The second session of the FAO Committee on Forest Development in the Tropics, held at Rome 21-24 October 1969, asked that this paper be given wide circulation.
AT THE RECENT FAO /IUFRO Second World Consultation on Forest Tree Breeding held at Washington, D.C. in August 1969, considerable attention was devoted to tropical tree breeding. Emphasis was given to the need for most tropical countries to develop their own seed sources of suitable species and provenances. To do this an individual country must first test a range of potential species and their provenances. Some tropical countries have already initiated extensive programmes of testing and afforestation but many have not. The present paper is intended to act as a basic checklist of considerations for rational provenance testing, assuming that information is available to indicate suitable species.
The term provenance refers to the geographic source of seed or plant material or to the plants from such a source. It is generally true that tree species with a wide geographical distribution exhibit considerable provenance variation in anatomy, morphology and physiology.
Genecological studies of tree species provide information on the pattern and extent of phenotypic and genotypic variation. The variability may be related to the distribution of continuous or disjunct environmental factors such as soil type, and altitude, exposure or latitude with their associated factors of precipitation, temperature and photoperiod. The relative contribution of genetics and environment to variation (both in classical taxonomic characters and in more recently studied physiological traits such as photo- or thermoperiodic response) may be evaluated by raising seedlings from various seed sources under relatively uniform conditions as in growth chambers, greenhouses, nurseries or field tests.
These genecological studies, although essential to a rational programme of afforestation with an untried species, are expensive in time, labour, skill, and equipment, and should therefore be left to suitable research organizations. Provenance trials, on the other hand, must be carried out by each country wishing to test a given species. The main practical object of such trials is to identify those provenances the seed of which will produce well-adapted and productive. forests. Additionally, information is required on the range of variation between individual trees within provenances.
Ideally, each country wishing to test a new species would collect its own seed. However, many species are indigenous in areas to which access is politically or geographically difficult, and some form of centralized, international seed collecting agency is desirable. This is particularly the case with species that are of widespread importance as exotics, such as Gmelina arborea, Tectona grandis, Pinus caribaea, P. kesiya, P. merkusii and many Eucalyptus species. Agencies already concerned with international seed collection and field testing for tropical species include the Forest Research Institute, Canberra, Australia; Commonwealth Forestry Institute, Oxford, England; Centre technique forestier tropical, Nogent-sur-Marne, France; Danish-Thailand Teak Breeding Centre, Lampang, Thailand. These efforts are being coordinated and, in some cases, financially supported by FAO, through its Panel of Experts on Forest Gene Resources.
If such agencies are able to collect sufficient representative seed from a representative series of provenances it is desirable that they should initiate cooperative, comparative, international trials using the same provenances in. several countries. Combined analysis of these yield more information than would be obtained from each collaborating country separately, but it is desirable that the combined analysis and interpretation be undertaken by the initiating agency. Examples of such trials are those organized during 1967-63 by the Commonwealth Forestry Institute at Oxford (up to 11 provenances of P. merkusii were allocated to 20 countries, and up to 14 provenances of Cedrela spp. were distributed. to 14 countries), and by FAO's Mediterranean Forestry Research Committee (Eucalyptus camaldulensis and E. dalrympleana). It is essential that results of individual and collaborative trials be disseminated as soon as possible by publication in appropriate journals or bulletins.
To coordinate and standardize provenance research and to facilitate dissemination of results, the International Union of Forest Research Organizations set up a Working Group on -Provenance Research and Testing in 1962. This working group presented to the IUFRO Congress at Munich, 1967, the second draft of their report, " Standardization of Methods for Provenance Research and Testing." ¹ This report includes general recommendations for provenance sampling and seed procurement; design, layout and analysis; techniques and assessments in the nursery and plantation stages; and publication of results. It is of general application but special problems of the tropics require further examination.
(¹In Proceedings of the fourteenth session of the IUFRO) Congress, section 22 p. 672-718.)
The IUFRO report has been widely circulated and, on the assumption that FAO, IUFRO, and the various other agencies exemplified above will organize authenticated seed collection and distribution, and international collation of results, the present paper concentrates on the reasons for and methodology of provenance trials which could be carried out by an individual tropical country.
It is neither possible nor desirable to be specific about methods. There is a great range among tropical countries in skills, resources, sites, species of interest, and objectives. This paper therefore outlines some basic concepts that should be considered by any tropical country embarking on provenance research.
The main practical object of provenance trials is to locate as quickly and as economically as possible those provenances yielding well-adapted and productive forests. Productivity itself may not always imply rapid growth; important criteria could be survival, resistance to adverse environmental factors or pests, wood quality, seed production. The second major object is to establish local seed production stands. Spare seed may be sown in separate provenance plots or as a bulk mixture but in either case the best individuals should be selected for future breeding.
The exact objectives should be understood in planning any provenance trial and the acceptable degree of difference between provenances should be known for all traits of interest; information on this level of difference and on the precision of measurement is essential for designing experiments and for interpreting analytical statistical significance in practical terms. It is of little value to measure individual tree heights (or to have sufficient replications to estimate provenance mean heights) to the nearest millimetre if the critical difference between two provenance mean values is 1 metre. Generally, meaningful differences are in the range of 5 to 10 percent of the mean.
In defining the objects of a trial the number of variables and the frequency of measurement must be clearly stated. Some immediately useful traits are discussed below but it should be stressed here that in many provenance trials too many measurements are made on too many variables.
It is tempting, particularly in the case of a completely untried species or of a species for which the provenance representation would be difficult to repeat, to attempt to measure every possible character as intensively and as frequently as possible. However, the difficulties of data processing, analysis and interpretation increase rapidly with increasing numbers of variables and the ratio of benefit to cost is extremely low. Commonly data are collected but not analysed. The increasing availability of high-speed computers facilitates the use of multivariate analytical techniques; these may bring out relationships between provenances that are not obvious from univariate analysis but often unplanned analyses are undertaken " to see what happens."
Even if specialized institutions are prepared to offer their staff and equipment for complex analyses, the resources of the tropical country establishing the trial may limit the data collected. Priorities must then be assigned to characters that are of immediate practical value and to those that are of use only as descriptive or biosystematic discriminators. Taxonomic traits could well be evaluated at least once on suitable samples by specialized research units, particularly by those initiating internationally comparable provenance trials.
Various types and stages of provenance trial may be necessary to locate the best provenance or provenances of a species for a given site, country, or region. The choice depends on the current information available, the extent of natural variation in the species, and the variation among the potential planting sites. The choice will be influenced also by the objects of the trial series, by the resources available, and by the importance of the species in the national programme. It may be possible to compress some of the stages into one, or to combine provenance testing of several similar species, or to run several of the stages on parallel. Additionally, in preceding species trials, two or more provenances of some species may be compared.
With a wide-ranging species it is unlikely that one experiment could indicate the best provenance even for one site type. The following stages may therefore be used on each major site type:
1. range-wide sampling phase;
2. restricted sampling phase;
3. proving phase (crop performance phase).
RANGE:-WIDE SAMPLING PHASE
This stage is intended to indicate broad regions in which provenances may be found to have reasonable adaptation and productivity. As stressed in the IUFRO report, they may reveal general areas from which further large-scale seed imports should be avoided or permitted pending results or second stage tests. However, their main function is to indicate procedures for the second stage. It would be valuable to develop for each species some systematized (preferably quantified) method of homoclimal classification. Precise definition of site is often unavailable.
Range-wide sampling studies should be brief and simple. If sufficient seed and resources are available large blocks of each provenance may be planted elsewhere for yield estimates at full rotation age and individual tree selection, provenance hybridization, seed production, and maintenance of the gene pool. However, the main purpose of this phase is to eliminate obviously unpromising provenances as early as possible, thus reducing the provenances to a manageable number for more critical testing. Depending on the range of the species and/or seed availability, the number of provenances tested may be large. Totals of 25 and 31; provenances of Pinus elliottii and P. taeda are currently under test at six sites in Malawi. Some 150 provenances are represented in some plantings of the latest Pseudotsuga menziesii trial organized by IUFRO and over one thousand seed collections of Picea abies have been collected.
The duration of this phase should be from one quarter to one half of the commercial rotation and, because interplant competition will be relatively small up to this stage, small plots are acceptable. Square plots of 16 to 25 trees are most commonly used but plots of 1 to 9 trees have been recommended particularly for experiments with large (more than 50) numbers of provenances, and where facilities are available for undertaking the resultant complex analyses. These extremely small plot sizes have less value in the tropics where fast growth rates produce early competition between individual trees. They have the disadvantage of requiring many replications with consequent difficulties of design, layout and recording.
Plots need not always be square in number of trees or in shape. If small numbers are used line plots may be adequate; however, the object of provenance trials is usually to assess provenance performance free from competition with other provenances, so that square or rectangular plots are preferable. The shape may be influenced by the intended management practices but it is largely dependent upon the pattern of site variability. Where marked linear gradients of site factors occur, rectangular plots should be laid out with their long axes parallel to the contours of slope.
One-tree plots are rarely used because normal mortality produces excessive numbers of missing plots and emphasizes interprovenance competition. Provenances that would otherwise be acceptable in pure plantation may be eliminated because of slow initial growth. After survival, height is usually the principal production variable assessed in range-wide studies because it is less influenced than diameter by tree competition.
RESTRICTED SAMPLING PHASE
This phase of provenance trials is designed to identify smaller regions and finally individual provenances that have the greatest productivity. The number of provenances tested may be between 5 and 10 and it may be valuable to represent each by seed from superior and randomly selected parent trees. This depends upon the resources of the seed collecting agency. Apart from financial, space, and plant resources, two conflicting considerations affect plot size at; this stage. There will be less variation between plots of selected provenances, plot interactions will be smaller, and therefore plot; size could be decreased. (Large plots are only required where plot interactions are expected to be great.) On the other hand, these plots should be maintained to full rotation age and could provide yield information
(height and diameter) prior to that obtained in the proving phase. Larger plot sizes are therefore desirable and plots of between 49 and 169 trees are commonly used. A one- or two-row surround may be desirable, particularly if yield information is not to be biased by edge effects.
A small number of probably useful provenances are tested at this stage under normal plantation conditions. Whereas in the first two phases careful attention must be given to experimental design, in the proving phase this aspect is less important. Replication of sample plots within and between proving plots is essential. Proving plots are large enough to support stander/1 mensurational assessments (0.5 to 1.0 hectare) or to facilitate costing and silvicultural studies (2 to 5 hectares). As with the earlier stages these should be repeated on the :major site types in each country. If yield can be correlated with some simpler traits such as height, plot size and duration of the proving phase could be reduced. Attention should be drawn. here to the desirability of developing early- test methods for all stages of species, provenance and progeny testing. It may prove possible for specialized institutions to conduct growth chamber research on seedling material that will facilitate predictions of performance under tropical plantation conditions.
In designing provenance trials the research worker should obtain the advice of a qualified and experienced statistician if possible, particularly if local conditions indicate that some departures from common designs will be necessary. The purpose of experimental design is to minimize environmental heterogeneity and to increase the precision of treatment (provenance) comparisons. It is possible to minimize experimental error by choosing a uniform site although a typical uniformity should be avoided in siting provenance trials. It is impossible to remove all sources of experimental error and the use of valid experimental designs facilitates the separation of experimental error from other sources of variation.
A randomized plot design could be useful for the range-wide sampling phase but for all stages the most favoured design is the :randomized complete block layout (RCB) in which each replication contains one plot of each provenance. It is simple, so that the layout can be planted and assessed by relatively untrained staff; preliminary analysis of variance of plot means can be undertaken easily with desk calculators. It is flexible, so that the field layout can be adjusted to local conditions. It is statistically robust, so that loss of some plots or even entire replications will not invalidate the use of remaining data. The number of replications depends upon the variability of the site, the variability of the provenances, the precision of measurement, and the size of plot, in addition to resource considerations. Repetition in separate years may be necessary if annual climatic fluctuations are great. The RCB design should be used wherever possible provided that the number of provenances and/or the size of plots do not require replication area to be too great.
When large numbers of provenances are tested or when large plots are used, increased environmental heterogeneity may reduce precision; some form of incomplete block design is then desirable. Treatments are arranged in small blocks and blocks are laid out so that intra-block environments are homogeneous. The balanced types are preferable because each treatment occurs once in a replication with all other treatments so that the entire experiment can be analysed as an RCB design if necessary. This allows early analysis by local staff and it may become the only possible analytical method if several plots fail or become damaged, or if only the better provenances are assessed for economy; further, it permits estimation of the increased precision obtained by using the incomplete block design.
Balanced incomplete block designs that are currently being analysed by the Commonwealth Forestry Institute, Oxford, for tropical species or provenance trials include rectangular lattices for 12, 20 and 30 treatments, and triple lattices for 9, 16, 25 and 36 treatments. These are proving more efficient than RCB designs but they are more difficult to supervise, lay out, assess and record and should not be used unless absolutely essential.
Whatever design is chosen for the plantation stage, it is desirable to use the same design in the nursery. Plants for a given field replication should be taken from the same nursery replication. In this way nursery and field replication effects are confounded and within-plot variance is minimized.
When a provenance trial series is planned for several sites, there must be a choice between the use of one central nursery and several separate nurseries. The decision must be based on local considerations of labour and transport costs, likely plant damage by transport, chance of losing entire experiment if based on one nursery, and likely future practice. As with all aspects of nursery and field management, the techniques applied should be typical of what is currently practiced or anticipated for the future for each site. The IUFRO report discusses many of the controversial aspects of nursery and field stage management.
The trial series is usually designed to locate the best provenance for each site under that site's conditions of natural environment and practice or anticipated cultural treatment. In any combined analysis over all sites, the estimate of provenance-site interaction will reflect the failure of provenances to perform consistently on all sites and will indicate the repeatability or genetic stability of the test material. The likelihood of provenance reaction to local site conditions is even greater in the case of international trials. Here similar material may be raised at different times of the year under widely varying climatic, edaphic, and management conditions. It would clearly be impracticable to raise the plants at one nursery even if quarantine regulations permitted shipping them.
It is not possible to generalize on the management and assessment of provenance trials but details should be included in each trial plan. It is difficult to determine in advance whether silvicultural operations such as thinning and pruning will be undertaken at prescribed times for the trial as a whole or for provenances differentially. Thinning may be systematic, random or selective, depending on the objects of the trial. Generally, however, management should be similar to that currently used or anticipated to be optimum for the test material.
The types and frequency of assessment will vary with the stage of provenance testing but they should be specified in the plan. At the end of the nursery stage the relative performance of the provenances can be examined as a self-contained study. This may yield biosystematic information or predictive information for the plantation stage. Nursery characters, particularly plant dimensions, can be used as co-variates in plantation analyses. Other nursery traits of value are seed germination rate and amount, plant survival, uniformity, cotyledon and leaf characteristics.
As stressed above, tropical countries with limited resources must assign priorities to productive, descriptive, and taxonomic traits. Productive and descriptive characters that can be assessed usefully in both the range-wide and restricted sampling phases include height, diameter, bark thickness, crown depth and width, branch number, angle and length, survival, uniformity; straightness of stem should be assessed either by some quantified continuous measurement or by categorical classification. These traits should be evaluated at three-to five-year intervals throughout the test period with more frequent (half-yearly or annual) sampling in the first two or three years.
The most useful measurements of yield will be made at the crop performance phase by assessing height, diameter, form, taper, and bark thickness. Wood quality comparisons can best be made at the end of this stage although preliminary information can be made in the earlier testing phases. Other important traits that should be assessed in all phases are phonology of growth and flowering, and seed productivity which may require intensive observation over several years. Reaction to adverse environmental and biotic factors must be recorded throughout all stages; when biotic damage is observed, the causal organism must be identified.
Attention is concentrated on early measurements for four reasons:
1. It is urgent to estimate juvenile-mature correlations, particularly for countries establishing large areas of plantation with incompletely tested species.
2. It may be necessary to make decisions on choice of provenance before information is available on mature crop performance or on juvenile-mature correlation. In this case increased safety should be obtained by multivariate analysis of several juvenile performance indicators, for example height and diameter at planting and after 6, 12, 18 and 24 months.
3. Part of the variation in mature characteristics may be explicable in terms of juvenile traits by covariance analysis. Nursery and postplanting dimensions may be valuable but attention must be given to possible subsistence immediately after planting. This could cause an apparent decrease in tree height. An assessment six months after planting may be preferable although an initial establishment survey is essential immediately after planting. Cultivation, particularly mechanical ploughing or disking, may cause an apparent increase in tree height although the effect of this decreases proportionally as tree height increases.
4. It will be desirable to compare the total growth of provenances over the entire test rotation. Where sigmoid growth curves are involved, great changes in slope and precision occur early in the rotation and greater sampling intensity is needed to obtain uniform precision of provenance comparison over the entire growth curve.
EGON GLESINGER RETIRES
Egon Glesinger, who joined the present Forestry and Forest Industries Division when it was first formed in 1946, has retired from FAO as an Assistant Director-General.
Now aged 62, Dr. Glesinger was born in the Silesian town of Teschen which was Austrian territory until 1919, when it was divided between Czechoslovakia and Poland. He studied law and economics at Geneva, Prague, and the London School of Economics, and international law at Vienna. Following the publication of his doctorate thesis on timber, he was invited by the Secretary-General of the League of Nations to organize a timber conference. From this conference grew the nongovernmental International Timber Committee of which he became secretary-general. The interests of this body were later assumed by FAO.
Author of The coming age of wood, Dr. Glesinger was rapporteur of the Technical Committee on Forestry and Primary Forest Products set up by the United Nations Interim Commission on Food and Agriculture in December 1943. The report of this committee led to the inclusion of forestry and primary forest products in the scope and constitution of the Food and Agriculture Organization when it came into being at Quebec in 1945.
As Director of the FAO/ECE Timber Division at Geneva, Egon Glesinger started the series of FAO regional timber trends studies. The first European study was published in 1953. His name will also long be associated with the FAO Mediterranean Development Project dating from the same period.
He was appointed FAO'S Director of Forestry in 1959 when Marcel Leloup retired. Four years later he was promoted to Assistant Director-General in charge of the Department of Public Relations and Legal Affairs.