Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
DOI: 10.31876/er.v4i34.750
49
Multivariate analyses to determine fungicide ecacy
on Ecuadorian bananas for consumption
Análisis multivariantes para determinar la ecacia de fungicidas en banano
ecuatoriano para consumo
José Ascencio-Moreno*, Miriam Vanessa Hinojosa-Ramos, Francisco Vera
***
Omar Ruiz-Barzola
****
,
María Isabel Jiménez-Feijoó
*****
, María Puricación Galindo-Villardón
******
, Miriam Ramos-Barberán
*******
Recibido: 18 de noviembre de 2019.
Aprobado: 02 de mayo de 2020.
Cite this:
Ascencio-Moreno, J. et al. (2020). Multivariate
analyses to determine fungicide efficacy on
Ecuadorian bananas for consumption. Espirales.
Revista Multidisciplinaria de investigación cientíca,
4(34), 49-66
*
Engineer. Escuela Superior Politécnica del Litoral,
Guayaquil, Ecuador.
E-mail: josdasce@espol.edu.ec.
ORCID: 0000-0002-6883-7195.
Google Scholar
**
Biologist. Escuela Superior Politécnica del Litoral,
Guayaquil, Ecuador.
E-mail: mvhinojo@espol.edu.ec.
ORCID: 0000-0002-4100-5284.
Google Scholar
***
PhD in Statistics. Escuela Superior Politécnica del
Litoral, Guayaquil, Ecuador.
E-mail: fvera@espol.edu.ec.
ORCID: 0000-0001-6541-7243.
Google Scholar
****
PhD in Education. Escuela Superior Politécnica del
Litoral, Guayaquil, Ecuador.
E-mail: oruiz@espol.edu.ec.
ORCID: 0000-0001-8206-1744.
Google Scholar
*****
Phd in Education. Escuela Superior Politécnica del
Litoral, Guayaquil, Ecuador.
E-mail: mjimenez@espol.edu.ec.
ORCID: 0000-0002-1961-5123.
Google Scholar
******
Phd in Education. Universidad de Salamanca,
Salamanca, España.
E-mail: pgalindo@usal.es.
ORCID: 0000-0001-6977-7545.
Google Scholar
*******
Master in Productivity and Quality Management.
Escuela Superior Politécnica del Litoral, Guayaquil,
Ecuador.
E-mail: mvramosb@espol.edu.ec.
ORCID: 0000-0002-8915-6938.
Google Scholar
Abstract
Half maximal effective concentration EC
50
is considered
the main reference for evaluating the ecacy of the
products in any plantation using doses and inhibition
percentages from laboratory data. However, EC
50
is not
the best representation when other relevant variables and
their relationships could be involved. As an agricultural
case study, fungicide sensitivity of Pseudocercospora
jiensis, the causal agent of black sigatoka, was evaluated
on bananasplantations in three provinces of Ecuador.
In this study, multivariate statistical process control
was adjusted to a fungicide ecacy evaluation case
considering multiple data tables from different locations
and years at the same time. The threshold conveyed by
inhibition percentages, related to the EC
50
, along with
locations and years allowed the multivariate analyses
carried out in the proposal. The multivariate statistical
control techniques applied were Multilinear Principal
Component Analysis (MPCA) and Dual STATIS-Parallel
Coordinates approach (DS-PC). A comparison was
developed and showed that both methods discriminate
correctly between the normal and anomalous conditions
within plantations along years, validating the ability of
the novel method DS-PC for exhibiting better signaling
of anomalous plantations and performing variable-wise
analysis to nd out possible causes of this behavior in an
easier time-saving graphical framework.
Key words: Bananas, fungicide ecacy, black sigatoka,
MPCA, DS-PC.
50
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Resumen
La concentración efectiva media máxima CE
50
se considera la
referencia principal para evaluar la ecacia de los productos en
cualquier plantación, utilizando dosis y porcentajes de inhibición a partir
de datos de laboratorio. Sin embargo, CE
50
no es la mejor representación
cuando otras variables relevantes y sus relaciones podrían estar
involucradas. Como estudio de caso agrícola se evaluó la sensibilidad
a los fungicidas de Pseudocercospora jiensis, el agente causal de
la sigatoka negra, en las plantaciones de banano en tres provincias
del Ecuador. En este estudio, el control estadístico multivariante de
procesos se ajustó a un caso de evaluación de ecacia de fungicida
al considerar múltiples tablas de datos de diferentes ubicaciones y
años al mismo tiempo. El umbral transmitido por los porcentajes de
inhibición, relacionados con la CE
50
junto con las ubicaciones y los
años, permitieron los análisis multivariados realizados en la propuesta.
Las técnicas de control estadístico multivariantes aplicadas fueron el
análisis de componentes principales multilineal (MPCA) y el STATIS
Dual-coordenadas paralelas (DS-PC). Se desarrolló una comparación
que mostró que ambos métodos discriminan correctamente entre las
condiciones normales y anómalas dentro de las plantaciones a lo largo
de los años, validando la capacidad del novedoso método DS-PC para
exhibir una mejor señalización de plantaciones anómalas y realizando
un análisis enfocado sobre variables para descubrir posibles causas
de este comportamiento en un marco gráco que ahorra tiempo.
Palabras clave: banano, ecacia de fungicida, sigatoka negra, MPCA,
DS-PC.
Introduction
Cultivated species of the genus Musa, bananas and plantains, are among the products of major
consumption and food relevance worldwide after rice, wheat and milk. Banana and plantain
production areas are located in more than 100 countries in tropical and subtropical regions
and cover approximately 5.6 million hectares in Ecuador, Philippines, Costa Rica, Colombia, and
Guatemala. In Ecuador, bananas are considered the second main non-petroleum exportation
product and aimed mostly to household consumption. Globally, the banana industry generates
51
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
around 8 billion dollars per year, with a growing demand and economic impact in third world
countries, where it represents 75 % of the monthly income of small-scale farmers (FAO, 2018;
Marín et al., 2003).
Banana production is mainly threatened by the ascomycete fungus Pseudocercospora
jiensis, which manifests with symptoms such as spots and chlorotic streaks (Marín et al.,
2003; Stover, 1980). The most effective disease management is through the application of
fungicides according to their different modes of action. The fungicides work by interfering in
vital processes of the pathogen (Bolaños, 2006; Luna-Moreno et al., 2019; Martínez-Bolaños
et al., 2012; Martínez et al., 2011; Romero & Sutton, 1997).
Chemical control is the main tool for black sigatoka management, including alternating
applications of protectant and systemic fungicides or mixtures of the two types of fungicides
(Brent & Hollomon, 2007). Protectant fungicides present low or no risk of resistance including
mancozeb and chlorothalonil, while systemic fungicides exhibit moderate to high risk of
resistance and include benzimidazoles, amines, triazoles, strobilurins, anilinopyrimidines,
carboxamides and guanidines. Pseudocercospora jiensis has developed resistance to
benzimidazoles, triazoles and strobilurins, which has reduced fungicide effectiveness and
limited its use in crops (Guzmán, Orozco-Santos & Pérez Vicente, 2013; Martínez et al.,
2011). The development of resistance to the fungicides based on the aforementioned
groups has increased the use of amines and anilinopyrimidines, which is seen as a risk,
due to the increased selection pressure on the pathogen. Therefore, the use of systemic
fungicides in banana must comply with the Fungicide Resistance Action Committee (FRAC)
recommendations, which consider a maximum of applications per year: triazoles (8), amines
(15), strobilurins (3), anilinopyrimidines (8), benzimidazoles (3), carboxamides (4) and
guanidines (6) (Guzmán, Orozco-Santos & Pérez Vicente, 2013; Martínez & Guzmán, 2011;
Martínez et al., 2011; Russell, 2004).
Half maximal effective concentration EC
50
is a standard measurement that has been widely
used in sensitivity analyses for chemical products in pathogen control, especially in black
sigatoka disease, which belong to the most important annual bioassays for decision makers
to keep tested fungicides in market. To examine the effectiveness of a fungicide and changes
in sensitivity, P. jiensis response is commonly analyzed under the EC
50
(Bolaños, 2006; Luna-
Moreno et al., 2019; Martínez-Bolaños et al., 2012; Martínez et al., 2011; Romero & Sutton, 1997).
Traditional analyses on fungicide ecacy take into account doses and inhibition percentages
for a regression model from which a dose response curve is determined. The EC
50
of a dose
response curve represents the fungicide concentration that causes the 50 % spores inhibition
or the relative growth of pathogen colonies (Bolaños, 2006; Velásquez et al., 2014).
Conventional screening and disease control studies focus exclusively in the EC
50
standard for
fungicide ecacy without considering its relationship with climate and geographic conditions
52
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
that affect the black sigatoka growth and dispersion (Jacome & Schuh, 1992; Pérez-Vicente
et al., 2000). In this sense, decisions made on a single number, such as EC
50
, throw away
potentially useful information. Considering multiple numbers (variables) at the same time
may give certain insights on the data, unavailable within a univariate analysis. That is where
multivariate analysis come into play in many areas of life sciences, which assess multiple
variables problems (Ali, 2011; Saed-Moucheshi et al., 2013).
In agriculture, one of the major areas of interest for the improvement of productivity in terms
of quality and quantity, is the selection of types of fungicides to apply according to pathogens
response (Aguirre, Piraneque & Rodríguez, 2015; Barakat et al., 2017; Godoy et al., 2016; Larsolle
& Hamid Muhammed, 2007). Hence, our proposal enriches fungicide ecacy analysis by taking
into account EC
50
among other related variables such as location and time. The suggested
multivariate approach allows monitoring all the variables at once and their relationships over
a four-year period for the evaluation of three fungicides in different plantations distributed in
three provinces of major banana production in Ecuador (Guayas, El Oro and Los Ríos). The
multivariate statistical techniques applied in this study are Multilinear Principal Component
Analysis (MPCA) and Dual STATIS-Parallel Coordinates (DS-PC) strategy, which aim to combine
the available data and conclude a single status about fungicide ecacy in terms of normal
(expected) versus abnormal (unexpected) P. jiensis sensitivity to the three products evaluated.
Materials and Methods
The database used in this study was structured by tables with microscope measurements of
the germ tube longitude derived from fungicide sensitivity analyses. Three products used to
control black sigatoka in banana plantations were evaluated with their corresponding active
ingredients: Boscalid, Fenpropimorph and Pyraclostrobin. Assays were performed on several
locations from three Ecuadorian provinces (Guayas, El Oro and Los Ríos) in four years from
2014 to 2017 as part of the country surveillance for fungicide ecacy analysis. Due to time,
costs and environmental conditions, certain locations were not considered in specic years
and less product concentrations were evaluated, registering missing values in data tables.
Fungicide sensitivity data
Samples were obtained from treated areas with a conventional management program (based
on chemical fungicides). For each analysis, the method suggested by the FRAC for ascospore
germ tube elongation version 1 was applied, with few adaptations specied below.
Sampling: Dry necrotic tissue in grade 6 according to the Stover’s scale was collected from
one block per farm that was exhibiting high disease severity. The necrotic tissue was cut into
small pieces (1-2 cm
2
) and disposed it into paper bags for its transportation to the laboratory,
properly labeled.
53
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Incubation: 5 to 9 pieces of tissue were set using staples, onto lter paper of 9 cm diameter
with the lower leaf surface facing up and put in humidity chambers at 26 ºC for 48 hours.
From every location, a minimum of ve lter papers were prepared: one for control (without
fungicides), and four corresponding to the product concentrations evaluated in parts per
million (C0 = 0, C1 = 0.01, C2 = 0.1, C3 = 1 and C4 = 10).
Culture media: A 2 % water agar was prepared (Bacto Agar Difco® 2%. 20g/L), amended with
the different concentrations of fungicides and poured into petri plates.
Ascospore discharge: Each lter paper with incubated ascospores was placed on the top of
its corresponding petri plate. One hour was allowed for ascospore discharge (extra time may
increase the amount of contaminants), after that, the lter paper was removed and the plates
were incubated for 48 hours at 26 ºC under dark conditions.
Germ tube measurement: A minimum of 30 ascospores were examined for every fungicide
concentration using an inverted microscope with micrometric scale and their germ tube
lengths (micrometers) were registered in tables with the structure shown at gure 1.
Figure 1. Germ tube lengths table. Structure of the tables used to register the germ tube lengths in each essay.
Concentrations are measured in parts per million (ppm) and longitudes in micrometers (µm). Source: author’s
own elaboration.
Data from these tables was used to perform univariate fungicide ecacy analysis in every
year, considering the mode of action of each evaluated product. Results of these analyses are
not part of the investigation subject, therefore, are not shown.
Multivariate statistical analyses
To eliminate the effect over the analysis of ascospores’ natural growth from every location,
mean inhibition percentages were calculated for each concentration per product by comparing
54
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
means of the germinative tube longitudes () with the means of control (). Referred percentages
were computed using the following expression:
Locations were coded with the initials of the Province and a number accompanied as shown
in Table 1.
For organizational purposes, an “assay set” was dened as the collection of matrices with
inhibition data (concentrations and mean inhibition) from the three products evaluated for a
specic year and location. Assay sets conducted in the same location along different years
were considered as different locations labeled as location-years (LYs). A total of 38 assay
sets were dened, indicated in Table 1. Based on empirical criteria, assay set of the locations-
years from 1 to 28 were used as reference, 29 and 30 to test normal behavior, and 31 to 38
for anomalous behavior.
Table 1. Locations-Years codes and provinces. Codes are established by the year, with the initial of the province
and a number ID
Code Province Code Province Code Province
1 2014-EO 01 El Oro 15 2016-LR 05 Los Ríos 29 2015-EO 04 El Oro
2 2014-LR 01 Los Ríos 16 2017-EO 01 El Oro 30 2016-EO 02 El Oro
3 2014-LR 02 Los Ríos 17 2017-EO 04 El Oro 31 2014-EO 03 El Oro
4 2015-EO 02 El Oro 18 2017-EO 02 El Oro 32 2014-LR 12 Los Ríos
5 2015-EO 03 El Oro 19 2017-EO 05 El Oro 33 2014-LR 03 Los Ríos
6 2015-GY 01 Guayas 20 2017-GY 05 Guayas 34 2014-LR 05 Los Ríos
7 2015-GY 02 Guayas 21 2017-GY 06 Guayas 35 2014-LR 13 Los Ríos
8 2015-GY 03 Guayas 22 2017-LR 03 Los Ríos 36 2015-GY 05 Guayas
9 2015-LR 03 Los Ríos 23 2017-LR 07 Los Ríos 37 2015-LR 10 Los Ríos
10 2015-LR 04 Los Ríos 24 2017-LR 08 Los Ríos 38 2017-LR 12 Los Ríos
11 2015-LR 05 Los Ríos 25 2017-LR 04 Los Ríos
12 2016-EO 04 El Oro 26 2017-LR 09 Los Ríos
13 2016-GY 04 Guayas 27 2017-LR 10 Los Ríos
14 2016-LR 06 Los Ríos 28 2017-LR 11 Los Ríos
Source: author’s own elaboration.
55
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Data from every matrix was plotted as sequential dots and connected by straight lines, then
a geometric Chaikin smoothing was performed (Chaikin, 1974; Rankin, 2013). This algorithm
produces an interpolating curve that allows estimating the mean inhibition in a range with
dose values according to the analyzed product, as shown in Table 2. Then, for each product, a
vector was generated (Figure 2) with estimated values expected to be around 50 % of inhibition.
Table 2. Ranges of concentrations evaluated by product. Every product is associated to its chemical group.
Minimum and Maximum concentrations in the range are presented
Product Chemical group/Group name Target site Minimum Maximum
P1 Carboxamide Succinate-Dehydrogenase Inhibitors (SDHIs) 1 2.14
P2 Amine Sterol Biosynthesis Inhibitors (SBIs) 2 4
P3 Strobilurin Quinone outside Inhibitors (QoIs) 1 2.5
Source: author’s own elaboration.
Dose values in the referred ranges, having elements within the interval, were generated as a
succession according to the following expression:
D = {d
1
,d
2
, … ,d
i
, … ,d
n
} , d
i
= exp (ln(a) + ((i-1) / (n-1))( ln(b)-ln(a)))
Binding the vectors from each product in every assay set, this algorithm creates a matrix of
Inhibition vs product, for every location-year (Figure 2). All generated matrices were stacked
one after another to conform a three-way data block used for statistical analysis (Figure 2).
Figure 2. Three-way structuration. Processing of original matrices from the different products creates -vectors
of LY tables, which conform the three-way array used in the analysis. Source: author’s own elaboration.
56
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
MPCA & DS-PC: Both methods are suitable for processing and generating graphic
representations of three-way data arrays conformed as sets of tables including mean inhibition
values previously computed from the dataset available in the established time frame (2014
to 2017). Every table was preprocessed, rst centering with the global mean, then scaling with
global standard deviation and normalizing dividing by the number of observations; global mean
and standard deviation per variables were calculated from reference information. Later, MPCA
and DS-PC methods were computed (Abdi et al., 2012; Inselberg & Dimsdale, 1990; Nomikos
& MacGregor, 1994, 1995; Ramos-Barberán et al., 2018; Rousseeuw, Ruts & Tukey, 1999).
Comparison: In order to validate and compare DS-PC strategy, MPCA is used as a reference. In
these methods, the three ways often are: observations/times, variables and batches; analogous,
the data block in this study has: observations, products and location-years. As mentioned in
the multiway structuration, three groups of LYs were conformed: one reference set to state
the typical behavior, and the others to represent “normaland “anomalous” behavior to test
the discriminative ability of the methods.
Principal components factor scores for LYs from reference set were computed in MPCA and
Interstructure from DS-PC according to Lu et al. (2008) and Ramos-Barberán et al. (2018)
respectively. Then, the bagplot-based control charts (CC) were set up as in DS-PC method
for both cases; after that, the normal and anomalous LYs were projected to examine their
positions in the CCs. Additionally, CCs for each product were dened and LYs were projected
using the STATIS Dual intrastructure. Finally, Parallel Coordinates plots for the normal and
anomalous LYs were created to conrm the behavior of the LY and to investigate the mean
inhibition values by products. All calculations were performed in the open source statistical
programming software R.
Results
MPCA
The Principal Components scores plot from MPCA and its control region (Figure 3) shows that
every reference and normal LYs are inside; on the contrary, anomalous locations are outside
the region. Normal LYs stand for expected P. jiensis sensitivity while anomalous locations
refer to unexpected sensitivity due to larger distances to the centroid in comparison to the
reference ones.
57
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Figure 3. MPCA control chart. Color coding for projected LYs is the following: yellow for reference, blue for
normal and red for anomalous. Source: author’s own elaboration.
DS-PC strategy
Figure 4 shows DS-PC Interstructure. Anomalous LY are clearly different from the reference
ones. The distances to the centroid, are larger than those observed in MPCA. Also, as in MPCA,
normal and reference LYs remain inside the control region, while anomalous LYs fall outside.
Spatial conguration of anomalous LYs are similar, comparing to MPCA results.
Figure 4. Interstructure control chart. To the left, a zoom image of the reference region. To the right, general
view of all the projected LYs; PC1 and PC2 values in the chart from the right is scaled to match with the gure
3. Color coding for projected LYs is the following: yellow for reference, blue for normal and red for anomalous.
Source: author’s own elaboration.
58
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Most of the LYs outside belong to Los Ríos province, which is likely because of prevailing
temperatures between 25-30 ºC and relative humidity around 95 % favored the life cycle of
P. jiensis, which is strongly determined by weather and microclimate (de Jesus et al., 2008;
Marín et al., 2003). Like many foliar fungal pathogens, P. jiensis ascospores are dispersed
by wind and require a wet leaf surface or very high relative humidity to germinate and infect
the leaf, and the rate of germination and infection during wet or humid periods is governed by
temperature (Uchôa et al., 2012). These environmental conditions, associated to location and
time, promote an ideal scenario for banana plants development and black sigatoka growth
and dispersion (Jacome & Schuh, 1992; Pérez-Vicente et al., 2000).
When it comes to the products CCs (Figure 5), all anomalous LYs remain outside, which means
unexpected P. jiensis sensitivity to at least one of the fungicides tested. Nevertheless, the
distances to the centroid are different. In P1 CC and P3 CC, LYs 35 and 36 stay far from the
centroid, but in P2 CC, stay close, still outside, but relatively close. LY 37 is far in P1 CC and
P2 CC, but in P3 CC gets a little closer. In general, anomalous LYs tend to get far in a specic
direction. However, in P2 CC, LY 31 and 38 appear in the opposite direction, in comparison to
the other anomalous LYs, although still outside.
59
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Figure 5. Intrastructure control charts, by product (From top to bottom: P1CC, P2CC, P3CC). All anomalous
LYs appear outside the control regions, nevertheless, the distribution observed is different. Color coding for
projected LYs is the following: yellow for reference, blue for normal and red for anomalous. Source: author’s
own elaboration.
Parallel Coordinates show mean inhibition values by product from every LY in black, and
reference LYs in gray. A red line connects the global means by product. In gure 9, normal LYs
(29 and 30) are inside the gray zone, as expected. LY 29 is above the mean for products P1
and P2; on the other hand, LY 30 is under the mean for product P1.
Figure 6. Parallel coordinates for normal LYs. Black lines, which represent normal LYs, stay inside the zone
conformed by gray lines of the reference LYs. Source: author’s own elaboration.
Anomalous LYs are clearly different, as observed in gure 10. LY 34 has the greater dispersion
in product P2, but still stays inside the gray region. Most of the anomalous LYs (except 33
and 37), show normal values for product P2. In addition, LY 31 and 38 have inhibition values
above the global mean. For product P1, all except LY 33, present low inhibition values. And all
the eight anomalous LYs have low inhibition values for product P3.
60
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Figure 7. Parallel coordinates for anomalous LYs. Black lines, which represent anomalous LYs, fall outside the
zone conformed by gray lines of the reference LYs. Source: author’s own elaboration.
Analysis by product (variable-wise analysis from DS-PC strategy) shows LYs more far or
close depending on the difference with reference values as shown by Parallel Coordinates,
and also, a shift in the direction of projected points related to the kind of differences in the
values (higher or lower), which is the case of LYs 31 and 38, whose positions in P2 CC are in a
different direction due to the high values presented in product P2. LYs like 35 and 36, which
stay closer to the control region in P2 CC, rather than in P1 CC and P3 CC, present the same
behavior in Parallel Coordinates, this conrms that the LYs is anomalous, but the product P2
isn’t the cause of this anomaly.
61
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Discussion
The threat for resistance development in important pathogens to FRAC groups is one of
the biggest concerns in banana crops. The resistance of P. jiensis in banana plantations to
fungicides provides compelling evidence that integrated crop management should include
fungicide ecacy analysis and other components related to locations and years. Nevertheless,
the widely used of EC
50
allows to evaluate fungicide ecacy only in one dimension, losing other
components behaviors needed to develop effective strategies for the management of black
sigatoka in banana crops (de Jesus et al., 2008; Guzmán et al., 2013; Marín et al., 2003; Martínez
et al., 2011). MPCA and DS-PC performed a correct discrimination of the location-years with
normal and anomalous behavior, which validates the capability of the multivariate statistical
analyses to monitor black sigatoka in terms of fungicide ecacy within banana plantations.
Discriminative ability of DS-PC strategy is better than MPCA, comparing the patterns shown
in the interstructure control charts (IS CCs), where normal LYs (expected sensitivity) exhibited
a consistent grouping tendency and abnormal LYs (unexpected sensitivity) displayed greater
distance from the control region. Although MPCA has shown better performance in variables
proles depending on time, instead of observations, DS-PC analysis possibilities are wider
than MPCA in the present study where intrastructure control charts (CO CCs) and parallel
coordinates revealed tested fungicides performance for each LY (location-year) (Nomikos &
MacGregor, 1994; Ramos-Barberán et al., 2018).
The main advantage of the multivariate approach (MPCA and DS-PC) discussed herein for
fungicide ecacy analysis is the possibility of capture the variability between locations, years,
products and different observations to conclude a single status of “normalor “anomalous
(expected versus unexpected P. jiensis sensitivity). Furthermore, the stated use of several
inhibition values is more accurate than studying a single estimated score, which leave out the
relationships of sensitivity with other associated variables that provide a better understanding
of pathogens and diseaseslocal dynamics over time. Similarly, it would be like comparing the
information provided by the full video versus a frame.
Lack of abnormal trend along the years allows to conclude that anomalous behavior of
the fungus can’t be generalized because genetic material affecting banana plantations in
Ecuador is continuously renewing. For instance, location LR 10 belonging to Los Ríos province
exhibited in 2015 abnormal behavior changing to normal condition 2 years later. Thus, present
analysis couldn’t compile enough statistical evidence to assure that low sensibility trend is
persistent, therefore, it can’t be said that resistance is showing up in the studied areas. Yet, if
we take a deeper look into anomalous location-years through parallel coordinates, in 7 cases
carboxamides and strobilurins didn’t perform eciently, meaning that black sigatoka showed
decreased sensitivity to these fungicides, while amines exhibited the expected response. If
this trend continues, the behavior stated by previous studies regarding amines would remain
similar in Ecuador (Marín et al., 2003; Martínez & Guzmán, 2011; Martínez et al., 2011).
62
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
In Ecuador, studies about triazoles and organic fungicides ecacy evaluation have been carried
out, by means of percentages of infection and lethal concentration-50 (LC
50
) (Caicedo, 2015;
Chávez, 2012; Diaz-Trujillo et al., 2018; Sabando, 2015), therefore it is not possible to make
comparisons with the present study. Monitoring of ecacy can thus be used as an indicator
of the possible development of resistance, with any reductions in ecacy below an agreed
threshold value (Russell, 2004). For further research, it is necessary that Ecuador, as a major
banana exporter, starts collecting evidence of ecacy data to build a baseline in terms of
conventional univariate analysis but also, multivariate ones.
Conclusion
In this study, a fungicide sensitivity of Pseudocercospora jiensis, the causal agent of black
sigatoka, was evaluated on bananasplantations in three provinces of Ecuador: Guayas, El Oro,
and Los Ríos. This approach took into account the threshold conveyed by inhibition percentages,
related to the EC
50
, along with locations and years allowing the multivariate analyses. Both
methods MPCA and DS-PC discriminate correctly between the normal and anomalous
conditions within plantations along years. The novel application of these multivariate analyses
posed a signicant advantage in terms of monitoring and control within plantations. However,
the ability of the DS-PC strategy for exhibiting better signaling of anomalous plantations and
performing variable-wise analysis was evidenced in an easier time-saving graphical framework.
References
Abdi, H. et al. (2012). STATIS and DISTATIS: optimum multitable
principal component analysis and three way metric
multidimensional scaling. Wiley Interdisciplinary
Reviews: Computational Statistics, 4(2), 124-167.
Aguirre, S.E., Piraneque, N.V. & Rodríguez, J. (2015). Relationship
between the nutritional status of banana plants
and black sigatoka severity in the Magdalena
region of Colombia. Agronomía Colombiana,
33(3), 348-355.
Ali, A. (2011). Analysis of Multivariate Agricultural Data BT.
In Lovric, M. (Ed.), International Encyclopedia
of Statistical Science (pp. 41-44). Heidelberg,
Germany: Springer Verlag GmbH.
63
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Barakat, A. et al. (2017). Multivariate analysis and GIS-based
soil suitability diagnosis for sustainable intensive
agriculture in Beni-Moussa irrigated subperimeter
(Tadla plain, Morocco). Modeling Earth Systems
and Environment, 3(1), 3.
Bolaños, A. (2006). Evaluación de dos fungicidas
sistémicos para el combate de la sigatoka negra
(Mycosphaerella jiensis var difformis) en una
plantación de banano currare (Musa AAB) en la
zona de San Carlos. Cartago, Costa Rica: Instituto
Tecnológico de Costa Rica.
Brent, K.J. & Hollomon, D.W. (2007). Fungicide Resistance
in Crop Pathogens: How can it be managed?
Brussels, Belgium: Fungicide Resistance Action
Committee.
Caicedo, E. (2015). Fungicidas orgánicos para el control de la
Sigatoka Negra (Mycosphaerella jiensis), en el
cultivo de banano (Musa AAA) Valencia-Los Ríos.
Quevedo, Ecuador: Universidad Técnica Estatal
de Quevedo.
Chaikin, G.M. (1974). An Algorithm for High-Speed Curve
Generation. Computer Graphics and Image
Processing, 3, 346-349.
Chávez, T. (2012). Análisis genético de la resistencia a
triazoles en aislados de Mycosphaerella jiensis
para poblaciones de Ecuador. Guayaquil, Ecuador:
Escuela Superior Politécnica del Litoral.
de Jesus, W. et al. (2008). Worldwide geographical distribution
of black sigatoka for banana: Predictions based
on climate change models. Scientia Agricola,
65(Special Issue), 40-53.
Diaz-Trujillo, C. et al. (2018). A new mechanism for reduced
sensitivity to demethylation-inhibitor fungicides
in the fungal banana black Sigatoka pathogen
64
Multivariate analyses to determine fungicide ecacy on Ecuadorian bananas for consumption
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Pseudocercospora fijiensis. Molecular Plant
Pathology, 19(6), 1491-1503.
FAO. (2018). Banana Facts and Figures. Retrieved from http://
www.fao.org/economic/est/est-commodities/
bananas/bananafacts/en/#.XHlSJohKjD4.
Godoy, R.C.B. et al. (2016). Physico-chemical characterization
of banana varieties resistant to black leaf streak
disease for industrial purposes. Ciência Rural,
46(9), 1514-1520.
Guzmán, M., Orozco-Santos, M. & Pérez Vicente, L. (2013).
Las enfermedades de las manchas de las hojas
del banano: dispersión, impacto y evolución de
las estrategias de manejo en América Latina
y El Caribe. In XX Reunião Internacional da
Associação para a Cooperação em Pesquisa
e Desenvolvimento Integral das Musáceas
(Bananas e Plátanos) (pp. 98-116). Fortaleza,
Brasil: ACORBAT.
Lu, H. et al. (2008). MPCA: Multilinear Principal Component
Analysis of Tensor Objects. IEEE Transactions on
Neural Networks, 19(1), 18-39.
Inselberg, A. & Dimsdale, B. (1990). Parallel Coordinates: A
Tool for Visualizing Multi-Dimensional Geometry.
In Proceedings of the First IEEE Conference on
Visualization, San Francisco, USA.
Jacome, L.H. & Schuh, W. (1992). Effects of Leaf Wetness
Duration and Temperature on Development
of Black Sigatoka Disease on Banana Infected
by Mycosphaerella fijiensis var. difformis.
Phytopathology, 82, 515-520.
Larsolle, A. & Hamid Muhammed, H. (2007). Measuring crop
status using multivariate analysis of hyperspectral
eld reectance with application to disease severity
and plant density. Precision Agriculture, 8(1), 37-47.
65
José Ascencio-Moreno, Miriam Vanessa Hinojosa-Ramos, Francisco Vera, Omar Ruiz-Barzola,
María Isabel Jiménez-Feijoó, María Puricación Galindo-Villardón, Miriam Ramos-Barberán
Espirales. Revista multidisciplinaria de investigación cientíca, vol 4, No. 34
Julio-septiembre 2020. e-ISSN 2550-6862. págs 49-66
Luna-Moreno, D.