The site description data is collected at the beginning of the trial establishment and should include the following:
Name of the site: ________________________________________
Country: ________________________________________
Province: ________________________________________
District: ________________________________________
Latitude (degrees and minutes): ________________________________________
Longitude (degrees and minutes): ________________________________________
Altitude (m above sea level): ________________________________________
Managing office/institution: ________________________________________
Owner: ________________________________________
Distance to nearest office responsible for management (km): ____________________
Distance to nearest villages/towns (km): ____________________
Number of inhabitants in the nearest villages/towns: ____________________
Type of area (e.g. research station, managed forest, etc.): ________________________________________
Add map(s) (see section 3,2. above).
Name of the station: ________________________________________
Latitude (degrees and minutes): ____________________
Longitude (degrees and minutes): ____________________
Altitude (m a.s.l.): ____________________
| Climatic data1 | Jan. | Feb. | Mar. | Apr. | May | June | July | Aug. | Sept. | Oct. | Nov. | Dec. | Year |
| Rainfall (mm) | |||||||||||||
| Temp mean (ºC) | |||||||||||||
| Temp. mean max.2 (ºC) | |||||||||||||
| Temp. mean min.3 (ºC) | |||||||||||||
| Evapotranspiration 4 (mm) |
1Period of observations: ____________________ (specify years)
2Average of daily maximum temperatures
3Average of daily minimum temperatures
4Potential evapotranspiration (ETP) - Penman's formula
Number/type of seasons:
one: _________ two: _________ even: _________ irregulars: _________
Period(s): ______________________________ (specify months)
No. of intermediate days proximate to growing season: ____________________
No. of wet days in growing season: ____________________
Dry months (number month per year of < 50 mm rain):: ______________________________
Frost (number of days/year): ____________________
Prevailing wind (direction, period, speed): ____________________
__ Flat/gentle (0-8 %) __ Intermediate (9-30 %) __ Steep (>30 %) < /P>
Please tick the following soil description
| Soil texture | Soil depth | Soil drainage/ Waterlogging | Gravel content, topsoil | ||||
|---|---|---|---|---|---|---|---|
| 1. Light/sandy | 1. Shallow (< 50 cm) | 1. Well drained | 1. None (< 15 %) | ||||
| 2. Medium/Loamy | 2. Deep (50-100 cm) | 2. Seasonal | 2. Gravelly (15-35 %) | ||||
| 3. Heavy/clayey | 3. Very deep (> 100 cni) | 3. Permanent | 3. Stony (> 35 %) | ||||
| Organic matter content | Reaction (pH) | Soil salinity | Ground water | ||||
| 1. Poor (< 2 % DM) | 1. Acid (pH < 6.5) | 1. None | 1 1 - Shallow (< 50 cm) | ||||
| 2. Medium (2-5 % DM) | 2. Neutral (6.5-7.5) | 2. Moderate | 12. Deep (50 - 150 cm) | ||||
| 3. Rich (> 5 %) | 3. Alkaline (pH> 7.5) | 3. High | 13. Very deep(>150 cm) | ||||
Natural (original) vegetation type:
________________________________________
______________________________________________________________________
Dominant natural (original) genera/species: ________________________________________
Land use history: ________________________________________
______________________________________________________________________
______________________________________________________________________
Apparent major problems of site (pests, diseases, etc.): ________________________________
______________________________________________________________________
Economics data, viz farm gate prices for fodder, fuel wood, small size timber, extractives, and other minor produce :
______________________________________________________________________
______________________________________________________________________
______________________________________________________________________
Specifically the objectives are i) to examine if there is statistically significance for substantial, i.e. not trivial, differences between seedlots in respect of adaptability (survival, health and flowering/fruiting) and growth. Further, ii) to examine any substantial differences observed between seedlots over a variety of environmental conditions, i.e. examine interactions of seedlot and site.
The design chosen must ensure that the objectives are fulfilled, i.e. any substantial differences are exposed as statistically significant. To what extent this will be possible depends on the trial lay-out.
For each trial, a set of unbiased, estimated seedlot-mean-values and corresponding standard errors are needed. This can be obtained using a variety of trial designs.
It has at this stage not been clarified what 'substantial differences' mean. If a trial is so efficient that small differences for a characteristic are exposed as statistically significant, a further evaluation should tell if those differences are of economic importance.
Plot: experimental unit containing one or more planting positions for one treatment, here seedlot; block: a compact piece of land of uniform conditions and containing a group of plots (treatments), which can be compared under similar environmental conditions; replicate: in complete block designs, replicate is synonymous with block which contains all treatments in the experiment, in incomplete block designs, treatments are arranged in blocks which are smaller than a complete replicate, i.e. a replicate includes a group of blocks.
Estimated error variance per plot, s2: the residual sum of squares in the analysis of variance; Standard error per plot, s,. the square root of the error variance; Estimated error variance for difference between means of two seedlots, 2s2/b: two times the standard error per plot divided by the number of blocks in the experiment.
Precision: How close a measurement is to the average of a long series of measurements under same conditions; precision is expressed by the standard error, or the confidence interval around the mean value. Accuracy: How close a measurement is to the true value.
Relative Efficiency: If the error variance per plot, s2, is the same in two experiments, the error variance for the difference between two treatment means, 2s2/b, is evidently smallest in the experiment with most replicates, or the relative efficiency is higher for that experiment.
For an overall statistical analysis of data from several trials, including seedlot & trial-site interaction, it would be easiest and less error prone if all trials included the same treatments, were of the same plot-size and had the same number of replicates. However, experience shows that even in such balanced cases error and interaction often differ considerably from trial to trial. This can usually be handled, although there may be problems of analysis and evaluation.
When the experiments differ in size and structure, a combined analysis can usually be made provided that the same treatments are found in all experiments. However, a number of additional complications have to be tackled. The problems concern: i) weighting for different error varlances, ii) F-test of interaction against the error mean square, iii) F-test of treatment mean square against the interaction mean square, and iv) the type of distribution of the interaction mean square causes problems with the interpretation of the ordinary F-test. Most of these problems can be managed with modern computer programmes, although there would always be problems of interpretation of results.
Contrary to this there is no real efficient remedy for fully improving on situations where environmental conditions differ from plot to plot within blocks. Although there are several methods available to make adjustments for such differences, they should be considered as methods for salvage of information from trials that for some reason have non-optimal designs, rather than as methods to consider to employ ad priori.
See (5) for a review on the problems a'nd possible solutions for the combined analysis of trials of similar or different structure.
It must be emphasized that the best quality of the results from the analysis of each trial and from the overall analysis will depend on the design of each trial being chosen so it fits the local conditions of environment and resources available.
A poor design will limit or nullify possibilities of obtaining good information. It is difficult, sometimes impossible, to separate environmental variation from genetical variation in poor experiments, whereas modem statistical computer based programmes can manage many of the problems mentioned previously regarding different error variances and imbalance.
Give therefore higher prioritv to efforts to make the optimal design for the actual site conditions than efforts to maintain uniformity and size of design from trial to trials (8)!
The randomized block design has several advantages. It is usually the most simple to lay out, to administer and to analyze. However, when there are many treatments, this design can result in the formation of large blocks. The larger the blocks, the more possibility of too much environmental variation within blocks. Heterogenity within blocks is a normal and major source of potential problem in the analysis of single experiments. and in the interaction analysis!
Other designs, viz incomplete block designs could be useful where there is much environmental variation in the trial area, where there are very many treatments, or where the plant material, land, labour or other resources are limited. Incomplete block designs have not been used much in the past because of analysis problems especially when there is imbalance. There are now, however several computer based programmes available that facilitate analysis of different types of incomplete block designs. Because of the type of environmental variation mostly found in tree trials, the incomplete block designs will basically be more efficient for the same size of experimental area, or in other words, the same precision can be obtained for a trial size (area) that is smaller than for the RCB-design. Good accounts are found in (5) and (9). See also (4) for a comparison of different designs. An example of how to construct incomplete block designs is given in (9).
Block shape, orientation and size must correspond to the environmental variation. Environmental conditions must be uniform within blocks, and different between blocks, as far as possible. This is a critical and important consideration for obtaining good results.
For example, where there is a clear environmental gradient in a particular direction, then make long blocks that are oriented perpendicularly to that gradient; never along the gradient. Blocks or replicates need not be of regular or the same shape, nor do blocks need to be neighbouring.
The number of blocks will influence the precision that can be achieved in an experiment. The more blocks the higher the precision. The number of blocks should match the precision wanted. In the choice of number of blocks it must also be ensured that there is a sufficient total number of trees representing a seedlot in the experiment. A minimum of some 20-30 trees of each seedlot in a trial are required to give reasonable estimates of seedlot mean values, provided that sampling has been proper in every step. These trees can then be grouped into plots of different sizes, and the number of blocks will then for the same number of total trees vary accordingly; see below under plot size. See (5) and (4) for the methodology of finding the optimal number of replicates. See also below 'Size of experiment'.
In addition to a higher precision, more blocks would, and possibly more important, provide a higher insurance against loss of whole provenances in the case of partial occurrence of fire, flooding, human and livestock disturbance, etc. Such calamities have actually in many cases made it impossible to obtain good information from trials with few blocks. More blocks are a relatively simple insurance against calamities.
Many and small blocks (including small plots) could in fact also safeguard the longevity of a trial. Until competition begins, data would be collected and analyzed without complications. When competition starts, ordinary thinnings would be carried out until only one-two trees remain per plot. If the trial is to be kept and assessed for so long that competition starts again, then the possibility exists with a RCB design with many replicates to thin away all trees of whole provenances in each block and then merge plots and blocks, for example as in the following, where 4 blocks have been reduced to 2 (the small number of seedlots in the blocks are for illustration only):
| INITIAL LAY-OUT | |||
|---|---|---|---|
| BLOCK A | BLOCK B | ||
| Tl | T4 | Tl | T3 |
| T2 | T3 | T2 | T4 |
| T4 | T3 | Tl | T4 |
| Tl | T2 | T3 | T2 |
| BLOCK C | BLOCK D | ||
| AFTER FINAL THINNING | |||
| BLOCK 1 | BLOCK 11 | ||
| Tl | T3 | T2 | T3 |
| T4 | T2 | Tl | T4 |
This kind of design and thinning would need careful thinking in order to ensure an existing balance of seedlots, and a planned removal of seedlots may compromise the assumption of randomization of seedlots within blocks.
Finally, many replicates would make it possible to use some of the blocks for destructive assessment, for example for biomass studies.
An increasing plot size minimizes the effects of abnormal individual trees. Large plots can be used to exclude the effect of competition between adjacent plots by excluding the outer row from measurement and assessment, and are generally considered useful for trials that are to last for a long time after competition between plots has started. Small plots may be critical for a proper evaluation of certain characteristics as survival, stem form, branching habit.
The Plot size should be so that an adequate plot exists throughout the life of an experiment, and should not be extended beyond that requirement. Generally, the statistical efficiency of a trial decreases with an increasing number of trees per plot, see (4) for more details.
The smallest plot size is obtained using single-tree plots. These are very efficient in sampling. However, the main disadvantages are the complicated establishment (many blocks and plots!) and the problems of keeping the identity of plots when survival is less than 100 percent. Additional problems arise when thinnings are to be done. Single tree plots should only be used if the responsible organization is capable of handling these issues.
An important question in connection with small plots is about correlation between treatment effects, or independency of errors. There would often be average positive correlations between neighbouring plots due to similar environments early in the life of a trial, whereas correlations on the average would be negative due to competition between trees in neighbouring plots later in life (6). Proper randomization would ensure that valid statistical analysis in most cases still can be made, but the interpretation of the results differs with type of correlation.
When environmental variation is small within blocks (as it should be), the first type of correlation is of no concern, since treatments are compared under the same conditions. The correlation due to competition between plots increases with time, and the smaller the plots, the earlier will treatments be inter-dependent. It appears that generally treatment variances are inflated, i.e. we get too many significant differences, whereas treatment ranking is not affected. In the case of provenance trials, where the objective is to explore the ranking, not necessarily the absolute performance of each seedlot, the advantage of small plots may outweigh the slight problem of too many significant differences.
If there, nevertheless, is a strong wish to keep a trial with large plots for a longer period of time, the problem of competition can be overcome by excluding the trees in the outer rows from the assessment and analysis. The risk remains that within-block environmental variation may be too big. The solution in such a case is to use an incomplete block design where the large plot size can be maintained, while block size is kept small.
The smaller the plots, the shorter the period of time where the trial could yield good information. However, most of the useful information will be obtained early in the life of the trials. Experience shows that, in many cases, the possibility of retaining trials intact for an extended period of time is not high.
There would likely be a different number of plants available for the various seedlots, and some would have a critically low number. If plot size is small, then more seedlots would have a sufficient number of plants for inclusion in the trial. Higher priority should be given to include as many seedlots as possible rather than to maintain large plots! Below is given an example of utilizing most of the plants available.
21 seedlots are represented in the nursery. Assuming that all seedlings may be usable, the following can be made:
A small complete and balanced trial using 21 seedlots in 4 blocks and using 8-tree plots. The trees may be arranged in rows of 1 x 8 or 2 x 4 depending on the situation. The number of blocks is in the lower end, but all seedlots are tested against each other.
| Plants available in nursery - before taking plants for small complete trial | Surplus plants - after taking plants for small, complete trial | ||||||
| 34 | 44 | 2 | 12 | ||||
| 61 | 63 | 68 | 29 | 31 | 36 | ||
| 92 | 94 | 60 | 62 | ||||
| 123 | 146 | 152 | 91 | 114 | 120 | ||
| 176 | 177 | 177 | 178 | 144 | 145 | 145 | 146 |
| 196 | 198 | 201 | 218 | 164 | 166 | 169 | 186 |
| 244 | 256 | 364 | 212 | 224 | 332 | ||
Using surplus plants from establishing this trial (see table above), an additional trial can be established with 16 seedlots in 6 blocks using 9-tree plots. The trees may be arranged in 3 rows x 3 columns, or in line plots as deemed necessary. This can make an efficient trial.
The two trials may be placed adjacent, or in different places. Placing them adjacent under similar conditions would give some idea of efficiency of trials, and in different places, an estimate of interactions can be obtained.
Alternatively, Vietnam has suggested making 4 blocks with 14 24-tree plots, and the remaining plots with a different number of trees according to the number of available plants. In the analysis of variance seedlot means would then be estimated with different errors, but apart from this there would be no statistical objection to this arrangement. This is the most simple arrangement, but possibly not the most efficient.
The area of the first two trials using a spacing of 3.5 x 3.5 meters would be: 0.82 ha + 1.06 ha = 1.88 ha. The area of the fast arrangement would be (2 plots of 8 trees, 3 of 15, 2 of 20 and 14 of 25 trees in each block): 2.21 ha, or slightly larger than the total of the two other trials.
The shape of plots would depend on the type of variation found at the trial site. A plot must sample the present within-block environmental variation as completely as possible. Compact, e.g. square, plots are useful specifically where the environmental variation is of a patchy type, while line plots usually are best on an area with a gradient type of variation. Line plots are oriented along the gradient. The smaller the gradient, the less trees in the line. Line plots are generally much more efficient than square plots in as much as environmental variation is difficult to predict. See (6) for examples and further references.
Finally, the shape of the plots should ideally vary from block to block according to differences in types of variation. However, it would be easier for administrative reasons if plots were all of the same shape in the whole experiment, so it is advised to keep same plot shape and size within an experiment.
The distance between trees naturally influences the plot size. It was originally recommended to be 4 x 4 metres. It is left to the experience and discretion of the trial host organization to determine spacing, which would most likely vary between 3 x 3 metres and 4 x 4 metres. Basically, the better the growth conditions, the wider the spacing should be used, and vice-versa. On poorer sites (see below) it may be argued that competition between trees for water and nutrients would necessitate a wider spacing. However, all experience shows that invariably crown closure is delayed for too long, and weed competition increases severely, leading to increased costs of maintenance and loss of benefit of wide spacing.
The size of the total experimental area naturally influences the costs of establishment and maintenance. A trial of 20 seedlots, 25-tree plots, 4 replicates, and spacing of 4 x 4 metres would cover 3.2 hectares, exclusive of any buffer lines. The table opposite shows the net trial area (i.e. without buffer lines) for 20 seedlots for various combinations of plot-size, number of blocks, and spacings. For an area of approximately 3 hectares, the relative efficiency (see above) of trials with 25, 16 and 9 trees per plot is 1, 1.5, and 2.5, respectively. This means also that, for a specified precision, the experimental area may be reduced considerably by using small plots. If funds for establishment and maintenance are limited, a reduction in experimental area should be considered. Alternatively, the available funds may be for establishment of more trials.
It is often argued that trials may ultimately be used for procurement of material for propagation of the best seedlot numbers for planting purposes, and that design may be modified to suit such a purpose. However, it is recommended to use a trial design that is most efficient for reaching the initial objectives of the trial.
| Number of Trees in Plot | Number of Blocks | Spacing of Trees (metres) | ||
| 3 x 3 | 3.5 x 3.5 | 4 x 4 | ||
| Area in hectares | ||||
| 9 | 6 | 1.0 | 1.3 | 1.7 |
| 9 | 8 | 1.3 | 1.8 | 2.3 |
| 9 | 10 | 1.6 | 2.2 | 2.9 |
| 9 | 12 | 1.9 | 2.6 | 3.5 |
| 16 | 6 | 1.7 | 2.4 | 3.1 |
| 16 | 8 | 2.3 | 3.1 | 4.1 |
| 16 | 10 | 2.9 | 3.9 | 5.1 |
| 16 | 12 | 3.5 | 4.7 | 6.1 |
| 25 | 4 | 1.8 | 2.5 | 3.2 |
| 25 | 5 | 2.3 | 3.1 | 4.0 |
| 25 | 6 | 2.7 | 3.7 | 4.8 |
| 25 | 8 | 3.6 | 4.9 | 6.4 |
When the soil in the site of the field trial is moist to a depth of 25-30 cm, or according to local experience, planting may commence.
From each provenance in the nursery select systematically, as described in section 4 on page 4 the appropriate number of healthy seedlings. They should be 'hardened off' 1-2 weeks before they are transported to the site. This can be done by gradually reducing the watering and shading.
The seedlings should be well watered before transportation. They should be packed in wooden or plastic boxes to avoid damage during transportation. Ensure that the seedlings are not packed loosely. Avoid wind damage during transport by erecting appropriate screens on the truck.
After arrival at the planting site, place the seedlings in a protected, shaded area until planting. Water them thoroughly daily.
Planting holes should be 25 x 25 x 25-30 cm. Cultivate the topsoil around the hole to a distance of some 15 cm. Place locally available well processed compost in the bottom of the hole and cover with soil. Also, application of a pesticide is recommended in termite infested areas at planting time. This is best done by watering around the plants with an emulsion after planting
Planting should be done in the morning and/or evening, not in the hot part of the day. Slit the potting bag with a sharp knife or razor blade and carefully remove the plastic bag without breaking the soil or damaging the roots. If one or more roots have encircled the mass of roots, then cut such roots with a sharp knife. Set the seedling in the hole with its root collar level with the ground surface. Fill in the soil around the roots of the seedling and pack the soil well down around the roots to avoid large air pockets in the soil. Do not stamp and do not use heel. Mulch with dry grass. If possible, water seedlings should be watered with 2 lts of water. Replace dead seedlings with new ones within 1-2 months after planting.
To qualify for stumps, seedlings need to have the following dimensions: minimum 10 mm ideally 20 mm - at root collar; stem length from 40 cm to an ideal 100 cm.
Plants are lifted - using a well sharpened spade. The tap root may be pruned in the process, but, at least 30-40 cm of the root must always be left.
Immediately when the plants have been lifted from the bed, they are taken to a shady place and if necessary further protected from desiccation. The stumps are prepared as soon as possible as follows:
Stumps prepared this way may be stored for a few days in the shade at, or near, the planting site.
Stumps are planted preferably using a crowbar which is dumped repeatedly into the soil until a sufficiently deep hole has been made. Do not widen the hole by pulling the crowbar from side to side since this will leave a space at the bottom where the roots cannot get into contact with the soil.
Alternatively plants are lifted with a ball of earth, and the stems are pruned to leave 30-60 cm, but no root pruning is done). Depending on climate, leaves may be trimmed fully or partially. Planting is done as for pots, except that the hole should be adequately deep.
When seedlings produced in open-root beds are too small for stumps, they may as early as possible, and at least one month before planting to the field, be transplanted to pots of a suitable size to avoid damaging the root system too much. After a thorough watering, the plants are lifted with a ball of earth and planted in the pots with additional soil as required. There is no need to trim stem and leaves, but irrigation of the plants is essential.
The seedlings should be kept free from weed competition during the initial stages of development. Manual weeding is recommended. Frequency and method of weed control will depend upon particular site conditions. Intercropping should be avoided since this may cause differences in the treatments of plots.
Every 2-3 months during the first year, pull all weeds within 50 cm of each seedling, and cut all those in the remainder of the experimental area to prevent competition. If there are sufficient resources a complete clearing of all weeds is recommended. For succeeding years, weed as necessary.
Conduct regular weeding and removal of residue to ensure that fuel does not accumulate within the experimental area, especially during the dry season. Prepare a firebreak, 8-12 m wide, around the experimental area before the beginning of the dry season. Cut and remove vegetation or plough within the firebreak regularly during the dry season.
The experimental area should be fenced to prevent damage, particularly damage caused by animals. To help prevent human damage to trees, communicate to all neighbours the study objectives and long-term importance of the results to the local population. This will help gain support from the local community.
Monitor pests and diseases, especially ants, crickets,
or other insects. Appropriate control measures should be taken as necessary
up to age 1 year. After that, taking effective measures may be difficult
and costly.
