, J., Miguez, F.,

Riebl, H., Singmann, H., 2022. Package

“

Emmeans

”

.

packages/emmeans/emmeans.pdf

Li, Y., Guan, K., Yu, A., Peng, B., Zhao, L., Li, B., Peng, J., 2019. Toward building a

transparent statistical model for improving crop yield prediction: modeling rainfed
corn in the U.S. Field Crops Res. 234 (January), 55

–

65.

https://doi.org/10.1016/

j.fcr.2019.02.005

https://doi.org/10.1111/ejss.12770

Libohova, Z., Seybold, C., Adhikari, K., Wills, S., Beaudette, D., Peaslee, S., Lindbo, D.,

Owens, P.R., 2019. The anatomy of uncertainty for soil pH measurements and
predictions: implications for modellers and practitioners. Eur. J. Soil Sci. 70 (1),
185

–

199.

.

Minasny, B., McBratney, A.B., 2018. Limited effect of organic matter on soil available

water capacity. Eur. J. Soil Sci. 69 (1), 39

–

47.

https://doi.org/10.1111/EJSS.12475

.

Moebius-Clune, B.N., Moebius-Clune, D.J., Gugino, B.K., Idowu, O.J., Schindelbeck, R.R.,

Ristow, A.J., van Es, H.M., Thies, J.E., Shayler, H.A., McBride, M.B., Kurtz, K.S.M.,
Wolfe, D.W., Abawi, G.S., 2016. Comprehensive Assessment of Soil Health. Cornell
University.

http://www.scs.cals.cornell.edu

.gov/Environmental-Health-Toxicology/Standardized-Precipitation-Evapotranspirat
ion-Inde/6nbv-i

National Environmental Public Health Tracking Network, 2018. Standardized

Precipitation Evapotranspiration Index, 1895-2016 | Data | Centers for Disease
Control and Prevention. Centers for Disease Control and Prevention.

https://data.cdc

ﬁ

b

.

Nelson, D.W., Sommers, L.E., 1996. Total carbon, organic carbon, and organic matter.

Methods of Soil Analysis 5 (Issue 5), 961

–

1010.

https://doi.org/10.2136/

sssabookser5.3.c34

https://www.ncei.noaa.gov/access/monitoring/climate-at-a-glance/count

Part 3

: Chemical Methods.

NOAA, 2018. Climate at a Glance. National Centers for Environmental Information

(NCEI).

y/mapping

https://doi.org/10.1002/saj2.20244

Nunes, M.R., Veum, K.S., Parker, P.A., Holan, S.H., Karlen, D.L., Amsili, J.P., van Es, H.M.,

Wills, S.A., Seybold, C.A., Moorman, T.B., 2021. The soil health assessment protocol
and evaluation applied to soil organic carbon. Soil Sci. Soc. Am. J. 85 (4),
1196

–

1213.

https://doi.org/10.3389/fsufs.2019.00127

Olayemi, O.P., Kallenbach, C.M., Schneekloth, J.P., Calder

on, F.J., Vigil, M.F.,

Wallenstein, M.D., 2020. From factory to

ﬁ

eld: effects of a novel soil amendment

derived from cheese production on wheat and corn production. Front. Sustain. Food
Syst. 3, 127.

Old

ﬁ

eld, E.E., Bradford, M.A., Wood, S.A., 2019. Global meta-analysis of the relationship

between soil organic matter and crop yields. SOIL 5 (1), 15

–

32.

https://doi.org/10.1016/J.BIOSYSTEMSENG.2007.10.006

10.5194/soil-5-15-2019

.

Parent, A.C., B

elanger, M.C., Parent, L.E., Santerre, R., Viau, A.A., Anctil, F.,

Bolinder, M.A., Tremblay, C., 2008. Soil properties and landscape factors affecting
maize yield under wet spring conditions in eastern Canada. Biosyst. Eng. 99 (1),
134

–

144.

https://doi.org/10.1016/j.geoderma.2010.02.003

Pribyl, D.W., 2010. A critical review of the conventional SOC to SOM conversion factor.

Geoderma 156 (3

–

4), 75

–

83.

.

Rhoton, F.E., Shipitalo, M.J., Lindbo, D.L., 2002. Runoff and soil loss from midwestern

and southeastern US silt loam soils as affected by tillage practice and soil organic
matter content. Soil Tillage Res. 66 (1), 1

–

11.

https://doi.org/10.1016/S0167-

1987(02)00005-3

.

Thaler, E.A., Larsen, I.J., Yu, Q., 2021. The extent of soil loss across the US Corn Belt.

Proc. Natl. Acad. Sci. U.S.A. 118 (8), e1922375118.

https://doi.org/10.1073/

PNAS.1922375118/SUPPL_FILE/PNAS.1922375118.SAPP.PDF

.

USDA, 2018. USDA - National Agricultural Statistics Service - Data and Statistics. United

States Department of Agriculture.

https://www.nass.usda.gov/Data_and_Statistics/

.

https://www.nass.usda.gov/Data_and_Statistics/index.php

.

Vendig, I., Guzman, A., De La Cerda, G., Esquivel, K., Mayer, A.C., Ponisio, L.,

Bowles, T.M., 2023. Quantifying direct yield bene

ﬁ

ts of soil carbon increases from

cover cropping. Nat. Sustain. 6 (September).

https://doi.org/10.1038/s41893-023-

01131-7

https://doi.org/10.1175/2009JCLI2909.1

Vicente-Serrano, S.M., Beguería, S., L

opez-Moreno, J.I., 2010. A multiscalar drought

index sensitive to global warming: the standardized precipitation evapotranspiration
index. J. Clim. 23 (7), 1696

–

1718.

.

Wade, J., Culman, S.W., Logan, J.A.R., Poffenbarger, H., Demyan, M.S., Grove, J.H.,

Mallarino, A.P., McGrath, J.M., Ruark, M., West, J.R., 2020. Improved soil biological
health increases corn grain yield in N fertilized systems across the Corn Belt. Sci. Rep.
10 (1), 1

–

9.

https://doi.org/10.1038/s41598-020-60987-3

https://doi.org/10.1016/j.geoderma.2018.07.026

Wiesmeier, M., Urbanski, L., Hobley, E., Lang, B., Lützow, M. Von, Marin-spiotta, E.,

Wesemael, B. Van, Rabot, E., Lie

ß

, M., Garcia-franco, N., Wollschl

€

ager, U., Vogel, H.,

K

€

ogel-knabner, I., 2019. Soil organic carbon storage as a key function of soils - a

review of drivers and indicators at various scales. Geoderma 333 (July 2018),
149

–

162.

.

Williams, A., Hunter, M.C., Kammerer, M., Kane, D.A., Jordan, N.R.,

Mortensen, D.A., Smith, R.G., Snapp, S., Davis, A.S., 2016. Soil water holding
capacity mitigates downside risk and volatility in US rainfed maize: time to
invest in soil organic matter? PLoS One 11 (8).

https://doi.org/10.1371/

JOURNAL.PONE.0160974

.

Wood, S.A., Blankinship, J.C., 2022. Making soil health science practical: guiding

research for agronomic and environmental bene

ﬁ

ts. Soil Biol. Biochem. 172 (April),

108776.

https://doi.org/10.1016/j.soilbio.2022.108776

.

Xu, T., Guan, K., Peng, B., Wei, S., Zhao, L., 2021. Machine learning-based modeling of

spatio-temporally varying responses of rainfed corn yield to climate, soil, and
management in the U.S. Corn Belt. Frontiers in Arti

ﬁ

cial Intelligence 4 (May), 1

–

21.

https://doi.org/10.3389/frai.2021.647999

S. Mahmood et al.

Soil

&

Environmental Health 1 (2023) 100048

8

Zhang, L.X., Li, S.X., Zhang, H., Liang, Z.S., 2007. Nitrogen rates and water stress effects

on production, lipid peroxidation and antioxidative enzyme activities in two maize
(Zea mays L.) genotypes. J. Agron. Crop Sci. 193 (6), 387

–

397.

10.1111/j.1439-037X.2007.00276.x

.

Zhou, W., Guan, K., Peng, B., Tang, J., Jin, Z., Jiang, C., Grant, R., Mezbahuddin, S., 2021.

Quantifying carbon budget, crop yields and their responses to environmental
variability using the ecosys model for U.S. Midwestern agroecosystems. Agric. For.
Meteorol. 307 (July), 108521.

https://doi.org/10.1016/j.agrformet.2021.108521

.

Zipper, S.C., Soylu, M.E., Booth, E.G., Loheide, S.P., 2015. Untangling the effects of

shallow groundwater and soil texture as drivers of sub

ﬁ

eld-scale yield variability.

Water Resour. Res. 51 (8), 6338

–

6358.

https://doi.org/10.1002/2015WR017522

.

Zuber, S.M., Veum, K.S., Myers, R.L., Kitchen, N.R., Anderson, S.H., 2020. Role of

inherent soil characteristics in assessing soil health across Missouri. Agricultural and
Environmental Letters 5 (1).

https://doi.org/10.1002/ael2.20021