The need for an efficient evaluation method
Evaluation results have often limited applicability, due to the considerable genotype x environment interaction. Extensive multilocational or multiseasonal evaluation would be needed to assess plant characteristics, but this expensive procedure can often not be applied to large germplasm collections. Moreover, extensive evaluations may still not lead to unequivocal results. Therefore, an efficient method is developed that allows the assessment of plant characteristics of large numbers of accessions under various environmental conditions, utilizing a limited amount of evaluation results.
Collection of durum wheat landraces
A total of 185 landrace populations of durum wheat [ Triticumturgidum L. var. durum (Desf.) MK] was collected in the Syrian Arab Republic in 1987 and 1988. During the collection missions, attention was paid to recording passport and collection information, and providing an agro-ecological description of each site. Farmers were asked for information on the landraces and their farm management practices.
Grouping of the 166 collection sites, based on four climatological variables, resulted in 14 relatively homogeneous regions of origin with respect to agro-ecological characteristics. Fifty-nine populations had their putative origin at or in the vicinity of their collection sites, and were considered representative for the respective environments.
A brief morphological description was given of the landrace groups as distinguished by farmers. Each group has distinct morphological characteristics, and is referred to by a local name that is often derived from a striking character or supposed origin. Some landrace groups are widely distributed over most of the country, whereas most others are regionally concentrated, either in a limited region or a few villages.
Thirty-eight populations were evaluated for agronomic performance at four locations, characterized by different long-term rainfall averages, over two seasons, and at two levels of nutrient availability. Actual season 1 rainfall was below average (viz. below 300 mm), although at the location with highest average yield, residual moisture had to be assumed.
Considerable differences in average yield were observed among experiments. Performance per population varied highly among locations, and populations showing superior performance under all conditions could not be identified. Modern varieties (incorporated in the experiments as check varieties) showed superior performance under best growing conditions, and landraces appeared competitive under more marginal conditions. Both modern varieties and landraces were not able to maintain yield levels under adverse growing conditions. Fertilizer application resulted in many cases in decreased kernel density, and delay of anthesis by one day, which could be an indirect effect of nitrogen shortage through increased canopy temperature.
Analysis of variance
Analysis of variance showed that most year, location, population and fertilizer effects were significant, as were interactions between year and location. Yield stability, which is an important characteristic of landraces, was indicated by the absence of significant effects for grain yield between years, within sites, and by the absence of a population effect for total dry matter production.
Through ANOVA, sources of variation are identified. This technique is therefore useful in preliminary, qualitative analysis of evaluation trials. However, it does not contribute to an understanding of processes underlying variation, and has therefore little capabilities in exploring plant behaviour under other environmental conditions.
Simulation of growth and development
Analysis of growth and development with a simulation model showed that total dry matter production was higher at higher total, and higher spring rainfall, and that at higher levels of moisture availability, more aboveground dry matter was produced per unit rainfall. Moisture and nitrogen availability interacted: at very low levels of moisture availability, this was the dominant growth limiting factor, whereas at higher levels, nitrogen recovery increased, and nitrogen availability became an additional growth limiting factor.
Under favourable environmental conditions, differences in weather conditions are reflected in the source size, i.e. the residual reserve carbohydrates at the end of grain fill. Genotypical differences in the balance between maintenance requirements and assimilation after anthesis determine to what extent the reserves are distributed to the sink. Under adverse growing conditions, severe drought stress after anthesis results in rapid senescence and cessation of grain fill, so that the reserve pool is not depleted. Under these conditions, higher grain yields are the result of increased remobilization of reserve carbohydrates, mainly due to higher kernel density.
The crop growth simulation model presented a comprehensive tool to increase insight in plant growth processes. Despite some shortcomings, the model simulated a recognizable durum wheat crop and reproduced in a consistent way genotype x environment interactions and their effects on yields. The model could be used in exploring crop behaviour under different environmental conditions, and replace extensive multilocational and multiseasonal evaluations.
It appeared possible to relate some collection site and plant characteristics at evaluation, which could be used to select germplasm on the basis of knowledge on the environment in the region of provenance. As this approach provides only general understanding, it is useful in preliminary evaluations that are aimed to identify groups of accessions in which further selections best can be made.
The germplasm was additionally evaluated for early and late season frost tolerance. All populations were well adapted to moderate early season frosts, which is characteristic for the Syrian climate. However, tolerance to late season frosts, which is uncharacteristic, was highest in populations from the inland, where minimum winter temperatures are lowest. Application of this relation is straightforward: in search for late season frost tolerance, breeding material can best be selected from inland populations, where chances of success are highest.
For fungal disease resistances, however, relations were not always cases clear. Only in the case of two of the three evaluated diseases, host resistance was highest in germplasm originating from regions with environmental characteristics favourable for development and incidence of the concerned disease. An efficient evaluation method is therefore difficult to formulate. Moreover, as the frequency of plants with specific vertical resistance can be very low, it may remain necessary to evaluate an entire collection.
It is useful to evaluate at least part of a germplasm collection at several locations over a number of seasons. This will provide data to analyze the variation patterns with ANOVA, validate a crop growth model, and determine possible relations between plant and collection site characteristics.
Subsequently, two evaluation methods can be followed, dependent upon the significance of the relationship between the observed plant character and environmental characteristic of the collection site. If such relationships are significant and causal, then preliminary selection of germplasm can be based upon collection site characteristics. Example are tolerance to late season frosts, and to a lesser extent some fungal disease resistance. If such relationships are difficult to establish, as for instance yield, then application of a crop growth simulation model appears a suitable technique in analysis of evaluation results of extensive germplasm collections, and for exploring growth and development under various environmental conditions. Whereas only a limited part of a large germplasm collection is evaluated multi locationally and multiseasonally, crop growth analysis can be applied to an entire collection. This would alleviate quantitative and qualitative limitations to evaluation programmes.
Landrace groups were generally more homogeneous for plant characteristics than regions of origin. Therefore, a pre-selection of germplasm could be made on the basis of landrace groups to restrict the number of accessions to evaluate. Of each group a small number of populations could be chosen, randomly or on the basis of origin, and evaluated, which would provide an indication of the characteristics of certain landrace groups. If necessary, this could be completed by a more detailed evaluation of promising material.
Genetic diversity was observed at four levels: within field populations, among populations at farm or village level, and belong to the same or different landrace groups, among regions, and as species mixtures in mountainous areas.
The heterogeneity of landrace populations complicates evaluation. Crop growth simulation models at the population level do not explain the consequences of heterogeneity in different environments. However, simulation at the population level was sufficiently accurate to justify application in interpretation of preliminary evaluation. In subsequent evaluation and selection, it may be necessary to quantify intrapopulation variation, to determine breeding potentials of best performing populations.
It is argued that landrace groups have been domesticated or cultivated for a long period, within relatively small areas with specific climatic conditions, followed by dispersion of some groups without losing characteristics. This probably has occurred only recently, as otherwise dispersed landrace groups would have been less uniform in agronomic characteristics due to environmental pressure.
Both regions and landrace groups must be sampled representatively. Hence, per agro-ecological region all landrace groups should be collected. Moreover, from each population a sufficient number of genotypes must be obtained. Although intrapopulation variation may be relatively low, it can be high in absolute terms, as in germplasm from mountainous areas.
Collection missions require a regional classification, such that the regions are relatively homogeneous with respect to environmental characteristics. In large zones, environmental conditions may be too heterogeneous to determine relations between environmental and plant characteristics. Therefore, large zones are not helpful in elucidating distribution and variation patterns, and setting priorities, whereas classification in many small regions neither provides much clarification, as neighbouringsmallregions are then likely to be similar, which limits the possibilities for discrimination.
Implications for genebank management
If a crop growth model is to be used in analysis of evaluation results, the evaluations themselves should meet some specific demands. Data on plant characteristics for which model performance is sensitive must be obtained. Some can easily be recorded in routine preliminary evaluations, if necessary for all accessions. Characteristics that require much effort to record, can be determined on few selected genotypes, and subsequently interpreted such that they are representative for the entire collection.
Appropriate computer hard- and software is required, and simulation models need to be available. Although for this study a fairly complex model was used, it may be more appropriate to apply summary models. These are small, comprehensive, and easy to parameterize for other plant characteristics or other crops. Training in modelling and simulation will be necessary for curators who are unacquainted with this.
Extensive data sets containing weather data and detailed soil characterization of locations in various agro-ecological regions have to be available at the genebank. Germplasm documentation has to be reorganized, as the more complex information on plant characteristics has to be pro
cessed and presented.
Simulation results need to be confirmed by field evaluation of a part of the collection. Therefore, a genebank will need access to evaluation sites characterized by distinct environmental conditions.
|Doctor of Philosophy
|18 Nov 1992
|Place of Publication
|Published - 18 Nov 1992
- computer simulation
- simulation models
- gene banks
- genetic resources
- resource conservation
- plant genetic resources