PAIRWISE ASSOCIATION AS A CRITERION FOR THE SELECTION OF COLLECTION SITES OF NATURAL ENEMIES OF THE CASSAVA GREEN MITE, Mononychellus tanajoa (BONDAR)

PAIRWISE ASSOCIATION AS A CRITERION FOR THE SELECTION OF COLLECTION SITES OF NATURAL ENEMIES OF THE CASSAVA GREEN MITE, Mononychellus tanajoa (BONDAR)

G.S. RODRIGUES¹; L.C. PARAÍBA¹; G.J. de MORAES²

¹CNPMA/EMBRAPA, C.P. 69, CEP: 13820-000 - Jaguariúna, SP.

²Depto. de Zoologia-ESALQ/USP, C.P.9, CEP: 13418-900 - Piracicaba, SP.

ABSTRACT: Climatic similarity has been the primary parameter considered in the selection of sites for the collection and release of natural enemies in classical biological control programs. However, acknowledging the relevance of the composition of biological communities can be essential for improving the record of successful biocontrol projects, in relation to the proper selection of collection sites. We present in this paper an analysis of the plant and mite assemblages in cassava fields of northeastern Brazil. Such analysis is suggested as an additional criterion for the selection of collection sites of mite predators of the cassava green mite, Mononychellus tanajoa (Bondar), in an international biological control program. Contingency TABLES were built using Dice's index as an indicator of significant associations between pairs of species. This analysis enabled the identification of plant and mite species typically found together, indicating interspecific interactions or similar ecological requirements. Finally, a cluster analysis was used to group sites containing similar assemblages. These sites exhibit comparable chances of harboring a given species. Applied at the species-group level, the analysis may assist in better defining sites for the collection of natural enemies to be released in a given region, improving the chances of establishment.

Key Words: biological control, cluster analysis, cassava green mite, Mononychellus tanajoa

ANÁLISE DE CO-OCORRÊNCIA DE ESPÉCIES COMO CRITÉRIO PARA SELEÇÃO DE ÁREAS DE COLETA DE INIMIGOS NATURAIS DO ÁCARO VERDE DA MANDIOCA, Mononychellus Tanajoa (BONDAR)

RESUMO: Similaridade climática é normalmente o principal parâmetro considerado na seleção de áreas de coleta e liberação de inimigos naturais de pragas em programas de controle biológico. Muitas vezes, contudo, a composição das comunidades biológicas presentes nessas áreas representa fator decisivo para a sobrevivência de um determinado organismo, melhorando ou dificultando seu estabelecimento. Este trabalho apresenta uma análise das sub-comunidades de ácaros e plantas que ocorrem em plantações de mandioca no nordeste brasileiro. A análise representa um critério adicional para a seleção de áreas de coleta de ácaros predadores do ácaro verde da mandioca, M. tanajoa (Bondar), como suporte para um projeto internacional de controle desta praga no continente africano. Tabelas de contingência foram elaboradas empregando o índice de Dice como indicador das possíveis associações entre pares de espécies encontradas em levantamentos de campo, permitindo a identificação de espécies que tipicamente co-ocorrem. Espécies assim associadas são possíveis indicadoras de associações interespecíficas ou requisitos ecológicos similares. Finalmente, uma análise de "cluster" foi utilizada para agrupar áreas apresentando comunidades similares. Estas áreas apresentam probabilidades comparáveis de abrigar uma dada espécie. Aplicando-se este conceito ao nível de grupo de espécies, a análise pode ajudar na definição de áreas para a coleta de inimigos naturais a serem liberados em uma dada região, aumentando as possibilidades de estabelecimento.

Descritores: controle biológico, análise de "cluster", ácaro verde da mandioca, Mononychellus tanajoa

INTRODUCTION

This paper focuses on the biological control of an introduced pest, the cassava green mite, M. tanajoa, in Africa (Herren, 1987). The mite is native to South America, and it was found for the first time in East Africa in the early 1970's. From there it spread across most of the tropical African cassava belt (Yaninek & Herren, 1988). The severe damage caused by this pest in Africa led to a study of its natural enemies in South America, in order to select promising species for introduction into Africa. Major effort has been dedicated to searching for prospective natural enemies in northeastern Brazil and northwestern South America, because of agroecological similarities with areas where M. tanajoa is a problem in Africa.

In general, the selection of collection and release sites for biological control agents has been based primarily on abiotic comparisons of the climate of the involved areas. Analyses of the biotic community structure in the collection and release areas are rarely considered as indicators of the chance of success for the introduction of natural enemies. Nevertheless, the presence of alternative plant substrates and prey may be as important for survival of a species as is microclimate (Pimentel, 1961; Root, 1973).

Consideration of the composition of the biological community in the area from which an organism will be drawn for release into another area and the association between the species present may provide essential information on how to manage the area of introduction. This is particularly true for agroecosystems where, for instance, weeds may enhance the survivorship of natural enemies (Risch et al., 1983; Andow, 1988).

Significant pairwise associations can be beneficial or detrimental, and may be so differentiated by the coefficients of association. These coefficients will indicate respectively greater or smaller chances of finding one given species in a field, once another species has been found. Significant associations may also indicate similarity or disparity in relation to other environmental conditions, e.g., climate. In such a case, even associations between disparate groups, such as mites and plants, may be of interest in the characterization of the habitat.

In the present paper we examine the pairwise coefficient of association for all mite species sampled and their respective substrate vegetation type in cassava fields of northeastern Brazil, and the coefficient of association between mites and each one of the species of plant sampled. This analysis provided an organized picture of the assemblages present throughout the region, as well as an indication of which associations could be essential for each species.

A cluster analysis was then performed based on the community structure defined by these coefficients of association. This analysis revealed the fields or sites with comparable likelihood of harboring a given species, providing an ecological criterion for future collections of natural enemies, as well as for the selection and categorization of release areas.

MATERIALS AND METHODS

The data were collected through a survey carried out between May and June 1990 in 33 cassava fields, in 28 municipalities spread throughout 5 States in northeastern Brazil. The sampling procedure consisted of examining during 15 minutes several specimens of each of the most abundant plant species in a given cassava field, including other cultivated plants on the edges of the field. A total of 137 plant and 51 mite species were examined.

Two types of sampling units were applied in the analyses. First, the data were organized in contingency TABLES having field as the sampling unit, with plant and mite species appearing binomially distributed (present or absent) in the TABLES (Hubalek, 1982). A second sampling unit was defined as each plant species. Similarly, contingency TABLES containing every mite species appearing on each plant species were built. A chi-square test was then applied to the TABLES (Dice, 1945; Ludwig & Reynolds, 1988) in order to test the significance of the pairwise associations.

When two species are found together more frequently than would be expected by chance alone, after considering their relative frequencies in the sample series, a significant chi-square results. Whenever the expected value for a cell in the table was too low, due to few observations for that particular species, a corrected chi-square was used. This aimed at ensuring approximation to the continuous chi-square distribution. Yate's correction formula (Ludwig & Reynolds, 1988) was applied for such correction.

After sorting out the significant associations, the Dice index (Dice, 1945; Ludwig & Reynolds, 1988) was used to express the strength of association between each pair of species. This index is a computation of the number of occurrences of either species A or B in the total sample series, and the proportion of samples in which both occur together. The index also defines the tendency of any significant association, or in other words, whether the species are positively or negatively associated, and to what extent.

When each cassava field was taken as a sampling unit, 17,578 possible pairwise associations between mites and plants and between mites and mites were examined. When each plant species was considered as a sampling unit, 1,275 possible associations between mite species resulted.

A cluster analysis was applied to the communities as defined in the contingency TABLES in order to classify the 28 municipalities according to the occurrence of mites and plant species with which these mites were associated. Sorensen's index was used to calculate a similarity matrix and the distance

dij = (2(1-sij))^1/2,

where sij is the similarity between municipality "i" and "j".

Sorensen's measure is very similar to Dice's measure and was first used by Czekanowski in 1913 (Krebs, 1989). This coefficient weighs matches in species composition between the two samples more heavily than mismatches. Because in the present sample series many species are present in a community but not present in a particular sample from that community, it may be useful to use Sorensen's coefficient rather than other measure (Krebs, 1989).

The SPASSOC.BAS (Ludwig & Reynolds, 1988) statistic routine and the SAS (1985) statistics package were used for the association and the cluster analysis, respectively.

RESULTS

Fields as Sampling Units: Thirty five significant associations between mites and plants and 18 significant associations between mite species with Dice's index higher than 0.5 (positive interactions) were identified when fields were taken as the sampling unit. All significant associations between mites and plants were positive (TABLE 1), which indicates that the presence of a particular plant species suggests the presence of the corresponding mite species in a given field, though not necessarily on that plant species. The table also shows an indication of the feeding habit of each mite species, so that the associations may be interpreted under this light. It should not be forgotten, however, that the associations correspond only to co-occurrences on the sampling unit considered, and not necessarily effective interaction between any two organisms or species.

The association between M. tanajoa and cassava was not significant regardless of the fact that the former feeds almost exclusively on that plant or on plants of the same genus. This result reflects the observation that several cassava fields were not infested by M. tanajoa. The predatory mite Amblyseius aripo DeLeon was significantly associated with cassava and Borreria verticillata.

Amblyseius manihoti Moraes was not significantly associated with cassava, although until now it has been found almost exclusively on that plant. This mite species was not found in many of the fields sampled. A. manihoti also was significantly associated with B. verticillata.

Three of the associations between mite species were negative, indicating that the presence of a particular species implied that the absence of the other species was more likely in that specific field (TABLE 2).

Aceodromus convolvuli Muma, Amblyseius idaeus Denmark & Muma and A. manihoti were negatively associated with A. idaeus, A. manihoti, and Phytoseius guianensis DeLeon, respectively. A. aripo was positively associated with A. manihoti, and A. idaeus was positively associated with Euseius alatus DeLeon.

Plant Species as Sampling Units: When the same type of analysis was applied using plants as sampling units, we found more significant associations among mite species. This indicates their occurrence on a same plant species, although not necessarily at the same time (TABLE 3). All of the significant associations were positive, indicating that the presence of one species increases the likelihood of the presence of the other. None of the common predators on cassava had Dice indexes higher than 0.5. A. aripo was significantly associated with M. tanajoa, but had a Dice value of 0.444, whereas A. idaeus was significantly associated with Tetranychus sp., with a Dice value of 0.387

Cluster Analysis: In relation to mite and plant assemblages, six groups of sites could be identified by cluster analysis: group I - five sites from south-central Bahia, one from central Piauí and one from western Pernambuco; group II - five sites from coastal or north-central Bahia, one from southern Ceará and two from coastal and northwestern Piauí; group III - five sites from central Piauí and three from northern and northeastern Maranhão; group IV - one site from central Piauí; group V - three sites from central Maranhão; group VI - one site from south-central Bahia (Figure 1).

Figure 1
- Grouping of collection sites in northeastern Brazil according to the occurrence of mites and plant species with which the former were associated, as determined by a cluster analyses.

Although group IV consists of a single site, it is closely related to groups II and III. The site corresponding to group VI, however, is distinctly different from all other sites.

Although, as expected, greater similarities were found between sites located closer to each other, both groups I and II included fields located hundreds of kilometers apart, as observed in Figure 1. This indicates the recurrence of mite and plant assemblages in these areas.

DISCUSSION

This work exemplifies the application of species association analysis in the definition of ecological similarities between different sites, based on the assemblages represented by mite and plant species within cassava fields in northeastern Brazil. After determining significant associations between the organisms, collection sites were classified according to their similarities in terms of co-occurring species.

These assemblages of mite and plant species may also give basis for the comparison of collection and release areas, enabling us to address natural enemies to sites more similar to their natural habitats, enhancing the chances of establishment. Furthermore, the analysis showed certain interspecific associations that may be important in the autoecology of the candidate species to be collected and released.

It must be kept in mind, however, that the association between two species, as indicated by a significant, large Dice index does not imply dependence or interaction directly. It just means that both species are co-occurring in the sampling unit more often than they would by chance alone (Dice, 1945; Ludwig & Reynolds, 1988). Ascertaining interaction (symbiosis) is only possible with more autoecological information about the involved species.

The use of two different sampling units served two purposes. First, using fields as sampling unit allowed the evaluation of associations between mite and plant species in terms of a broader, habitat level basis. As field may be considered a relatively large scale sampling unit for mites, it was not surprising that all associations observed were positive. This indicates that, by and large, most mite species occurring in the region would appear in such a large sampling unit. Very high Dice indexes, however, may indicate interactions worth considering in terms of adapting a mite species to a new area. For instance, the significant positive association between the predatory mites A. aripo and A. manihoti with the plant B. verticillata may indicate that the presence of this plant species or its close relatives may be important in the release sites. It is possible that B. verticillata plays a significant role in the ecology of these mites, which could be confirmed in a specifically tailored study involving those organisms.

The negative associations between certain mite species, at the field level, may be an indication of broad differences in the ecological requirements of those species, as in the case of A. convolvuli, A. idaeus and A. manihoti been negatively associated with A. idaeus, A. manihoti and P. guianensis, respectively.

The second approach to the analysis, taking plant species as sampling unit, enabled the assessment of associations between mites at a smaller scale, related to their substrates. At this level, though, the Dice indexes very much leveled off, reaching at the maximum only 0.7 in only two occasions. This led us to the consideration that true, obligate association (symbiosis) was not demonstrated, but that in all cases at least one of the components of each mite pair may survive independently of the presence of the other.

Plant and mite species composition may vary between seasons in any given environment. In this study, field data were collected during the wet season in all fields. Other seasons should also be considered, because alternative prey or plant substrates could be important in maintaining the predators in the cassava fields in different, less favorable conditions. In addition, mite species composition could vary from season to season.

It is to be expected that cluster analysis will assign a high level of similarity to neighboring fields, unless the study is conducted in a highly variable region. Similarities between distant fields are of greater interest because they indicate similar chances for those fields to successfully harbor an organism. The positions of the groups of most similar sites determined in this study are concurrent with a bioclimatic zoning of Brazil. This was expected because of the effect of climate on the kinds of organisms inhabiting a given area. However, available climatic data for some understudied areas may not be sufficiently detailed as desirable in the process of selecting areas where to look for or to release prospective natural enemies. The analysis proposed in this study may provide a finer method for the selection of specific collection or release sites within a broad area.

An application of this type of study could be the concurrent analysis of a given assemblage in a place where M. tanajoa is a pest in Africa, and in a series of prospective collecting sites in South America. Alternatively, a given source of predators of M. tanajoa could be concurrently compared with a series of prospective release sites in Africa.

It seems advisable, however, that associations be considered at the species-group level rather than at the single-species level, because of the number of exclusive indigenous species of each continent that may belong to species-groups occurring in both continents.

ACKNOWLEDGMENTS

To Dr. David Pimentel (Cornell University), Dr. Maria Alice Garcia (UNICAMP), Dr. Moacyr Dias Filho (EMBRAPA/CPATU) and Dr. Robert J. O'Neil (Purdue University), for their suggestions.

Recebido para publicação em 30.01.96

Aceito para publicação em 07.09.96

• ANDOW, D.A. Management of weeds for insect manipulation in agroecosystems. In: ALTIERI M.A.; LIEBMAN M., (Eds.). Weed management in agroecosystems: ecological approaches. Boca Raton: CRC Press, 1988. p.265-301.
• DICE, L.R. Measures of the amount of ecological association between species. Ecology. v.26, p.297-302, 1945.
• HERREN, H.R. A review of objectives and achievements. Insect science and application. v.8, n.(4/5/6), p.837-840, 1987.
• HUBALEK, Z. Coefficients of association and similarity based on binary (presence-absence) data: an evaluation. Biological Reviews, v.57, p.669-689, 1982.
• KREBS, C.J. Ecological methodology. New York: Harper Collins Publishers, 1989. 654p.
• LUDWIG, J.A.; REYNOLDS, J.F. Statistical ecology: a primer on methods and computing. New York: John Wiley, 1988. 337p.
• PIMENTEL, D. Species diversity and insect population outbreaks. Annals of the Entomological Society of America, v.54, p.76-86, 1961.
• RISCH, S.J.; ANDOW, D.; ALTIERI, M.A. Agroecosystem diversity and pest control: data, tentative conclusion and new research directions. Environmental Entomology, v.12, p.625-629, 1983.
• ROOT, R.B. Organization of a plant-arthropod association in simple and diverse habitats: the fauna of collards (Brassica oleracea). Ecological monographs. v.43, p.95-104, 1973.
• SAS User's Guide: Statistics, Version 5 Edition. Cary, SAS Institute, 1985.
• YANINEK, J.S.; HERREN, H.R. Introduction and spread of the cassava green mite, Mononychellus tanajoa (Bondar) (Acari: Tetranychidae), an exotic pest in Africa and the search for appropriate control methods: a review. Bulletin of Entomological Research, v.78, p.1-13, 1988.

Publication Dates

• Publication in this collection
  10 Feb 1999
• Date of issue
  May 1996

History

• Received
  30 Jan 1996
• Accepted
  07 Sept 1996