Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Soil & Tillage Research 191 (2019) 1–10

Contents lists available at ScienceDirect

Soil & Tillage Research


journal homepage: www.elsevier.com/locate/still

Delineation of management zones in a peach orchard using multivariate and T


geostatistical analyses

Henrique Oldonia, Viviane Santos Silva Terrab, Luís Carlos Timmc, , Carlos Reisser Júniord,
Alex Becker Monteiroe
a
Department of Rural Engineering, College of Agricultural Sciences, São Paulo State University, Rua José Barbosa de Barros, n. 1780, Portaria II, CEP: 18610-307,
Botucatu, São Paulo, Brazil
b
Center of Technological Development, Water Resources Engineering, Federal University of Pelotas, Campus Porto, Rua Gomes Carneiro, n. 01, CEP: 96010-610, Pelotas,
Rio Grande do Sul, Brazil
c
Department of Rural Engineering, Faculty of Agronomy, Federal University of Pelotas, Campus Universitário s/n, CEP: 96010-900, Capão do Leão, Rio Grande do Sul,
Brazil
d
Embrapa – Brazilian Agricultural Research Corporation, BR392, Km 78, CEP: 96001-970, Pelotas, Rio Grande do Sul, Brazil
e
Faculty of Agronomy, Federal University of Pelotas, Campus Universitário s/n, CEP: 96010-900, Capão do Leão, Rio Grande do Sul, Brazil

A R T I C LE I N FO A B S T R A C T

Keywords: The soil nutritional management in peach orchards in Southern Brazil is still based on conventional and uniform
Prunus persica recommendations, without taking into consideration the spatial variability of the physical environment. This
Soil and tree attributes might result in over-application of fertilizers in areas with high nutrient levels and under-application in areas
Spatial variability with low nutrient levels, which will cause high heterogeneity in the peach production and fruit quality resulting
in a low profit of farmers. The objectives of this work were to characterize the spatial variability of soil and tree
attributes, in a peach orchard using classical statistics and geostatistical tools as well as to delineate potential
management zones based on the soil and tree attributes using the principal component and fuzzy c-means
clustering analyses. A 1.8 ha peach orchard, in Morro Redondo, State of Rio Grande do Sul, Brazil, was selected
as the study area and a grid of 101 trees was established. Close to each of these trees, soil samples were collected
in the 0−0.10 m and 0.10−0.20 m layers to determine sand, silt and clay fractions, soil bulk density, volumetric
water content, pH in water, organic matter, available phosphorus as well as exchangeable potassium, magne-
sium and calcium contents. During the years of 2010, 2011 and 2012, peach yield, number of fruits per tree,
average fruit weight per individual tree, flesh firmness and total soluble solids content per tree were determined.
All data sets were submitted to descriptive statistics and the Kolmogorov–Smirnov test as well as to geostatistical
analysis. Principal component and fuzzy c-means clustering analyses were used to delineate the management
zones based on the optimal number of clusters identified employing the fuzziness performance index and the
modified participation entropy. Results showed high spatial variability of yield and quality of peaches. It was
found that the optimal number of management zones for the peach orchard was two and most of the soil and tree
attributes presented statistical differences in each defined management zone. The defined management zones
based on soil and tree attributes open the opportunity to farmers for site-specific management in the peach
orchard, aiming at precision agriculture and directing soil sampling design as well.

1. Introduction among the stone fruits, with a global production of 22,795,855 tons in
2014, in a cultivated area of 1,494,836 ha (FAOstat, 2017), evidencing
Peach crop [Prunus persica (L.) Batsch], originated in China and in their worldwide socio-economic importance. China had the greatest
temperate climate, is nowadays spread all over the world and is re- production in 2014, followed by Spain and Italy. Brazil was the thir-
cognized for producing juicy fruit, of pleasant taste and high nutritional teenth country in this ranking (FAOstat, 2017).
value. Peach and nectarine represent the fourth most important crops Fruit production has become one of the agricultural activities with


Corresponding author.
E-mail addresses: henriqueoldoni@gmail.com (H. Oldoni), vssterra10@gmail.com (V.S. Silva Terra), lctimm@ufpel.edu.br, lcartimm@yahoo.com.br (L.C. Timm),
carlos.reisser@embrapa.br (C.R. Júnior), alexbeckermonteiro@gmail.com (A.B. Monteiro).

https://doi.org/10.1016/j.still.2019.03.008
Received 15 May 2017; Received in revised form 30 January 2019; Accepted 14 March 2019
Available online 27 March 2019
0167-1987/ © 2019 Elsevier B.V. All rights reserved.
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

great importance in the Brazilian economy over the last years due to its maps can serve as an effective tool for farm managers and policy ma-
profitability in small areas, playing a relevant role in the generation of kers in site-specific nutrient management.
income and employment in rural areas, mainly regarding family-based Studies on the use of spatial distribution maps of soil and perennial
agricultural production systems. The production of temperate climate crop attributes for the delineation of management zones are still scarce,
fruit crops outstands in the Brazilian socio-economic scenario, mainly mainly for site-specific management of peach orchards aiming at pre-
in the States of Rio Grande do Sul, Santa Catarina, São Paulo and Minas cision agriculture. In this sense, the objectives of this work were: i) to
Gerais, either for in natura commercialization or for the industrial assess the spatial variability of soil and tree attributes using geostatis-
market (Fachinello et al., 2011). This agricultural activity is very im- tical tools in a peach orchard located in the region of Pelotas – RS,
portant in Southern Brazil, since it increases the profitability and the Brazil, and ii) to delineate specific management zones based on soil and
generation of income for family-based agricultural production systems tree attributes, through principal component and fuzzy c-means clus-
(economic scenario). It also allows the generation of employment in tering analyses.
farms and families to stay in rural areas (social scenario).
The State of Rio Grande do Sul (RS) is the main peach producer in 2. Material and methods
Brazil, with a harvested area of 12,574 ha and a total production of
about 129 thousand tons in 2015 (IBGE, 2016). The main industrial 2.1. Study area and soil and tree attributes
park of peach cannery is located in Pelotas, southern Rio Grande do Sul
State, which is responsible for producing over 60 thousand tons of this The study was carried out in a 1.8 ha peach orchard, Esmeralda
fruit. This manifests a relevant activity for the socio-economic devel- cultivar, located in the municipality of Morro Redondo, in the south of
opment of this region, since most of the production systems are based Rio Grande do Sul state, Brazil (31° 31′ 55.30″ S and 52° 35′ 37.87″ W).
on small family systems, and this activity opens an opportunity to The climate in the region, according to the Köppen’s classification is of
generate income and allows families to stay in rural areas. the “Cfa” type, i.e., humid temperate with hot summers. The region has
The soil nutritional management in peach orchards in the region of an annual mean temperature of 18 °C and an annual mean precipitation
Pelotas is still based on conventional and uniform recommendations to of 1,509 mm. The annual mean air relative humidity is 78.8%. The soil
the whole orchard area, not taking into consideration the variability of in the experimental area was classified as an Aquertic Hapludalf (Soil
the physical and chemical or nutritional environment where the trees Survey Staff, 2010).
develop. This might result in over-application in areas with high nu- An experimental grid was established in the peach orchard which
trient levels and under-application in areas with low nutrient levels, was constituted of 18 rows of trees (1.5 m far from each other and 6.0 m
resulting in higher heterogeneity in the peach production and fruit between rows), totaling 1,450 trees. 101 trees were selected randomly,
quality and lower profitability to farmers. In addition, soils are highly each one being georeferenced using a navigation GPS, model GPSmap
variable due to the combined effect of physical, chemical and biological 76CSx (Fig. 1).
processes that operate at different intensities and scales (Goovaerts, In July 2010, disturbed and undisturbed soil samples were collected
1998). Therefore, the use of multivariate statistical analysis (e.g., in the 0−0.10 m and 0.10−0.20 m layers, in a trench opened next to
principal component analysis and clustering analysis) has the potential each of the 101 trees. In each soil sample, the following soil hydro-
to define the soil homogeneous management zones based on crop and physical and chemical attributes were determined: sand by using the
soil spatial variability patterns in the field, and becomes promising for Sieving method and silt and clay fractions by using Pipette method (Gee
the rational management of nutrients and water in peach orchards. and Or, 2002); soil bulk density (Bd), using the Volumetric Ring
Davatgar et al. (2012) defined homogeneous management zones of Method (Grossman and Reinsch, 2002); volumetric soil water content
physical and chemical properties, through geostatistical, principal (θ), at the moment of the soil sampling in triplicate, was calculated by
component and fuzzy clustering analyses, aiming at providing in- multiplying the gravimetric soil water content by the soil bulk density
formation for a variable application rate of nutrients in the soil. These (Dirksen, 1999); pH in water (pH), organic matter content (OM),
authors concluded that clustering analysis reduced the variability of soil available phosphorus (P), and exchangeable potassium (K+), magne-
properties within each delimited zone providing the farmers the op- sium (Mg2+) and calcium (Ca2+) were determined following meth-
portunity of adopting a site-specific management of the nutrients of the odologies described by Claessen (1997). In order to represent the 0-
soil. Valente et al. (2012), evaluating soil physical and chemical prop- 0.20 m soil layer, the mean value of each soil attribute was calculated
erties, used geostatistical analysis combined with fuzzy c-means clus- from the 101 sampling points. Sampling in two soil layers (0-0.10 m
tering analysis to define management zones within coffee (Coffea ara- and 0.10−0.20 m) was carried out to represent the average of the
bica L.) fields. The authors concluded that more than one soil variable 0−0.20 m layer.
should be used to define management zones. In an apple orchard in The following attributes of each one of the 101 peach trees were
Greece, Aggelopooulou et al. (2013) used a combination of multivariate evaluated in 2010, 2011 and 2012: i) yield per tree (Y, kg tree−1): was
geostatistics with a non-parametric clustering approach to delineate determined by weighing of all fruits per tree; ii) number of fruits per
management zones through soil and tree attributes. From that study, tree (NF): was determined by counting the total fruits per tree; iii)
Aggelopooulou et al. (2013) divided an apple orchard into four zones, average individual fruit weight per individual tree (AFW, g) was de-
conditioning the differentiated management of soil fertility, irrigation termined by the ratio between Y and NF; iv) flesh firmness (FF, lb):
and cultural practices. measured with a hand penetrometer (FT 327, T.R. Turoni, Forli, Italy)
In five commercial fields of wheat production (Triticum aestivum L.) with an 8 mm tip in one fruit per tree and harvest date; v) total soluble
in Argentina, Peralta et al. (2015) concluded that the delineation of solid content (TSS, ºBrix): evaluated in the same fruit used to determine
management zones increased the efficiency of the use of nitrogen based FF, by extracting the juice and reading in a digital refractometer (Atago
fertilizers in three of all five evaluated fields, allowing the application PR-101, Atago Co Ltda., Japan). The number of harvest dates in 2010,
of nitrogen fertilizers at variable rates. Also, in a study carried out in the 2011 and 2012 were 3, 5 and 4, respectively. Fruits were harvested
south of India, Vasu et al. (2017) identified critical zones of nutrient manually. Seeking to reduce the possible isolated effects, which might
deficiency (soil pH, organic carbon, soil available nitrogen, phos- have occurred in the peach orchard in each year of evaluation, the
phorous, potassium and sulphur) from the elaboration of spatial dis- mean value of each tree attribute was used in this study.
tribution maps using geostatistical tools (semivariogram and ordinary
kriging) and delivered nutrient management recommendations based 2.2. Classical statistical and geostatistical analyses
on soil test results to farmers for adopting need based variable rate of
fertilizer application. These authors also pointed out that the generated All data sets were submitted to descriptive statistics, calculating the

2
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

Fig. 1. Location of the study area in the Brazil (A) and Rio Grande do Sul (B) maps and the sampling experimental grid established in the peach orchard constituted of
101 georeferenced trees (C).

mean, median, minimum (Min) and maximum (Max) values, the stan- principal components (PCs) (Jolliffe, 2002). PCA is a dimension re-
dard deviation (s), coefficient of variation (CV) and skewness and duction technique that uses correlated attributes, or variables, and
kurtosis coefficients as well. Data dispersion around the mean was identifies an orthogonal linear recombination of the attributes that
classified according to Warrick and Nielsen (1980) as: low – CV < summarize the principal sources of variability in the data (Xin-Zhong
12%; moderate – 12% ≤ CV < 60%; and high – CV ≥ 60%. The et al., 2009). In the present study, PCs were selected based on the fact
Kolmogorov-Smirnov (K–S) non-parametric test, at 5% significance that they explain at least 70% of the total variance of each data set and
level, was applied to check the normality of each data set’s distribution. that each calculated eigenvalue was > 1.0 (Jolliffe, 2002; Li et al.,
To select which soil and tree attributes could be used to delineate 2007). The factor score matrix of the selected PCs used to clustering
the management zones, Pearson correlation coefficients among all soil analysis was determined by multiplying the eigenvector matrix and the
and tree attributes were calculated. Through the t test, at 5% and 1% standardized interpolated data matrix of the selected soil and tree at-
significance levels, only soil attributes that presented significant cor- tributes, similarly to that used by Yao et al. (2014). The scores of each
relation with tree attributes were selected. selected principal component were plotted to obtain each component’s
The spatial variability structure of each selected soil and tree at- contour map.
tribute was characterized and quantified using isotropic experimental The delineation of the specific management zones in the peach
semivariograms, calculated through the Matheron estimator (Journel orchard was carried out through clustering analysis, using the fuzzy c-
and Huijbregts, 1978). To each experimental semivariogram, theore- means algorithm (Bezdek et al., 1984), applied to the factor score
tical semivariogram models (Nielsen and Wendroth, 2003) were ad- matrix of the selected PCs. The fuzzy c-means clustering analysis al-
justed and their respective adjustment parameters were obtained lows, through an iterative process, the best division of a data set into
(nugget effect: C0, sill: C0 + C and range: A). The degree of spatial different clusters with similar characteristics, calculating the center of
dependence [DSD (%) = (C0/(C + C0)×100] of each soil and tree at- each cluster (centroid) and the membership value between each ob-
tribute was calculated and classified according to Cambardella et al. servation of the data set and each cluster centroid. Each observation
(1994): DSD ≤ 25% – strong; 25 < DSD ≤ 75% – moderate; and will belong to the cluster to which it presents the highest membership
DSD > 75% – weak. To evaluate the quality of the semivariogram, the value, i.e., defuzzication procedure (Guastaferro et al., 2010). The fol-
cross-validation technique was used. When a spatial dependence lowing definitions were used to carry out the clustering analysis: i)
structure of a given variable was well adjusted by a theoretical semi- Euclidian distance as a norm to express the distance between the ob-
variogram model, a contour map was constructed by using ordinary servations and the cluster centroids; ii) value of the fuzziness weighting
kriging (Webster and Oliver, 2007). exponent = 1.5; iii) maximum number of iterations = 150, this allowed
The classical statistical and the Pearson’s correlation analyses were a maximum difference of 0.0001 for the convergence criterion of the
carried out using the program R, version 3.3.1 (R Core Team, 2016), adopted interactive process (Odeh et al., 1992); and iv) minimum
through the “Rcmdr” package (Fox, 2005, 2017). The K–S test was number of clusters = 2 and maximum number of clusters = 6.
applied using the “stats” package (R Core Team, 2016). The geostatis- The choice of the best number of clusters (number of management
tical analysis and the construction of contour maps were carried out zones) to be adopted was in function of the fuzziness performance index
aided by the GS + software, version 7.0 (Gamma Design Software, (FPI) and the modified partition entropy (MPE). The FPI indicates the
2004). degree of separation between the observations and the clusters gener-
ated (Fridgen et al., 2004). The MPE indicates the disorder degree
2.3. Multivariate statistical analysis created between the clusters. Its values range between 0 and 1, and the
closer to 0 it is, the better the division of clusters is. From the choice of
Previous to the multivariate statistical analyses, each data set was the best number of clusters, the map of management zones was ela-
standardized (mean = 0 and standard deviation = 1) aiming to trans- borated and significant differences between zones evaluated using the
form each original data to the same order of magnitude. Because the variance analysis.
variables present different orders of magnitude the use of the standar- The analyses were carried out using the program R - version 3.3.1,
dized variables requires that the variance and covariance matrixes are with the packages: “base” (R Core Team, 2016) for the principal com-
the same as the correlation matrix. ponent analysis; “e1071” (Meyer et al., 2015) for the fuzzy c-means
The principal component analysis (PCA) was applied to the selected clustering analysis; and “Rcmdr” (Fox, 2005, 2017) for the variance
soil and tree data sets aiming at transforming correlated original vari- analysis.
ables into a smaller number of non-correlated variables, named

3
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

Table 1
Descriptive statistics and the nonparametric Kolmogorov–Smirnov test for all evaluated soil hydro-physical and chemical attributes in the 0−0.20 m soil layer of an
Aquertic Hapludalf.
Attribute Mean Median Minimum Maximum s CV (%) Skewness Kurtosis d

Sand (%) 58 61 33 72 8.63 14.8 −0.64 −0.48 0.14N


Clay (%) 21 20 13 36 5.15 24.9 0.73 −0.10 0.13N
Silt (%) 21 21 14 31 4.57 21.7 0.45 −0.71 0.09N
Bd (g cm−3) 1.22 1.23 1.08 1.36 0.06 4.6 −0.09 −0.10 0.05N
θ (%) 21.4 20.8 15.7 28.1 2.91 13.6 0.40 −0.53 0.10N
OM (%) 1.92 1.85 1.15 3.05 0.42 21.7 0.39 −0.61 0.11N
pH in water 6.1 6.1 5.2 7.7 0.43 7.1 0.35 1.06 0.06N
P (mg dm−3) 5.4 3.5 0.3 30.7 5.69 105.9 2.00 4.86 0.19NN
K+ (mg dm−3) 85.9 80.0 26.0 232.5 29.98 34.9 1.35 4.58 0.09N
Ca2+ (cmolc dm−3) 1.7 1.5 0.1 5.3 1.10 63.4 0.96 0.78 0.11N
Mg2+ (cmolc dm−3) 1.4 1.3 0.7 2.9 0.44 32.2 0.82 0.66 0.12N

Sand, Clay, Silt: contents of sand, clay and silt in the 0−0.20 m depth layer, respectively; Bd: soil bulk density; θ: volumetric soil water content; OM: organic matter;
P: available phosphorus; K+: exchangeable potassium; Ca2+: exchangeable calcium; Mg2+: exchangeable magnesium; s: standard deviation of each data set; CV:
coefficient of variation of each data set; d: calculated values of the Kolmogorov-Smirnov test (K–S critical value is 0.14 at 5% of probability); N: normal distribution of
data and NN: non-normal distribution according to the Kolmogorov-Smirnov test at 5% probability.

3. Results and discussion orchard was from 5.6–22.0 kg tree−1, values corresponding to 23 and
89% of this variety potential, respectively (Raseira et al., 2014)
Average values of sand, clay and silt contents indicate that the soil is (Table 2). A lower peach yield was obtained in 2010 due to an occur-
a sandy clay loam in the 0–0.20 m soil layer, acidity (pH in water = rence of a high wind event (72.4 km h−1) in the experimental area on
6.1) close to that considered suitable (pH = 6.0) for the development October 31, 2010, which caused a fruit fall during the growing fruit
of peaches. The mean contents of organic matter (OM), available P and period. The mean value of TSS content (12.6°Brix, Table 2) is within the
exchangeable Ca2+ classified this soil as low, and exchangeable K+ and interval expected for the Esmeralda cultivar, which is between 12 and
Mg2+ classified as high (Table 1) (CQFS/SBCS, 2004). 14°Brix according to Raseira et al. (2014).
All data set distributions presented close mean and median values, Based on the coefficient of variation´s (CV) classification proposed
except for Sand, P and K+ distributions (Table 1). The data dispersion by Warrick and Nielsen (1980), NF (CV = 20.4%) and Y (CV = 22.1%)
around the mean was considered high for P and Ca2+ data sets data distributions were classified as being of moderate variability (12%
(CV ≥ 60%), low for distributions of Bd and pH in water (CV < 12%) ≤ CV < 60%), while those of FF, TSS and AFW were classified as of
and moderate for the remaining data sets (Warrick and Nielsen, 1980). low dispersion (CV < 12%). Liguori et al. (2013), studying the varia-
The degree of asymmetry of a data distribution is characterized by the bility of fruit quality in nectarine [Prunus persica (L.) Batsch] reported
skewness coefficient, while the extent of peakedness or flatness of a that the data variability of NF, TSS and AFW tree attributes increases
distribution is quantified by the kurtosis coefficient (Nielsen and with ripening and delay in the fruit harvest. This shows that the fruit
Wendroth, 2003). A normal probability distribution has values of both harvest carried out alternately (on different days of each evaluated
coefficients equal to zero. The coefficients of skewness ranged from year), might have influenced the low CV values for NF, TSS and AFW
-0.64 (for Sand) to +0.96 (for Ca2+), except for data distributions of P data set distributions in the present study (Table 2). All tree attribute
(+2.00) and K+ (+1.35), which indicates that these elements have a distributions showed a tendency toward normality according to K–S test
local distribution, that is, high values were found for these elements at at 5% probability.
some points, but most values were low (Grego et al., 2006). The same The majority of the soil attributes presented statistically significant
tendency was observed for the coefficients of kurtosis (Table 1). Based correlations with tree attributes, except for silt, Bd, pH and Ca2+
on this, it can be seen that most of the data distributions tended to be (Table 3). The tree attributes FF and NF did not present correlation with
normal (except for P and K distributions that tended to be lognormal), the soil attributes under evaluation. Based on that, the soil attributes:
so the mean value of each data set can be considered the center of Sand, Clay, θ, OM, P, K+ and Mg2+ and the tree attributes: TSS, AFW
distribution (Nielsen and Wendroth, 2003). However, only for P data and Y were used as input data for the delineation of the orchard
distribution this fact was not confirmed by applying the nonparametric management zones.
test of Kolmogorov-Smirnov at the 5% probability level (Table 1). A positive correlation between TSS and Sand variables and a ne-
All data set distributions of tree attributes presented proximity be- gative correlation of TSS with Clay, θ, OM and K+ variables was found
tween mean and median values (Table 2). The mean value of peach at least at the 5% probability level (Table 3). This can be ascribed to a
yield per tree (Y) for all three years (2010, 2011 and 2012) was 14.1 kg dilution effect, since the higher water availability for the peach crop
tree−1 which can be considered normal, since the variation within the can be related to the increase of clay and OM contents in the soil layer

Table 2
Descriptive statistics and the nonparametric Kolmogorov–Smirnov test for all evaluated peach attributes.
Attribute Mean Median Min Max s CV (%) Skewness Kurtosis d

FF (lb) 7.3 7.3 5.9 9.8 0.61 8.3 2.53 0.84 0.11N
TSS (ºBrix) 12.6 12.6 10.2 14.2 0.76 6.0 0.60 −0.38 0.08N
NF 133.0 135.3 56.0 202.3 27.20 20.4 0.06 −0.19 0.05N
AFW (g) 108.0 107.4 73.2 146.8 10.93 10.1 1.65 0.20 0.07N
Y (kg tree−1) 14.1 14.2 5.6 22.0 3.12 22.1 0.23 −0.21 0.04N

FF: flesh firmness; TSS: total soluble solids content; NF: number of fruits per tree; AFW: average fruit weight per individual tree; Y: peach yield per tree; s: standard
deviation of each data set; CV: coefficient of variation of each data set; d: calculated values of the Kolmogorov-Smirnov test (K–S critical value is 0.14 at 5% of
probability); N: normal distribution of data according to the Kolmogorov-Smirnov test at 5% probability.

4
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

Table 3
Matrix of Pearson’s correlation coefficients between soil properties and the mean values of peach attributes during the three years of evaluation.
Attribute Sand Clay Silt Bd θ OM pH P K+ Ca2+ Mg2+ FF TSS NF AFW Y

Sand 1.00
Clay −0.90** 1.00
Silt −0.87** 0.58** 1.00
Bd 0.16 −0.10 −0.18 1.00
θ −0.84** 0.81** 0.69** −0.41** 1.00
OM −0.77** 0.78** 0.57** −0.24* 0.79** 1.00
pH 0.28** −0.40** −0.07 −0.17 −0.23* −0.14 1.00
P 0.22* −0.15 −0.25* 0.21* −0.27** −0.09 −0.28** 1.00
K+ −0.25* 0.25* 0.19 −0.16 0.25* 0.14 −0.03 0.10 1.00
Ca2+ −0.17 0.08 0.23* −0.23* 0.17 0.20* 0.28** −0.27** 0.11 1.00
Mg2+ −0.69** 0.66** 0.57** −0.19 0.66** 0.73** 0.18 −0.22* 0.30** 0.29** 1.00
FF 0.06 −0.05 −0.05 0.13 −0.14 −0.12 −0.11 0.09 −0.11 −0.10 −0.09 1.00
TSS 0.28** −0.30** −0.19 0.08 −0.31** −0.21* 0.08 0.18 −0.30** 0.12 −0.16 0.18 1.00
NF −0.01 0.04 −0.02 −0.00 0.02 0.16 0.04 0.04 −0.05 −0.06 0.15 0.04 −0.07 1.00
AFW −0.13 0.20* 0.02 −0.06 0.23* 0.16 0.02 −0.23* 0.21* −0.01 0.13 −0.09 −0.64** 0.08 1.00
Y −0.11 0.18 0.01 −0.02 0.14 0.25* 0.02 −0.06 0.08 −0.07 0.23* 0.01 −0.32** 0.91** 0.43** 1.00

Sand, Clay, Silt: sand, clay and silt contents in the 0-0.20 m soil layer, respectively (%); Bd: soil bulk density (g cm−3); θ: volumetric soil water content (%); OM:
organic matter content (%); P: available phosphorus (mg dm−3); K+: exchangeable potassium (mg dm−3); Ca2+: exchangeable calcium (cmolc dm−3); Mg2+:
exchangeable magnesium (cmolc dm−3); FF: flesh firmness (lb); TSS: total soluble solids content (ºBrix); NF: number of fruits per tree; AFW: average fruit weight per
individual tree (g); Y: peach yield per tree (kg tree−1); * and **: significant correlation at 5% and 1% probability level through the t test, respectively.

under evaluation, increasing the proportion of liquids in the fruits and Table 4
decreasing the content of soluble solids. Theoretical semivariograms and their respective parameters (C0, C0+C and A),
The results from various studies showed consistently that the in- statistics of the fitted models (r2), spatial dependence degree (DSD) and results
crease in water content with the addition of organic matter is higher at of the cross-validation procedure (CR and R2) for all selected soil and tree at-
field capacity than that at permanent wilting point (Hudson, 1994; tributes.
Minasny and Mcbratney, 2018; Obour et al., 2018). This effect is more Attribute Model C0 C0 + C A DSD r2 Cross-validation
pronounced for sand soils with 0–20% clay content, and becomes
smaller from 20 to 40% clay and insignificant when clay is greater than CR R2

40% (Minasny and McBratney, 2018). According to Saxton and Rawls Sand Spherical 0.100 40.83 32.7 0.2 0.85 1.01 0.75
(2006), OM effects are similar to those of clay, thus those textures with Clay Gaussian 6.400 46.14 134.1 13.9 0.98 1.05 0.66
high clay content mask the effects of increased OM, which reduces the θ Spherical 0.430 5.20 27.3 8.3 0.96 1.00 0.61
soil potential to retain water at high water tension. This result can also OM Spherical 0.028 0.22 103.2 12.7 0.87 1.00 0.56
ln P Spherical 0.279 2.39 135.3 11.7 0.91 0.93 0.30
be corroborated by the strong negative correlation between TSS and
K+ Spherical 1.000 881.40 25.1 0.1 0.88 0.38 0.05
AFW variables (-0.64, Table 3). Lopez et al. (2010), in a study carried Mg2+ Spherical 0.036 0.26 127.2 14.0 0.90 0.96 0.40
out in a peach orchard in Spain, found a positive effect of reduced soil TSS Spherical 0.005 0.57 20.5 0.9 0.78 0.37 0.03
water availability on fruit quality. These authors concluded that trees AFW Gaussian 0.100 124.30 21.1 0.1 0.94 0.92 0.30
Y Spherical 1.598 9.63 29.0 16.6 0.90 0.71 0.19
with lower water availability during the last stage of the fruit devel-
opment yielded fruits with lower weight but higher TSS values. Terra
Sand and Clay: sand and clay contents, respectively (%); θ: volumetric soil
et al. (2014) studied the relation among soil and tree attributes in a water content (%); OM: organic matter content (%); ln P: natural logarithm of
peach orchard by applying multivariate cross-canonical correlation available phosphorus values (mg dm−3); K+: exchangeable potassium (mg
analysis. The negative relation between K+ and TSS values found in our dm−3); Mg2+: exchangeable magnesium (cmolc dm−3); TSS: total soluble solids
study (Table 3) is not common in the literature as it can be seen in the (ºBrix); AFW: average fruit weight per individual tree (g); Y: peach yield per
studies of Ruiz (2006); Ben Mimoun et al. (2009) and Dbara et al. tree (kg tree−1); C0: nugget effect; C0 + C: sill; A: range (m); r2: determination
(2016). Therefore, other factors (for instance, the availability of soil coefficient values for the adjusted theoretical semivariograms; CR and R2: re-
water during the high fruit growth phase which is close to the harvest) gression and determination coefficients related to the cross-validation proce-
can be acting on the relation between both variables in the orchard. dure, respectively.
Previous to the geostatistical analysis, P data were log-transformed
(ln-based) since their distribution indicated non-normality (Table 1). of the total variance (C0 + C) that is explained randomly and the
The structures of spatial variability of most of the soil variables were proportion of the structured variance (C) explained percentually. A
modeled by the spherical model, except for clay content which was strong dependence (DSD ≤ 25%) means that most of the variance takes
described with the gaussian model (Table 4). The spherical model is, part in C and a small part of the nugget effect (C0) (Bitencourt et al.,
according to McBratney and Webster (1986); Cavalcante et al. (2007) 2016). DSD values were classified as strong (DSD ≤ 25%) for all se-
and Liu et al. (2015), the most frequently adjusted one to describe the lected soil attributes (Table 4), indicating that the estimates of the
structure of spatial variability of soil attributes. variances during kriging will be better and the maps more precise, so
The spatial range (A) varied from 25.1 m (K+ variable) to 135.3 m that homogeneous management zones for soil fertilization are better
(ln P variable) (Table 4). The greater the range of the spatial depen- defined. The determination coefficient values r2 related to the fitted
dence is, the greater will be the distance in which there will be simi- theoretical semivariograms varied from 0.85 to 0.98 indicating that the
larity between neighbors, indicating a higher length of the spatial precision of all adjustments is suitable.
continuity of the attribute under study. On the other hand, smaller The adjusted theoretical semivariogram must adequately represent
ranges show that the variance tends quickly to a random behavior, the values of the estimated semivariance as a function of the distance
indicating a smaller length of the spatial continuity of the attribute within the spatial dependence range for the use of the geostatistical
under study. DSD is a non-dimensional measure of the spatial depen- kriging interpolator. According to Nielsen and Wendroth (2003), this
dence, proposed by Cambardella et al. (1994) to express the proportion means that part of the semivariogram reflects the structured variation

5
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

Fig. 2. Spatial distribution maps for all selected soil hydro-physical and chemical attributes in the 0-0.20 m soil layer: (A) sand content; (B) clay content; (C)
volumetric soil water content (θ); (D) soil organic matter (OM); (E) available phosphorus (P); (F) exchangeable potassium (K+); and (G) exchangeable magnesium
(Mg2+).

Fig. 3. Spatial distribution maps for the selected peach attributes (A) total soluble solids (TSS); (B) average fruit weight per individual tree (AFW); and (C) peach
yield per tree (Y).

of the attribute under study. The cross-validation procedure has been texture and structure of a soil (Reichardt and Timm, 2016). The spatial
one of the most used methods to judge the goodness of the fitted distributions of OM (Fig. 2D) and Mg2+ variables (Fig. 2G) were si-
semivariogram models (Webster and Oliver, 2007). From Table 4, it can milar, with greater concentrations in the east part of the experimental
be seen that the quality of the fitted semivariogram models is suitable area. It can also be seen that the K+ spatial distribution (Fig. 2F) was
for the majority of soil attributes, except for the K+ spatial structure in more heterogeneous as compared to those of OM, P and Mg2+ attri-
which the regression coefficient (CR) of 0.38 was not close to one and butes. The spatial distribution of P (Fig. 2E) was opposite to those of
the determination coefficient (R2) of 0.05 was low leading to greater clay, OM and Mg2+ distributions in the 0−0.20 m soil layer (Fig. 2B, D
deviation between measured and estimated values through kriging. and G, respectively). This result can be associated with management
The spherical model described the spatial distributions of TSS and Y factors. The increase in P concentrations in the northwest part of the
tree attributes, while the AFW distribution was modeled by the orchard (Fig. 2E) might have occurred because of the difficulty of
Gaussian model (Table 4). The spatial range (A) values varied from maneuvering the fertilizer spreader mounted to the tractor at the be-
20.5 m (TSS variable) to 29.0 m (Y variable). DSD values for spatial ginning of tree rows. This caused an accumulation of fertilizers in this
distributions of TSS, AFW and Y attributes were classified as strong part, which leads to incorrect accumulation of P, since it has low mo-
(DSD ≤ 25%) (Cambardella et al., 1994). The quality of the fitted bility through the soil profile when compared to other nutrients.
semivariogram model for the TSS variable, as expressed by the cross- The spatial distributions of TSS, AFW and Y tree attributes presented
validation procedure results, was lower than those ones for the other high heterogeneity in the orchard area (Fig. 3) as compared to those for
two variables. all selected soil attributes, since the spatial distributions of all soil at-
The largest sand concentrations in the 0-0.20 m soil layer are lo- tributes, in general, presented homogeneous and well-defined zones
cated in the west part of the experimental area (Fig. 2A). The opposite with greater extension in the experimental area (Fig. 2A to G). This
occurs with the spatial distribution of clay contents (Fig. 2B) and θ shows that the spatial variability of the tree attributes is not limited to
(Fig. 2C), since θ is a soil attribute that integrates factors related to the their complex interrelation with the selected soil attributes, but that it

6
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

depends on several other factors such as microclimate, health and plant attributes. A negative influence of TSS on PC2 was found (Table 5). PC3
genetic quality. explained 11.07% of the total variance, receiving greater positive in-
The north and northeast parts of the orchard area are limited by fluence of P and K+ soil variables and can be considered the soil
native forest, which interferes in sunlight incidence in these parts macronutrient component.
during the first hours of the day. This decreases soil temperature and, The FPI and MPE values as a function of the number of management
consequently, reduces the speed of organic compound degradation in zones are presented in Fig. 5. The optimal number of management
the soil (Erhagen et al., 2013) (higher OM contents; Fig. 2D). As a zones for each computed index is when the index is at the minimum,
consequence, lower TSS values were found in these parts (Fig. 3A). representing the least membership sharing (FPI) or greatest amount of
Alcobendas et al. (2013), assessing the quality of peach fruits in a organization (MPE) as a result of the clustering process (Odeh et al.,
commercial orchard in Spain, found lower TSS values when the fruits’ 1992; Fridgen et al., 2004). FPI and MPE minimum values were ob-
exposure to sunlight was lower. tained when the number of management zones was 2 (Fig. 5). It means
In order to reduce the dimensionality of both groups, the principal that the data set of selected measured properties present the lowest
component analysis (Jolliffe, 2002) was applied to the selected soil variance, the highest degree of membership sharing among classes and
(Sand, Clay, θ, OM, P, K+ and Mg2+ variables) and tree (TSS, AFW and reflects strong organization in the classification inside each delimited
Y variables) attributes. Principal component analysis is a dimension management zone. The farmer’s ability to manage the created number
reduction technique that uses correlated attributes, or variables, and of management zones needs to be considered in order to use the existing
identifies orthogonal linear recombination of the properties that sum- data, in conjunction with his knowledge to understand the yield pro-
marize the principal sources of variability in the data (Li et al., 2007; cesses in the field. Therefore, a more efficient orchard management can
Xin-Zhong et al., 2009; Yao et al., 2014). The first three components be carried out.
(PC1, PC2 and PC3) presented eigenvalues > 1.0 and explained 73.40% The defuzzication procedure was applied by using the membership
of the total variance in the data sets (Table 5). Therefore, PC1, PC2 and values from the 2 management zones, caused each observation (inter-
PC3 were used to define the orchard potential management zones. polated point) to be classified to a unique management zone
The first principal component (PC1) explained 44.95% of the total (Guastaferro et al., 2010). Therefore, each observation received a value
data variance (Table 5), being that Clay, θ, OM and Mg2+ attributes z equal to 1 or 2 (zones 1 and 2) and, from these values, the map of
presented higher negative influence to this component, while the Sand management zones was elaborated (Fig. 6).
content presented high positive influence to PC1, as indicated by the Through the variance analysis, it was possible to identify significant
PC1 component loadings (Table 5). PC1 can be considered the com- differences between management zones for most of the attributes
ponent linked to the availability and retention of water and organic (Sand, Clay, θ, OM, Mg2+, TSS and AFW, considering the p-value < 1%
matter in the soil. Soil organic matter is an indicator of soil quality and and < 5%; Table 6). Based on the mean values of soil properties
affects soil physical, chemical and biological properties (Castro et al., (Table 6), zone 2 presented the highest values of Clay, θ, MO and Mg2+,
2015). As a key variable for a number of climatic, ecological, hydro- showing greater soil aggregation and higher plant water and nutrient
logical, and nutrient-based processes operating in different intensities retention potential (Aggelides and Londra, 2000; Celik et al., 2004). For
and, at different spatio-temporal scales, soil organic matter is highly tree attributes, zone 2 presented the highest AFW values and the lowest
variable in time and space (Nielsen and Wendroth, 2003) across land- SST values.
scapes. Although the peach yield per tree (Y) did not present a significant
The PC1 spatial distribution map (Fig. 4A) is similar to that of the difference between the management zones (Table 6), knowing the dif-
Sand (Fig. 2A), as a result of the positive relation between them. On the ference between zones of the AFW attribute can serve as a basis to
other hand, spatial distributions of PC1 and those of Clay, θ, MO and define the peach productive potential and profitability of the peach
Mg2+ attributes were also similar but they are negatively correlated. trees in each zone. The peach fruit size is regarded a high fruit quality
The second principal component (PC2) explained 17.37% of the attribute because it is the attribute with the highest economic relevance
total variance of the data sets, and can be named as the plant compo- (Lopez et al., 2010).
nent, which received greatest positive influence of AFW and Y tree The division of the orchard area in two management zones allows

Table 5
Results of the principal component analysis applied to the soil (sand and clay contents, soil water content, soil organic matter, available phosphorus, exchangeable
potassium and exchangeable magnesium) and tree attributes (total soluble solid contents, average fruit weight per individual tree and peach yield per tree).
Principal Eigenvalues Variance (%) Cumulative
component variance (%)

PC 1 4.495 44.95 44.95


PC 2 1.737 17.37 62.33
PC 3 1.107 11.07 73.40
PC 4 0.894 8.95 82.34
PC 5 0.647 6.47 88.82
PC 6 0.359 3.59 92.40
PC 7 0.316 3.16 95.56
PC 8 0.217 2.17 97.73
PC 9 0.145 1.45 99.19
PC 10 0.081 0.82 100.00

Principal component loadings for each attribute


Sand Clay θ OM P K+ Mg2+ TSS AFW Y
PC 1 0.902 −0.899 −0.899 −0.860 0.280 −0.361 −0.804 0.463 −0.384 −0.328
PC 2 0.258 −0.194 −0.171 −0.242 −0.174 0.207 −0.206 −0.702 0.799 0.546
PC 3 0.016 0.039 −0.069 −0.004 0.787 0.688 0.019 −0.056 −0.065 0.002

Sand and Clay: sand and clay contents in the 0-0.20 m soil layer, respectively (%); θ: volumetric soil water content (%); OM: organic matter content (%); P: available
phosphorus (mg dm−3); K+: exchangeable potassium (mg dm−3); Mg2+: exchangeable magnesium (cmolc dm−3); TSS: total soluble solids content (ºBrix); AFW:
average fruit weight per individual tree (g); Y: peach yield per tree (kg tree−1).

7
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

Fig. 4. Spatial distribution maps for the scores of the first three principal components (PCs): (A) PC 1; (B) PC 2; and (C) PC 3.

monitor soil water content can be installed in different zones, and the
crop water demand be attended according to the needs of the trees in
each zone. With regard to soil fertility, fruit quality and yield man-
agement in the peach orchard can be improved with an adequate nu-
tritional management (Chatzitheodorou et al., 2004; Ben Mimoun et al.,
2009). Doses of fertilizers can be applied to each management zone
depending on its demands, as suggested in several studies
(Aggelopooulou et al., 2013; Yao et al., 2014; Tripathi et al., 2015).
Cultural practices, likewise, can be carried out in a differentiated
basis in each management zone. As for example, the harvest date can be
postponed in trees cultivated in Zone 1. Fruit weight is associated to the
fruit ripening time in the tree, which tends to increase when the harvest
is delayed (Liguori et al., 2013). The higher is the soil water content,
the higher is the plant vegetative growth. Therefore, the pruning in
Fig. 5. Calculated fuzziness performance index (FPI) and modified partition
Zone 2 could be carried out seeking higher sunlight exposure inside the
entropy (MPE) against cluster numbers.
canopy in relation to that in the Zone 1, providing higher TSS content in
the fruits (Motisi et al., 2006; Alcobendas et al., 2013; Liguori et al.,
2013).
The combined use of statistical tools from the geostatistical and
multivariate analysis allowed the reduction of the variability of soil and
tree attributes within each one of the two outlined zones providing an
opportunity to adopt specific management of each individual soil or
tree attribute or of each soil/tree group zone. However, the low cor-
relation between soil and tree attributes in this present study, mainly
due to the high spatial variability of tree attributes, indicates that for
taking better advantage of the real potential of the peach orchard, other
factors should be taken into consideration, such as microclimate, health
and seedling rootstock variability. The use of canopy sensors, for ex-
Fig. 6. Spatial distribution map for the resulting management zones (zones 1
ample, which has allowed high sampling density at data survey of the
and 2).
vegetation index in vineyards in precision viticulture (Trought and
Bramley, 2011; King et al., 2014), can be a future alternative to identify
an economically viable management, easy for the producer, for ex- in more details the variability of peach canopy characteristics in the
ample, to apply fertilizers, irrigation and cultural practices as well. The area. From this, a better understanding of the correlation between yield
management of each variable should be based on the expert knowledge and fruit quality variables can be find. Therefore, the delineation of
of the farmer and technical manuals that guide the response of each management zones is a promising alternative to improve the profit of
attribute to the most appropriate expected level. The division in man- peach orchards for small and medium producers.
agement zones leads the farmer to reduce the number of samplings with
adequate representativeness. Based on the sampling, the peach orchard
can be irrigated, fertilized or conducted in a similar way in each zone, 4. Conclusions
aiming to create the conditions to the highest economic return.
The irrigation management directly influences both yield and fruit - The structure of the spatial variability of the majority of the
quality of peaches (Alcobendas et al., 2013). Therefore, sensors to evaluated soil and tree attributes was best modeled by the spherical
semivariogram model and the degree of spatial dependence of all

Table 6
Mean values of each management zone and variance analysis for all selected soil and tree attributes.
Zone n Sand Clay θ OM P K+ Mg2+ TSS AFW Y
(%) (mg dm−3) (cmolc dm−3) (ºBrix) (g) (kg tree−1)

1 60 64.17 17.32 19.48 1.66 6.46 81.82 1.13 12.81 105.25 13.72
2 41 49.40 25.67 24.10 2.30 3.79 91.89 1.67 12.25 112.03 14.77
Pr (> F) < 0.001 < 0.001 < 0.001 < 0.001 0.020 0.097 < 0.001 < 0.001 0.002 0.097

n: number of sampling points; Sand and Clay: sand and clay contents in the 0-0.20 m soil layer, respectively; θ: volumetric soil water content; OM: organic matter
content; P: available phosphorus; K+: exchangeable potassium; Mg2+: exchangeable magnesium; TSS: total soluble solids content; AFW: average fruit weight per
individual tree; Y: peach yield per tree.

8
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

attributes was classified as strong. Chapman and Hall/CRC Press.


- The spatial variability of tree attributes was more heterogeneous as Fridgen, J.J., Kitchen, N.R., Sudduth, K.A., Drummond, S.T., Wiebold, W.J., Fraisse, C.W.,
2004. Management zone analyst (MZA): software for subfield management zone
compared to those of soil attributes in the orchard area. delineation. Agron. J. 96, 100–108.
- Yield and quality of peach fruits were affected by texture, water Gamma Design Software, 2004. GS+: Geostatistics for the Environmental Sciences.
content, organic matter, exchangeable potassium and magnesium of the Gamma Design Software, Plainwell.
Gee, G.W., Or, D., 2002. The solid phase: particle-size analysis. In: Dane, J.H., Topp, G.C.
0−0.20 m soil layer. (Eds.), Methods of Soil Analysis. Physical Methods. Soil Science Society of America,
- The optimal number of management zones for the peach orchard Madison, pp. 255–293.
was two and the most of soil and tree attributes presented statistical Goovaerts, P., 1998. Geostatistical tools for characterizing the spatial variability of mi-
crobiological and physico-chemical soil properties. Bio. Fert. Soils 27, 315–334.
differences in each defined management zone. Grego, C.R., Vieira, S.R., Lourenção, A.L., 2006. Spatial distribution of Pseudaletia sequax
- The delineation of management zones based on the spatial varia- Franclemlont in triticale under no-till management. Sci. Agric. 63, 321–327.
bility of soil and tree attributes opens the opportunity to the farmer to Grossman, R.B., Reinsch, T.G., 2002. The solid phase: bulk density and linear ex-
tensibility. In: Dane, J.H., Topp, G.C. (Eds.), Methods of Soil Analysis: Physical
manage site-specifically soil fertility, irrigation, and cultural practices,
Methods. Soil Science Society of America, Madison, pp. 201–228.
aiming at the increase in yield and fruit quality in peach orchard. Guastaferro, F., Castrignanò, A., De Benedetto, D., Sollitto, D., Troccoli, A., Cafarelli, B.,
2010. A comparison of different algorithms for the delineation of management zones.
Acknowledgements Precision Agric. 11, 600–620.
Hudson, B.D., 1994. Soil organic matter and available water capacity. J. Soil Water
Conserv. 49, 189–194.
To Embrapa Clima Temperado - RS for the infrastructure; to the IBGE, 2016. Área colhida e quantidade produzida da lavoura permanente. Instituto
producer Marcus Fiss for granting the experimental area, and to the Brasileiro de Geografia e Estatística, Rio de Janeiro. http://www.ibge.gov.br/
estadosat/.
CNPq for granting scholarships. This study was financed in part by the Jolliffe, I.T., 2002. Principal Component Analysis, second ed. Springer, New York.
Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brazil Journel, A.G., Huijbregts, J.C.H., 1978. Mining Geostatistics. Academic Press, London.
(CAPES) - Finance Code 001" King, P.D., Smart, R.E., McClellan, D.J., 2014. Within-vineyard variability in vine vege-
tative growth, yield, and fruit and wine composition of Cabernet Sauvignon in
Hawke’s Bay, New Zealand. Aust. J. Grape Wine Res. 20, 234–246.
References Li, Y., Shi, Z., Li, F., Li, H., 2007. Delineation of site-specific management zones using
fuzzy clustering analysis in a coastal saline land. Comput. Electron. Agric. 56,
174–186.
Aggelides, S.M., Londra, P.A., 2000. Effect of compost produced from town wastes and
Liguori, G., Farina, V., Gullo, G., Inglese, P., 2013. Tree and orchard variability of Silver
sewage sludge on the physical properties. Bioresour. Technol. Rep. 71, 253–259.
King nectarine (Prunus persica (L.) Batsch) fruit quality components. Hort. Sci.
Aggelopooulou, K., Castrignanò, A., Gemtos, T., De Benedetto, D., 2013. Delineation of
(Prague) 40, 72–77.
management zones in an apple orchard in Greece using a multivariate approach.
Liu, C.L., Wu, Y.Z., Liu, Q.J., 2015. Effects of land use on spatial patterns of soil properties
Comput. Electron. Agric. 90, 119–130.
in a rocky mountain area of Northern China. Arab. J. Geosci. 8, 1181–1194.
Alcobendas, R., Mirás-Avalos, J.M., Alarcón, J.J., Nicolás, E., 2013. Effects of irrigation
Lopez, G., Behboudian, M.H., Vallverdu, X., Mata, M., Girona, J., Marsal, J., 2010.
and fruit position on size, colour, firmness and sugar contents of fruits in a mid-late
Mitigation of severe water stress by fruit thinning in ‘O’ Henry’ peach: implications
maturing peach cultivar. Sci. Hort. 164, 340–347.
for fruit quality. Sci. Hort. 125, 294–300.
Ben Mimoun, M., Ghrab, M., Ghanem, M., Elloumi, O., 2009. Effects of potassium foliar
McBratney, A.B., Webster, R., 1986. Choosing functions for semi-variograms of soil
spray on olive, peach and plum, Part 2: peach and plum experiments. Optimizing
properties and fitting them to sampling estimates. J. Soil Sci. 37, 617–639.
Crop Nutrition. International Potash Institute. E-ifc 19, 14–17. http://www.
Meyer, D., Dimitriadou, E., Hornik, K., Weingessel, A., Leisch, F., Chang, C.C., Lin, C.C.,
ipipotash.org/fr/eifc/2009/19/4.
2015. Misc Functions of the Department of Statistics, Probability Theory Group
Bezdek, J.C., Ehrlich, R., Full, W., 1984. FCM: the fuzzy c-means clustering algorithm.
(Formerly: E1071). e1071. TU Wien. https://CRAN.R-project.org/package=e1071.
Comput. Geosci. 10, 191–203.
Minasny, B., Mcbratney, A.B., 2018. Limited effect of organic matter on soil available
Bitencourt, D.G.B., Barros, W.S., Timm, L.C., She, D., Penning, L.H., Parfitt, J.M.B.,
water capacity. Eur. J. Soil Sci. 69, 39–47.
Reichardt, K., 2016. Multivariate and geostatistical analyses to evaluate lowland soil
Motisi, A., Marra, F.P., Gullo, G., Mafrica, R., Zappia, R., 2006. Relationship between
levelling effects on physico-chemical properties. Soil Till. Res. 156, 63–73.
canopy architecture and fruit quality of ‘Rich May’ peach grafted onto ‘Penta’ and
Cambardella, C.A., Moorman, T.B., Nocak, J.M., Parkin, T.B., Karlen, D.L., Turco, R.F.,
‘GF677’rootstocks. Acta Hort. 713, 365–371.
Konopka, A.E., 1994. Field-scale variability of soil properties in central Iowa soils.
Nielsen, D.R., Wendroth, O., 2003. Spatial and temporal statistics: sampling field soils and
Soil Sci. Soc. Am. J. 58, 1501–1511.
their vegetation. Catena Verlag, Reiskirchen.
Castro, G.S.A., Crusciol, C.A.C., Calonego, J.C., Rosolem, C.A., 2015. Management im-
Obour, P.B., Jensen, J.L., Lamandé, M., Watts, C.W., Munkholm, L.J., 2018. Soil organic
pacts on soil organic matter of tropical soils. Vadose Zone J. 14 (1), 1–8. https://doi.
matter widens the range of water contents for tillage. Soil Till. Res. 182, 57.
org/10.2136/vzj2014.07.0093.
Odeh, I.O.A., McBratney, A.B., Chittleborough, D.J., 1992. Soil pattern recognition with
Cavalcante, E.G.S., Alves, M.C., Souza, Z.M., Pereira, G.T., 2007. Variabilidade espacial
fuzzy-c-means: application to classification and soil-landform interrelationships. Soil
de atributos químicos do solo sob diferentes usos e manejos. R. Bras. Ci. Solo 31,
Sci. Soc. Am. J. 56, 505–516.
10–18.
Peralta, N.R., Costa, J.L., Balzarini, M., Franco, M.C., Córdoba, M., Bullock, D., 2015.
Celik, I., Ortas, I., Kilic, S., 2004. Effects of compost, mycorrhiza, manure and fertilizer on
Delineation of management zones to improve nitrogen management of wheat.
some physical properties of a Chromoxerert soil. Soil Till. Res. 78, 59–67.
Comput. Electron. Agric. 110, 103–113.
Chatzitheodorou, I.T., Sotiropoulos, T.E., Mouhtaridou, G.I., 2004. Effect of nitrogen,
R Core Team, 2016. R: a Language and Environment for Statistical Computing. R
phosphorus, potassium fertilization and manure on fruit yield and fruit quality of the
Foundation for Statistical Computing, Vienna. https://www.R-project.org.
peach cultivars ‘Spring Time’ and ‘Red Haven’. Agron. Res. 2, 135–143.
Raseira, M.C.B., Pereira, J.F.M., Carvalho, F.L.C. (Eds.), 2014. Pessegueiro. Embrapa,
Claessen, M.E.C., 1997. Manual de métodos de análise de solo (Org.). second ed. rev.
Brasília.
atual Embrapa-CNPS, Rio de Janeiro.
Reichardt, K., Timm, L.C., 2016. Solo, Planta e Atmosfera: conceitos, processos e
CQFS/SBCS Comissão de Química e Fertilidade do Solo, 2004. Manual de adubação e de
aplicações, third ed. Manole, Barueri.
calagem para os Estados do Rio Grande do Sul e de Santa Catarina, tenth ed.
Ruiz, R., 2006. Effects of different potassium fertilizers on yield, fruit quality and nutri-
Sociedade Brasileira de Ciência do Solo, Núcleo Regional Sul, Porto Alegre.
tional status of ‘Fairlane’ nectarine trees and on soil fertility. Acta Hort. 721,
Davatgar, N., Neishabouri, M.R., Sepaskhah, A.R., 2012. Delineation of site specific nu-
185–190.
trient management zones for a paddy cultivated area based on soil fertility using
Saxton, K.E., Rawls, W.J., 2006. Soil water characteristic estimates by texture and organic
fuzzy clustering. Geoderma 173–174, 111–118.
matter for hydrologic solutions. Soil Sci. Soc. Am. J. 70, 1569–1578.
Dbara, S., Gader, T., Ben Mimoun, M., 2016. Improving yield and fruit quality of peach
Soil Survey Staff, 2010. Keys to Soil Taxonomy, eleventh ed. USDA-Natural Resources
cv. ‘Flordastar’ by potassium foliar spray associated to regulated deficit irrigation. J.
Conservation Service, Washington, DC.
New Sci. Agric. Biotechnol 28, 1631–1637.
Terra, V.S.S., Valgas, R.A., Reisser Júnior, C., Timm, L.C., Pereira, J.F.M., Carvalho,
Dirksen, C., 1999. Soil Physics Measurements. Catena Verlag, Reiskirchen.
F.L.C., Oldoni, H., 2014. Multivariate analysis applied to the study of the relationship
Erhagen, B., Öquist, M., Sparrman, T., Haei, M., Ilstedt, U., Hedenstrom, M., Schleucher,
between soil and plant properties in a peach orchard. R. Bras. Ci. Solo 38, 755–764.
J., Nilsson, M.B., 2013. Temperature response of litter and soil organic matter de-
Tripathi, R., Nayak, A.K., Shahid, M., Lal, B., Gautam, P., Raja, R., Mohanty, S., Kumar,
composition is determined by chemical composition of organic material. Glob.
A., Panda, B.B., Sahoo, R.N., 2015. Delineation of soil management zones for a rice
Change Biol. Bioenergy 19, 3858–3871.
cultivated area in eastern India using fuzzy clustering. Catena 133, 128–136.
Fachinello, J.C., Pasa, M.S., Schmtiz, J.D., Betemps, D.L., 2011. Situação e perspectivas da
Trought, M.C.T., Bramley, R.G.V., 2011. Vineyard variability in Marlborough, New
fruticultura de clima temperado no Brasil. Rev. Bras. Frutic. Special 109–120.
Zealand: characterising spatial and temporal changes in fruit composition and juice
FAOstat, 2017. Food and Agricultural Organization Statistical Database. FAO, Rome.
quality in the vineyard. Aust. J. Grape Wine Res. 17, 79–89.
http://www.fao.org/faostat/en/#data/QC.
Valente, D.S.M., Queiroz, D.M., Pinto, F.A.C., Santos, N.T., Santos, F.L., 2012. Definition
Fox, J., 2005. The R commander: a basic statistics graphical user interface to R. J. Stat.
of management zones in coffee production fields based on apparent soil electrical
Softw. 9, 1–42.
conductivity. Sci. Agric. 69, 173–179.
Fox, J., 2017. . Using the R Commander: A Point-and-Click Interface or R. Boca Raton FL.
Vasu, D., Singh, S.K., Sahu, N., Tiwary, P., Chandran, P., Duraisami, V.P., Ramamurthy,

9
H. Oldoni, et al. Soil & Tillage Research 191 (2019) 1–10

V., Lalitha, M., Kalaiselvi, B., 2017. Assessment of spatial variability of soil properties Wiley, Chichester.
using geospatial techniques for farm level nutrient management. Soil Till. Res. 169, Xin-Zhong, W., Guo-Shun, L., Hong-Chao, H., Zhen-Hai, W., Qing-Hua, L., Xu-Feng, L.,
25–34. Wei-Hong, H., Yan-Tao, L., 2009. Determination of management zones for a tobacco
Warrick, A.W., Nielsen, D.R., 1980. Spatial variability of soil physical properties the soil. field based on soil fertility. Comput. Electron. Agric. 65, 168–175.
In: Hill, D. (Ed.), Applications of Soil Physics. Academic Press, New York, pp. Yao, R.J., Yang, J.S., Zhang, T.J., Gao, P., Wang, X.P., Hong, L.Z., Wang, M.W., 2014.
319–344. Determination of site-specific management zones using soil physico-chemical prop-
Webster, R., Oliver, M.A., 2007. Geostatistics for Environmental Scientists, second ed. erties and crop yields in coastal reclaimed farmland. Geoderma 232–234, 381–393.

10

You might also like