n the developing world, where yield stability is important and expensive inputs are impracticable, new crop lines must be bred for their ability to adapt to a harsh environment and low-input techniques. That is the only way to food security .
        But how do you know you've bred a line which will grow well in the fields of the target area? The answer lies in the science of biometrics.  Crudely defined, biometrics is the science of statistics related to biological phenomena: what do the figures really mean?
        There is an image of the agricultural researcher as a man with a clipboard, a waterproof coat and a pair of rubber boots, crouching in a windswept field, frowning with concentration as he examines his latest experiment for signs of stripe rust or ascochyta blight. This is not the whole picture. Experimental results must be analyzed before you know what they are actually telling you. Biometrics is a key part of this process, and biometricians dispense with the muddy boots; instead, they manipulate statistics using an ever-growing arsenal of sophisticated software.
        ICARDA's biometrician, Dr. Murari Singh, explains how he and his colleagues tackled the problem of measuring adaptability to farmers' fields. "What we are talking about is G x E interaction, where G is the genotype--  that is, the plant material, or variety, you have bred and are testing--and E is the environment. What you have to measure is varietal sensitivity to the environment.
        "In fact, there has been considerable research on using statistics to measure that sensitivity. What we set out to do in 1991 was to take that a step further, and examine the potential of a variety for its transferrability within a given environment. Because even within a given ecoregion, exact circumstances are going to vary."
       A quick glance at the rainfall for ICARDA's research stations bears this out. Breda, Boueidar, Ghrerife and Jindiress are all in the same general region as the main headquarters research station at Tel Hadya, yet in September 1996 their accumulated rainfall for the season was 5.4, 1.6, 1.2 and 7.1mm, respectively. The figure for Tel Hadya itself was 22.9mm. Substantial variations may also be found in mean averages of minimum and maximum temperature.
        "Practically speaking, you can't breed for every conceivable variation in climate and soil, so you have to breed instead for adaptability. And if you have a variety adaptable to a number of  environments, you may like to examine if the same is adaptable/transferrable to a new (or an additional) environment," says Dr Singh.
        "To obtain an index of inter-site transferrability, we model the response of a variety, using linear regression, based on data from all the test sites except the target site, to which it is intended to be transferred; we then compare observed performance with the predicted performance at that site. This process is repeated for each site, considering it as a target site. The inter-site transferrability index is a function of the squared differences between observed and the predicted response, and the plot residual sum of squares with each site.

        "It is, in fact, more complicated than that because each value of the transferrability index obtained is judged against an environmental index, which itself is an integration of various conditions which could affect the transferrability of a variety to a particular site. This would certainly include rainfall, but it also takes account of soil type, mean maximum and minimum as well as average temperatures, incidence of pests and diseases and other factors. You have not just to include these, but also assess the way in which they act upon each other. Remember that we are trying to breed for adaptability to all stresses, including biotic ones such as insects as well as abiotic stresses like drought, moisture stress, heat and cold."
        Interim conclusions were published in Inter-site transferrability of crop varieties: another approach for analyzing multi-locational variety trials (Euphytica 89:305-311) in 1996. The authors were Dr Singh; Dr S. K. Yau of ICARDA's Germplasm Program; Dr John Hamblin, formerly of ICARDA and now at CLIMA in Australia; and Dr Enrico Porceddu of the Institute of Agricultural Biology in Viterbo, Italy, who is a Professor of Genetics and a former member of ICARDA's Board of Trustees.
        "A number of statistics to measure varietal sensitivity to environmental variation have been introduced...based on the response of a genotype to a changing environment...These statistics play an important role in studying G x E interaction, but this in itself is not the final step in the process of developing, evaluating and releasing a variety. The present approach takes the varietal assessment process further to the level where the potential of the variety could  be estimated for its transferrability within the target domain," they write.
        Since 1991, Dr Singh and the ICARDA breeders have been testing the viability of the statistical model they have obtained.
        It is an example of the sort of work that must be done using modern software tools if experiment results are to bring maximum benefit to the farmers. Biometrics is, in fact, part of the Computer and Biometrics Support Unit (CBSU) of ICARDA, which, besides maintaining and developing the Center's computer hardware and software, also carries out research computing; it is making growing use of Geographic Information Systems (GIS) for identification of water sources, environmental and degradation hazards, land-use patterns and sources of biodiversity, and has also developed research application tools such as the Trials Management System (TMS), a Windows-based program that helps scientists plan, manage and analyze data and report results, manage meterological information through the METDB (Meteorological Data Base) system and keep track of their experiments from the desktop.
        It is hoped that these developments will be covered in future issues of Caravan.
        In the meantime, anyone who feels that the work done by Dr Singh and his colleagues could have a useful application to their own trials should contact him at ICARDA, or by email to M.SINGH@CGIAR.ORG. He promises not to bring muddy boots into your laboratory...