PhD Thesis
Publisher: University of Edinburgh
Abstract:
(a) The overall objectives of this thesis were
to show how the pattern of geographical variation in wild cocoa might determine germplasm collecting strategies and
to investigate the population biology of cocoa in its natural habitat,
using data acquired in the course of a major cocoa germplasm collecting project, the London Cocoa Trade Amazon Project (LCTAP) ;
(b) The planning of germplasm collection involves decisions about the choice of target region and about the sampling strategy to be used within the target region ;
(c) Target regions have typically been selected using basic data on variation throughout the geographical range of cocoa although practical and political considerations have often been of overriding importance. Most attention has been directed at the Upper Amazon region, which appears to be a centre of diversity for cocoa. I emphasize that this should not automatically be taken to mean that the Upper Amazon is also the centre of origin for cocoa ;
(d) Within a target region, one widely accepted criterion for a sampling strategy is that collecting should capture at least 95% of alleles having frequency > .05. This presupposes that data are available to enable the delineation of populations within the target region ;
e) Cost-effective collecting also requires information on the allocation of variation within and between localities within the target region, and on the 'cost ratios', which are based on the relative costs of travelling to additional localities or collecting more samples at the same localities ;
f) Wild cocoa populations in the Amazon Region of Ecuador were systematically collected and characterized using a set of morphological descriptors. Partial data were also obtained for some Peruvian material, collected by Pound, and held in the Pichilingue genebank in Ecuador. The descriptor data were analyzed as frequency distributions for six areas, and by mapping the geographical distribution of groups of accessions generated by cluster analysis ;
g) Within Ecuador, variation was surprisingly limited, and each of the groups of accessions identified via cluster analysis was distributed throughout the region. In sampling strategy terms, it would be reasonable to treat the Amazon Region of Ecuador as a single population, occupying an area some 400 km (E-W) by 500 km (NS). On this basis, the number of accessions successfully established at the San Carlos genebank (about 280) would mean that the probability of failing to capture four alleles with frequency .05 at a random locus would be less than 10-11 , far exceeding the criterion stated in (d) above ;
h) On the Rio Caqueta in Colombia in an area about 300 km (E-W) by 100 km (N-S), there were two distinct pod types with only limited geographical separation. The 34 accessions at San Carlos correspond to a probability of only 88.3% for capturing four alleles having frequency .05 at a random locus ;
i) The available Peruvian material was a small and unrepresentative sample of the original populations. The Pound accessions were most clearly distinguished from the Ecuadorian population on the seed, flush and flower pigmentation descriptors ;
j) Relationships between geographical accession groups were further explored by principal components analysis which provided a concise summary of the similarities and differences noted above, and which generated a series of 'composite descriptors' ;
k) The composite descriptors were used to estimate within- and between-area components of variation for two levels of geographic subdivision The larger geographic subdivisions (Level 1 areas) covered areas of 30,000km2 or more, each corresponding to one or two entire provinces in Ecuador. The smaller geographic subdivisions (level 2 areas) covered10,000-30,000km2 ,and were based on groups of collecting localities within the Level 1 areas. It was concluded that level 1 areas contributed about the same amount of variation as Ievel 2 areas, while Level 2 areas contributed about five times as much variation per tree as individual trees within Level 2 areas ;
(l) The corresponding cost ratios were estimated from information on LCTAP expenditure. Cost ratios were of the order of 1, for sampling level 2 areas within level I areas, and 5, for trees within Ievel 2 areas ; the latter ratio was reduced to about 2 if post collection costs were included, and would be reduced still further by including quarantine and evaluation costs. It was concluded that relatively small sample sizes would be most cost-effective ;
(m) The population biology of wild cocoa was investigated using observations on a study population of about 50 trees at San Carlos, as well as data from trees in the San Carlos genebank and data from collecting trips. In general, cocoa was a widespread component of the understory of undisturbed forest, with typical densities of around 5 trees/ha. It was not found in secondary forest ;
(n) Observed rates of reproduction and mortality were low ; it is suggested that cocoa is a very long-lived species which can maintain population numbers with minimal recruitment. It could be characterized as a shade-bearing, K-adapted species, although short-term seed production and growth rates were increased by shade removal: (o) Seed dispersal by mammals resulted in most seeds germinating close to the parent tree and the resulting seedling clusters suffered high mortality within the first three months ; subsequent seedling mortality was caused mainly by mechanical damage ;
o) Population age structures were deduced from short term growth rate data from trees of varying sizes ; problems with this method were discussed. The age distribution both for the study population and for trees encountered on collecting trips showed a peak in the age class 60-80 years, with fewer older or younger trees. Evidence relating to causes of disturbance which might have affected these populations was reviewed ;
q) The widespread occurrence of cocoa in the Upper Amazon forests, the evidence relating to its reproduction under forest conditions, and the fact that the indigenous inhabitants of the region do not treat cocoa as a crop species, make it very likely that T. cacao is a genuinely wild species whose present-day distribution in the Upper Amazon has been only marginally affected by human activities ;
r) It appears most likely that cocoa is distributed as a wild species in both South and Central America. It is known that cocoa was first cultivated in Central America, but it does not appear that cocoa has evolved any characteristics related to cultivation by which it might be termed a domesticated species. As the practice of cultivation spread through the range of the wild species, it appears that local populations have been added to the set of cultivated types. On this model, the Upper Amazon cannot be described as the centre of origin, and, indeed, the concept of such a centre is not appropriate for cocoa. Rather, the potential value of Upper Amazon cocoa populations to cocoa breeders derives from the fact that, with the exception of the Pound collections, they comprise a wild gene pool not previously incorporated into cultivated cocoa populations. Adaptations for forest conditions, which appear to include very low rates of reproduction and mortality, and slow vegetative growth in low light levels, may explain why wild genotypes sometimes perform badly under plantation conditions.