Laub et al. (2025). Maize yield responses to soil organic carbon under integrated soil fertility management in tropical environments

; Erenstein et al.

) as well

as the strongest population growth of all continents (Lutz

et al.

2018

). Most agricultural soils in sub-Saharan Africa

are highly weathered (von Fromm et al.

; Reichen-

bach et al.

) and are characterized by low nutrient

levels and organic matter status (Kihara et al.

). This

severely limits crop productivity under current manage-

ment and has led to stagnating crop yields (FAO

As a potential solution to these challenges, integrated soil

fertility management (ISFM; Vanlauwe et al.

2010

) has

been postulated to simultaneously increase crop yields and

soil fertility by enhancing soil organic matter and nutrient

supply. Since soil organic carbon (SOC) constitutes about

half of soil organic matter (Berryman et al.

), it serves

as a reliable proxy for assessing the overall status of soil

organic matter.

While several field experiments have demonstrated the

crop yield benefits of ISFM or similar practices (Chivenge

et al.

; Mutuku et al.

; Cardinael et al.

MacLaren et al.

; Laub et al.

), long-term

experiments in sub-Saharan Africa have also revealed

instances where these practices led to a decline in SOC

over time (Sommer et al.

; Cardinael et al.

et al.

), even when yields were increasing (Laub

et al.

2023b

). This prompts the question of whether SOC is

indeed a driving factor of long-term crop yield increases.

Studies that show evidence of the long-term yield benefits

of sustainable intensification practices are scarce, with

even fewer assessing the underlying mechanisms of

crop yield benefits. However, such research is essential

to corroborate or refute the frameworks of sustainable

intensification, as well as the hypothesized long-term

relationships between SOC and yields (Vanlauwe et al.

2010

Sustainable intensification of crop production in a

changing climate can only be achieved if we understand

the factors that determine the long-term productivity of

crops, identify those controllable by humans, and adjust

crop management practices accordingly. While many

studies have indicated the importance of changes in both

SOC (Tully et al.

; Ndung’u et al.

; Cardinael

et al.

; Thierfelder et al.

; Ma et al.

) and

weather patterns (Block et al.

; Jägermeyr et al.

)

for maize productivity, there are few studies with datasets

of sufficient duration to determine the effect of gradual

changes in SOC on crop yield, especially in sub-Saharan

Africa. Usually, relationships between yield and SOC

are established through large-scale correlations (Oldfield

et al.

2019

; Ma et al.

), which may be confounded

by factors such as net primary productivity, influencing

both SOC levels and crop yields. Therefore, such studies

cannot inform about the influence that SOC-altering

management practices have on yields. Similarly, studies

examining the effects of weather patterns on crop yield

usually have limited information on soil fertility, relying

either on aggregated county-level data (Schlenker and

Roberts

2009

) or on short-term field trials, often taking

soil fertility as a given (Lobell et al.

). There is

therefore a need for comprehensive analyses of the relative

importance of weather patterns versus soil-quality-

improving management practices on crop yield.

This study first examined the relationship between

management induced changes in SOC and changes in maize

crop yield. Our experimental approach, which controls for

variations in soil and climatic properties, provides more

robust insights into the SOC-yield relationship compared

to global large-scale correlation studies. We focused on

continuous maize cropping systems because demand for

maize is high across sub-Saharan Africa (Erenstein et al.

), and it is the main staple food in eastern Africa

(Ngeno

2024

); consequently, cultivation of maize on the

same land year after year is widespread (Badu-Apraku

and Fakorede

2017

). Second, we assessed the relative

importance of controllable (i.e., management practices)

versus uncontrollable (i.e., weather variability) factors that

determine maize yield in any given season. We analyzed a

dataset comprising maize yield records and SOC data from

four experiments in Kenya, conducted over 16 to 19 years,

covering a range of soil types (sandy to clayey) and climates

(semi-arid to humid). These experiments investigated

ISFM practices involving different combinations of organic

amendments and mineral N fertilizer. Two recent studies

(Laub et al.

) reported the long-term trends in

maize yields and SOC in these experiments. Interestingly,

the treatments that best maintained or increased SOC across

all sites were also associated with higher maize yields. Yet,

the magnitude of treatment effects was site specific, and

these studies did not analyze the relationship between maize

yield and SOC changes, nor the relationship between yield

and weather variability. Therefore, in the current study, we

specifically wanted to explore: (1) to what extent do ISFM-

induced SOC trends affect long-term maize yield trends,

and how consistent are effects across different sites, and

(2) what is the relative contribution of ISFM management

practices compared to season-to-season weather variability

in explaining maize yield variation?

To address these questions, we conducted two types of

analyses. First, we investigated whether long-term trends in

maize yields were correlated with treatment-related long-

term trends in SOC within each site. We further explored

Page 3

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 3 of 15

how this relationship differed in response to mineral N

fertilizer application and how it was affected by the long-

term trends in rainfall, temperature, and soil pH. Second,

we evaluated the extent to which fluctuations in maize

yields between the seasons at each site could be attributed

to ISFM treatment factors, in comparison to seasonal

weather variability. In this analysis we used the ISFM

treatment factors as covariates instead of SOC, because

SOC measurements were only available every few years.

Furthermore, we examined significant interactions between

treatment factors and primary weather patterns (rainfall,

temperature, and season length).

The four experimental sites had identical treatments, with

organic amendments of low to high C:N ratios, i.e., farmyard

manure, two types of green manures (

Tithonia diversifolia

and

Calliandra calothyrsus

), maize stover and sawdust, at

application rates of 1.2 and 4 t C ha

. Using a split plot

design, half of each plot received 120 kg N ha

of mineral

fertilizer per growing season (+N and -N treatments; see

Methods for further details). Two sites, Aludeka and Sidada,

were situated in humid western Kenya, comprising 16 years

of data. Two other sites, Embu and Machanga, were in the

sub-humid and semi-arid areas of the central region of

Kenya, respectively, comprising 19 years of data. All four

sites were characterized by two maize growing seasons per

year, corresponding to the long and the short rainy season.

2 Materials and methods

2.1 Study sites

The four experimental sites, located in central and western

Kenya, were under continuous maize monocropping with

two growing seasons per year. The long rainy season lasts

from March until August/September and the short rainy

season from September/October until January/February. The

two experiments at Embu and Machanga (central Kenya)

were initiated in 2002, while those at Sidada and Aludeka

(western Kenya) started in 2005. The four sites were selected

to represent different temperatures, levels of rainfall, and

soil conditions. With a higher amount of annual rainfall, the

sites at Sidada (1730 mm, 675 mm in season; 22.6°C) and

Aludeka (1660 mm, 600 mm in season; 24.4°C) represent

a more favorable climate for maize than the sites in central

Kenya, Embu (1175 mm, 475 mm in season; 20.1°C) and

Machanga (795 mm, 290 mm in season; 23.7°C). The soils

at all four sites are heavily weathered, with the Aludeka and

Machanga sites having coarse-textured soils with low SOC

contents (both had < 15% clay, and 7 and 8 g SOC kg

soil,

respectively, at the start of the experiment), while the soils at

Sidada and Embu are fine-textured with relatively high clay

and SOC contents (both had > 55% clay, with 26 and 31 g

SOC kg

soil, respectively, at the start of the experiment).

Initial soil pH in water was 5.3 at Machanga, 5.5 at Aludeka

and 5.4 at the other two sites. The soils were classified

according to the World Reference Base for Soil Resources

(IUSS Working Group

) as a Humic Nitisol at Embu,

a Humic Ferralsol at Sidada, a Haplic Acrisol at Aludeka

and a Ferric Alisol at Machanga. All sites have an almost

flat surface, but at the Embu site this has been achieved by

terracing the field (5% slope). More details on the sites have

recently been published (Laub et al.

2.2 Experimental setup

All four experiments were set up with an identical experi-

mental design. This was a split plot design with three rep-

licates. The main treatments (i.e., main plots) consisted of

the addition of five different types of organic resources,

each applied at two different rates: 1.2 t C ha

and 4 t C

. Besides, a control treatment without any input of

organic resources was established, leading to 11 main plots.

The plots had a size of 12 × 6 m (12 × 5 m at Embu). The

subplots consisted of the application of 120 kg of mineral

fertilizer N ha

per season in the +N treatment (calcium

ammonium nitrate) compared to the absence of mineral N

input in the -N treatment, leading to 22 treatments (i.e., split

plots) tested over three replicates per site. The mineral N

application was split; 40 kg N ha

was applied at planting

and the rest as top dressing about 1.5 months later. All plots

received a blanket application of 60 kg P ha

as triple super-

phosphate and 60 kg K ha

as muriate of potash at planting

in each growing season. Organic resources were applied

only once a year, specifically at planting in the long rainy

season. They were incorporated to about 15 cm soil depth

with a hand hoe. Maize residues were removed at harvest,

so the roots of maize were the only source of C input apart

from external C inputs related to the treatments. The applied

organic resources were chosen in a way to include all four

quality classes of organic resources as defined by Palm et al.

(

2001

). They represented a gradient of C:N ratios, lignin

and polyphenol contents: pruned leaves including stems of

<2 cm thickness from

Tithonia diversifolia

- (C:N of 12,

9% lignin, 2% polyphenols), pruned leaves including small

stems from

Calliandra calothyrsus

(C:N of 14, 11% lignin,

11% polyphenols), stover of

Zea mays

(C:N of 59, 5% lignin,

1% polyphenols), sawdust from

Grevillea robusta

trees (C:N

of 199, 17% lignin, 1% polyphenols), and locally available

farmyard manure (C:N of 12, 20% lignin, 1% polyphenols;

Figure

). The initial assumption behind the organic resource

selection was that organic resources with a low C:N ratio

would be best for supplying plant nutrients, while those

rich in lignin and polyphenols would be better for maintain-

ing SOC (Woomer and Swift

1994

). However, this proved

Page 4

M. Laub et al.

Page 4 of 15

incorrect, because low C:N resources were most effective

at maintaining both maize yield and SOC, while high C:N,

lignin- and polyphenol-rich organic resources neither main-

tained yield nor SOC (Laub et al.

2.3 Datasets A and B

We used two different types of datasets in this study: dataset

A to study the determinants of the seasonal maize yields,

and dataset B to study the determinants of the long-term

maize yield trends. Dataset A consists of the maize grain

yields recorded in each season at the subplot (mineral N

× organic resource treatment × block) level. Dataset B

comprises the long-term trends in maize grain yields and

SOC at the mineral N × organic resource treatment level,

derived from linear mixed models applied to dataset A in

earlier studies (Laub et al.

https:// power. larc. nasa. gov/ docs/ metho

). Dataset B also includes

the long-term trends in soil pH, temperature, rainfall, and the

number of dry days per season, while dataset A incorporates

aggregated weather data per season as described in the

following section. The dependent variable in dataset A is

the maize yield recorded in each season in each subplot at

each site, while in dataset B, it is the modelled long-term

trend of maize yield of each treatment at each site.

2.4 Seasonal covariate data in dataset A

The climate data used as explanatory covariates in this study

were recorded with weather stations at each of the sites, but

some data gaps had to be filled. At Embu and Machanga,

daily minimum and Maximum temperature and rainfall

were available from 2002 until the end of 2007 and after

2017. From 2008 until 2017, only recorded rainfall was

available. Embu had further gaps in rainfall data in 2008

to 2012, 2014 and 2016. At Aludeka and Sidada, manual

recordings of daily minimum and maximum temperature

and rainfall were available for all years from 2005 to 2017,

but Sidada had no temperature data in 2019 and 2020, and

Aludeka had data gaps from July 2017 to April 2018 and

after July 2019. The data gaps were filled using the NASA

POWER product (

dology/

) after bias correction. For this, a linear regression

with measured data as dependent variable (y) and NASA

POWER data as independent variable (x) was conducted.

For temperature, the slopes were not different from 1, but

intercepts for maximum temperature (−0.3°C, −0.4°C,

+3°C, and +6°C for Embu, Machanga, Sidada and Aludeka,

respectively) and minimum temperature (−0.25°C, −0.5°C,

−3°C and +1°C for Embu, Machanga, Sidada, and Aludeka,

respectively) were applied. The final covariate used was the

daily mean temperature, calculated from minimum and

maximum temperature. No bias correction for rainfall was

done because of nonsignificant slopes and intercepts.

The daily primary covariates were then aggregated

to the season-specific covariates, based on the recorded

planting and harvesting dates of maize at each site. Those

were seasonal mean temperature, cumulative rainfall, and

season length. Secondary weather covariates were calculated

from the primary ones. Those were maximum consecutive

dry days in each season, the cumulative growing degree

Figure 1

Illustration of the

effect of soil fertility on maize

growth. Pictures display the

treatment receiving 4 t C

organic resource inputs ha

−1

year

−1

and 120 kg N ha

−1

season

−1

(left) versus the con-

trol receiving no organic inputs

and no mineral N fertilizer

(right). Pictures were taken at

the Machanga experimental

site in July 2023, Marking the

year since the start of the

experiment (Photocredit: Moritz

Laub).

Page 5

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 5 of 15

days, number of days without rainfall, and the standardized

precipitation index (Svoboda and Fuchs

2017

). Further,

ratios of rainfall and growing degree days in the first month

of the growing season (i.e., the first 30 days after planting)

were calculated as proxies for abnormal rainfall and

temperature distributions within the season. Finally, dataset

A contained site, block, and treatment as covariates. They

can be seen as covariates that represent the SOC state of

each treatment, because SOC was only assessed once every

few years and thus not available at seasonal resolution.

2.5 Long‑term trend data in dataset B

Trends in soil pH, temperature, rainfall, and the number of

dry days per season were determined by linear regression

on the seasonal averages per site, with year as covariate.

The long-term trends in SOC and maize yield per site and

experimental treatment were from two recent publications

(Laub et al.

). In brief, mixed linear models were

fitted to maize yield data of all seasons across all years or

all available measurements for SOC. Maize grain yield

data was available from every season across all years, and

were standardized to 12% moisture content. SOC data was

obtained from 2 mm-sieved soil samples taken with a gauge

auger at 0-15 cm soil depth and analyzed by dry combustion

using an elemental analyzer (CHN628; LECO Corporation,

Michigan, USA). SOC measurements were done in the first

experimental year, and in the years 2017, 2018, 2019, and

2021. In Embu and Machanga, additional soil sampling was

conducted every two to three years between 2002 and 2017,

while in Sidada and Aludeka, budget constraints did not

allow sampling from 2005 to 2017.

Mixed linear models were applied to account for

autocorrelation in the data, with a random effect for the main

plot nested in block, each growing season and site in the

yield model, and a random effect for the main plot nested in

block, sampling campaign, and site in the SOC model. The

initial fixed effects were organic resource treatment, mineral

fertilizer N treatment, site and time (years since the start

of the experiment), all possible interactions (to four-way

interactions) were initially included and then eliminated,

until only significant effects and interactions remained

(Zuur et al.

2009

). From this model, the treatment- and site-

specific slopes of the time effect were extracted to derive the

treatment-specific temporal trends in maize yield and SOC.

2.6 Statistical analyses

We used the R software version 4.0.5 (R Core Team

), standard linear regressions, and the nlme package

(Pinheiro et al.

2016

) for mixed linear effects models

to conduct the statistical analyses of the two dependent

variables in this study. For both datasets, we explored

the importance of individual covariates and groups of

covariates in explaining the variability in the dependent

variable, by assessing the reduction in explanatory power

and Akaike Information Criterion) when removing

them from the model.

Dataset B was analyzed with a simple linear effects

model, since autocorrelation in the data had already

been accounted for by the mixed models that the

trends originated from. Dataset A was analyzed by two

approaches. First, we used a mixed linear model across

sites with the established autocorrelation structure of

the model from which the long-term trends were derived

(see section above; details are in Laub, et al.

2023b

;

for which the model was initially developed). To this

model, all weather covariates were added to derive the

fullest model, and then selected groups of covariates (i.e.,

temperature-related, rainfall-related, treatment-related)

were removed to assess their explanatory power. The

pseudo R

for the mixed model was calculated based on

the likelihood (Nagelkerke

1991

). In the second approach,

linear models without random effects were constructed for

each site individually to evaluate the site specificity of

weather- vs. treatment-related covariates on maize yield.

For these, we added a fixed intercept effect to account

for the effects of the experimental blocks. In the site-

specific models, the response variable was square root

transformed to achieve normality and homoscedasticity

of residuals. To delineate the individual effects of weather-

related covariates, site-specific linear models with only the

mineral N fertilizer treatment, organic resource treatment,

time, and all possible two-way interactions between these

three covariates, as well as the additional block effect

were built (base model). To this base model, single

weather-related covariates were added to estimate their

overall effect. Finally, all possible two-way interactions

between mineral N fertilizer treatment, organic resource

treatment, and the primary weather-related covariates

(seasonal mean temperature, in-season rainfall, season

length) were tested to understand how one covariate

affected the maize grain yield depending on the level of

another covariate. For each possible combination of these

five covariates, the significance of their interaction was

assessed by comparing a base model that included the

pair of covariates without interaction to a base model that

included the pair of covariates and their interaction. For

those interactions that were found to be significant, post-

hoc estimates were conducted to understand the nature

of their interactions. Specifically, the maize grain yields

for the range of conditions of both factors were predicted

and were displayed in plots, where maize grain yield was

the y-axis, one covariate (e.g., rainfall) was the x-axis,

Page 6

M. Laub et al.

Page 6 of 15

and the other covariate (e.g., seasonal mean temperature)

was displayed at several levels as the color coding. These

post-hoc predictions were only produced for the observed

range of each covariate at each site. This was done to avoid

extrapolation beyond the data.

3 Results

3.1 Long‑term trends in SOC and maize yield

A gain in SOC over time was only observed at one site

(i.e., Aludeka) and only in the treatment with farmyard

Manure at a rate of 4 t C ha

−1

. At all other sites, SOC

either declined by up to 0.6 g kg

−1

at the clayey sites

(which started at 26 to 31 g SOC kg

−1

soil) and up to 0.2

g kg

−1

at the sandy sites (which started at only 7 to 8

g SOC kg

−1

soil) or was maintained at best, depending on

the treatment. Although both sites started with similarly

low SOC levels, SOC at Machanga could, in contrast to

Aludeka, not be increased by any treatment. The range of

trends in maize grain yields were between increases of 0.2

t ha

−1

in the best treatment at Sidada and losses of

0.15 ha

−1

at Machanga. Maize yields declined in most

treatments at Embu and Machanga, in about half of the

treatments at Aludeka, and in some treatments at Sidada.

However, only Machanga experienced a maize yield

decline in all treatments, while at the other sites yields

remained stable or even increased in the best treatments.

The treatment that was the most effective in maintaining

or increasing SOC and maize yields across sites was the

combination of farmyard manure at a rate of 4 t C ha

−1

with mineral N fertilizer. On the other hand, the

organic resources with high C:N ratio, i.e., maize stover

and sawdust, hardly had any effect on both maize yield

and SOC, whilst

Tithonia

and

Calliandra

residues had a

limited positive effect on both.

3.2 Determinants of long‑term trends in maize

yields

At the humid sites in western Kenya, Aludeka and Sidada,

treatment-specific long-term maize yield trends were sig-

nificantly (p < 0.01) and positively correlated with the

trends in SOC content. Experimental treatments that expe-

rienced the strongest SOC loss also exhibited the strongest

yield loss, whilst those increasing or maintaining SOC

showed increased yield (Figure

). At the sub-humid site

of Embu, the correlation between SOC loss and yield loss

was weaker (p < 0.05) and only observed in the +N but not

in the -N treatments. At the semi-arid site of Machanga,

no significant (p > 0.5) relationship was found between

the long-term trends in yield and SOC, despite significant

differences in SOC content and maize yields between treat-

ments (Laub et al.

Overall, the strongest correlation between temporal

trends in maize yield and SOC was observed at Aludeka,

characterized by a humid climate and a coarse-textured

soil. A mean loss of 0.1 g kg

−1

SOC per year at Aludeka

translated to a grain yield loss of 68 kg ha

−1

per year

in the +N treatment and 38 kg ha

−1

per year in the -N

treatment. In contrast, at Sidada, characterized by a humid

climate and a clayey soil, the -N treatment experienced the

strongest effect, and SOC losses of up to about 0.3 g kg

−1

per year did not lead to grain yield losses. A loss (or gain)

of 0.1 g kg

−1

SOC per year translated to a grain yield loss

(or gain) of 56 kg ha

−1

in the -N treatment and 28 kg ha

−1

in the +N treatment.

Overall, the interaction between site, trend in SOC,

and mineral N fertilizer treatment explained 84% of the

variability in maize yield trends across sites (Table

). Trends

in soil pH, while statistically significant (p < 0.05), added

limited explanatory value (increase to 86% of variability

explained). Temporal trends in seasonal rainfall were not

significantly different from zero, while trends in temperature

were significant at Embu and Aludeka (p < 0.05; Figure S1).

Despite this, neither rainfall nor temperature trends were

significant in explaining the maize yield trends. Notably,

site alone explained 56% of the variability in the temporal

maize yield trends, while the interaction between site and

SOC trend explained 74% of variability (Table

3.3 Determinants of season‑to‑season variability

in maize yields

The significance of weather-related covariates in explaining

the season-to-season maize yield variability across sites

was high (Table S1). A model with site, organic resource

treatments and mineral N fertilizer treatments explained

only 51% of the season-to-season maize yield variability,

which increased to 60% when adding only temperature-

related covariates, to 69% when adding only rainfall-related

covariates, and to 74% when adding both temperature- and

rainfall-related covariates.

However, the degree to which season-to-season vari-

ability in maize yields was explained by each covariate

depended strongly on the site (Table

). Weather vari-

ability played a minor role at Aludeka, as indicated by

the experimental treatments explaining 41% of the maize

yield variability, a value close to that of the model with

all covariates (49%). At Embu, in contrast, experimental

treatments alone only explained 17% of the variability,

while weather variability alone explained 47%, and all

covariates combined 60%. Sidada was similar to Embu,

with treatments alone explaining only 18%, weather alone

Page 7

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 7 of 15

explaining 37% and all covariates combined explaining

57% of the variability in maize yield. At Machanga, treat-

ments (32%) and weather alone (28%) held similar impor-

tance, while 59% of maize yield variability was explained

by all covariates combined.

Individual assessment of weather-related covariates

was conducted by adding each one to the model with only

organic resource and mineral N fertilizer treatments and

their interactions with time. In-season rainfall, standardized

precipitation index, and growing degree days per season had

significant positive effects (p < 0.01) on maize yield at all

sites (Table

), with the length of the growing season show-

ing significance (p < 0.001) at all sites, except Aludeka. As

an indicator of unusually cold or hot mid- and/or late-season

conditions, the ratio of growing degree days in the first 30

days of the season to the total of the entire season had a

significant (p < 0.01) negative effect at all sites. Interest-

ingly, some weather-related covariates had a positive or

insignificant effect at the sites in western Kenya but a nega-

tive effect at the sites in central Kenya. Seasonal mean tem-

perature had a significant negative effect at the drier sites,

Embu and Machanga, but a significant positive effect at the

more humid site, Aludeka. Similarly, the ratio of rainfall in

the first 30 days of the season to the total of the entire sea-

son had a significant negative effect at the drier sites, Embu

and Machanga, but a significant positive effect at the more

humid site, Sidada. The total number of days without rain

had a significant positive effect at the drier sites, Embu and

Figure 2

Relationship between the long-term changes in soil organic

carbon (SOC) and long-term trends in maize yields at the four experi-

mental sites in Kenya, displayed as mean annual change estimated by

a linear mixed model per site and N fertilizer treatment. The cropping

system in the experiment is monoculture of maize with two crop-

ping cycles per year and different treatments of addition of organic

resources and mineral nitrogen (N) fertilizer (see text for details).

Further displayed are the regression of SOC changes on yield trends,

its formula, correlation coefficient and significance level. The yellow

dots represent treatments with the addition of 120 kg mineral N ha

−1

per season, the black dots represent treatments without N fertilizer.

An asterisk behind the site of Sidada indicates a significant differ-

ence between the regression estimates for the +N and -N treatment

of that site. Panels correspond to Aludeka in the -N (

), +N (

), and

+/-N treatments combined (

), Embu in the -N (

), +N (

), and +/-N

treatments combined (

), Machanga in the -N (

), +N (

), and +/-N

treatments combined (

), and Sidada in the -N (

), +N (

), and +/-N

treatments combined (

Page 8

M. Laub et al.

Page 8 of 15

Machanga, but a significant negative effect at Sidada. Addi-

tionally, the effect of in-season rainfall and the standardized

precipitation index on maize yield was greater at the drier

sites in central Kenya compared to the more humid sites

in western Kenya, despite the fact that maize yields were

generally higher in western Kenya. For example, the effect

of in-season rainfall on maize grain yield at Embu (2.7 kg

−1

) and Machanga (1.9 kg ha

−1

) was higher

than at Sidada (1.6 kg ha

−1

) and much higher than

at Aludeka (0.2 kg ha

−1

; note that all values were

back-transformed from their square root, applying the mean

in-season rainfall level of each site).

Table 1

Explanatory strength

of different linear models

across all sites with the trend

in maize yield as the dependent

variable and with different

covariates used. Displayed are

the adjusted R

, the Akaike

Information Criterion (AIC) and

the reduction in R

compared

to the model with the highest

. Abbreviations: SOC: soil

organic carbon; N TRT: mineral

nitrogen fertilizer treatment; dry

days: number of dry days per

season.

Model of yield trend

Adjusted

AIC

Reduction in adjusted R

com-

pared to model with highest

Site only

0.56

−262 −0.3

SOC trend only

0.04

−194 −0.82

N TRT only

0.02

−193 −0.84

pH trend only

0.00

−183 −0.86

Temperature trend only

0.01

−191 −0.85

Rainfall trend only

0.07

−198 −0.79

Dry days trend only

0.00

−190 −0.86

Site × SOC trend

0.74

−304 −0.12

Site × SOC trend × N TRT

0.84

−341 −0.02

Site × SOC trend × N TRT + pH trend

0.86

−335 0

Site × SOC trend × N TRT + temperature trend 0.84

−341 −0.02

Site × SOC trend × N TRT + rainfall trend

0.84

−341 −0.02

Site × SOC trend × N TRT + dry days trend

0.84

−341 −0.02

Table 2

Explanatory strength

of different linear models by

experimental site, with maize

yield per season, treatment and

block as the dependent variable

and with different covariates

used. Displayed are the adjusted

, the Akaike Information

Criterion (AIC) and the

reduction in R

compared to the

full model with all covariates

(highest R

). Treatment

covariates here refer to both the

organic resource and mineral

nitrogen fertilizer treatments.

Site

Model of yields in all years by site

Adjusted R

AIC Reduction in adjusted R

compared to fullest model

Aludeka

Model with only treatment and time

0.41

3346 −0.08

Full model with all covariates

0.49

3009

Model with treatment, time, and

temperature

0.43

3235 −0.06

Model with treatment, time, and rainfall

0.44

3198 −0.05

Model with weather covariates only

0.04

4460 −0.45

Embu

Model with only treatment and time

0.17

4882 −0.43

Full model with all covariates

0.60

3044

Model with treatment, time, and

temperature

0.26

4526 −0.34

Model with treatment, time, and rainfall

0.54

3334 −0.06

Model with weather covariates only

0.47

3669 −0.13

Machanga Model with only treatment and time

0.32

4479 −0.27

Full model with all covariates

0.59

3160

Model with treatment, time and temperature 0.48

3742 −0.11

Model with treatment, time, and rainfall

0.53

3490 −0.06

Model with weather covariates only

0.28

4496 −0.31

Sidada

Model with only treatment, and time

0.18

3316 −0.39

Full model with all covariates

0.57

1623

Model with treatment, time, and

temperature

0.32

2786 −0.25

Model with treatment, time, and rainfall

0.49

2061 −0.08

Model with weather covariates only

0.37

2589 −0.2

Page 9

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 9 of 15

Several significant interactions were identified between

the primary explanatory covariates, namely seasonal mean

temperature, in-season rainfall, and season length. These

interactions were observed both between individual covari-

ates and treatments, and between covariates. Interactions

between in-season rainfall and seasonal mean temperature

Table 3

Effect size of individual weather-related covariates by exper-

imental site, when added to a model that only contains the organic

resource and mineral nitrogen fertilizer treatments, time and their

two-way interactions as covariates. The effect sizes are the β coeffi-

cients of the respective numerical covariates from linear models built

by site, with the square root of maize yield per season as the depend-

ent variable. Only covariates that were significant when adding them

to the treatments-only model are displayed. Abbreviations:

response

variable is the square root of yield (t ha

−1

season

−1

)

0.5

, p <0.05; **, p

<0.01; ***, p <0.001;

not significant.

Covariate

Unit

Aludeka

Embu

Machanga

Sidada

Seasonal mean temperature

°C

0.080***

−0.082***

−0.339***

0.008

Season length

Days

0.001

0.012***

0.014***

0.020***

Cumulative rainfall

0.0004***

0.002***

0.001***

Maximum consecutive dry days

Days

−0.007***

0.002

−0.005***

−0.011***

Cumulative growing degree days

°d

0.0001**

0.0013***

0.0008***

0.0004***

Ratio of rainfall in 1

30 days

0.093

−2.100***

−0.536***

1.041***

Ratio of GDD in 1

30 days

−0.252**

−7.106***

−4.325***

−3.008***

Days without rainfall

Days

−0.0001

0.0041***

0.0084***

−0.0015***

Standardized precipitation index

0.083***

0.408***

0.384***

0.069***

Figure 3

Interactive effects between in-season rainfall and organic

resource input treatment on maize yield per season for the sites

Aludeka (

), Embu (

), Machanga (

), and Sidada (

). Treatments

were the additions of 4 t C ha

−1

of either farmyard manure,

Cal-

liandra

leaves,

Tithonia

leaves, maize stover, saw dust, compared to a

control treatment with no inputs. Displayed are the post-hoc predic-

tions using a base linear models per site to which the in-season rain-

fall and its interaction with organic resource treatments were added.

The base linear models consisted of the organic resource treatment,

mineral N fertilizer treatment, time since experiment start and their

interactions, as well as a block effect. Note that the response vari-

able in the linear model was root-transformed and post-hoc predic-

tions were transformed back. Ribbons indicate the 95% confidence

interval. Abbreviations: **/***, interaction significant at that site (at

p < 0.01/p < 0.001); n.s., interaction not significant at that site.

Page 10

M. Laub et al.

Page 10 of 15

were significant at all sites (p < 0.001; Figure S2). At

Aludeka and Machanga, the effect of in-season rainfall on

maize yield was negative at seasonal mean temperatures up

to 22°C, turning positive at 23 °C and above. At Embu and

Sidada, the effect of in-season rainfall on maize yield was

consistently positive. However, the impact of in-season rain-

fall increased with higher seasonal mean temperatures at

Embu while it weakened at Sidada. The effect of in-season

rainfall also interacted significantly with organic resource

treatment at Aludeka and Machanga (p < 0.01; Figure

It was pronounced in the farmyard manure,

Calliandra

, and

Tithonia

treatments (at 4 t C ha

−1

), but much weaker in

the maize stover, sawdust, and control treatments. A signifi-

cantly stronger effect of in-season rainfall in the +N treat-

ment than in the –N treatment (p < 0.001) was only observed

at Embu and Machanga (Figure S3). Finally, site-specific

significant interactions at all sites also existed between sea-

son length and seasonal mean temperature (p < 0.001; Fig-

ure S4) and between the season length and in-season rainfall

(p < 0.001; Figure S5).

4 Discussion

4.1 Long‑term maize yield trends are strongly

linked to SOC trends

With the interaction of site characteristics and temporal

trends in SOC predominantly explaining the long-term

trends in maize yield, our study provides clear evidence of

the strong association between maintaining SOC and sus-

taining maize yields under tropical conditions, but only

under conditions generally suitable for maize cultivation.

This corroborates the validity of the principles behind

ISFM (Vanlauwe et al.

2010

), namely that SOC serves as

an indicator of soil fertility (Lal

2004

; Janzen et al.

)

and is effective in sustaining crop yields (Tully et al.

Mhlanga et al.

; Thierfelder et al.

). Higher levels

of SOC imply higher amounts of mineralizable soil nutrients

for plant uptake (Bashir et al.

). They further enhance

the soil cation exchange capacity (Soares and Alleoni

2008

particularly in clay-poor soils (Bashir et al.

), which

may explain the stronger relationship between long-term

trends of SOC and yield in the +N compared to the -N

treatment at the clay-poor Aludeka site. The lack of a sig-

nificant regression intercept at Aludeka for both the -N and

+N treatments, unlike at the Sidada and Embu sites that are

characterized by clayey soils (Figure

), further underscores

the site-specific nature of the effect of SOC on maize yields

(Vanlauwe et al.

). It supports the notion that SOC plays

a more critical role in coarse-textured soils that inherently

have limited nutrient (and water) holding capacities (Bashir

et al.

; Thierfelder et al.

It is important to acknowledge in this context that

correlation does not guarantee causation, and, therefore,

it cannot be excluded that other factors besides SOC may

have contributed to changes in maize yields in the long

term. For example, additional micronutrients provided by

organic inputs (Mucheru-Muna et al.

2014

; Xiao et al.

or organic N, P, K and other essential nutrients that are

co-stabilized with SOC (Manzoni et al.

2012

; Sinsabaugh

et al.

2016

) may have also played an important role.

Furthermore, positive feedback loops between increased

SOC and increased crop biomass productivity may exist

(Couëdel et al.

2024

In general, our study aligns with two recent global stud-

ies, showing that higher levels of SOC are associated with

higher maize yields, and vice versa. However, the SOC

levels at which the highest yields were estimated differed

considerably between these two studies, with estimates

being 20 g C kg

soil (Oldfield et al.

2019

) and 43 g C kg

soil (Ma et al.

). This difference, together with the

site-specific effects of SOC trends on yield trends observed

in our study, as well as the interactions with mineral N

fertilizer, suggests that the effect of SOC on yield should

ideally be determined independently for contrasting site

conditions, such as varying rainfall regimes or soil tex-

tures. For example, the decrease in average maize yield

related to decreasing SOC was primarily observed at

Aludeka and Sidada, sites with favorable climatic condi-

tions for maize production (an average of about 600 mm of

in-season rainfall). This finding suggests a clear benefit of

increasing or maintaining SOC under such conditions. In

contrast, at the Machanga and Embu sites, characterized

by in-season rainfall ranging from 300 to 450 mm and

frequent dry spells, SOC changes had little to no associa-

tions with long-term maize yield trends. This indicates

a limited benefit of additional SOC under such rainfall-

limited conditions, somewhat contradicting recent claims

that SOC universally enhances climate change adaptation

(Rumpel et al.

Although SOC can to some extent enhance soil water

holding capacity (Ussiri and Lal

2019

), the overall effect

is limited and often negligible at realistic rates of SOC

change (Minasny and McBratney

2018

). Thus, changes

in SOC likely have limited effects on long-term maize

yield trends in regions with poor rainfall conditions where

water, rather than nutrients, is the main limiting factor

for crop growth. In that sense, our results provide one

possible explanation for the strong spatial variability in

the effectiveness of ISFM practices in sub-Saharan Africa

(Mhlanga et al.

Nevertheless, this should not be interpreted as evidence

that ISFM is only effective under high rainfall conditions,

but rather as a crop-specific limitation related to maize.

For example, a study on sorghum showed that even in

Page 11

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 11 of 15

arid conditions, an increase in yields over time is pos-

sible with ISFM applied to the right cropping system

(Adams et al.

). Further, a meta-analysis showed that

sites with arid conditions and sandy soils had among the

strongest relative increase in maize yield from ISFM com-

pared to control treatments (Chivenge et al.

), most

likely because of the water retention by the freshly added

organic materials that allowed plant growth and fertilizer

uptake. These points demonstrate that our results should

not be generalized to crops other than maize, or to situa-

tions where complementary management strategies, such

as water harvesting, are used. ISFM may, therefore, still

be a valuable component for improving crop yields across

diverse environmental conditions, including those with

limited rainfall.

4.2 Effects of weather on season‑to‑season maize

yield variability are site specific

The importance of weather-related covariates in explaining

seasonal maize yield across the sites in our study is

consistent with findings from a recent global study (Ma

et al.

) and several other studies suggesting in-season

rainfall as a main limiting factor for maize yield in sub-

Saharan Africa (Lobell et al.

Mhlanga et al.

; Simanjuntak et al.

; Rezaei et al.

). However, our site-specific analyses revealed a more

nuanced picture. Because the crop growth-defining factors,

i.e., solar radiation and genotypes, were near optimal at all

four sites and all received best-practice pest management,

the differences in the relative importance of treatments and

weather covariates reflect the dominant growth-limiting

factors for maize productivity at each site (van Ittersum and

Rabbinge

1997

). At Aludeka, which has favorable weather

conditions but poor soil fertility, nutrients are the main

growth-limiting factor (Mubanga and Steyn

). ISFM

treatments therefore have the strongest effects on maize

yields, fostering a strong relationship between long-term

yield and SOC trends. The positive maize yield response

to temperature and the low effect of rainfall at this site are

likely due to generally sufficient soil moisture availability.

Additionally, as Aludeka has low soil clay content, excessive

rainfall may lead to nutrient leaching and thus yield losses

(Weil and Brady

2016

). These complex interactions could

also be an explanation for the overall low importance of

weather covariates in the linear models for this site. At

Machanga, where both weather and soil fertility are growth-

limiting, ISFM treatments and weather covariates hold

similar importance. However, because Machanga is at the

rainfall margin for maize cultivation, long-term SOC trends

are not significantly related to yield trends. The treatment

effects may instead reflect the short-term benefits of added

organic resources, which can improvewater holding capacity,

especially from those with high C:N ratio and high lignin

content (Zhang et al.

). At sites with clay-rich soils,

such as Embu and Sidada, soil fertility is not the major

limitation to plant growth if soil acidity is properly managed

(Zhang et al.

). Thus, weather becomes the main

limiting factor. Nevertheless, the strong correlation between

yield trends and SOC trends at Sidada suggests that SOC

remains important for achieving maximum maize yields

when weather conditions are favorable. This is especially

the case when no mineral N is applied, as evidenced by the

higher slope in the –N treatment (Figure

The more pronounced positive effect of rainfall on maize

yields at Embu and Machanga (e.g., effect sizes; Table

)

aligns with the fact that these sites are near or below the

rainfall margin (about 450 mm of seasonal rainfall) for maize

cultivation. Embu, with an average seasonal rainfall of 476

mm, is close to the lower threshold required for viable maize

yields, while Machanga, with just 290 mm, is well below it

(Ngetich et al.

2014

; Tayel et al.

). Low maize yields at

these sites are thus primarily driven by moisture limitations.

This interpretation is further corroborated by the significant

interactions between ISFM inputs and rainfall at both sites.

Specifically, in seasons with adequate rainfall, maize yields

responded positively to the application of organic resources

with a low C:N ratio (at Machanga; Figure

) and mineral N

fertilizer (at both sites; Figure S3).

Additional evidence for this response comes from the

significant negative effect of a high ratio of rainfall in

the first 30 days to the total in-season rainfall at Embu

and Machanga. Field observations by the site manager,

confirmed that early rainfall followed by dry spells led to

premature germination and subsequent severe drought stress

or even plant death.

Regarding the effect of temperature, we initially wanted

to explore whether daily temperatures above 30 °C would

negatively influence maize yields, as shown in the study by

Lobell et al. (

), conducted using over 20.000 historical

maize trials in sub-Saharan Africa. However, such tempera-

tures occurred in only two seasons (both at Aludeka), mak-

ing it impractical to include this covariate in our analysis.

Still, we found significant negative effects of seasonal mean

temperatures on maize yield at Embu and Machanga, likely

due to interactions with water stress. This is supported by the

greater sensitivity of maize yields to in-season rainfall under

higher temperatures (at all sites except Sidada; Figure S2).

This finding is thus consistent with the finding of Lobell

et al. (

), that high temperatures have 1.7 times stronger

negative effects on maize yields under drought conditions

than under adequate moisture conditions. Given the pro-

jected increases in temperature in Kenya without increases

in rainfall (Mumo et al.

; Ojara et al.

), coping

strategies such as irrigation (Rezaei et al.

) may become

necessary to sustain maize yields in the future.

Page 12

M. Laub et al.

Page 12 of 15

4.3 Local adaptation and targeting of ISFM must

address the site‑specific limitations

By combining the results of the long-term trends analysis

and the season-to-season yield variations, our study

demonstrates that addressing growth-limiting factors

in maize production has both a long-term aspect, i.e.,

maintaining or improving soil fertility, and a short-term

aspect, i.e., ensuring sufficient water availability, especially

under high temperatures. Long-term SOC changes translated

into maize yield changes only at Aludeka and Sidada, where

rainfall is not a growth-limiting factor. In contrast, at the

rainfall-constrained sites of Embu (in the -N treatment) and

Machanga, there was no clear association between SOC

and long-term yield changes, showing a more complex

relationship under water-limited conditions. Consequently,

we posit that the main benefit of maintaining or increasing

SOC through sustainable intensification practices (e.g.,

ISFM) lies in relieving nutrient limitations. Importantly, in

environments, where water is the main limiting factor, SOC

improvements alone may not increase crop yields, making

water management the priority. Effective local adaptation

of ISFM (Vanlauwe et al.

) therefore requires targeting

site-specific constraints. For example, applying low C:N

ratio organic resources together with mineral nitrogen

fertilizer under rainfed conditions appears beneficial only

where in-season rainfall exceeds approximately 475 mm.

In drier areas, several strategies could enable effective use

of ISFM: implementing water harvesting (Kebenei et al.

), or irrigation, which also significantly reduces the

effect of temperature stress and increases the yield per

amount of N applied (Lobell et al.

; Vanlauwe et al.

), transitioning to more drought-tolerant crops such as

sorghum (Adams et al.

), or applying lignin-rich organic

resources with a high C:N ratio (Zhang et al.

), that

enhance water retention. These approaches may allow ISFM

to be effective even in moisture-constrained environments.

A practical challenge to ISFM is the limited availability

of manure, the most effective external organic resource for

maintaining SOC (Laub et al.

https:// doi. org/ 10. 1007/ s13593- 025- 01054-x

). In many farming sys-

tems, crop and livestock production are not well integrated.

As a result, pastoral areas often have surplus manure, while

cropping areas may face shortages or lack proper manure

collection due to free-roaming livestock (Sileshi et al.

2025

Therefore, promoting better integration of crop-livestock

systems, either on the same farms, or through the develop-

ment of value chains for manure trade, could help address

this imbalance. Together with further research on adapt-

ing ISFM to regions with rainfall-limited conditions, these

efforts could unlock the potential of ISFM for a wider range

of agroecological conditions.

5 Conclusion

We presented a unique assessment of the relation between

long-term trends in soil organic carbon and maize yield,

along with season-to-season yield variability, in four long-

term experiments conducted in Kenya, with the goal of

providing nuanced insights into the interacting factors that

affect maize yield in tropical cropping systems. Significant

positive correlations between temporal trends in maize yield

and soil organic carbon showed the benefits of maintaining

soil organic carbon for sustaining maize yields, but only at

sites with adequate seasonal rainfall (475-600 mm in-season

rainfall). In contrast, in regions, such as Machanga, with

around 300 mm in-season rainfall, moisture is a major

limiting factor (especially under rising temperatures), and

maintaining soil organic carbon has less influence on maize

yield outcomes. Next to the long-term trends, seasonal

variability of maize yields was strongly influenced by mean

seasonal temperature and total seasonal rainfall at three of

the four sites, but the magnitude of these weather effects

varied between sites. Therefore, sustainable intensification

strategies should be tailored to local climate conditions,

particularly rainfall, and prioritize interventions based on

the most limiting crop-growth factor.

Supplementary Information

The online version contains supplemen-

tary material available at

Acknowledgements

The AI language model “DeepL Write” has been

used to improve the grammar of the manuscript.

Authors'contributions

BV and DM established the long-term

experiments that this research was conducted in. BV, JS, MC and ML

designed this study. MWMM, DM, SMN, RY and WW managed and

maintained the long-term experiment over the years. SMN, WW, BV,

RY, JS, and ML were involved in the various sampling campaigns. MC,

BV and JS acquired funding for the research. ML summarized the data,

did the statistical analysis in collaboration with CML, and prepared the

original draft. All co-authors contributed to the writing and editing of

the final submitted article.

Funding

Open access funding provided by Swiss Federal Institute

of Technology Zurich. This research has been supported by the

Schweizerischer Nationalfonds zur Förderung der Wissenschaftli-

chen Forschung (grant no. 172940), the European Union’s Horizon

2020 framework (LANDMARC; grant no. 869367), the DSCATT

project “Agricultural Intensification and Dynamics of Soil Carbon

Sequestration in Tropical and Temperate Farming Systems” (grant

nos. AF 1802-001 and FT C002181), supported by the Agropolis

Foundation (“Programme d’Investissement d’Avenir” LabEx Agro,

grant no. ANR-10-LABX-0001-01), TOTAL Foundation (within

a patronage agreement) and the CGIAR Excellence in Agronomy

(EiA) Initiative.

Data availability

The datasets used for the calibration of this study are

available in the IITA data repository, available at

https:// doi. org/ 10. 25502/ be9y- xh75/d

25502/ wdh5- 6c13/d

for SOC and

for yields and biomass.

Page 13

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 13 of 15

Code availability

The data analyses applied the R software to create

standard linear (mixed) models which are easily reproducible from

the description in the text. No specialized software or code was used.

Declarations

Conflict of interest

The authors declare no competing interests.

Open Access

This article is licensed under a Creative Commons Attri-

bution 4.0 International License, which permits use, sharing, adapta-

tion, distribution and reproduction in any medium or format, as long

as you give appropriate credit to the original author(s) and the source,

provide a link to the Creative Commons licence, and indicate if changes

were made. The images or other third party material in this article are

included in the article’s Creative Commons licence, unless indicated

otherwise in a credit line to the material. If material is not included in

the article’s Creative Commons licence and your intended use is not

permitted by statutory regulation or exceeds the permitted use, you will

need to obtain permission directly from the copyright holder. To view a

copy of this licence, visit

http://creativecommons.org/licenses/by/4.0/

References

Adams AM, Gillespie AW, Dhillon GS, Kar G, Minielly C, Koala S,

Ouattara B, Kimaro AA, Bationo A, Schoenau JJ, Peak D (2020)

Long-term effects of integrated soil fertility management practices

on soil chemical properties in the Sahel. Geoderma 366:114207.

https:// doi. org/ 10. 1016/j. geode rma. 2020. 114207

Badu-Apraku B, Fakorede MAB (2017) Maize in Sub-Saharan Africa:

Importance and Production Constraints. In: Badu-Apraku B,

Fakorede MAB (eds) Advances in Genetic Enhancement of Early

and Extra-Early Maize for Sub-Saharan Africa. Springer Interna-

tional Publishing, Cham, pp 3–10

Bashir O, Ali T, Baba ZA, Rather GH, Bangroo SA, Mukhtar SD,

Naik N, Mohiuddin R, Bharati V, Bhat RA (2021) Soil Organic

Matter and Its Impact on Soil Properties and Nutrient Status. In:

Dar GH, Bhat RA, Mehmood MA, Hakeem KR (eds) Microbi-

ota and Biofertilizers, vol 2. Ecofriendly Tools for Reclamation

of Degraded Soil Environs. Springer International Publishing,

Cham, pp 129–159

Berryman E, Hatten J, Page-Dumroese DS, Heckman KA, D’Amore

DV, Puttere J, SanClements M, Connolly SJ, Perry CH (Hobie),

Domke GM (2020) Soil Carbon. In: Pouyat RV, Page-Dumroese

DS, Patel-Weynand T, Geiser LH (eds) Forest and Rangeland

Soils of the United States Under Changing Conditions: A Com-

prehensive Science Synthesis. Springer International Publish-

ing, Cham, pp 9–31. ISBN: 978-3-030-45216-2

Block S, Haile B, You L, Headey D (2021) Heat shocks, maize

yields, and child height in Tanzania. Food Secur 14:93–109.

https:// doi. org/ 10. 1007/ s12571- 021- 01211-6

Cardinael R, Guibert H, Kouassi Brédoumy ST, Gigou J, N’Goran

KE, Corbeels M (2022) Sustaining maize yields and soil carbon

following land clearing in the forest–savannah transition zone of

West Africa: results from a 20-year experiment. Field Crop Res

275:108335.

https:// doi. org/ 10. 1016/j. fcr. 2021. 108335

Chivenge P, Vanlauwe B, Gentile R, Wangechi H, Mugendi D, van

Kessel C, Six J (2009) Organic and mineral input manage-

ment to enhance crop productivity in Central Kenya. Agron

J 101:1266–1275.

https:// doi. org/ 10. 2134/ agron j2008. 0188x

Chivenge P, Vanlauwe B, Six J (2011) Does the combined applica-

tion of organic and mineral nutrient sources influence maize

productivity? A meta-analysis. Plant Soil 342:1–30.

org/ 10. 1007/ s11104- 010- 0626-5

Couëdel A, Falconnier GN, Adam M, Cardinael R, Boote K, Justes

E, Smith WN, Whitbread AM, Affholder F, Balkovic J, Basso B,

Bhatia A, Chakrabarti B, Chikowo R, Christina M, Faye B, Fer-

chaud F, Folberth C, Akinseye FM, Gaiser T, Galdos MV, Gay-

ler S, Gorooei A, Grant B, Guibert H, Hoogenboom G, Kamali

B, Laub M, Maureira F, Mequanint F, Nendel C, Porter CH,

Ripoche D, Ruane AC, Rusinamhodzi L, Sharma S, Singh U,

Six J, Srivastava A, Vanlauwe B, Versini A, Vianna M, Webber

H, Weber TKD, Zhang C, Corbeels M (2024) Long-term soil

organic carbon and crop yield feedbacks differ between 16 soil-

crop models in sub-Saharan Africa. Eur J Agron 155:127109.

https:// doi. org/ 10. 1016/j. eja. 2024. 127109

Erenstein O, Jaleta M, Sonder K, Mottaleb K, Prasanna BM (2022)

Global maize production, consumption and trade: trends and

R&D implications. Food Secur 14:1295–1319.

10. 1007/ s12571- 022- 01288-7

FAO (2023) FAOSTAT Online Database.

https:// www. fao. org/ faost

at/ en/# data/ QCL

. Accessed on 28 Apr 2023

IUSS Working Group (2015) World reference base for soil resources

2014, update 2015. International soil classification system for

naming soils and creating legends for soil maps. World Soil

Resources Reports No. 106. FAO, Rome

Jägermeyr J, Müller C, Ruane AC, Elliott J, Balkovic J, Castillo O,

Faye B, Foster I, Folberth C, Franke JA, Fuchs K, Guarin JR,

Heinke J, Hoogenboom G, Iizumi T, Jain AK, Kelly D, Khabarov

N, Lange S, Lin T-S, Liu W, Mialyk O, Minoli S, Moyer EJ,

Okada M, Phillips M, Porter C, Rabin SS, Scheer C, Schneider

JM, Schyns JF, Skalsky R, Smerald A, Stella T, Stephens H, Web-

ber H, Zabel F, Rosenzweig C (2021) Climate impacts on global

agriculture emerge earlier in new generation of climate and crop

models. Nat Food.

https:// doi. org/ 10. 1038/ s43016- 021- 00400-y

Janzen HH, Janzen DW, Gregorich EG (2021) The ‘soil health’ meta-

phor: Illuminating or illusory? Soil Biol Biochem 159:108167.

https:// doi. org/ 10. 1016/j. soilb io. 2021. 108167

Kebenei MC, Mucheru-Muna M, Muriu-Ng’ang’a F, Ndung’u CK

(2021) Zai Technology and Integrated Nutrient Management for

Improved Soil Fertility and Increased Sorghum Yields in Kitui

County, Kenya. Front Sustain Food Syst 5:714212.

org/ 10. 3389/ fsufs. 2021. 714212

Kihara J, Bolo P, Kinyua M, Nyawira SS, Sommer R (2020) Soil health

and ecosystem services: lessons from sub-Sahara Africa (SSA).

Geoderma 370:114342.

https:// doi. org/ 10. 1016/j. geode rma. 2020.

114342

Lal R (2004) Soil carbon sequestration to mitigate climate change.

Geoderma 123:1–22.

https:// doi. org/ 10. 1016/j. geode rma. 2004.

01. 032

Laub M, Corbeels M, Couëdel A, Ndungu SM, Mucheru-Muna MW,

Mugendi D, Necpalova M, Waswa W, Van de Broek M, Vanlauwe

B, Six J (2023a) Managing soil organic carbon in tropical agro-

ecosystems: evidence from four long-term experiments in Kenya.

SOIL 9:301–323.

https:// doi. org/ 10. 5194/ soil-9- 301- 2023

Laub M, Corbeels M, Mathu Ndungu S, Mucheru-Muna MW, Mugendi

D, Necpalova M, Van de Broek M, Waswa W, Vanlauwe B, Six J

(2023b) Combining manure with mineral N fertilizer maintains

maize yields: evidence from four long-term experiments in Kenya.

Field Crop Res 291:108788.

https:// doi. org/ 10. 1016/j. fcr. 2022.

108788

Lobell DB, Bänziger M, Magorokosho C, Vivek B (2011) Nonlinear

heat effects on African maize as evidenced by historical yield

trials. Nat Clim Change 1:42–45.

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

ate10 43

Lobell DB, Burke MB, Tebaldi C, Mastrandrea MD, Falcon WP, Nay-

lor RL (2008) Prioritizing climate change adaptation needs for

Page 14

M. Laub et al.

Page 14 of 15

food security in 2030. Science 319:607–610.

https:// data. europa. eu/ doi/ 10. 2760/ 835878

1126/ scien ce. 11523 39

Lutz W, Stilianakis N, Stonawski M, Goujon A, Samir KC (2018)

Demographic and human capital scenarios for the 21st century:

2018 assessment for 201 countries. Publications Office of the

European Union.

Ma Y, Woolf D, Fan M, Qiao L, Li R, Lehmann J (2023) Global crop

production increase by soil organic carbon. Nat Geosci 16:1159–

1165.

https:// doi. org/ 10. 1038/ s41561- 023- 01302-3

MacLaren C, Mead A, van Balen D, Claessens L, Etana A, de Haan

J, Haagsma W, Jäck O, Keller T, Labuschagne J, Myrbeck Å,

Necpalova M, Nziguheba G, Six J, Strauss J, Swanepoel PA, Thi-

erfelder C, Topp C, Tshuma F, Verstegen H, Walker R, Watson C,

Wesselink M, Storkey J (2022) Long-term evidence for ecological

intensification as a pathway to sustainable agriculture. Nat Sustain

5:770–779.

https:// doi. org/ 10. 1038/ s41893- 022- 00911-x

Madembo C, Mhlanga B, Thierfelder C (2020) Productivity or stabil-

ity? Exploring maize-legume intercropping strategies for small-

holder conservation agriculture farmers in Zimbabwe. Agr Syst

185:102921.

https:// doi. org/ 10. 1016/j. agsy. 2020. 102921

Manzoni S, Taylor P, Richter A, Porporato A, Ågren GI (2012) Envi-

ronmental and stoichiometric controls on microbial carbon-use

efficiency in soils. New Phytol 196:79–91.

1111/j. 1469- 8137. 2012. 04225.x

Mhlanga B, Ercoli L, Pellegrino E, Onofri A, Thierfelder C (2021)

The crucial role of mulch to enhance the stability and resilience

of cropping systems in southern Africa. Agron Sustain Dev 41:29.

https:// doi. org/ 10. 1007/ s13593- 021- 00687-y

Mhlanga B, Pellegrino E, Thierfelder C, Ercoli L (2022) Conservation

agriculture practices drive maize yield by regulating soil nutrient

availability, arbuscular mycorrhizas, and plant nutrient uptake.

Field Crop Res 277:108403.

https:// doi. org/ 10. 1016/j. fcr. 2021.

108403

Minasny B, McBratney AB (2018) Limited effect of organic matter on

soil available water capacity. Eur J Soil Sci 69:39–47.

org/ 10. 1111/ ejss. 12475

Mubanga DKH, Steyn PMJ (2020) Rainfall is not the most limiting

factor to maize (

Zea mays

L.) production in intermediate rainfall

regions of Zambia. Lessons from Choma District. J Agric Policy

3:18–40.

https:// doi. org/ 10. 47941/ jap. 377

Mucheru-Muna M, Mugendi D, Pypers P, Mugwe J, Kung’u J, Van-

lauwe B, Merckx R (2014) Enhancing maize productivity and

profitability using organic inputs and mineral fertilizer in central

Kenya small-hold farms. Exp Agr 50:250–269.

1017/ S0014 47971 30005 25

Mumo L, Yu J, Ojara M, Lukorito C, Kerandi N (2021) Assessing

changes in climate suitability and yields of maize and sorghum

crops over Kenya in the twenty-first century. Theor Appl Climatol

146:381–394.

https:// doi. org/ 10. 1007/ s00704- 021- 03718-6

Mutuku EA, Roobroeck D, Vanlauwe B, Boeckx P, Cornelis WM

(2020) Maize production under combined conservation agricul-

ture and integrated soil fertility management in the sub-humid and

semi-arid regions of Kenya. Field Crop Res 254:107833.

https://

doi. org/ 10. 1016/j. fcr. 2020. 107833

Nagelkerke NJD (1991) A note on a general definition of the coef-

ficient of determination. Biometrika 78:691–692.

10. 2307/ 23370 38

Ndung’u M, Ngatia LW, Onwonga RN, Mucheru-Muna MW, Fu R,

Moriasi DN, Ngetich KF (2021) The influence of organic and

inorganic nutrient inputs on soil organic carbon functional groups

content and maize yields. Heliyon 7:e07881.

1016/j. heliy on. 2021. e07881

Ngeno V (2024) Profit efficiency among Kenyan maize farmers. Heli-

yon 10:e24657.

https:// doi. org/ 10. 1016/j. heliy on. 2024. e24657

Ngetich KF, Diels J, Shisanya CA, Mugwe JN, Mucheru-Muna M,

Mugendi DN (2014) Effects of selected soil and water conserva-

tion techniques on runoff, sediment yield and maize productivity

under sub-humid and semi-arid conditions in Kenya. CATENA

121:288–296.

https:// doi. org/ 10. 1016/j. catena. 2014. 05. 026

Ojara MA, Yunsheng L, Ongoma V, Mumo L, Akodi D, Ayugi B,

Ogwang BA (2021) Projected changes in East African climate

and its impacts on climatic suitability of maize production areas

by the mid-twenty-first century. Environ Monit Assess 193:831.

https:// doi. org/ 10. 1007/ s10661- 021- 09547-4

Oldfield EE, Bradford MA, Wood SA (2019) Global meta-analysis of

the relationship between soil organic matter and crop yields. Soil

5:15–32.

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

Palm C, Gachengo CN, Delve RJ, Cadisch G, Giller KE (2001)

Organic inputs for soil fertility management in tropical agroeco-

systems: application of an organic resource database. Agr Eco-

syst Environ 83:27–42.

https:// doi. org/ 10. 1016/ S0167- 8809(00)

00267-X

Pinheiro J, Bates D, DebRoy S et al (2016) {nlme}: Linear and Non-

linear Mixed Effects Models. R package version 3.1-127.

https://

doi. org/ 10. 32614/ CRAN. packa ge. nlme

R Core Team (2021) R: A Language and Environment for Statisti-

cal Computing. R Foundation for Statistical Computing, Vienna,

Austria

Reichenbach M, Fiener P, Hoyt A, Trumbore S, Six J, Doetterl S (2023)

Soil carbon stocks in stable tropical landforms are dominated by

geochemical controls and not by land use. Glob Change Biol

29:2591–2607.

https:// doi. org/ 10. 1111/ gcb. 16622

Rezaei EE, Webber H, Asseng S, Boote K, Durand JL, Ewert F, Martre

P, MacCarthy DS (2023) Climate change impacts on crop yields.

Nat Rev Earth Environ 4:831–846.

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

s43017- 023- 00491-0

Rumpel C, Amiraslani F, Chenu C, Garcia Cardenas M, Kaonga M,

Koutika L-S, Ladha J, Madari B, Shirato Y, Smith P, Soudi B,

Soussana J-F, Whitehead D, Wollenberg E (2020) The 4p1000

initiative: opportunities, limitations and challenges for imple-

menting soil organic carbon sequestration as a sustainable devel-

opment strategy. Ambio 49:350–360.

https:// doi. org/ 10. 1007/

s13280- 019- 01165-2

Sanchez PA (2015) En route to plentiful food production in Africa.

Nat Plants 1(1):14014.

https:// doi. org/ 10. 1038/ nplan ts. 2014. 14

Schlenker W, Roberts MJ (2009) Nonlinear temperature effects indicate

severe damages to U.S. crop yields under climate change. P Natl

Acad Sci USA 106:15594–15598.

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

09068 65106

Sileshi GW, Stewart ZP, Odhong J, Mhlanga B, Amede T, Aynekulu

E, Thierfelder C, Marenya P, Dittmer KM, Aliyu KT, Chikowo

R, Chiduwa M, Ngoma H, Snapp S (2025) A review of organic

inputs to inform soil health advice for African smallholder farm-

ers: localization matters. npj Sustain Agric 3:1–15.

10. 1038/ s44264- 025- 00063-3

Simanjuntak C, Gaiser T, Ahrends HE, Ceglar A, Singh M, Ewert

F, Srivastava AK (2023) Impact of climate extreme events and

their causality on maize yield in South Africa. Sci Rep 13:12462.

https:// doi. org/ 10. 1038/ s41598- 023- 38921-0

Sinsabaugh RL, Turner BL, Talbot JM, Waring BG, Powers JS, Kuske

CR, Moorhead DL, Follstad Shah JJ (2016) Stoichiometry of

microbial carbon use efficiency in soils. Ecol Monogr 86:172–

189.

https:// doi. org/ 10. 1890/ 15- 2110.1

Soares MR, Alleoni LRF (2008) Contribution of soil organic carbon to

the ion exchange capacity of tropical soils. J Sustain Agr 32:439–

462.

https:// doi. org/ 10. 1080/ 10440 04080 22573 48

Sommer R, Paul BK, Mukalama J, Kihara J (2018) Reducing losses

but failing to sequester carbon in soils – the case of conservation

agriculture and integrated soil fertility management in the humid

Page 15

Maize yield responses to soil organic carbon under integrated soil fertility management in…

Page 15 of 15

tropical agro-ecosystem of Western Kenya. Agr Ecosyst Environ

254:82–91.

https:// doi. org/ 10. 1016/j. agee. 2017. 11. 004

Svoboda MD, Fuchs BA (2017) Handbook of Drought Indicators and

Indices. In: Wilhite D, Pulwarty RS (eds) Drought and Water Cri-

ses. Integrating Science, Management, and Policy, 2nd edn. CRC

Press, Boca Raton, pp 155–207

Tayel MY, Mansour HA, Pibars SK (2015) Performance of maize under

bubbler irrigation system. In: Goyal MR, Mansour HAA (eds)

Closed Circuit Trickle Irrigation Design, 1st edn. Apple Academic

Press, New York, pp 135–147

Thierfelder C, Paterson E, Mwafulirwa L, Daniell TJ, Cairns JE,

Mhlanga B, Baggs EM (2022) Toward greater sustainability:

how investing in soil health may enhance maize productivity in

Southern Africa. Renew Agr Food Syst 37:166–177.

org/ 10. 1017/ S1742 17052 10004 42

Tully K, Sullivan C, Weil R, Sanchez P (2015) The state of soil deg-

radation in sub-Saharan Africa: baselines, trajectories, and solu-

tions. Sustainability-Basel 7:6523–6552.

https:// doi. org/ 10. 3390/

su706 6523

Ussiri DAN, Lal R (2019) Soil Degradation in Sub-Saharan Africa:

Challenges and Opportunities for Restoration. In: Lal R, Stewart

AB (eds) Soil Degradation and Restoration in Africa. CRC Press,

Boca Raton, pp 1–24

van Ittersum MK, Rabbinge R (1997) Concepts in production ecology

for analysis and quantification of agricultural input-output com-

binations. Field Crops Res 52:197–208.

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

S0378- 4290(97) 00037-3

Vanlauwe B, Bationo A, Chianu J, Giller KE, Merckx R, Mokwunye U,

Ohiokpehai O, Pypers P, Tabo R, Shepherd KD, Smaling EMA,

Woomer PL, Sanginga N (2010) Integrated soil fertility manage-

ment: operational definition and consequences for implementation

and dissemination. Outlook Agric 39:17–24.

5367/ 00000 00107 91169 998

Vanlauwe B, Descheemaeker K, Giller KE, Huising J, Merckx R,

Nziguheba G, Wendt J, Zingore S (2015) Integrated soil fertility

management in sub-Saharan Africa: unravelling local adaptation.

Soil 1:491–508.

https:// doi. org/ 10. 5194/ soil-1- 491- 2015

von Fromm SF, Hoyt AM, Lange M, Acquah GE, Aynekulu E, Berhe

AA, Haefele SM, McGrath SP, Shepherd KD, Sila AM, Six J,

Towett EK, Trumbore SE, Vågen T-G, Weullow E, Winowiecki

LA, Doetterl S (2021) Continental-scale controls on soil organic

carbon across sub-Saharan Africa. Soil 7:305–332.

10. 5194/ soil-7- 305- 2021

Weil RR, Brady NC (2016) The nature and properties of soils, 15th

edn. Pearson London, London, UK

Woomer PL, Swift MJ (1994) The biological management of tropical

soil fertility. John Wiley, Chichester, UK

Xiao Q, Huang Y, Wu L, Tian Y, Wang Q, Wang B, Xu M, Zhang W

(2021) Long-term manuring increases microbial carbon use effi-

ciency and mitigates priming effect via alleviated soil acidification

and resource limitation. Biol Fert Soils 57:925–934.

org/ 10. 1007/ s00374- 021- 01583-z

Zhang J, Fu G, Jin Z, Chi L, Xu G, Yue D (2022) Optimization of

organic mulches thickness improves soil moisture retention under

controlled conditions. Phyton 91:841–857.