Development stage-dependent susceptibility of cocoa fruit to pod rot

0 downloads 0 Views 246KB Size Report
Sep 23, 2012 - to pod rot caused by Phytophthora megakarya. P. Takam Soh & M. ..... easy to observe) and then >6 (because after that num- ber the lesions ...
Eur J Plant Pathol (2013) 135:363–370 DOI 10.1007/s10658-012-0092-4

Development stage-dependent susceptibility of cocoa fruit to pod rot caused by Phytophthora megakarya P. Takam Soh & M. Ndoumbè-Nkeng & I. Sache & E. P. Ndong Nguema & H. Gwet & J. Chadœuf

Accepted: 10 September 2012 / Published online: 23 September 2012 # KNPV 2012

Abstract Pod rot causes up to 30 % losses in world cocoa production. In order to predict the risk evolution of disease, it is important to take into consideration the developmental stage of fruits. In fact, it has been shown that the risk of attack by pod rot depends amongst others on the developmental stage of fruits. We proposed here to estimate the susceptibility at different stages. Susceptibility of fruit to disease was investigated at three fruit developmental stages (cherelle, young pod and adult pod); disease severity was assessed in laboratory conditions, on detached, artificially inoculated fruits, while disease incidence was assessed in the field, under natural inoculum pressure. In both assessment fruits at the cherelle stage were the most susceptible whereas the young and adult fruits were equally susceptible. The vertical position of the fruits on the tree did not influence their susceptibility. Estimates of the fruit susceptibility and of the infectious potential were derived from the severity and incidence measurements, using a model P. T. Soh : M. Ndoumbè-Nkeng IRAD, BP 2123, Yaoundé, Cameroon P. T. Soh : E. P. N. Nguema : H. Gwet ENSP, BP 8390, Yaoundé, Cameroon P. T. Soh : J. Chadœuf (*) INRA, Biostatistique et Processus Spatiaux, 84914 Avignon, France e-mail: [email protected] I. Sache INRA, UR Bioger-CPP, 78850 Thiverval-Grignon, France

assuming that the number of spores on a fruit follows a Poisson distribution with the mean, the density of spores per fruit as the parameter. The estimated parameter values allowed the evaluation of the probability of attack of a fruit by the disease, which could be implemented in a disease warning system. Keywords Infectious potential . Model . Poisson variable

Introduction Pod rot caused by the oomycete Phytophthora megakarya is the most devastating disease of the cocoa plant, claiming up to 30 % of the world cocoa harvest (Ndoumbe-Nkeng et al. 2009). In a recent review, Acebo-Guerrero et al. (2012) stressed that harvest losses due to pod rot may range from 20 % to 30 %, although individual farms may suffer from 30 % to 90 % harvest loss; the disease can be especially severe in West and Central Africa, causing up to 64 % of losses in plantations. In experimental situations, pod loss can be efficiently reduced by fungicide control of the disease (Akrofi et al. 2003). Recommendations to farmers have been issued accordingly (Berry 1999). However, a multidisciplinary survey of pesticide use in Cameroon villages supported the conclusion that the current use of fungicide is not a successful disease management strategy for most farmers (Sonwa et al. 2008). The authors of the survey advocated for a

364

change in management practices and highlighted “the need to further develop locally acceptable Integrated Pest Management (IPM) strategies”. In this context, the number of fungicide sprays should be reduced by targeting applications to critical stages of crop and disease development. In a field study, complementing the already documented evidence of developmental stage-dependent fruit susceptibility to pod rot (Ndoumbe-Nkeng 2002), Deberdt et al. (2007) have shown that: (i) there was a link between the fruit developmental stages and pod rot incidence; and (ii) immature fruits were more susceptible. However, the study was not designed to perform a quantitative, accurate assessment of the fruit susceptibility on different stages. In field conditions, fruit susceptibility to pod rot would depend on the genetic composition of the pathogen population (Ducamp et al. 2004), on the climatic conditions (Ndoumbe-Nkeng et al. 2009), and on the infectious potential. The aim of this study is to quantify the susceptibility of the cocoa fruit at different developmental stages submitted to artificial or natural inoculum load, combining experimentation and modelling. Susceptibility of a fruit to a disease is usually assessed in controlled conditions as the infection efficiency that is the ratio of lesion density to inoculum density (e.g., Xu and Robinson 2010). However, the assessment of infection efficiency is tedious in controlled conditions and hardly feasible in the field. Therefore, we proposed a model that allowed the estimation of the non-measured susceptibility of cocoa fruits to pod rot disease as a function of the developmental stages using disease severity or incidence observations. In this model, we assumed that: (i) the number of spores deposited on a fruit follows a Poisson variable of mean the infectious potential, that is the density of spores per fruit; and (ii) a spore on a fruit can attack it with a probability that represents the fruit susceptibility to disease (Soubeyrand et al. 2007). We first estimated the model using disease severity data obtained under laboratory controlled conditions. Secondly, we estimated the model using disease incidence data obtained in the field. Thirdly, we compared the susceptibility obtained in laboratory and field data. Finally, we discuss the implications of our results for cocoa protection and the wider interest of our modelling framework for plant disease epidemiology.

Eur J Plant Pathol (2013) 135:363–370

Material and methods Plant material All fruits used in the study were collected from amelonado trees, in a farmland located in Mbankomo (3°78′0″N, 11°38′0″E; altitude, 730 m above sea level), Centre region of Cameroon. Characterized by two dry seasons and two rainy seasons, this region is a predominantly P. megakarya endemic area (Nyasse et al. 1999). Laboratory assessment of disease severity Ninety fruits were sampled on the 30th of June 1999, 17th of July 1999 and 7th of August 1999, thus 30 fruits per day. Each series of fruits comprised 10 cherelles (fruits of less than 4 cm in length), 10 young pods (fruits of more than 4 cm in length but not at adult stage yet) and 10 adult pods (fruits with adult length, but not yet ripe). The fruits were placed in sealed bags and brought immediately to the laboratory. Twenty four hours after harvest, fruits were rinsed twice with distilled water to remove spores likely to be present on their surface. The fruits were allocated, according to a fully randomized design, inside inoculation tanks. Sheets of tissue filter paper were inserted inside each tank to maintain a homogenous humidity. A moderately aggressive isolate of P. megakarya maintained in the Plant Pathology Laboratory at the IRAD Nkolbisson Research Centre (Cameroon), was used in this study. It was isolated from a cocoa plantation situated in the Centre region of the country, and it has been maintained in the laboratory by successive transfers on 1.5 % pea-based agar medium. To maintain its aggressiveness, the isolate has been periodically inoculated onto cocoa pods (Efomban et al. 2011; Nyassé et al. 1999). Five days after incubation in the closed tanks, the number of lesions per fruit was counted and fruits were ranked according to the following scale: 0 0 no lesion, 1 0 1–6 lesions, 2 0 >6 lesions per fruit. Field assessment of disease incidence In the same plot, 15 trees were randomly selected during two years (1999 and 2001), and each tree was divided into three levels (bottom: 0–0.5 m, middle: 0.5–1.5 m, top: >1.5 m). In each level, four randomly

Eur J Plant Pathol (2013) 135:363–370

365

selected fruits were tagged; only fruits at the cherelle stage were considered. The size of the fruit cohort was then 159 in 1999 and 169 in 2001. Observations were performed on the tagged fruits each Monday of every week for 20 weeks, starting on 26 April 1999 and 23 April 2001, respectively. At each observation day, the developmental stage of each fruit (cherelle, young pod, adult pod), and its disease status (0 0 healthy fruit, 1 0 fruit having at least one pod rot lesion) were recorded. On each observation day, infected fruits were removed so that the observed fruits on a given day were those that had not been found infected the week before. Modeling the number of lesions on a fruit We denote i for a fruit and j for the series of observations, 1≤ j≤D, where D03 for laboratory data and D0 20 for field data. For a fruit i belonging to a series j, we denote Ei, j its developmental stage (1 0 cherelle, 2 0 young pod, 3 0 adult pod) and Ni,j its number of lesions. We assume that: (i) the number of spores on a fruit follows a Poisson distribution with mean the infectious potential Uj, that is the density of spores per fruit; and (ii) a spore on a fruit i of series j can penetrate the fruit and causes disease (hereafter summarized as “attack”), with a probability SEi; j which represents the stagedependent fruit susceptibility. It follows from (i) and (ii) that the number Ni, j of spores that attacks fruit i belonging to the series j follows a Poisson variable with mean sEi; j Uj . For a fruit i belonging to a series j and of stage Ei, j, the probability to observe a class of lesions represented by the interval [n1, n2[ is thus given by: PrðNi;j 2 ½n1 ; n2 ½Þ ¼

p¼n 2 1 X

expðsk Uj Þ

p¼n1

ðsk Uj Þp : p!

ð1Þ

For the special case where there is no lesion on a fruit, this probability reduces to: PrðNi; j ¼ 0Þ ¼ expðsk Uj Þ:

Tibshirani 1993). We used the bootstrap method to avoid the approximation from asymptotic maximum likelihood estimation theory because the sample size was small. We used the likelihood-ratio test (DacunhaCastelle and Duflo 1992) to test: (i) whether the fruit susceptibility actually depends on the developmental stage of the fruit; (ii) whether the fruit susceptibility to attacks depends on the tree level; (iii) whether the infectious potential depends on the series of observation; (iv) whether the fruit susceptibility from the laboratory data is different from that from the field data; and (v) whether the fruit susceptibility depends on the year of observation. To validate the model estimated from laboratory data we computed, for each series and for each stage, 1000 simulations of severity classes to get simulated frequencies of each severity class for each stage and for each series under the model. The p-value of each observed severity class for each stage and for each series was calculated under the model by using simulated frequencies (Peter and Gill 2007). In a further step, we validated globally the model using the chisquare distance test (Monfort 1971) by comparing the observed and simulated severity frequency classes. Let us recall here that the same model was applied on the field data by using only two classes of lesions: 0 (for 0 lesion) and 1 (for more than one lesion). In that case, the model took into consideration the level of the tree and was given by PrðNi; j ¼ 0Þ ¼ expðskh Uhj Þ

ð3Þ

where h denoted the level of tree in which the fruit was selected. The results were obtained using the R environment (R Development Core Team 2008). We used the nlm function to maximize the likelihood function.

Results

ð2Þ Laboratory assessment of disease severity

We used the likelihood method (Monfort 1971) to estimate the susceptibility and the infectious potential k¼3 P under the normalization condition sk ¼ 1, which k¼1

ensures the identifiability of the model. The confidence intervals for susceptibility and infectious potential were estimated by using parametric bootstrap (Efron and

In the laboratory experiment, all inoculated fruits except one adult pod exhibited disease symptoms (Fig. 1). Most cherelles (28/30) had more than six lesions per fruit, the maximum severity class used in this study. The fruits at the two other stages were equally distributed among the two severity classes across the experiment.

366

Eur J Plant Pathol (2013) 135:363–370

Fig. 1 Classified severity of cocoa pod rot (three classes: no lesion, 1–6 lesions, >6 lesions per fruit) on fruit inoculated in the laboratory at three developmental stages (CH: cherelle, YP: young pod, AP: adult pod), three series of 30 fruits each

However, there was a strong series effect on disease severity, both in young and adult pods. Within a series, most young pods were in the same severity class (for instance, 8/10 in the highest class in series 1 vs. 9/10 in the lowest class in series 3). This effect was less pronounced in adult pods for instance, in the third series, fruits were equally distributed among the two severity classes.

During the first five weeks, the evolution of disease incidence was similar for the three levels with a mean maximal incidence of around 20 % (Fig. 3). Young pods were present at the beginning of observations and their number peaked at week 5. Disease incidence exhibited a high peak at the end of the study for each level, when only small adult pods remained.

Field assessment of disease incidence

Estimation of fruit susceptibility and of infectious potential

In 1999, the observed duration of the cherelle stage was two-times shorter (max. five weeks) than for the two other stages (max. 10–11 weeks; Fig. 2). Moreover, the dynamics of fruit development differed among the tree levels; in the bottom levels, the distribution of young pods showed a peak more pronounced than in the middle and the top levels. Many fewer adult pods were produced in the bottom level. During the first five weeks, the evolution of disease incidence was similar for the three tree levels with a maximal incidence of 20 % (Fig. 3). For the next ten weeks, disease was mostly recorded on young pods (there were no longer cherelles, but a few adult pods were already present in the two upper levels); disease incidence increased with decreasing tree level. In the last stages of the epidemic, disease incidence on mature pods was particularly high in the bottom level. In 2001, the observed duration of the cherelle stage was around five weeks (Fig. 2). The dynamics of fruit development in young pods stage did not differ among the three levels. Conversely, the dynamics of fruit development in adult pods stage differed among the levels. In the bottom levels, the distribution showed a pronounced peak at week 6; in the middle and the top levels, the peak in the distribution was delayed to week 10.

Estimation with laboratory data Fruit susceptibility (sk) estimated from the laboratory data was significantly different between the fruit stages (P 1.5m 60

60

60

level 1 : 0 − 0.5m

CH YP AP

50

50

50

CH YP AP

5

10

15

20

40 30 10 0

0 0

20

Number of heslty fruits

40 30 10

20

Number of heslty fruits

40 30 20 0

10

Number of heslty fruits

CH YP AP

0

5

weeks

10

weeks

15

20

0

5

10

15

20

weeks

Fig. 2 Number of healthy cherelles (CH), young pods (YP) and adult pods (AP) recorded each week at three tree levels in the field experiment for two years (top, 1999; bottom, 2001)

effect of the position of the tree on susceptibility and on infectious potential. Estimation with laboratory and field data Fruit susceptibility estimated from the field and laboratory data was significantly different between the fruit stages (P0.05) from the susceptibility of adult pods (0.295, s.d. 0.095). In 1999, the infectious potential increased from the beginning to the middle of the campaign, peaking at week 11 and then decreasing until the 20th and last day of observation (Fig. 4). In 2001, the infectious potential peaked at the last week of observation (Fig. 4).

Discussion In this study, the susceptibility of cocoa fruits depended on the developmental stage but not on the position of the fruit on the tree. Other studies have shown that the

susceptibility of cocoa fruit depended on the position of the fruit on the tree (Martijn ten Hoopen et al. 2012). This contradiction might be explained by the age of the trees. Our data were collected in a young farm whereas the two previous studies used fruits from older farms. On very old trees, increasing amounts of inoculum are deposited on the tree trunk. The presence of this inoculum could then influence the risk of attack at each level of tree. In the present study, no difference of attack was detected between the three levels because the data came from relatively young farms. Moreover, the pod susceptibility did not depend on the year of data collection. The estimates of the susceptibility showed that, when submitted to artificial or natural inoculum, (i) cherelles were more susceptible to pod rot than young and adult pods and (ii) there was no significant difference in susceptibility between the young pods and adult pods. These results suggest that the disease status of a fruit does not strongly depend on environmental conditions. The laboratory data were obtained on fruits classified according to the number of lesions. In fact, the best way to estimate our model would be to use the

368

Eur J Plant Pathol (2013) 135:363–370

5

10

15

20

100 80 60

Fruit disease incidence

20

CH YP AP

0

20 0

0 0

40

80 60

Fruit disease incidence

60 40

CH YP AP

40

80

CH YP AP

20

Fruit disease incidence

level 3: 1 − 1.5m

100

level 2: 0.5m − 1m

100

level 1 : 0 − 0.5m

0

5

weeks

10

15

20

0

5

weeks

10

15

20

weeks

Fig. 3 Cocoa pod rot incidence (percentage of infected fruits) on cherelles (CH), young pods (YP) and adult pods (AP) recorded each week at three tree levels in the field experiment for two years (top, 1999; bottom, 2001)

lesion counts. However, it was not possible to count all the lesions on fruits with sufficient precision. In such a case, it is usually recommended to classify the fruits in three classes: 0 lesion, 1–6 lesions (which are easy to observe) and then >6 (because after that number the lesions could superimpose on each other). The estimation of susceptibility could have been biased if we had used only the field data, due to the presence of other diseases in the field that can mask attacks by pod rot. Furthermore, fruits observed in the field were submitted to a non-controlled, natural

inoculum and experienced more climatic variations (Deberdt et al. 2007). Accordingly, we have not estimated the susceptibility using the data collected in year 2000 because that year, 32 % of fruits were wilted while only one fruit was actually attacked by pod rot. Our conclusions confirmed preliminary results obtained by Ndoumbe-Nkeng (2002) using a nonparametric Kruskal-Wallis test. However, such a nonparametric test requires a large sample, which is not easy to obtain in most experimental conditions. Our parametric model, validated on the data set obtained in

Table 1 Chi-square assessment (p-value) of the model of cocoa pod rot severity on fruits at the cherelle.young pod and adult pod development stages.three series of laboratory experiments (s1. s2. s3). The model is not rejected when P>0.05 Lesions

Developmental stages Cherelle

Young pod

Adultpod

s1

s2

s3

s1

s2

s3

s1

s2

s3

0 lesion

0.5

0.5

0.5

0.51

0.5

0.56

0.01

0.5

0.56

1 to 6 lesions

0.64

0.51

0.3

0.95

0.24

0.1

0.21

0.74

0.9

>6 lesions

0.36

0.49

0.7

0.04

0.76

0.88

0.92

0.25

0.08

Eur J Plant Pathol (2013) 135:363–370 Year : 2001

5 4 3 2 0

1

estimated infectious potential

4 3 2 1

estimated infectious potential

5

Year : 1999

0

Fig. 4 Time course of infectious potential of Phytophthoramegakarya as estimated by the model (solid line: estimated values; dashed lines: 95 % confidence band)

369

5

10

weeks

laboratory conditions, allowed us to estimate fruit susceptibility to pod rot in field conditions, assuming that fruits were submitted to the same infectious potential at a given date. Moreover, our model allowed us to estimate the fruit susceptibility and infectious potential, which are biological parameters of epidemiological significance not easy to measure. A simple, practical recommendation to be drawn from this work is to focus the use of fungicides at the beginning of the farming season, when most fruits are at the cherelle stage, the most susceptible to disease. Fungicide sprays could therefore be recommended early in the season, when infection levels are not too high; later on, the removal of attacked fruits (phytosanitary pod removal: Soberanis et al. 1999; Ndoumbe-Nkeng et al. 2004; Opoku et al. 2007) could be recommended to decrease the amount of inoculums in the plots. Our model was based on the assumption that spores can infect the fruit independently of each other according to a Poisson distribution; this assumption is not specific to cocoa pod rot and can be transferred to other airborne fungal diseases. Fruit susceptibility has been reported to depend on fruit age (Ando et al. 2009; Xu et al. 2009) or maturity (Chillet et al. 2007; Moral et al. 2008; Xu and Robinson 2010) in several other diseases. An epidemiological model should therefore consider fruits with variable susceptibility, especially depending on the fruit developmental stage. In our model, based on the same assumptions whatever the data, the susceptibility and infectious potential parameters have the same biological meaning and their estimates are consistent in the laboratory and in

15

20

5

10

15

20

weeks

the field. Accordingly, depending on the required level of precision and practical conditions, these parameters can be estimated, either in the laboratory or on the field, from standard assessment of disease incidence or severity. The proposed model allows the estimation of the probability to be attacked at each developmental stage of fruit. But it does not take into account the time spent at each developmental stage. This work can then be considered as a preliminary step for a proportional hazard model that would take into account the time spent at each stage in order to estimate the risk of attack by disease knowing the developmental stage. The estimation of lifetime on our data (field data) needs more attention because the data do not allow use of the standard method of Kaplan and Meier (1958) in order to estimate the lifetime at each developmental stage. In fact due to the weekly observations, the data present three specificities: (i) the entry date at each stage is not observed; (ii) the changing date at each stage is not observed; and (iii) for such fruits, the disease occurred in the interval in which the fruits change the stage and both events were observed. The estimation of lifetime by taking into account these specificities is actually done by the same authors of this work. Furthermore, even if the lifetime was estimated at each developmental stage, using a Cox proportional model (Cox 1972), it would need the transformation of the laboratory data in two classes (1 0 0 lesions and 2 0 more than one lesion). But in this case, we would lose information on the laboratory data and the method would be less efficient.

370 Acknowledgments IRAD (Institute of Agricultural Research for Development, Cameroon), INRA (National Institute of Agricultural Research, France), ENSP (National Polytechnic Institute, Cameroon) and SCAC (Service for Cooperation and Cultural Action, French Embassy in Cameroon) are acknowledged for the support provided. Sincere thanks are expressed to the staff of the Plant Pathology Laboratory of IRAD who carried out the field and laboratory trials, as well as to Mr. Awah N. Richard who critically reviewed the manuscript.

References Acebo-Guerrero, Y., Hernandez-Rodriguez, A., Heydrich-Pérez, M., El Jaziri, M., & Hernandez-Lauzardo, A. N. (2012). Management of black pod rot in cocoa (Theobroma cocoa L.): a review. Fruits, 67, 41–48. Akrofi, A. Y., Appiarh, A. A., & Opokua, I. Y. (2003). Management of Phytophthora pod rot disease on cocoa farms in Ghana. Crop Protection, 22, 469–477. Ando, K., Hammar, S., & Grumet, R. (2009). Age-related resistance of diverse cucurbit fruit to infection by Phytophthoracapsici. Journal of the American Society for Horticultural Science, 134, 176–182. Berry, D. (1999). Lutte chimique raisonnée et techniques culturales. In D. Mariau (Ed.), Les maladies des cultures pérennes et tropicales (pp. 282–294). Montpellier: Cirad. Chillet, M., Hubert, O., & Bellaire, L. D. (2007). Relationship between physiological age, ripening and susceptibility of banana to wound anthracnose. Crop Protection, 26, 1078–1082. Cox, D. R. (1972). Regression models and life tables. Journal of the Royal Statistical Society, B, 34, 187–220. Dacunha-Castelle, D., & Duflo, M. (1992). Probabilités et Statistiques: problèmes à temps fixe. Paris: Masson. Deberdt, P., Mfegue, C. V., Tondje, P. R., Bon, M. C., Ducamp, M., & Hurard, C. (2007). Impact of environmental factors, chemical fungicide and biological control on cocoa pod production dynamics and black pod disease (Phytophthoramegakarya) in Cameroon. Biological Control, 44, 149–159. Ducamp, M., Nyasse, S., Grivet, L., Thévenin, J. M., Blaha, G., Despréaux, D., & Cilas, C. (2004). Genetic diversity of cocoa tree Phytophthora pathogens. In C. Cilas & D. Despréaux (Eds.), Improvement of cocoa tree resistance to Phytophthora disease (pp. 45–75). Montpellier: CIRAD. Efomban, M. I. B., Biyesse, D., Nyasse, S., & Eskes, A. B. (2011). Selection for resistance to Phytophthora pod rot of cocoa (Theobroma cocoa L.) in Cameroon: repeatability and reliability of screening tests and field observations. Crop Protection, 30, 105–110. Efron, B., & Tibshirani, R. J. (1993). An introduction to the bootstrap. London: Chapman and Hall. Kaplan, E. L., & Meier, P. (1958). Nonparametric estimation from incomplete observations. Journal of the American Statistical Association, 53, 451–481. Martijn ten Hoopen, G., Deberdt, P., Mbenoun, M., & Cilas, C. (2012). Modelling cacao pod growth: implications for disease control. Annals of Applied Biology, 160, 260–272.

Eur J Plant Pathol (2013) 135:363–370 Monfort, A. (1971). Cours de statistique mathématique. Paris: Economica. Moral, J., Bouhmidi, K., & Trapero, A. (2008). Influence of fruit maturity, cultivar susceptibility, and inoculation method on infection of olive fruit by Colletotrichum acutatum. Plant Disease, 92, 1421–1426. Ndoumbe-Nkeng, M. (2002). Incidence des facteurs agro-écologiques sur l’épidémiologie de la pourriture brune des fruits du cacaoyer au Cameroun: contribution à la mise en place d’un modèle d’avertissement agricole. Paris, France: INAPG, PhD thesis. Ndoumbe-Nkeng, M., Cilas, C., Nyemb, E., Nyasse, S., Bieysse, D., Flori, A., & Sache, I. (2004). Impact of removing disease pods on cocoa black pod caused by Phythopththoramegakarya and on cocoa production in Cameroon. Crop Protection, 23, 415–424. Ndoumbe-Nkeng, M., Efoumbagn, M. I. B., Nyasse, S., Nyemb, E., Sache, I., & Cilas, C. (2009). Relationships between cocoa Phythophthorapod rot disease and climatic variables in Cameroon. Canadian Journal of Plant Pathology, 31, 309–320. Nyasse, S., Grivet, L., Risterucci, A. M., Blaha, G., Berry, D., Lanaud, C., & Despréaux, D. (1999). Diversity of Phytophthora megakarya in Central and West Africa revealed by isozyme and RAPD markers. Mycological Research, 103, 1225–1234. Opoku, I. Y., Akrofi, A. Y., & Appiah, A. A. (2007). Assessment of sanitation and fungicide application directed at cocoa tree trunks for the control of Phytophthora black pod infections in pods growing in the canopy. European Journal of Plant Pathology, 117, 167–175. Peter, M., & Gill, W. (2007). Efficient calculation of p-values in linear-statistic permutation significance tests. Journal of Statistical Computation and Simulation, 77, 55–61. R Development Core Team. (2008). R: a language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. Soberanis, W., Rios, R., Arévalo, E., Zuniga, L., Cabezas, O., & Krauss, U. (1999). Increased frequency of phytosanitary pod removal in cacao (Theobroma cacao) increases yield economically in eastern Peru. Crop Protection, 18, 677– 685. Sonwa, D. J., Coulibaly, O., Weise, S. F., Adesina, A. A., & Janssens, M. J. J. (2008). Management of cocoa: constraints during acquisition and application of pesticides in the humid forest zones of southern Cameroon. Crop Protection, 27, 1159–1164. Soubeyrand, S., Sache, I., Lannou, C., & Chadœuf, J. (2007). A frailty model to assess plant disease spread from individual count data. Journal of Data Science, 5, 67–83. Xu, X. M., & Robinson, J. D. (2010). Effect of fruit maturity and wetness on the infection of apple fruit by Neonectriagalligena. Plant Pathology, 59, 542–547. Xu, X. M., Robinson, J. D., & Berrie, A. M. (2009). Infection of blackcurrant flowers and fruits by Botrytis cinerea in relation to weahter conditions and fruit age. Crop Protection, 28, 407–413.