You are on page 1of 13

Pedosphere 22(2): 230–242, 2012

ISSN 1002-0160/CN 32-1315/P


c 2012 Soil Science Society of China
Published by Elsevier B.V. and Science Press

Spatial Distribution of Soil Organic Matter Using Geostatistics: A Key


Indicator to Assess Soil Degradation Status in Central Italy∗1

A. MARCHETTI1 , C. PICCINI1,∗2 , R. FRANCAVIGLIA1 and L. MABIT2


1
Research Centre for the Soil-Plant System, CRA-RPS, Via della Navicella 2–4, 00184 Rome (Italy)
2
Soil and Water Management & Crop Nutrition Laboratory, FAO/IAEA Agriculture & Biotechnology Laboratory, IAEA
Laboratories Seibersdorf, PO Box 100, Wagramerstrasse 5, A-1400 Vienna (Austria)
(Received March 16, 2011; revised November 5, 2011)

ABSTRACT
Soil organic matter (SOM) content is one of the main factors to be considered in the evaluation of soil health and fertility.
As timing, human and monetary resources often limit the amount of available data, geostatistical techniques provide a valid
scientific approach to cope with spatial variability, to interpolate existing data and to predict values at unsampled locations
for accurate SOM status survey. Using geostatistical and geographic information system (GIS) approaches, the spatial
variability of some physical and chemical soil parameters was investigated under Mediterranean climatic condition in the
Abruzzo region of central Italy, where soil erosion processes accelerated by human induced factors are the main causes of
soil degradation associated with low SOM content. Experimental semivariograms were established to determine the spatial
dependence of the soil variables under investigation. The results of 250 soil sampling point data were interpolated by means
of ordinary kriging coupled with a GIS to produce contour maps distribution of soil texture, SOM content related to texture,
and C/N ratio. The resulting spatial interpolation of the dataset highlighted a low content of SOM in relation with soil
texture in most of the surveyed area (87%) and an optimal C/N ratio for only half of the investigated surface area. Spatial
location of degraded area and the assessment of its magnitude can provide decision makers with an accurate support to
design appropriate soil conservation strategies and then facilitate a regional planning of agri-environmental measures in the
framework of the European Common Agricultural Policy.
Key Words: C/N ratio, ordinary kriging, soil organic carbon, soil quality, spatial variability

Citation: Marchetti, A., Piccini, C., Francaviglia, R. and Mabit, L. 2012. Spatial distribution of soil organic matter using
geostatistics: A key indicator to assess soil degradation status in central Italy. Pedosphere. 22(2): 230–242.

INTRODUCTION
zations, land degradation has moreover negative eco-
Land degradation is associated with the irretrieva- logical and socio-economic impacts. Worldwide eco-
ble loss of the basic soil resource, and thus is a major nomic cost of soil degradation from agricultural land
threat to water and biogeochemical cycles, biodiversity was estimated at US$ 400 billion per year (Pimentel,
and plant primary productivity. Accelerated soil degra- 2006). Extensive soil and land degradation processes
dation decreases soil fertility and productivity (on-site mostly human induced are occurring also in Europe
impacts) and increases related environmental pollution due to extreme intensification of agriculture and mis-
and sedimentation problems (off-site impacts). Current management of soil resource (EEA, 2002; Boardman
concerns about the effects related to accelerated soil and Poesen, 2006; CEC, 2006b). In 2003 for European
resource degradation generate an urgent need for ob- Union (EU), the total assessed cost of soil degradation,
taining reliable quantitative data on its extent world- i.e., erosion, organic matter decline, salinization, land-
wide (Pimentel, 2006). As stated by most of the United slides and contamination based on available data was
Nations organizations and non-governmental organi- up to e38 billion annually (Montanarella, 2010).

∗1
Supported by the Italian Ministry of Agricultural, Food and Forestry Policies (No. DM 19366).
∗2
Corresponding author. E-mail: chiara.piccini@entecra.it.
SPATIAL DISTRIBUTION OF SOM IN CENTRAL ITALY 231

Soil erosion is a selective process of soil redistri- Paustian, 2011; Farina et al., 2011).
bution in the landscape and thus contributes to the The problem of SOM depletion under Mediter-
spatial variability of soil quality (Verity and Anderson, ranean climatic conditions is of particular concern in
1990). Many researches highlight and demonstrate that central and southern Italy due to high temperature and
water erosion, wind erosion and tillage erosion are the reduced soil moisture that accelerate SOM decomposi-
main factors explaining the spatial variability of soil or- tion processes. In this part of Italy, the adoption of in-
ganic matter (SOM) content in cultivated fields (e.g., tensive mechanized agriculture accelerates SOM degra-
Mabit and Bernard, 1998; Fenton et al., 2005; Ritchie dation rate. Effectively, deep ploughing adopted to im-
et al., 2005; Polyakov and Lal, 2008). In addition, as prove soil structure, permeability and aeration, and
SOM represents one of the largest reservoir of organic to help crop growth, combined with non-conservative
C in the global carbon cycle (Lützow et al., 2006), agronomic practices, e.g., monoculture under conven-
evaluating soil carbon sequestration, storage and re- tional tillage without residues left on the soil surface to
distribution at various scales is of pressing concern for reduce rainfall erosivity, causes a SOM dilution in the
soil protection and in mitigation strategies for global arable soil layer due to the mixing with the underlying
warming. horizons poor in SOM.
SOM affects the chemical and physical properties The European Commission has set out common
of the soil and its overall health, and is actually a key principles for protecting soils across the EU, proposing
parameter of soil quality and a soil fertility indicator. a framework directive to combat European soil degra-
Besides providing nutrients and habitat to soil organi- dation (CEC, 2006a). The recently adopted Soil The-
sms, organic matter influences soil physical properties matic Strategy for Soil Protection by the European
in binding soil particles into aggregates, and in im- Commission provides the legal framework for EU mem-
proving soil water holding capacity (Lal, 2003; Bot and ber states to implement adequate responses in order
Benites, 2005; Lal, 2007). to reverse the negative trend of land and soil degrada-
In cultivated land without restoring the organic tion in Europe. This directive establishes a framework
matter and nutrient contents, nutrient cycles are bro- for the protection of soil, its sustainable use and the
ken, soil fertility declines and the balance in agro- preservation and restoration of the soil capacity to per-
ecosystems is impaired. Depending on cultivation prac- form as many as possible of the environmental, ecologi-
tices, plant cover, soil drainage and agro-climatic cal, socio-economical and cultural functions. In this
conditions, agricultural practices generally accelerate framework, the knowledge of the current state of the
SOM decomposition, and therefore increase arable resources should be the basis of any planning to pre-
land vulnerability and susceptibility to erosion pro- vent, or at least to limit, any negative effect that could
cesses (Bot and Benites, 2005). Combating this kind arise from anthropogenic land management. Further-
of degradation is highly desirable to ensure European more, additional pressure for soil conservation comes
arable soils sustainability, knowing that reversal of soil from the Water Framework Directive 2008 amendment
degradation trends through conversion to restorative which gives attention to the off-site impact of land
land use would enhance the SOM pool (Lal, 2009). degradation on downstream water quality in the con-
Warming temperatures are expected to speed-up text of sediment pollution.
natural processes of SOM decomposition, which should Therefore, current concerns to ensure a sustainable
increase the atmospheric CO2 level that contribute also use and management of European soil and water re-
to climate change (Davidson and Janssens, 2006). Cli- source generate an urgent need for obtaining reliable
mate change is projected either to increase or de- quantitative data on the extent and magnitude of soil
crease crop yields and soil quality in Mediterranean degradation. At the European level, as stated by EU
agroecosystems depending on site characteristics, crop Joint Research Centre, there is a serious gap of geo-
management and soil tillage practices. Results from referenced data available on SOM from systematic soil
different simulation models (e.g., Century and Epic) sampling programmes. Currently the most compre-
clearly indicated that management choices linked to hensive data set on SOM content of European soils is
crop intensification level and/or conservation tillage provided by the European Soil Database at a scale of
could affect the soil organic carbon balance more 1:1 000 000 (Jones et al., 2004, 2005).
strongly than climate change, suggesting the adoption To produce soil parameter redistribution map, one
of improved rotations, grassland conversion and no- needs a sufficient number of samples and a validated
tillage (Lugato and Berti, 2008; Álvaro-Fuentes and sampling strategy to ensure an accurate assessment of
232 A. MARCHETTI et al.

unsampled areas (Mabit et al., 2002). Also, soil (Fig. 1). This region is mainly hilly, crossed by rivers,
parameters are not distributed randomly in the land- and agricultural lands are predominant. The most eco-
scape; there is a spatial autocorrelation that can nomically important zone falls into the Controguerra
explain the spatial distribution of their magnitudes viticultural district, the DOCG (Denominazione di
(Mabit and Bernard, 2007). Geostatistics can be used Origine Controllata e Garantita, attesting the origin
to characterize the spatial behaviour and spatial dis- and the quality of a wine) area of Colline Teramane.
tribution of a parameter and to use this information The mean temperatures during winter and during sum-
to predict the value of this variable between sampled mer are 7.5 and 25 ◦ C, respectively, the mean annual
points and to minimize estimation error (Webster and rainfall is about 790 mm, and spring and autumn are
Oliver, 2001). For this reason, the estimation by geo- the rainy seasons. Temperate winter and dry summer
statistical techniques through moving average interpo- are typical of the Mediterranean climate.
lation such as kriging has been applied extensively in Considering the Soil Regions of Italy as defined by
soil science since the 80s (e.g. Burgess and Webster, the European Soil Bureau (ESB, 2000), this area is lo-
1980; Oliver and Webster, 1991; Goovaerts, 1997; Web- cated in the Soil Region 61.3, namely “Hills of central
ster, 2001; Kanevski and Maignan, 2004; Duffera et al., and southern Italy on Pliocene and Pleistocene ma-
2007). But still credible measurements of the SOM and rine deposits and Holocene alluvial sediments along the
its landscape dynamic at a range of spatial and tem- Adriatic Sea”. According to Costantini et al. (2004),
poral scales remain a challenge (Lal, 2009). the main soil units are: soils eroded and with reorga-
SOM distribution using geostatistical approaches nization of carbonates (Eutric and Calcaric Regosols,
has been mapped accurately at the field scale (e.g., Calcaric Cambisols and Haplic Calcisols), soils with
Mabit et al., 2008a; Mabit and Bernard, 2010) or clay accumulation (Haplic and Calcic Luvisols), soils
mapped using modelling and/or geographic informa- with vertic properties (Vertic Cambisols and Calcic
tion system (GIS) and remote sensing support at the Vertisols), and alluvial soils (Calcaric, Eutric and
country or continental scale (Jones et al., 2004, 2005; Gleyic Fluvisols). Severe and continuous soil erosion
Werner et al., 2009). However, reliable distribution of is one of the major causes for the low organic matter
SOM through geostatistical techniques at intermedi- content of many soils in these regions where the quality
ate scale providing policy makers with information at of agricultural soils is reducing.
a scale at which protection measures are likely to be A useful way to organize the knowledge about the
implemented and followed with efficient results, i.e., soils in a geographic area is the soil system, i.e., a re-
the watershed scale, are still scarce. curring group of soils that occupies specific landscape
The aims of this study conducted in central Italy positions as a result of the internal soil environment
were i) to determine the structure of spatial depen- produced by the interaction of stratigraphy, hydrology,
dence of different soil physical and chemical properties geomorphology and climate (Daniels et al., 1984). In
(i.e., soil textural parameters, SOM and C/N ratio), the area under investigation four main soil systems
ii) to evaluate these soil parameters distribution from are represented (Chiuchiarelli et al., 2006): A1, sandy
point-type data in non sampled locations by ordinary sediments close to the present sea shore, including the
kriging (OK), and iii) to map their spatial distribu- coastal zone; A2, ancient valley bottoms and terraces
tion using geostatistics and GIS facilities to define soil of the meso-Adriatic alluvial deposits, bordering the
quality status in identifying risk areas with depleted main rivers of the region; A3, high terraces of the meso-
SOM content, in order to address the regional plan- Adriatic Plio-Pleistocenic reliefs with sandy-gravelly
ning of agri-environmental measures under the Euro- substrate; and A4, meso-Adriatic Plio-Pleistocenic re-
pean Common Agricultural Policy. liefs with clay loam substrate. Within a soil system,
soil subsystems account for differences in physiogra-
MATERIALS AND METHODS
phy, lithology, morphometry, land use, vegetation and
Study area soil associations. In Fig. 1 a 1:250 000 soil subsystem
map of the area is reported (Chiuchiarelli et al., 2006),
The study area covering about 100 km2 is located
including eight different soil subsystems as described
in the northern part of the Abruzzo region, Italy, be-
in Table I.
tween the latitudes 42◦ 47 41 and 42◦ 53 38 N
and the longitudes 13◦ 46 06 and 13◦ 55 57 E; Soil analysis and data base building
it comprises four municipalities in Teramo Province:
Controguerra, Corropoli, Colonnella and Martinsicuro A set of 250 georeferenced samples were collected
SPATIAL DISTRIBUTION OF SOM IN CENTRAL ITALY 233

Fig. 1 Soil sampling locations according to soil subsystems of the study area (adapted from Chiuchiarelli et al., 2006).

by the Regional Agency for Agricultural Extension Ser- sampled locations were performed using geostatisti-
vices of Abruzzo Region (ARSSA) from the surface cal approach. Using two-dimensional space, the spatial
horizon (plough layer, about 50 cm depth) of agricul- structure of a variable can be visualized by maps. How-
tural soils (Fig. 1). ever, in order to produce spatial representation of
The physical and chemical analyses of the soil sam- dataset, we need to interpolate the values at unsam-
ples included the determination of particle size distri- pled locations. Various interpolation techniques exist
bution (Gee and Bauder, 1986), organic carbon content (e.g., inverse distance weighting, IDW, also termed in-
(OC) applying the modified Walkley-Black method verse distance to a power, triangulation with linear in-
(Nelson and Sommer, 1982), and total nitrogen content terpolation) which are “user-friendly”, but these exact
(N) using the Kjeldahl method (Bremner and Sparks, interpolators do not take in consideration the spatial
1996). According to Jackson (1958), the SOM content autocorrelation of data and thus oversimplify the re-
was evaluated from the organic carbon using the fol- ality (Isaaks and Srivastava, 1989; Robinson and Met-
lowing formula: ternicht, 2006; Mabit et al., 2008a).
Therefore OK, known as the best linear unbiased
SOM = OC × 1.724 (1) estimator (BLUE), was chosen as interpolation method
for each soil parameter to minimize the prediction er-
The SOM status of the sampled area was classified
ror variance. OK is by far the most common type of
into four different levels (i.e., very low, low, medium
kriging, consisting in a form of weighted averaging, and
and high) based on the USDA textural classes (see Ta-
is based on the concept of a variable z(x) that is both
ble II).
random and spatially autocorrelated (Heuvelink and
Data mapping optimization and validation using geo- Webster, 2001). The predictions are based on the fol-
statistical analysis lowing model:

Z(x) = μ + ε(x) (2)


After descriptive statistical analysis, the mode-
lling of spatial variability and the estimation at non- where μ is the constant stationary function (global
234 A. MARCHETTI et al.

TABLE I

Characteristics of the soil subsystems in Teramo Province, Italy (Chiuchiarelli et al., 2006)

Map unit Physiography and lithology Morphometry Land use and vegetation
A1a Coastal area with incoherent Altitude: 0–20 m a.s.l. (99%); Artificial surfaces: 59%; Agricultural are-
substrates, sandy and sandy- Slope: 85% from 0% to 5% as: 28% (arable crops 16%, permanent
gravelly crops 4%, heterogeneous areas 7%); Sea
shores, dunes and sand plains: 3%
A2a Valley bottoms of the main ri- Altitude: 0–200 m a.s.l. (98%); Artificial surfaces: 17%; Agricultural are-
vers, and alluvial terraces; sub- Slope: 88% from 0% to 5% as: 64% (arable crops 38%, permanent
strates of gravelly, sandy and crops 12%, heterogeneous areas 7%); Ri-
clay loam alluvial sediments parian formations: 13%
A2c Recent river terraces higher than Altitude: 0–200 m a.s.l. (91%); Artificial surfaces: 28%; Agricultural are-
the present valley bottom; sub- Slope: 80% from 0% to 5% as: 64% (arable crops 40%, permanent
strates of gravelly-sandy and crops 13%, heterogeneous areas 14%)
clay loam sediments, interca-
lated or underlying clay loam
colluvial sediments
A2d Ancient river terraces higher Altitude: 20–300 m a.s.l. (82%); Artificial surfaces: 10%; Agricultural are-
than the present valley bottom; Slope: 83% from 0% to 13% as: 83% (arable crops 36% and olive
substrates of gravelly-sandy se- groves 22%)
diments
A3b Top residual surfaces, narrow Altitude: 100–400 m a.s.l. (95%); Artificial surfaces: 16%; Agricultural are-
and stretched, reduced to sub- Slope: 67% from 0% to 13% as: 85% (vineyards 25%, olive groves 20%,
level crests; substrates of grave- complex cultivation patterns 17%, arable
lly-sandy sediments crops 16%)
A4b Linear slopes and secondarily Altitude: 50–300 m a.s.l. (84%); Agricultural areas: 79% (arable crops
unstable slopes with gully ero- Slope: 79% from 5% to 35% 29%, olive groves 22%, vineyards 11%)
sion; substrates of marine clay
loam sediments
A4d Linear slopes and secondarily Altitude: 100–300 m a.s.l. Artificial surfaces: 6%; Agricultural
unstable slopes with gully ero- (85%); areas: 78% (arable crops 40%, olive
sion; substrates of marine Slope: 76% from 5% to 35% groves 12%, annual crops associated with
coarse gravel sediments permanent crops 12%)
A4e Linear slopes and eroded sur- Altitude: 50–300 m a.s.l. (97%); Artificial surfaces: 5%; Agricultural areas:
faces of terraces with slightly Slope: 98% from 0% to 35% 91% (arable crops 47%, annual crops asso-
undulating morphology ciated with permanent crops 19%)

TABLE II

Soil organic matter (SOM) evaluation according to USDA textural classes (SILPA, 1999)

USDA texture class SOM content

Very low Low Medium High


%
Sand loamy, sand, sandy loam < 0.8 0.8–1.4 1.5–2.0 > 2.0
Loam, sandy clay, sandy clay loam, silty loam, silt < 1.0 1.0–1.8 1.9–2.5 > 2.5
Clay, clay loam, silty clay, silty clay loam < 1.2 1.2–2.2 2.3– 3.0 > 3.0

n

mean) and ε(x) is the spatially correlated stochastic
Ẑ(x0 ) = λi Z(xi ) (3)
part of variation. Estimation of Z at an unsampled i=1
point x0 , Ẑ(x0 ), is made by a weighted average of the
data: where λi is the kriging weight assigned to sampling site
SPATIAL DISTRIBUTION OF SOM IN CENTRAL ITALY 235

Z(xi ). The weights are allocated to the sample data riogram model are used to assign optimal weights for
within the neighbourhood of the point to be estimated interpolation by OK.
in such a way to minimize the estimation variance. Ac- An important property of admissible variogram
cording to Webster and Oliver (2001), to ensure that models is that any linear combination of admissible
the estimate is unbiased, the weights are made to sum variogram models with positive coefficients is also an
to 1: admissible model (nested structure). Nested structures
n indicate the presence of processes operating at different

λi = 1 (4) scales (Armstrong, 1998). Best fitted variogram models
i=1 can be obtained by minimizing the indicative goodness
of fit (IGF) criterion, which is a standardized measure
and the expected prediction error is
of the difference between observed and predicted values
based on a least squares estimator (Pannatier, 1996):
E[Ẑ(x0 ) − Z(x0 )] = 0 (5)

D(k)  γ(i) − γ̂(i) 2


N n(k)
To estimate the weights in an objective way, so that 1   P (i)
IGF = · · (7)
n(k) σ2
k=1 i=0 
they reflect the true spatial autocorrelation structure, N d(i)
P (j)
the so-called semivariances (differences between neigh-
j=0
bouring values) are used:
where N is the number of directional variograms, n(k)
1 is the number of lags relative to variogram k, D(k) is
γ(h) = E[Z(xi ) − Z(xi + h)]2 (6)
2 the maximum distance relative to variogram k, P (i) is
where Z(xi ) is the value of the target variable at sam- the number of pairs for lag i of variogram k, d(i) is
pled location i and Z(xi + h) is the value of the neigh- the mean pair distance for lag i of variogram k, γ(i)
bour at distance h. A set of n point observations yields is the experimental measure of spatial continuity for
n(n − 1)/2 pairs for which a semivariance can be cal- lag i, γ̂(i) is the modelled measure of spatial continui-
culated. ty for d(i), and σ 2 is the variance of the data for the
Once plotting all the semivariances versus their dis- variogram.
tances a variogram cloud is produced, and averaging Based on optimized variograms, contour map of
the values for a standard distance (lag) an experimen- soil texture (USDA classification) and contour map
tal variogram is obtained. In the variogram the spa- of SOM related to soil texture were generated using
tial dependence of the data is typically expressed by GIS facilities. After a preliminary estimation of sand,
a monotonic increase from the origin with increasing clay and SOM contents, these vectorial estimated maps
lag distance; the variogram is hence a function of the were converted into raster maps, with 20 m × 20 m
spatial autocorrelation of the sample. The semivari- cells; from these cells the values referred to the cen-
ances are typically smaller at shorter distance, and may tre of each cell (also termed centroids) were extracted.
reach, or asymptotically approach, an upper bound USDA texture classes were determined for each cen-
(sill) at a finite distance (range), beyond which there is troid by means of a dedicated PC program created for
no longer spatial autocorrelation (Heuvelink and Web- this purpose (in FORTRAN 90), and the SOM levels
ster, 2001). In fact, the values of a target variable are related to soil texture were evaluated. The final maps
more similar at a shorter distance, up to a certain dis- were derived from these outputs, and the map of C/N
tance where the differences between the pairs are more ratio—proxy indicator of SOM humification rate—was
or less equal to the global variance (Hengl, 2007). The also generated.
nugget variance, a positive intercept on the ordinate, Cross-validation procedure was conducted to eva-
is an uncorrelated component and indicator of short luate the accuracy of the model through three statisti-
distance variation which includes measurement error, cal measurements of the prediction error: mean error
sampling error, inter-sample error and unexplained and (ME), root mean square error (RMSE), and root mean
inherent variability. square standardized error (RMSSE) by
A mathematical model is then fitted to the experi- N
1  
mental variogram (Goovaerts, 1997) to minimize the ME = Ẑ(xi ) − Z(xi ) (8)
N i=1
variance of the errors, and the parameters of the va-
236 A. MARCHETTI et al.


 N suitable to normalize data distribution with coefficient
  2
 Ẑ(xi ) − Z(xi )
 of skewness higher than 1. Despite this transforma-
i=1
RMSE = (9) tion, the data distribution still did not reach a satis-
N factory degree of normality. Then, the analysis of the
 frequencies histograms of the different parameters al-
 N

2 lowed identifying outliers, representing less than 5%
 [Ẑ(xi ) − Z(xi )] σ(xi )
 of the total dataset. Their exclusion, considering also
i=1
RMSSE = (10) their spatial distribution, improved significantly the
N
normality degree of the soil parameters frequencies dis-
where σ(xi ) is the prediction standard error in location tribution, and then geostatistical approaches were ap-
xi . plied for their interpolation (Table IV).
Statistical analyses, geostatistical analyses and GIS The most common admissible variogram models
mapping were carried out with the following softwares: include spherical, exponential, Gaussian and linear
R 2.13.1 (R Development Core Team, 2011), VarioWin models:
2.21 (Pannatier, 1995), and ArcGIS 9.2 with the Geo-
statistical Analyst extension. For a given variable all Spherical model = nugget + sill[1.5(lag/range) −
kriged maps were kept on the same scale in order to 0.5(lag/range)3 ] (11)
allow easier comparisons.
Exponential model = nugget +
RESULTS AND DISCUSSION sill[1 − exp(−lag/range)] (12)

Descriptive statistical analysis of the soil parame- Gaussian model = nugget +


ters is presented in Table III that provides information sill[1 − exp(lag2 /range2 )] (13)
about the central tendency of the investigated varia- Linear model = nugget + sill × lag (14)
bles. Geostatistical techniques require a normal dis-
tribution of the parameter values under investigation Experimental variograms of the following soil pa-
(Webster, 2001). The data distribution of the different rameters: % sand, % clay, SOM and C/N were cal-
parameters tested is not normal, with a coefficient of culated using a lag distance of 575 m (Fig. 2). All
skewness far from zero and a kurtosis coefficient higher optimized variograms were omnidirectional (isotropic)
than 3 (Table III). with no directional dependence, meaning that the va-
A logarithmic transformation of the data is usually lues of the soil parameters vary similarly in all dire-

TABLE III

Descriptive statistics of the soil parameters


Variable No. of samples Minimum Maximum Mean Standard deviation Variance Skewness Kurtosis
Sand (%) 250 0.0 66.8 20.5 11.3 128.19 1.045 5.055
Clay (%) 250 3.7 45.3 31.6 6.8 47.06 −0.912 4.087
Silt (%) 250 14.2 63.3 48.0 7.2 51.14 −0.716 4.886
SOM (%) 250 0.07 5.96 1.78 0.98 0.96 1.770 7.483
C/N 250 0.39 41.57 9.50 4.47 19.97 2.147 14.724

TABLE IV

Descriptive statistics without outliers


Variable No. of samples Minimum Maximum Mean Standard deviation Variance Skewness Kurtosis
Sand (%) 241 0.0 46.2 19.2 9.2 85.32 0.287 3.207
Clay (%) 241 15.0 45.3 32.1 6.1 37.25 0.546 3.003
Silt (%) 241 28.5 63.3 48.2 6.7 44.26 −0.247 3.041
SOM (%) 238 0.07 3.57 1.62 0.67 0.45 0.379 3.051
C/N 238 0.37 20.49 9.10 3.51 12.35 0.251 3.360
SPATIAL DISTRIBUTION OF SOM IN CENTRAL ITALY 237

ctions, and that the semivariance depends only on the In Table V the parameters of the chosen variogram
distance between the sampling points (Burgos et al., models are reported with associated IGF values, know-
2006). Then, the models of spatial variability were ing that an IGF close to zero indicates a good fit of
fitted to the experimental variograms by minimizing the model in agreement with the hypothesis of a well
the sum of squares between the experimental values characterized and constant spatial structure.
and those of the model (Lark, 2000). The best vario- The preliminary estimated maps for sand, clay and
gram models and their parameters were determined to SOM, and that of the C/N ratio were generated by
validate the modelling of the spatial autocorrelation OK. The cross validation and associated prediction er-
through the variogram parameters optimization. rors for each map are reported in Table VI. For an
For C/N only one spatial structure was identified, unbiased prediction (centred on the true values) the
while for sand, clay and SOM two different nested ME should be near zero. The RMSE quantified the
models were fitted: the first one with a short range amount of difference in the model simulation deviat-
and the second one with a long range. The best fitted ing from the observation, presented in the unit of the
models are almost all spherical, except for clay content. observation. If the model accurately describes the data,

Fig. 2 Experimental variograms of sand, clay, and soil organic matter (SOM) contents and C/N ratio.

TABLE V

Parameters of the optimized variogram models


Variable Nugget Model 1 Sill Range Model 2 Sill Range IGFa)
m m
Sand (%) 7.584 8 Spherical 55.73 646 Spherical 18.17 5 684 0.000 6
Clay (%) 9.876 2 Spherical 18.62 646 Gaussian 8.36 5 684 0.002 9
SOM (%) 0.13 Spherical 0.275 601 Spherical 0.055 5 510 0.000 8
C/N 5.329 6 Spherical 8.968 3 075 - - - 0.001 6
a)
Indicative goodness of fit.
238 A. MARCHETTI et al.

TABLE VI

Cross validation and associated prediction errorsa)

Variable ME RMSE RMSSE


Sand (%) −0.0554 8.676 1.074
Clay (%) 0.0532 5.719 1.060
SOM (%) 0.0011 0.666 1.026
C/N 0.0093 2.838 0.994
a)
ME = mean error; RMSE = root mean square error;
RMSSE = root mean square standardized error.

the RMSE should be approximately equal to the stan-


dard deviation. This condition is verified for all the
considered variables. The RMSSE should be close to
1 for a good prediction, and also this condition is fully Fig. 4 Map of soil organic matter related to soil texture
verified. interpolated by the ordinary kriging (OK) method.
Fig. 3 shows the soil texture map of the investi-
gated area, which is derived from the preliminary sand the total surface. The few zones with high values (range
and clay maps. The prevailing texture class is the silty 2.50%–2.97%, average 2.62%) are located mainly in
clay loam occupying 58% of the total area, essentially the northern part of the studied area representing only
located inland; and 35% of the soils have a clay loam 0.1% of the total surface. The low SOM contents are
texture and are mainly located along the coastline and related to soil morphology, especially on hills where
in the southern part of the surveyed area. The re- erosive processes are exacerbated by the intensive agri-
maining soils are principally loam (about 7% of the cultural practices adopted in these areas (Fig. 5).
total area).

Fig. 5 Map of soil organic matter related to soil texture


draped over a hillshade of the area.
Fig. 3 Soil texture map of the investigated area.
The map presented in Fig. 6 shows the distribu-
Fig. 4 illustrates the spatial pattern of SOM content tion of the C/N soil ratio, that can be considered a
in relation to soil texture through four class levels. proxy indicator of SOM humification rate and soil N
About 87% of the total surface has a low SOM con- mineralization/immobilization potentials, differentia-
tent with values ranging from 1.00% to 2.19% for an ted through three value classes based on USDA-NRCS
average of 1.62%. Small surfaces, representing 6.4% of (1995) and Thomsen et al. (2008): C/N < 9, 9 ≤ C/N
the total area, have a very low content (range 0.44%– ≤ 12, and C/N > 12. When the ratio is close to 10,
1.19%, average 1.02%), while medium values (range SOM mineralization and immobilization processes are
1.81%–2.87%, average 2.08%) occupy around 6.5% of in equilibrium; soils with a value between 9 and 12
SPATIAL DISTRIBUTION OF SOM IN CENTRAL ITALY 239

have optimum conditions, with a stable and well hu- evaluate the impact of natural resource management
mified organic matter (USDA-NRCS, 1995). The ratio on the environment, and to perform integrated analy-
decreases as SOM mineralization increases, and vice ses of environmental policy issues.
versa. Estimated map shows that about 50% of the sur- This study defines the existing SOM status in an
veyed area has an optimum C/N, while 42% of the area area of central Italy and addresses a clear warning on
has a C/N ratio lower than 9, indicating that oxidation the SOM management in this region. The results of
processes prevail involving an accelerated decomposi- this investigation highlight the vulnerability of these
tion of SOM. In the remaining 8% of the surveyed area, arable lands to soil degradation processes accelerated
where C/N ratio is higher than 12, soil nitrogen is in- by a rapid mineralization of organic matter (dry con-
sufficient to ensure a good and acceptable humification ditions and high temperature) that need appropriate
process. management to ensure agricultural land sustainability.
In Italian agricultural land the average top soil or-
ganic carbon content is around 12 g kg−1 equivalent
to about 2% of SOM (Jones et al., 2004). The situ-
ation in the study area is then critical with a mean
SOM content of about 1.6%, 20% less than the coun-
try average, regardless of the prevalent land use or the
soil type. The establishment of baseline reference level
and accurate picture of SOM distribution will allow to
monitor future trends of SOM (i.e., re-sampling con-
cept) and also to assess the potential effectiveness of
soil conservation measures to enhance carbon in soil
that should be adopted to develop a more sustainable
soil management focusing on stabilizing and increasing
SOM content.
Future applied researches should focus on carbon
Fig. 6 Spatial representation of C/N ratio value in soils.
cycle (stability and turnover) and especially on carbon
The 2003 mid-term review of the Common Agri- sequestration as fighting tools to combat soil degra-
cultural Policy (CAP) and the resulting new Coun- dation. The relationships between land use and the
cil Regulation No. 1782/2003 explicitly specify obliga- various SOM pool dynamics can also be investigated,
tions for EU Member States to ensure good agricul- since the most important processes responsible for low-
tural and environmental condition of land. The pro- ering carbon levels in agricultural soils are erosion and
posed European Framework Directive on soil protec- the effect of tillage operations, coupled with low input
tion, with one of the priorities being to make a list of agricultural residues (Nieto et al., 2010). Then, the
of the soils at risk in Europe (CEC, 2006b), identi- next logical step of our research activities could be the
fies eight main soil threats considering SOM decline in soil erosion magnitude assessment with conventional
southern Europe one of the most serious soil degrada- and/or radioactive tracers such as fallout radionuclides
tion processes. Both the CAP and the European Soil (Mabit et al., 2008b) which could reinforce the infor-
Thematic Strategy to combat land degradation request mation gained on soil quality degradation in the area
an urgent need for obtaining reliable soil degradation as represented by the SOM distribution described in
data and related agro-environmental impact in a range the current contribution.
of European landscapes, to efficiently mitigate/control As soil conservation and management schemes
land degradation and thereby underpin sustainable must be based on robust scientific evidence, applica-
soil conservation strategies and policies supported by tion of geostatistics to map targeted soil parameters
a scientifically-sound management of land and water. such as SOM represents a valid tool allowing the as-
Therefore, current European research activities on na- sessment of soil degradation in a cost-effective and little
tural resources are firmly set in the context of renew- time-consuming way. Such maps provide a relatively
able and non-renewable resources sustainable develop- cheap information, that can be easily updated. This
ment with emphasis on the setting up of approaches can influence and support local land-use planners, can
and methods to assess the quality of ecosystems, to help farmers and hence provide a validated approach
240 A. MARCHETTI et al.

for science-based policy options to EU policy makers to ter (compost, manure, sewage sludge, etc.).
address key challenges facing European soil resources. Assessing the spatial variability and mapping of soil
quality indicators such as SOM content connected with
CONCLUSIONS soil textural classes and C/N ratio in agrosystems, and
then, localizing degraded areas are a first step towards
SOM decrease associated with changes in land use,
an efficient resource management policy to target a
soil cultivation and climate change are threatening the
successful implementation of agricultural soil conser-
health of European soils. Within the framework of the
vation practices.
European Directive for soil protection, the spatializa-
tion of SOM content can provide decision makers with
ACKNOWLEDGEMENTS
a useful and proper reference tool to identify degraded
areas with depleted fertility potential and then facili- The authors would like to acknowledge Dr. San-
tate and optimize the planning of agro-environmental tucci, S., Dr. Chiuchiarelli, I. and the Laboratorio
measures. Agrochimico ARSSA of Avezzano, Italy, for their help
As traditional statistical methods are unable to and support in the soil samples collection and physi-
treat this aspect, geostatistics (i.e., ordinary kriging) cal and chemical analyses during the implementation
was used to establish the spatial structure of SOM, soil of the project.
texture and C/N soil ratio and to create maps of SOM
distribution to identify degraded agricultural lands in
central Italy. REFERENCES
The different variables analyzed showed a strong Álvaro-Fuentes, J. and Paustian, K. 2011. Potential soil car-
spatial structure. The kriging-interpolated maps gene- bon sequestration in a semiarid Mediterranean agroe-
rated combining soil texture and SOM, and C/N ratio cosystem under climate change: Quantifying manage-
provide a meaningful spatial representation of SOM ment and climate effects. Plant Soil. 338: 261–272.
and a real decisional tool that establish a real assess- Armstrong, M. 1998. Basic Linear Geostatistics. Springer,
ment of the soil quality status in the Teramo Province, New York.
Boardman, J. and Poesen, J. (eds.). 2006. Soil Erosion in
as a simple SOM map does not provide enough accu-
Europe. Wiley, Chichester, UK.
rate information. In this study, the spatial estimation
Bot, A. and Benites, J. 2005. The Importance of Soil Or-
of SOM has shown a low content in relation to soil tex- ganic Matter: Key to Drought-Resistant Soil and Sus-
ture in 87% of the area coupled with an optimal C/N tained Food Production. Food and Agriculture Orga-
ratio in only half of the surveyed surface. nization of the United Nations, Rome.
Furthermore, the results of this research highlight Bremner, J. M. and Sparks, D. L. 1996. Nitrogen-Total. In
the potential of variography and geostatistics in combi- Methods of Soil Analysis. Part III. Chemical Methods.
nation with GIS to provide improved distribution pat- Soil Science Society of America Inc., Madison, USA.
terns and maps of various soil quality indicators. pp. 1085–1121.
In central Italy low SOM level is the result of ero- Burgess, T. M. and Webster, R. 1980. Optimal interpola-
tion and isarithmic mapping of soil properties: I. The
sive processes under Mediterranean climate on hills
semi-variogram and punctual kriging. J. Soil Sci. 31:
coupled with intensive and/or inappropriate agricul- 315–331.
tural practices. Based on the results of this study, Burgos, P., Madejón, E., Pérez-de-Mora, A. and Cabrera,
proper agronomical and environmental planning such F. 2006. Spatial variability of the chemical characteri-
as soil conservation strategy is highly required in cen- stics of a trace-element-contaminated soil before and
tral Italy to restore and increase the SOM content in after remediation. Geoderma. 130: 157–175.
agricultural soils, to combat soil erosion and to main- Chiuchiarelli, I., Paolanti, M., Rivieccio, R. and Santucci,
tain soil ecological and socio-economical functions and S. 2006. Soils and Landscapes of Abruzzo—Soil Map
productivity. Replenishing of SOM could be reached in of Abruzzo Region (in Italian). Agenzia Regionale per
I Servizi di Sviluppo Agricolo, Regione Abruzzo. Avez-
degraded areas (i.e., low SOM content and C/N ratio)
zano, Italy.
by adopting conservative practices such as conserva-
Commission of the European Communities (CEC). 2006a.
tion tillage or no-tillage (e.g., direct seeding), impro- Proposal for a Directive of the European Parlia-
ving land use rotations with forage crops, returning ment and of the Council Establishing a Frame-
crop residues to soil, growing green manure crops, and work for the Protection of Soil and Amending Dire-
supplying the soil with proper exogenous organic mat- ctive 2004/35/EC. Available online at http://eur-lex.
SPATIAL DISTRIBUTION OF SOM IN CENTRAL ITALY 241

europa.eu/LexUriServ/LexUriServ.do?uri=CELEX: Heuvelink, G. B. M. and Webster, R. 2001. Modelling soil


52006PC0232:EN:NOT (verified on October 28, 2011). variation: past, present and future. Geoderma. 100:
Commission of the European Communities (CEC). 2006b. 269–301.
Thematic Strategy for Soil Protection. Communica- Isaaks, E. H. and Srivastava, R. M. 1989. An Introduction
tion from the Commission to the Council, the Euro- to Applied Geostatistics. Oxford University Press, New
pean Parliament, the European Economic and Social York.
Committee, and the Committees of the Region—COM Jackson, M. L. 1958. Soil Chemical Analysis. Prentice-
(2006)231 Final. Available online at http://ec.europa. Hall., Inc., Englewood Cliffs, N. J., USA.
eu/environment/soil/pdf/com 2006 0231 en.pdf (veri- Jones, R. J. A., Hiederer, R., Rusco, E. and Montanarella,
fied on September 26, 2011). L. 2005. Estimating organic carbon in the soils of Eu-
Costantini, E. A. C., Urbano, F. and L’Abate, G. 2004. rope for policy support. Eur. J. Soil Sci. 56: 655–671.
Soil Regions of Italy. Available online at http://www. Jones, R. J. A., Hiederer, R., Rusco, E., Loveland, P. J. and
soilmaps.it/en/downloads.html (verified on September Montanarella, L. 2004. The Map of Organic Carbon in
26, 2011). Topsoils in Europe. Version 1.2. European Soil Bureau
Daniels, R. B., Kleiss, H. J., Buol, S. W., Byrd, H. J. and Research Report No. 17. Office for Official Publica-
Phillips, J. A. 1984. Soil Systems in North Carolina. tions of the European Communities, Luxembourg.
North Carolina Agricultural Research Service, North Kanevski, M. and Maignan, M. (eds.). 2004. Analysis
Carolina State University, Raleigh, NC, USA. and Modelling of Spatial Environmental Data. EPFL
Davidson, E. A. and Janssens, I. A. 2006. Temperature Press, Lausanne.
sensitivity of soil carbon decomposition and feedbacks Lal, R. 2003. Soil erosion and the global carbon budget.
to climate change. Nature. 440: 165–173. Environ. Int. 29: 437–450.
Duffera, M., White, J. G. and Weisz, R. 2007. Spatial varia- Lal, R. 2007. Farming carbon. Soil Till. Res. 96: 1–5.
bility of Southeastern U.S. Coastal Plain soil physical Lal, R. 2009. Challenges and opportunities in soil organic
properties: Implications for site-specific management. matter research. Eur. J. Soil Sci. 60: 158–169.
Geoderma. 137: 327–339. Lark, R. M. 2000. A comparison of some robust estimators
European Environmental Agency (EEA). 2000. Down to of the variogram for use in soil survey. Eur. J. Soil Sci.
Earth: Soil Degradation and Sustainable Development 51: 137–157.
in Europe: A Challenge for the 21st Century. Environ- Lugato, E. and Berti, A. 2008. Potential carbon sequestra-
mental Issue Report No. 16. Available online at http: tion in a cultivated soil under different climate change
//www.eea.europa.eu/publications/Environmental iss- scenarios: a modeling approach for evaluating promi-
ue series 16 (verified on September 26, 2011). sing management practices in north-east Italy. Agr.
European Soil Bureau (ESB). 2000. Georeferenced Soil Ecosyst. Environ. 128: 97–103.
Database for Europe. Manual of Procedures. Version Lützow, M. V., Kögel-Knabner, I., Ekschmitt, K., Matzner,
1.1. European Soil Bureau Research Report No.5. Of- E., Guggenberger, G., Marschner, B. and Flessa, H.
fice for Official Publications of the European Commu- 2006. Stabilization of organic matter in temperate soils:
nities, Luxemburg. mechanisms and their relevance under different soil
Farina, R., Seddaiu, G., Orsini, R., Steglich, E., Roggero, conditions—a review. Eur. J. Soil Sci. 57: 426–445.
P. P. and Francaviglia, R. 2011. Soil carbon dynamics Mabit, L., Benmansour, M. and Walling, D. E. 2008b.
and crop productivity as influenced by climate change Comparative advantages and limitations of Fallout ra-
in a rainfed cereal system under contrasting tillage u- dionuclides 137 Cs, 210 Pb and 7 Be for assessing soil ero-
sing EPIC. Soil Till. Res. 112: 36–46. sion and sedimentation. J. Environ. Radioact. 99:
Fenton, T. E., Kazemi, M. and Lauterbach-Barrett, M. A. 1799–1807.
2005. Erosional impact on organic matter content and Mabit, L. and Bernard, C. 1998. Relationship between soil
137
productivity of selected Iowa soils. Soil Till. Res. 81: Cs inventories and chemical properties in a small in-
163–171. tensively cropped watershed. C. R. Acad. Sci.-Series
Gee, G. W. and Bauder, J. W. 1986. Particle-size analysis. IIA-Earth Planet. Sci. 327: 527–532.
In Methods of Soil Analysis: Part I. Physical and Mine- Mabit, L. and Bernard, C. 2007. Assessment of spatial dis-
ralogical Methods. 2nd Edition. Soil Science Society tribution of Fallout radionuclides through geostatistics
of America, Madison, USA. pp. 383–411. concept. J. Environ. Radioact. 97: 206–219.
Goovaerts, P. 1997. Geostatistics for Natural Resource Mabit, L. and Bernard, C. 2010. Spatial distribution and
Evaluation. Oxford University Press, New York. content of soil organic matter in an agricultural field in
Hengl, T. 2007. A Practical Guide to Geostatistical Map- Eastern Canada, as estimated from geostatistical tools.
ping of Environmental Variables. Available online at Earth Surf. Proc. Land. 35: 278–283.
http://bookshop.europa.eu/uri?target=EUB:NOTICE: Mabit, L., Bernard, C. and Laverdière, M. R. 2002. Influ-
LBNA22904:EN:HTML (verified on September 30, ence of the sampling strategy on soil loss assessments
2010). from 137 Cs measurements. In Rubio, J. L., Morgan, R.
242 A. MARCHETTI et al.

P. C., Asins, S. and Andreu, V. (eds.) Man and Soil vironment for Statistical Computing. R Foundation
at the Third Millenium. Geoforma Ediciones, Logroño, for Statistical Computing, Vienna, Austria. Availa-
Spain. pp. 2083–2090. ble online at http://www.r-project.org (verified on 26
Mabit, L., Bernard, C., Makhlouf, M. and Laverdière, M. September 2011).
R. 2008a. Spatial variability of erosion and soil or- Ritchie, J. C., McCarty, G. W., Venteris, E. R. and Kas-
ganic matter content estimated from 137 Cs measure- par, T. C. 2005. Using soil redistribution to under-
ments and geostatistics. Geoderma. 145: 245–251. stand soil organic carbon redistribution and budgets.
Montanarella, L. 2010. Moving ahead from assessments to In Horowitz, A. J. and Walling, D. E. (eds.) Sediment
action: could we win the struggle with soil degradation Budgets 2. International Association of Hydrological
in Europe? Chapter 2. In Zdruli, P., Pagliai, M., Ka- Sciences Publication 292. IAHS Press, Wallingford,
pur, S. and Faz Cano, F. (eds.) Land Degradation and UK. pp. 3–8.
Desertification: Assessment, Mitigation and Remedia- Robinson, T. P. and Metternicht, G. 2006. Testing the
tion. Springer-Verlag, Berlin. pp. 15–23. perfomance of spatial interpolation techniques for map-
Nelson, D. W. and Sommers, L. E. 1982. Total carbon, ping soil properties. Compu. Electron. Agr. 50: 97–108.
organic carbon, and organic matter. In Page, A. L., Società Italiana dei Laboratori Pubblici di Agrochimica
Miller, R. H. and Keeney, D. R. (eds.) Methods of (SILPA). 1999. From Soil Analysis to the Fertilization
Soil Analysis. Part 2. Chemical and Microbiological Advice (in Italian). ASSAM, Agenzia Servizi Settore
Properties. Soil Science Society of American, Madison, Agroalimentare delle Marche, Regione Marche, Jesi,
WI. pp 539–579. Italy.
Nieto, O. M., Castro, J., Fernández, E. and Smith, P. 2010. Thomsen, I. K., Petersen, B. M., Bruun, S., Jensen, L. S.
Simulation of soil organic carbon stocks in a Mediter- and Christensen, B. T. 2008. Estimating soil C loss
ranean olive grove under different soil-management sys- potentials from the C to N ratio. Soil Biol. Biochem.
tems using the RothC model. Soil Use Manage. 26: 40: 849–852.
118–125. United States Department of Agriculture-Natural Re-
Oliver, M. A. and Webster, R. 1991. How geostatistics can sources Conservation Service (USDA-NRCS). 1995.
help you. Soil Use Manage. 7: 206–217. Soil Survey Laboratory Information Manual. Soil Sur-
Pannatier, Y. 1995. Software VarioWin 2.2. Institute of Mi- vey Investigations Report No. 45. National Soil Survey
neralogy, University of Lausanne, Switzerland. Avai- Center, Soil Survey Laboratory, Lincoln.
lable online at http://www-sst.unil.ch/research/va- Verity, G. E. and Anderson, D. W. 1990. Soil erosion ef-
riowin/index.html (verified on September 21, 2010). fects on soil quality and yield. Can. J. Soil Sci. 70:
Pannatier Y. 1996. Variowin, Software for Spatial Data 471–484.
Analysis in 2D. Springer Verlag, Berlin. Webster, R. 2001. Statistics to support soil research and
Pimentel, D. 2006. Soil erosion: A food and environmental their presentation. Eur. J. Soil Sci. 52: 331–340.
threat. Environ. Dev. Sustain. 8: 119–137. Webster, R. and Oliver, M. A. 2001. Geostatistics for En-
Polyakov, V. O. and Lal, R. 2008. Soil organic matter vironmental Scientists. Wiley & Sons Ltd., Chichester.
and CO2 emission as affected by water erosion on field Werner, L., Bahn, M. and Heinemeyer, A. 2009. Soil
runoff plots. Geoderma. 143: 216–222. Carbon Dynamics: An Integrated Methodology. Cam-
R Development Core Team. 2011. R: A Language and En- bridge University Press, UK.

You might also like