Mäkipää et al. (2024). Is the organic carbon-to-clay ratio a reliable indicator of soil health?

Geoderma 444 (2024) 116862

Available online 27 March 2024

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

, Eduardo Martínez-García , Tiina T

Is the organic carbon-to-clay ratio a reliable indicator of soil health?

Raisa M

¨

akip

¨

a

¨

a , Lorenzo Menichetti

*

Aleksi Lehtonen

Natural Resources Institute Finland (Luke), Latokartanonkaari 9, FI-00790 Helsinki, Finland

A R T I C L E I N F O

Handling Editor: L. Morgan Cristine

Keywords:

Soil organic carbon (SOC)

SOC:Clay ratio

European mineral soils

LUCAS soil survey

Forest soil, agricultural soil

Soil monitoring

A B S T R A C T

Climate action plans under the Paris Climate Agreement and other national commitments aimed at improving

soil-based ecosystem services require the operational monitoring of soil carbon (C). The European Union is

aiming to enhance soil health, and as part of the proposed Soil Monitoring Law, the European Commission

recommends the monitoring of the soil C loss indicator among other soil health indicators. In this study, we

evaluate the feasibility of the proposed soil C loss indicator by assessing its performance using the EU-wide 2009

LUCAS soil survey data. The proposed indicator is the soil organic carbon (SOC) to clay ratio, with a threshold

value of 1:13. The results are also compared with the C stock changes reported by countries to the climate

convention (UNFCCC). Our results reveal that the variation in SOC and clay content at European scale exceeds

that of the data used to develop the proposed indicator. We also found that the variation in the SOC content was

influenced not only by clay content but also by climate and land-use reflecting C input levels. Therefore, the

defined threshold is inadequate for detecting degraded soils if the SOC and clay content are beyond the con-

ditions used to establish the criteria. Furthermore, major discrepancies were observed between the soil carbon

stock changes reported by the national greenhouse gas (GHG) inventories and the proportions of degraded soils

identified by using the soil C loss indicator. We conclude that employing a single indicator such as SOC:Clay ratio

with one threshold value for all soils across various land covers, management practices, and climatic conditions,

as defined by the European Commission for the Soil Monitoring Law, is inappropriate for monitoring soil C loss.

1. Introduction

Anthropogenic activities have considerably depleted initial soil

organic carbon (SOC) stocks, thus directly and negatively affecting soil

health, its functioning, and capacity to deliver ecosystem services such

as food security, water security, climate regulation or biodiversity pro-

tection (

Kopittke et al., 2023

). Soil health refers to ‘the ability of the soil

to sustain the productivity, diversity, and environmental services of

terrestrial ecosystems

’

(

FAO ITPS, 2020

). Yet, the concept remains

elusive and cannot be quantitatively measured (

Janzen et al., 2021

). As

a result, considerable efforts have concentrated on identifying a set of

measurable biological, chemical, and/or physical indicators to facilitate

the assessment of soil health. These indicators should represent soil

processes, properties, and ecosystem services, but they must also be

sensitive to management and be feasible for practical use (

Doran and

Zeiss, 2000

). In this context, the SOC content is the most widely used

indicator (

Bünemann et al., 2018

). In developing a plan for restoring soil

health, the loss of SOC has been considered as one of the measures of soil

degradation by the European Commission in the proposed Soil Moni-

toring Law, which aims to turn all soils within the European Union into

healthy state by 2050 (

Panagos et al., 2022

). In the proposed law, the

SOC content and SOC:Clay ratio were suggested to be an indicator of the

soil C loss (

European Commission, 2023

IPCC, 2006; European Environment Agency, 2023

Monitoring of the soil C serves both international climate policy and

national targets set for soil productivity and C stocks. Currently, moni-

toring of the SOC stock changes is a part of the national greenhouse gas

inventories (

) and

countries may have included soil C sequestration in their climate action

plans as set in the Paris Climate Agreement (

UNFCCC, 2023

). Loss of

SOC on mineral soils can be measured by repeated sampling (e.g.,

Grüneberg et al., 2014; Ortiz et al., 2013

). However, most countries use

modeling to report changes in SOC stocks in their national greenhouse

gas (GHG) inventories following IPCC guidelines (

IPCC, 2019

), since

repeated measurements for soil carbon monitoring are both labor-

intensive and expensive (

). On the other hand, the

soil C stability and direction of change of the SOC stock might be

* Corresponding author.

E-mail address:

lorenzo.menichetti@luke.fi

(L. Menichetti).

Contents lists available at

ScienceDirect

Geoderma

journal homepage:

www.elsevier.com/locate/geoderma

https://doi.org/10.1016/j.geoderma.2024.116862

Received 16 December 2023; Received in revised form 8 March 2024; Accepted 15 March 2024

Geoderma 444 (2024) 116862

2

assessed using indicators based on measured SOC content and other soil

variables. According to

Drexler et al. (2022)

, the interpretation of

measured SOC contents as a C loss indicator relies on i) the comparison

of SOC values with a set of baseline values derived from representative

datasets (

De Rosa et al., 2023; De Vos et al., 2015; Nunes et al., 2021

),

and/or ii) the use of desirable SOC thresholds or reference values. These

thresholds can vary according to relationships between SOC and soil

properties (

Carter, 2002; Loveland and Webb, 2003

), the use of SOC

values of undisturbed control sites values (

Maharjan et al., 2020; Poe-

plau and Don, 2013

), and/or the use of proxies such the proposed SOC:

Clay ratio (

Johannes et al., 2017; Prout et al., 2021

Krause et al., 2018; Siebers et al., 2018

The amount of clay is proposed to be a major factor influencing the

ability of a soil to retain SOC (

Hassink, 1997

), related to the SOC-

stabilizing properties of clay (

).

Dexter et al. (2008)

suggested that a SOC:Clay ratio of 1:10 is the limit

for clay to protect SOC. This approach was further reassessed by

Johannes et al. (2017)

and

Prout et al. (2021)

defining thresholds for

SOC:Clay ratios of 1:8, 1:10, and 1:13 to indicate the boundaries be-

tween ‘very good

’

,

’

good

’

, ‘moderate

’

, and ‘degraded

’

levels of soil

structural conditions, respectively. In view of this, the European Soil

Monitoring Law proposes to use a SOC:Clay ratio of 1:13 as the threshold

for identifying degraded European mineral soils (

European Commission,

2023

).

In this study, we tested the performance of SOC:Clay ratio as an in-

dicator of soil C loss on the soil data obtained from the Soil Module of

2009 Land Use and Coverage Area Frame Survey (LUCAS) survey in

combination with other datasets for soil class, climate zone, and soil C

stock changes. The LUCAS soil survey dataset covers a wide range of

land covers, soil types, and SOC-clay content data across the EU. The

principal objective was to evaluate applicability of the SOC:Clay ratio

and its 1:13 threshold as a normalized ‘soil health

’

metric for identifying

carbon loss of European mineral soils and to compare this indicator with

the C stock changes reported by countries to the climate convention

(UNFCCC). Secondary objectives were to analyze distribution of the

degraded soils across land cover and climatic zones, and to examine the

potential of additional variables in predicting variation of the SOC

content.

2. Material and methods

We relied on the 2009 LUCAS soil survey (

https://esdac.jrc.ec.eu

ropa.eu/content/lucas-2009-topsoil-data

) comprising a total of 21,859

data points in Europe. Since soil texture data (incl clay content) were

collected during the LUCAS soil sampling in 2009, a more recent LUCAS

soil survey was not used in this study. The data points are georeferenced

and represent a composite soil sample of four subsamples to a depth of

approximately 20 cm (measured after removing the organic layer and

vegetation residues), collected in a radius of 100 m from the relative

LUCAS point. Further details can be found in the documentation of the

dataset (

The land cover data from each data point was obtained from the

primary 2009 LUCAS dataset (

https://ec.europa.eu/eurostat/web/lu

cas/data/primary-data/2009

). This dataset was cross-referred with

several datasets using spatial join, with the aim of determining both the

soil class and climatic zone of each point. For determining climate zones,

we relied on the current K

¨

oppen-Geiger climate classification from

Beck

et al. (2023)

. The soil classification, which was based on the World

Reference Base (WRB) for Soil Resources (

IUSS Working Group WRB,

2015

), was obtained from the European Soil Database v2.0 to define the

soil class (

https://esdac.jrc.ec.europa.eu/content/european-soil-databa

se-v20-vector-and-attribute-data

). In further selection, we excluded all

points with SOC content above 200 g kg

−

1

dw, identified as Histosols by

the WRB. This is because the high organic content in Histosols is likely a

result of either extremely high inputs or extremely reduced decompo-

sition, and as such, Histosols may not be sensitive to processes related to

clay protection mechanisms.

We considered coarse land cover classes, namely Artificial (A in 2009

LUCAS classification), Cropland (B), Forest (C), Shrubland (D), Grass-

land (G), Bare land (F), Water land (G), and Wetlands (H). We studied

the distribution of ‘healthy/degraded soils

’

in this dataset, defined by

the 1:13 (or 0.077) SOC:Clay ratio, according to land cover, countries,

and climatic zones. In addition, we assessed the explanatory potential of

the different variables included in the dataset to explain the SOC vari-

ance. For this, we trained a random forest model on 70 % of the data

points, with the remaining 30 % reserved for model validation. Subse-

quently, we used this model to determine the importance of each vari-

able in predicting SOC. The implementation used for this analysis was

the

paRF

command (

Zhang et al., 2021

) in the

BAGofT

package in R

software (version 4.3.1), and the metaparameter

mtry

(i.e., number of

variables to sample at each split) was optimized via

caret

. Three random

forests (RF) models were trained (Eqs. (1)-(3)):

SOC

≈

f

(

r

coarse

,

r

clay

,

r

sand

,

r

silt

,

pH

CaCl

2

,

CaCO

3

,

Country

,

LC

,

lat

,

long

,

)

(1)

SOC

≈

f

(

r

coarse

,

r

clay

,

r

sand

,

r

silt

,

pH

CaCl

2

,

CaCO

3

,

Country

,

LC

,

lat

,

long

,

WRB

)

(2)

SOC

≈

f

(

r

coarse

,

r

clay

,

r

sand

,

r

silt

,

pH

CaCl

2

,

CaCO

3

,

Country

,

LC

,

lat

,

long

,

CEC

)

(3)

where ‘r

’

is each specific soil fraction, ‘

pH

CaCl

2

’

is pH (measured in 0.01

M CaCl

2

), ‘CaCO

3

’

is calcium carbonate content, ‘Country

’

is the indi-

vidual country, ‘

LC’

is the land cover class, while ‘lat

’

and ‘long

’

are the

latitude and longitude, respectively. We note that variables ‘N

’

nitrogen,

‘P

’

phosphorous, ‘K

’

and potassium, which are also included in the 2009

LUCAS soil survey, were excluded from Eq. (1) because they are always

strongly correlated with SOC, being also elements present in organic

matter. The parameter WRB in Eq. (2) represents the soil class according

to WRB, while the parameter ‘CEC

’

in Eq. (3) represents cation exchange

capacity of soil.

Variable importance is calculated as mean decrease in impurity,

which is averaged across the trees in the forest. This is based on the

reduction in impurity (Gini impurity) that each feature brings to the

trees within the model. This metric quantifies the contribution of each

feature to the overall homogeneity of the nodes and leaves by averaging

the impurity decrease attributed to that feature across all trees in the

forest. The more a feature decreases impurity, the more important it is

considered for the model

’

s predictive accuracy. The metric has only a

value relative to each trained model.

Finally, we also compared the proportion of healthy/degraded soils

(based on criteria SOC:Clay with threshold of 1:13) in each country

according to land cover types with the emission data obtained from the

annual EU greenhouse gas inventory as submitted to the UNFCCC (

Eu-

ropean Environment Agency, 2023

). These data are developed based on

various methodologies. Greenhouse gas inventories report their soil C

stock changes based on methods that are in line with IPCC good practice

guidelines (

Penman et al., 2003

). These methods vary according to their

Tier levels from default emission factors to state-of-art soil models and to

repeated soil inventories (

European Environment Agency, 2023

).

3. Results

3.1. Land cover and indicated soil C loss

Over the two dimensions of SOC and clay content, which were pro-

posed to be a basis of the soil C loss indicator, the distribution of the data

points in what is classified as degraded space was clearly influenced by

land cover. In this analysis, we focused on the three main land cover

classes, i.e., cropland, grassland, and forest, which represent 44.6 %,

21.8 % and 28.6 % of the total points, respectively (see Supplementary

Table S1

). Our results show a much higher proportion of degraded soils

R. M

¨

akip

¨

a

¨

a et al.

Geoderma 444 (2024) 116862

3

among croplands in comparison to grasslands and forests (

). The

share of degraded soil increased in the order forests (4.2 %)

<

grasslands

(15.7 %)

<

croplands (51.0 %; Supplementary

Table S2

). Within other

land cover classes (Supplementary

Table S2

), bare lands showed the

highest proportion of degraded soils (69.7 %), while these soils were less

represented in shrublands and wetlands (18.8 % and 4.0 %,

respectively).

The median of the SOC content in cropland, grassland, and forest on

mineral soils (when observations above 200 g kg

−

1

dw were excluded)

were 14.2, 15.8, and 18.2 g kg

−

1

dw, respectively, whereas their median

of clay content 210, 150, and 70 g kg

−

1

dw, respectively. Thus, grassland

and forest contained more SOC in comparison to cropland, while the

clay content of the soil was highest in cropland (

, Supplementary

Table S3

).

3.2. SOC:Clay as soil C loss indicator and reported C losses by countries

According to the SOC:Clay ratio, it is estimated that on average 29.1

% of the mineral soils analyzed in the 2009 LUCAS survey are non-

healthy (i.e., degraded;

). However, there is a large variation in

the proportion of degraded soils between studied countries, ranging

from 1.2 % in Ireland to 68.5 % in Spain (

b).

The SOC:Clay ratio is proposed as an indicator of the soil C loss.

Therefore, the results that were based on the indicator values were

compared to the soil C stock changes reported in the national GHG in-

ventories (

a). Based on the last 10 years of United Nations

Framework Convention on Climate Change (UNFCCC) data for mineral

soils (and excluding countries within a

±

0.5 t C ha

−

1

which were

considered neutral), all grasslands and most of the forests (except

Ireland) acted as C sinks. Croplands in several countries (Belgium,

Finland, France, Germany, Netherlands, Poland, Portugal, and Slovenia)

acted instead as C sources. In those countries, proportion of healthy

croplands according to SOC:Clay ratio varied from 55 % in Slovenia to as

high as 71 % in Germany. Interestingly, our findings reveal that Spain

has neutral C soils in all reported land covers according to their GHG

reporting to the UNFCCC, but only around 30 % of soils were classified

as healthy with C loss indicator. In Estonia, the proportion healthy soils

vary between 20 and 50 % depending on land-use, but simultaneously

all these land-uses have been reported as carbon sinks by the UNFCCC

reporting. The comparison of results between reporting of soil C stock

changes to the UNFCCC and the proposed soil health indicator highlights

inconsistencies between measures that are aiming to monitor soil C

losses.

Our analysis also draws attention to the latitudinal gradient and its

correspondence with climatic conditions on the dominance of healthy

soils across the 2009 LUCAS survey. The healthiest soils are concen-

trated in northern European countries, while more degraded soils are

found in southern countries (

b). Among the climatic zones repre-

sented in the dataset (

), the colder climates, i.e., Continental

subartic (Dfc), and Continental humid with warm summers (Dfb), pre-

sented 98.5 % and 85.7 % of healthy soils, respectively. These regions

are situated in northern Europe. The northern part of the boreal region

has a Dfc climate, while the southern boreal region and the majority of

eastern Europe are characterized by Dfb climate (

Supplementary

Fig. S1

Temperate oceanic (Cfb) regions, mainly in UK, France, certain parts

in Germany, as well as mountain ranges in Italy, had 76 % of soils

classified as healthy (

). Moreover, the Mediterranean climate with

warm summers (Csb) had 68 % of the points classified as healthy. These

regions are characterized by a Mediterranean climate that has oceanic

influences, such as those found in northern Portugal and Spain. In

contrast, Mediterranean areas with hot summers (Csa) only 45 % of the

soils was classified as healthy. These regions are characterized by

Mediterranean macchia vegetation with some xeric adaptation, as in the

south-west part of Italy and Spain. Similarly, the Hot humid continental

climate (Dfa) only included 48 % of healthy soils. It should be noted that

soils from this climate are primarily present in Hungary, representing 98

% of all data points recorded for this climate class.

The Oceanic subtropical (Cfa) climate accounts for only 42 % of

healthy soils (

pean Environment Agency, 2023

). This climate is common in highly productive

agricultural regions around the Mediterranean basin, such as Pianura

Padana in Italy. While the climate is hot, it is also typically quite humid.

Finally, Arid (BWk and BWh) and Semi-arid steppe climates (BSk and

BSh) are predominantly situated in central Spain, with smaller pro-

portions in southern Italy and in Greece. These regions only included 30

%, 25 %, 20 %, and 0 % of healthy soils, respectively. They represent the

driest Mediterranean areas, such as most of Puglia in Italy and the

western part of central Spain, where agriculture requires significant

adaptation to cope with the arid conditions.

3.3. Importance of variables in explaining the SOC variance

The 10 most important predictors in the first model (Eq. (1),

Fig. 4

)

determining variation of the SOC content were: 1) clay, 2) pH, 3) if land

Fig. 1.

Distribution of the soil organic carbon (SOC) and clay content for different land covers (cropland, forest, and grassland). Density plots of SOC and clay

contents for each land cover are shown in panel a). Panel b) shows the 2-dimension density distribution of the data points of the 2009 LUCAS soil survey on

croplands, croplands, and grasslands in the same space. Dashed line in panels a) and b) represents the SOC:Clay threshold of 1:13. Less represented land-use classes

are not plotted for readability and SOC contents above 200 g kg

−

1

dry weight (dw) were excluded, since they were considered to represent organic soils where the

SOC:Clay content is not proposed to be a relevant indicator.

R. M

¨

akip

¨

a

¨

a et al.

Geoderma 444 (2024) 116862

4

cover was a cropland or any other class, 4) latitude, 5) longitude, 6) sand

fraction, 7) calcium carbonate (CaCO

3

) content, 8) coarse fraction, 9) silt

fraction, and 10) if land cover was grassland or any other class.

In terms of linear correlations (i.e., Pearson

’

s correlation and R

2

of a

linear regression), from the lowest correlation to the highest, the vari-

ables correlated with organic C as follows: Clay (Pearson

’

s

=

0, R

2

=

0),

silt (Pearson

’

s

=

-0.01, R

2

=

0), sand (Pearson

’

s

=

0.01, R

2

=

0), K

(Pearson

’

s

=

-0.01, R

2

=

0), P (Pearson

’

s

=

0.02, R

2

=

0), coarse

(Pearson

’

s

=

0.11, R

2

=

0.01), CaCO

3

(Pearson

’

s

=

-0.12, R

2

=

0.02),

Longitude (Pearson

’

s

=

0.12, R

2

=

0.01), CEC (Pearson

’

s

=

0.27, R

2

=

0.08), Latitude (Pearson

’

s

=

0.32, R

2

=

0.11), pH(CaCl

2

) (Pearson

’

s

=

-0.36, R

2

=

0.13), and N (Pearson

’

s

=

0.84, R

2

=

0.71).

The first random Forest (RF) model (Eq. (1)) could explain 50 % of

the variance on the independent validation dataset (

Supplementary

Fig. S2

), which even fall to 49 % when including the WRB soil classes as

predictors (Eq. (2)). However, the introduction of CEC (Eq. (3)) signif-

icantly contributed to explaining a large portion of the SOC variability,

as the only available information on C inputs to the soil were the land

cover classes. On the independent validation dataset, this model

explained 73 % of the variance.

Fig. 2.

Proportions of healthy (non-degraded) mineral soils by the three major land cover classes and country included in the 2009 LUCAS soil survey (a) and the

proportion of healthy/degraded soils by country (b). The colors in panel (a) show the carbon stock change for each land cover based on the last 10 years of

greenhouse gas inventory reporting under United Nations Framework Convention on Climate Change (UNFCCC) for mineral soils. Blue bars denote carbon sink, while

orange bars indicate carbon source. Pink bars denote

“

neutral

”

countries, within a

±

0.5 t C ha

−

1

change. Grey bars denote missing data. The small labels under each

bar denote the land covers cropland (C), forest (F), and grassland (G). The colors in panel (b) correspond to degraded (red, shaded) and healthy (green) soils.

Fig. 3.

Proportion of soil data classified as healthy (green) or degraded (red shaded) based on the SOC:Clay 1:13 threshold for different K

¨

oppen climatic zones (a),

with the density distribution of their latitude (b) and longitude (c). Density distributions are calculated on all the points considered with kernel density estimation,

darker color corresponds to higher point density.

R. M

¨

akip

¨

a

¨

a et al.

Geoderma 444 (2024) 116862

5

4. Discussion

4.1. The SOC:Clay ratio as indicator of soil carbon loss

Observed discrepancies between soil carbon stock changes reported

by the national GHG inventories and proportions of degraded soils

identified by using the soil C loss indicator showed that a simple indi-

cator such as SOC:Clay ratio with one threshold value for all soils across

land cover classes, management practices and climate conditions cannot

serve soil health monitoring as proposed by the European Commission

for the Soil Monitoring Law. Although, monitoring of a simple indicator

of the soil C loss cannot replace soil carbon inventory that builds on

repeated soil measurements or comprehensive soil C modelling (

Euro-

), the SOC:Clay ratio may be a valid

indicator for soil structural quality that protects SOC on local scale as

proposed in the studies where approach was developed (

Dexter et al.,

2008; Johannes et al., 2017; Prout et al., 2021

). In soils with SOC:Clay

<

1:13, majority of the SOC is assumed to be protected by mineral matter

and when the ratio is

>

1:13 there is assumed to be unprotected SOC

which may be relatively easily mineralized in case of management

change or changes in environmental conditions. For a soil health indi-

cator that is relevant on a larger scale, variation in the SOC and clay

content should be considered and threshold values specific to different

land cover classes, edaphic and climatic conditions should be developed

(

Rabot et al., 2024

). For example, in a dataset encompassing 3800 sites

of croplands, grasslands, and forest soils throughout the UK where in-

dicator based on SOC:Clay ratio was tested, the median clay content

varied from 242 g kg

−

1

dw in forest soils to 260 g kg

−

1

dw in permanent

grassland soils (

Prout et al., 2021

). In the European wide 2009 LUCAS

soil survey data, however, the median clay contents of forest and

grassland soils across Europe were 80 and 180 g kg

−

1

dw, respectively.

Thus, the SOC:Clay threshold values derived by

Prout et al (2021)

cannot be applicable for soils across different land-use classes in Europe.

The SOC:Clay ratio in the 2009 LUCAS soil survey data indicate that

a large proportion of the cropland soils is non-healthy (degraded), while

the soils of forests and grasslands exhibited higher SOC content and

lower clay content being therefore considered healthy if 1:13 threshold

is applied (

et al., 2013; Prout et al., 2021

). There are, however, several limitations to consider

when using the SOC:Clay ratio and its associated degradation threshold

as a soil health indicator across different land cover types. For instance,

clay content is not directly affected by the land cover (vegetation) and

management history, which are drivers of C inputs to the soil and

together with microbial activity determine SOC dynamics (

Wiesmeier

). Croplands tend to have, since they are

usually located in the bottom part of the soil catena, higher clay content

but simultaneously lower SOC content when compared to forests (

ogel-Knabner and Amelung, 2021

)

due to lower C inputs that forest and grassland soils. Due to their higher

inputs, forest and grassland soils have usually more C than cropland soils

(

Morais et al., 2019

). This suggests that land management is much more

intensive in agriculture, which has annually repeated agricultural

practices, compared to forestry, where there can be several decades

between operations (e.g., soil preparation, thinning, and final felling). In

any case, it appears clear the importance of C inputs to determine levels

of SOC stocks. On the other hand, relatively low clay content of forest

soils may relate to land cover history, i.e., areas with higher clay content

were more fertile and converted to croplands, while soils with lower clay

content remained as forests and grasslands. All this means that SOC:Clay

ratio alone without also considering C inputs to the soil, is not a

consistently sensitive indicator for detecting soil carbon loss on different

land-use classes. Consequently, it is necessary that such soil health in-

dicators are tailored to specific conditions and land covers. The problem

can of course be mitigated by subsetting the SOC variability into main

classes, for example cropland/forest/grassland, but still it is likely that

there will be a significant part of the SOC variability explained by C

inputs, which are variable also within each major class. The finer clas-

sification, based on some proxy of C inputs, we use, the more we would

mitigate the problems of the SOC:Clay indicator at the expense of

reducing its general applicability.

Our study found an influence of soil types (Supplementary

Table S4

),

but often in interaction with land cover and climate. It is well known

that different soil types inherently contain varying levels of SOC and

clay (

Prout et al., 2021

). According to our analyses, Vertisols were the

most degraded class, despite being rich in high-active clay minerals

(

K

¨

), likely due to cumulative loss of C

during their intensive and long-term use in agriculture. The second and

third most degraded soil classes where Calcisols and Solonchaks, which

are soils with high concentration of calcium and sodium salts (respec-

tively) and are found in arid and semi-arid regions, which suggests an

impact of climate and limited productivity on their degradation status.

On the other hand, Podzols, which are usually soils having sandy to

loamy sand texture and not particularly rich in clay, where the second

least degraded class. This is likely due to Podzols being usually char-

acterized by slow microbial activity in cool climate and forest

’

s high C

input, partly persistent towards decomposition, to the soil. Cryosols

were the less degraded class, and these soils are usually on marginal

lands such as tundra but here cold climate associated with cryogenic

processes (

Ping et al., 2008; K

¨

ogel-Knabner and Amelung, 2021

) is likely

to play a role slowing down decomposition. Thus, influence of soil

classes seemed to be consequence of many different processes.

According to applied SOC:Clay ratio as an indicator for soil C loss, the

proportion on non-degraded healthy mineral soils was highest in the

cool climate of northern Europe and Ireland, while in the Mediterranean

region a large proportion of the soils were degraded (

b and

Relying solely on the SOC:Clay ratio oversimplify the multiple processes

involved in soil C dynamics and penalizes regions or countries with land

covers and soil properties that lead to lower SOC:Clay ratios. Recently,

Poeplau and Don (2023)

carried out a study on German soils, in which

they criticized the SOC:Clay ratio as an inappropriate metric for

measuring SOC loss sensitivity. The authors argued that this indicator is

highly reliant on clay and is highly susceptible to strong bias, but it is

also being partially insensitive to SOC changes. Our research expands on

the analyses of the use of this indicator in the wider EU region, high-

lighting the limitations of the proposed indicator. We suggest that

sensitivity of a certain soil to C losses is more strongly influenced by land

cover, intensity of land management, and climate rather than clay

content (

and

Hallsworth and Wilkinson, 1958

). Soil C stocks are affected by land-use and land

management practices, since intensive management and frequent dis-

turbances reduce both mean biomass stocks over rotations and annual

organic matter inputs to soil (

Lorenz and Lal, 2010

). Climate and soil

fertility define productivity, which affects biomass and organic matter

inputs to soil. In addition, while the ability of soils to store organic C into

Fig. 4.

Relative importance of the first 10 explanatory variables defined by the

first Random Forest model (Eq. (1)). Note that the relative variable importance

is proportional to the fraction of explained variance for each variable, but not

the same thing. Colour is proportional to the relative importance and is added

only for readability purposes.

R. M

¨

akip

¨

a

¨

a et al.

Geoderma 444 (2024) 116862

6

soil particles is partly determined by their clay content, the type of clay

minerals present also affects this process (

Georgiou et al., 2022

). There

are regional patterns in this respect, e.g., northern EU regions are richer

than southern regions in vermiculite and less in smectite (

Ito and Wagai,

2017

). As soil formation proceeds, clay composition undergoes changes.

Soils that are enriched in subsoil by certain type of clay minerals are

considered in international soil classification systems (IUSS Working

Group WRB 2022). Thus, existing soil maps are connected to clay

mineralogy. However, it is hard to define a precise geographical pattern

of clay activity. Including clay mineralogy into the predictive models of

SOC might contribute to explain more variance. However, it is difficult

to speculate on the extent of this contribution without high spatial res-

olution soil data or clay mineralogy data related to each specific LUCAS

sample.

4.2. What variables are the most important in determining soil organic

carbon content?

Although the SOC content does not necessarily determine the

fragility of SOC and trends of SOC stock changes, the same variables that

impact SOC stocks can assist in identifying potential future trends and

the risk of SOC losses. Therefore, a good predictor of SOC content may

also be a good predictor of soil C health.

According to our analysis, the relationship between SOC and inde-

pendent variables appears to be rather complex and likely with in-

teractions between the main drivers: soil C protection/stabilization,

climate, and C inputs to the soil (and therefore land cover). We showed

that an important predictor of SOC was whether a sampling point is on

cropland or not, which is the most impacting land cover for soil C.

Based on the statistical analyses (random forest) conducted in this

study, the clay content was identified as one of the main predictors of the

variation of soil organic matter when CEC was not considered. However,

the model only accounted for 50 % of the total variance in the validation

dataset, so there is still a lot of unexplained variation. Moreover, in

terms of linear correlation, the explanatory power of clay fraction was

among the lowest of all the variable tested, suggesting that although clay

can indeed contain information about soil C, the relationship is not

linear and is influenced by other variables (e.g., land cover). Therefore,

although clay is known as a good proxy for some of the processes related

to soil C loss (in particular, all the processes influencing the protection of

soil C through adsorption;

Xu et al., 2016

), the clay fraction alone is a

poor indicator of soil C content.

When also cation exchange capacity (CEC) was included in the model

(Eq. (3)), it emerged as more important than clay, and the predictive

power of the model increased substantially to a 73 % of the variance of

the validation dataset explained by it. This might be related to clay

exchange sites, but may also be attributed to the ability of organic

matter to retain soluble ions because of its abundance of weak bonding

sites (

). This makes hard to determine if

the accumulation of SOC might be a result of the effect of organic matter

on CEC or the cause. The good predictive power of pH in all models

could have a similar explanation.

Silt, coarse, and sand fractions were all identified as good predictors,

among the texture descriptors, together with clay, suggesting that part

of the variance due to soil physical processes is related also to coarser

fractions. These may also provide information regarding hydrology or

other processes associated with soil C, which is not present in clay.

Additionally, another important predictor was latitude. Latitude has

a strong influence on temperature, which in turn strongly affects SOC

decomposition (

Lloyd and Taylor, 1994

). Similarly, longitude is also

important, particularly in combination with latitude as it contains in-

formation about the distance from the sea and/or large mountain ranges

and therefore the precipitation patterns. The interaction between lati-

tude and longitude collectively provides information on climate. Ac-

cording to our analysis, soil class (WRB) was not a good predictor of

SOC. We can only speculate about the reason for that, but it is possible

that the information contained in the soil classification, which is highly

multivariate and not quantitative, may not be suitable for describing the

processes associated with SOC protection in a general enough way.

Another possibility is soil edaphic variability, which is always a concern

when using soil maps due to their relatively coarse resolution compared

to soil variability, which may result in inaccurate data.

4.3. Limitations of the study: Representativity of the LUCAS sampling

The LUCAS soil survey has proven crucial in mapping the variability

of European soils (

Orgiazzi et al., 2017

). However, as an extensive

sampling effort requires coordination of multiple actors, the survey

presents potential sources of bias. For example, there is a documented

spatial bias in the exclusion of soils located at an altitude higher than

1500 m.a.s.l. (

Schmidt et al., 2018

). Furthermore, sampling protocol can

be challenged, since the organic layer of the boreal forest soils, which is

highly sensitive to disturbance induced loss of soil C, is not considered.

Additionally, a sampling protocol, where sampling of the organic layer is

not considered and separation of organic layer and uppermost mineral

soil layer is not specified, may result in varying quantities of organic

layer being included in the soil sample that represents the upper mineral

soil layer. This can lead to biased results on the soil carbon content of

that layer (see

compared the 2015

LUCAS soil with German national soil inventory data. They found much

higher variability and a bias towards higher values of C and N in the

LUCAS data, possibly because of imprecision in how the organic layer

was separated from the mineral layer.

Panagos et al. (2020)

had re-

ported the same issue, again concerning forest soils. Also,

reported that in some cases part of the organic layer might have

been sampled together with mineral soils in the 2009 sampling. The

identification of the organic layer in forest soils is difficult also for

trained soil scientists, and it is reasonable to imagine that the LUCAS

sampling, relying on more general professionals, could have faced this

problem. This might have increased the variability and reduced the

significance of any predictive relationship of SOC, including the SOC:

Clay ratio.

5. Conclusions

The SOC:Clay ratio with a threshold value of larger than 1:13 for

healthy soils as proposed by the European Commission in the Soil

Monitoring Law might be a very speditive indicator of soil C loss, but its

general applicability as such is only within narrow ranges of SOC values,

clay contents, land-use, and climate. Although there is evidence of the

linkage between SOC:Clay ratio and soil structural quality, which in-

dicates soil functioning and protection of SOC by mineral matter

(

Johannes et al., 2017; Prout et al., 2021

), this process may not be the

most important. By examining the points of the 2009 LUCAS soil dataset,

we found that the variation in the SOC content was influenced not only

by clay content but also by climate and land-use reflecting C input levels.

Thus, the defined threshold is inadequate for detecting degraded soils if

the SOC and clay content are beyond the conditions used to establish the

criteria.

More efficient monitoring of the soil C and the risk of soil degrada-

tion could be achieved by utilizing other potential indicators beside

SOC:Clay ratio, such as more strictly process-based approach that ex-

amines the ratio between particulate and mineral associated organic

matter (POM and MAOM, respectively) as a warning signal of soil C

fragility (

Whalen et al., 2022

). However, such a method requires rela-

tively expensive analyses and monitoring network. A more cost-efficient

strategy could rely on SOC models, which already account land-use and

climate effects, to determine the reference maximum achievable C

stocks for the entire EU. Thereafter, the measured levels may be

compared to the reference. However, the lessons learned from the

model-based setting of the forest reference level for 2021

–

2025 under

the Land Use, Land Use Change, and Forestry (LULUCF) regulation and

R. M

¨

akip

¨

a

¨

a et al.

Geoderma 444 (2024) 116862

7

then determining LULUCF 2030 target based on mean values of the re-

ported C sinks suggest that it may be challenging to develop and agree

method that could provide solid scientific basis for the threshold of the

healthy soil.

Currently, the national GHG inventories report SOC loss (or positive

change in the SOC stock) through repeated SOC stock measurements or

modelling (e.g.,

European Environment Agency, 2023

). Given the

inconsistency between the soil C stock change estimates of the GHG

inventories and the SOC:Clay ratio as an indicator of soil C loss (

org/10.1016/j.geoderma.2024.116862

),

we conclude that the ratio is unsuitable for the monitoring of the SOC

changes. Since many countries do not estimate soil C stock changes for

all land-use classes or they continue using default emissions factors in

their reporting under the UNFCCC, it is evident that both identification

of non-healthy soils with a set of indicators and reporting of soil C stock

changes under UNFCCC with higher Tier methods need repeated soil

measurements and/or advanced soil modelling. Therefore, our study

highlights the importance of implementing repeated soil measurements

by national soil monitoring frameworks and further developing soil

health indicators, scope, and relevant thresholds for SOC status evalu-

ation. This will provide solid data for the local implementation of the

European Soil Monitoring Law across different land-uses and biomes.

CRediT authorship contribution statement

Raisa M

¨

akip

¨

a

¨

a:

Writing

–

review

&

editing, Writing

–

original draft,

Supervision, Resources, Project administration, Funding acquisition,

Conceptualization.

Lorenzo Menichetti:

Writing

–

review

&

editing,

Writing

–

original draft, Visualization, Methodology, Formal analysis,

Conceptualization.

Eduardo Martínez-García:

Writing

–

review

&

editing, Writing

–

original draft.

Tiina T

¨

orm

¨

anen:

Writing

–

review

&

editing, Writing

–

original draft.

Aleksi Lehtonen:

Writing

–

review

&

editing, Writing

–

original draft, Funding acquisition,

Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial

interests or personal relationships that could have appeared to influence

the work reported in this paper.

Data availability

Data used in this study is available as OpenData as described in the

manuscript

Acknowledgements

We acknowledge the LUCAS Soil survey for providing the European

soil data and the national GHG inventories that created the UNFCCC

data on soil carbon stock changes by the countries and land-use types.

This study received funding from the European Union

’

s Horizon

2020 and Horizon Europe research and innovation programs under

Grant Agreements number 101000289 for the HoliSoils project and

number 101091010 for the Soil Mission project Benchmarks. The study

has been done with affiliation to the UNITE Flagship funded by the

Finnish Research Council (decision no 337655).

Appendix A. Supplementary data

Supplementary data to this article can be found online at

https://doi.

https://doi.org/10.1038/s41597-023-02549-6

References

Beck, H.E., McVicar, T.R., Vergopolan, N., Berg, A., Lutsko, N.J., Dufour, A., Zeng, Z.,

Jiang, X., van Dijk, A.I.J.M., Miralles, D.G., 2023. High-resolution (1 km) k

¨

oppen-

Geiger maps for 1901

–

2099 based on constrained CMIP6 projections. Sci. Data 10,

724.

https://doi.org/10.2134/agronj2002.3800

Bünemann, E. K., Bongiorno, G., Bai, Z., Creamer, R. E., De Deyn, G., de Goede, R.,

Fleskens, L., Geissen, V., Kuyper, T. W., M

¨

ader, P., Pulleman, M., Sukkel, W., van

Groenigen, J. W., Brussaard, L. (2018). Soil quality

–

A critical review. Soil Biol.

Biochem., 120, 105

–

125. 10.1016/j.soilbio.2018.01.030.

Carter, M.R., 2002. Soil quality for sustainable land Management. Agron. J. 94, 38

–

47.

De Rosa, D., Ballabio, C., Lugato, E., Fasiolo, M., Jones, A., Panagos, P., 2023. Soil

organic carbon stocks in european croplands and grasslands: how much have we lost

in the past decade? Glob. Chang. Biol. 00, e16992

De Vos, B., Cools, N., Ilvesniemi, H., Vesterdal, L., Vanguelova, E., Carnicelli, E., 2015.

Benchmark values for forest soil carbon stocks in Europe: results from a large scale

forest soil survey. Geoderma 251

–

252, 33

–

46.

https://doi.org/10.1016/j.geoderma.2008.01.022

geoderma.2015.03.008

.

Dexter, A.R., Richard, G., Arrouays, D., Czy

˙

z, E.A., Jolivet, C., Duval, O., 2008.

Complexed organic matter controls soil physical properties. Geoderma 144 (3

–

4),

620

–

627.

.

Doran, J.W., Zeiss, M.R., 2000. Soil health and sustainability: managing the biotic

component of soil quality. Appl. Soil Ecol. 15 (1), 3

–

11.

https://doi.org/10.1016/

S0929-1393(00)00067-6

.

Drexler, S., Broll, G., Flessa, H., Don, A., 2022. Benchmarking soil organic carbon to

support agricultural carbon management: a german case study. J. Plant Nutr. Soil

Sci. 185, 427

–

440.

https://doi.org/10.1002/jpln.202200007

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

Proposal for a DIRECTIVE OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL on

Soil Monitoring and Resilience (Soil Monitoring Law). Brussels, 5.7.2023, COM

(2023) 416 final, 2023/0232 (COD).

European Environment Agency (2023). Annual European Union greenhouse gas

inventory 1990

–

2021 and inventory report 2023. Submission to the UNFCCC

Secretariat. EEA/PUBL/2023/044. 2023.

FAO ITPS (2020). Towards a Definition of Soil Health. Soil Letters. 2020. Food and

Agriculture Organization of the United Nations (FAO) and Intergovernmental

Technical Panel on Soils (ITPS).

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

ˇ

skov

´

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

¨

om, A., Feng, W.,

Harden, J.H., 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, 3797.

.

Grüneberg, E., Ziche, D., Wellbrock, N., 2014. Organic carbon stocks and sequestration

rates of forest soils in Germany. Glob. Chang. Biol. 20, 2644

–

2662.

10.1111/gcb.12558

.

Hallsworth, E.G., Wilkinson, G.K., 1958. The contribution of clay and organic matter to

the cation-exchange capacity of the soil. J. Agric. Sci. 51, 1

–

3.

10.1017/S0021859600032743

.

Hassink, J., 1997. The capacity of soils to preserve organic C and N by their association

with clay and silt particles. Plant Soil 191, 77

–

87.

https://doi.org/10.1023/A:

1004213929699

IPCC (2006) 2006 IPCC Guidelines for National Greenhouse Gas Inventories. Available

at: http://www.ipcc-nggip.iges.or.jp/public/2006gl/index.htm.

IPCC (2019) Refinement to the 2006 IPCC Guidelines for National Greenhouse Gas

Inventories. Available at: https://www.ipcc-nggip.iges.or.jp/public/2019rf/index.

IUSS Working Group WRB (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 N

o

. 106, Rome. FAO.

Ito, A., Wagai, R., 2017. Global distribution of clay-size minerals on land surface for

biogeochemical and climatological studies. Sci. Data

4

, 170103.

10.1038/sdata.2017.103

Janzen, H.H., Janzen, D.W., Gregorich, E.G., 2021. The ‘soil health

’

metaphor:

Illuminating or illusory? Soil Biol. Biochem. 159, 108167

soilbio.2021.108167

Johannes, A., Matter, A., Schulin, R., Weisskopf, P., Baveye, P.C., Boivin, P., 2017.

Optimal organic carbon values for soil structure quality of arable soils. Does clay

content matter? Geoderma 302, 14

–

21.

https://doi.org/10.1016/j.geoderma.2020.114785

geoderma.2017.04.021

.

K

¨

ogel-Knabner, I., Amelung, W., 2021. Soil organic matter in major pedogenic soil

groups. Geoderma 384, 114785.

.

Kopittke, P.M., Minasny, B., Pendall, E., Rumpel, C., McKenna, B.A., 2023. Healthy soil

for healthy humans and a healthy planet. Crit. Rev. Environ. Sci. Technol. 52,

4308

–

4324.

https://doi.org/10.1080/10643389.2021.2024484

.

Krause, L., Rodionov, A., Schweizer, S.A., Siebers, N., Lehndorff, E., Klumpp, E.,

Amelung, W., 2018. Microaggregate stability and storage of organic carbon is

affected by clay content in arable Luvisols. Soil Till. Res. 182, 123

–

129.

https://doi.

org/10.1016/j.still.2018.05.003

.

Lloyd, J., Taylor, J.A., 1994. On the temperature dependence of soil respiration. Funct.

Ecol. 8, 315

–

323.

https://doi.org/10.2307/2389824

.

Lorenz, K., Lal R. (2010) Carbon sequestration in forest Ecosystem. Springer Press. .

Loveland, P., Webb, J., 2003. Is there a critical level of organic matter in the agricultural

soils of temperate regions: a review. Soil Tillage Res. 70 (1), 1

–

18.

10.1016/S0167-1987(02)00139-3

.

Maharjan, B., Das, S., Acharya, B.S., 2020. Soil health gap: a concept to establish a

benchmark for soil health management. Global Ecol. Conserv. 23, e01116

.

akkinen, M., Muukkonen, P., Peltoniemi, M., 2008. The costs of

monitoring changes in forest soil carbon stocks. Boreal Environ. Res. 13, 120

–

130

Morais, T.G., Teixeira, R.F.M., Domingos, T., 2019. Detailed global modelling of soil

organic carbon in cropland, grassland and forest soils. PLoS One 14 (9), e0222604

.

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

R. M

¨

akip

¨

a

¨

a et al.

Geoderma 444 (2024) 116862

8

protocol and evaluation applied to soil organic C. Soil Sci. Soc. Am. J. 85,

1196

–

1213.

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

.

Orgiazzi, A., Ballabio, C., Panagos, P., Jones, A., Fern

´

andez-Ugalde, O., 2017. LUCAS

soil, the largest expandable soil dataset for Europe: a review. Eur. J. Soil Sci. 69 (1),

140

–

153.

https://doi.org/10.1111/ejss.12499

Ortiz, C.A., Liski, J., G

¨

arden

¨

as, A.I., Lehtonen, A., Lundblad, M., Stendahl, J.,

Å

gren, G.I.,

Karltun, E., 2013. Soil organic carbon stock changes in swedish forest soils

–

a

comparison of uncertainties and their sources through a national inventory and two

simulation models. Ecol. Model. 251, 221

–

231.

Panagos, P., Montanarella, L., Barbero, M., Schneegans, A., Aguglia, L., Jones, A., 2022.

ecolmodel.2012.12.017

.

Panagos, P., Ballabio, C., Scarpa, S., Borrelli, P., Lugato, E., Montanarella, L. (2020) Soil

related indicators to support agro-environmental policies. EUR 30090 EN.

Publications Office of the European Union, Luxembourg. 10.2760/011194.

Soil priorities in the European Union. Geoderma Reg. 29, e00510

Penman, J., Gytarsky, M., Hiraishi, T., Krug, T., Kruger, D., Pipatti, R., Buendia, L., Miwa,

K., Ngara, T., Tanabe, K.

&

Wagner, F. (2003). IPCC Good practice guidance for land

use, land-use change and forestry, Institute for Global Environmental Strategies

(IGES), Japan.

Ping, C.L., Michaelson, G.L., Kimble, J.M., Romanovsky, V.E., Shur, Y.L., Swanson, D.K.,

Walker, D.A., 2008. Cryogenesis and soil formation along a bioclimate gradient in

Arctic North America. J. Geophys. Res. 113, G3.

https://doi.org/10.1029/

2008JG000744

.

Poeplau, C., Don, A., 2013. Sensitivity of soil organic carbon stocks and fractions to

different land-use changes across Europe. Geoderma 192, 189

–

201.

10.1016/j.geoderma.2012.08.003

.

Poeplau, C., Don, A., 2023. A simple soil organic carbon level metric beyond the organic

carbon-to-clay ratio. Soil Use Manag. 39, 1057

–

1067.

https://doi.org/10.1111/

sum.12921

.

Prout, J.M., Shepherd, K.D., McGrath, S.P., Kirk, G.J.D., Haefele, S.M., 2021. What is a

good level of soil organic matter? an index based on organic carbon to clay ratio.

Eur. J. Soil Sci. 72, 2493

–

2503.

https://doi.org/10.1111/ejss.13012

Rabot, E., Saby, N.P.A., Martin, M.P., Barr

´

e, P., Chenu, C., Cousin, I., Arrouays, D.,

Angers, D., Bispo, A., 2024. Relevance of the organic carbon to clay ratio as a

national soil health indicator. Geoderma 443, 116829.

https://doi.org/10.1002/jpln.201800128

geoderma.2024.116829

.

Schmidt, S., Ballabio, C., Alewell, C., Panagos, P., Meusburger, K., 2018. Filling the

European blank spot

–

Swiss soil erodibility assessment with topsoil samples. J. Plant

Nutr. Soil Sci. 181 (5), 737

–

748.

https://doi.org/10.1016/j.geoderma.2017.11.024

Siebers, N., Bauke, S.L., Tamburini, F., Amelung, W., 2018. Short-term impacts of forest

clear-cut on P accessibility in soil microaggregates: an oxygen isotope study.

Geoderma 315, 59

–

64.

.

T

´

oth, G., Jones, A., Montanarella, L. (2013). LUCAS Topsoil Survey. Methodology, data

and results. JRC Technical Reports. Luxembourg. Publications Office of the

European Union, EUR26102

–

Scientific and Technical Research series. Doi:

10.2788/97922.

UNFCCC (2023) Nationally determined contributions under the Paris Agreement.

Synthesis report by the secretariat, UN Climate Change Conference 14 Nov 2023

FCCC/PA/CMA/2023/12 https://unfccc.int/documents/632334.

Whalen, E.D., Grandy, A.S., Sokol, N.W., Keiluweit, M., Ernakovich, J., Smith, R.G.,

Frey, S.D., 2022. Clarifying the evidence for microbial- and plant-derived soil

organic matter, and the path towards a more quantitative understanding. Glob.

Chang. Biol. 28, 7167

–

7185.

https://doi.org/10.1111/gcb.16413

Wiesmeier, M., Hübner, R., Barthold, F., Sp

¨

orlein, P., Geu

ß

, U., Hangen, E., Reischl, A.,

Schilling, B., von Lützow, M., K

¨

ogel-Knabner, I., 2013. Amount, distribution and

driving factors of soil organic carbon and nitrogen in cropland and grassland soils of

southeast Germany (Bavaria). Agr. Ecosyst. Environ. 176, 39

–

52.

10.1016/j.agee.2013.05.012

Xu, X., Shi, Z., Li, D., Rey, A., Ruan, H., Craine, J.M., Liang, J., Zhou, J., Luo, Y., 2016.

Soil properties control decomposition of soil organic carbon: results from data-

assimilation analysis. Geoderma 262, 235

–

242.

https://doi.org/10.1080/01621459.2021.1979010

geoderma.2015.08.038

.

Zhang, J., Ding, J., Yang, Y., 2021. Is a classification procedure good enough? a

goodness-of-fit assessment tool for classification Learning. J. Am. Stat. Assoc. 118

(542), 1115

–

1125.