derstanding of gene function, which could provide opportu­nities to exploit these mechanisms in crop species (Foolad, 2004; Denby and Gehring, 2005). However, direct extrapo­lation of single gene responses, gained through Arabidopsis studies, to functional abiotic tolerance of cultivated crop species could continue to be limited by differences in gene sequence between Arabidopsis and crop species (Edmeades et al., 2004; White et al., 2004). Moreover, gene expression in Arabidopsis changes when exposed to field conditions (Miyazaki et al., 2004, as reviewed by White et al., 2004), as would be expected given the influence of genotype by en­vironment interactions. Genes for heat tolerance have been identified in a number of species, including rice, cowpea, and groundnut, which is likely to provide future opportuni­ties for heat-tolerance breeding.
        Attaining more effective use of genomics for abiotic stress-tolerance breeding will depend on closer integration of this discipline with physiology, which could lead to better understanding of how genes confer changes in whole-plant biological function and agronomic performance (genotype-to-phenotype relationships) (Edmeades et al., 2004; White et al., 2004). However, the current imbalance between ge-nomic research and field-based physiological studies, in fa­vor of the former, could undermine future AKST progress towards developing new stress-tolerant germplasm. Lastly, the scope of abiotic stress research needs to be extended to include more investigations of stress caused by mineral deficiencies and toxicities (Ishitani et al., 2004), as these factors strongly influence root development with implica­tions for tolerance to climatic extremes (Lynch and St. Clair, 2004). For example, many tropical agricultural soils have high levels of exchangeable Al which stunt root system de­velopment. Bringing mineral stress tolerance more closely into the realm of abiotic stress research, while increasing the complexity of the breeding challenge, could possibly avoid short-circuiting progress on drought, heat and salinity breeding efforts when scaling up to actual field conditions where multiple and complex stresses occur.
         Technological breakthroughs in breeding for abiotic stress tolerance could ultimately be limited by a potential loss of crop wild relatives to climate change. In the next 50 years, 16 to 22% of species that are wild relatives of pea­nut, potato, and cowpea could become extinct as a result of temperature increases and shifts in rainfall distribution, and most of the remaining species could lose over 50% of their range size (Jarvis et al., 2008). These three crops are important for food security in low-income countries, and their wild relatives are a vital genetic resource for develop­ing future drought and pest resistant crop varieties, as well as varieties with enhanced nutritional value. Greater efforts to collect seed for gene banks (ex situ conservation) and to target in situ conservation, such as through addressing habitat fragmentation, could help to mitigate these potential losses. Strengthening links between conservation, breeding, and farmers' groups is an important component of this ef­fort. However, diversity for its own sake is not useful, as farmers retain varieties for specific traits, not for the sake of conservation (Box 6-2).
        Agronomic and genetic improvement of underutilized (or "lost") crops could provide a good opportunity to en­hance   agricultural  diversification,  particularly  in  Africa

where approximately 2,000 underutilized food species are consumed (NRC, 1996). Crops such as the legume Bambara groundnut (Vigna subterranean) and the cereal fonio (Digi-taria exilis and Digitaria iburua) still figure prominently in the African diet. Fonio has very good prospects for semi-arid and upland areas because it is widely consumed, toler­ates poor soil and drought conditions, matures very quickly (6-8 weeks), and has an amino acid profile superior to to­day's major cereals (NRC, 1996). Unlocking the genetic potential of this cereal through conventional breeding and biotechnology to address low yields, small seeds, and seed shattering could help meet development and sustainability goals (Kuta et al., 2003; NRC, 1996). Similar potential ex­ists for Bambara groundnut (Azam-Ali, 2006; Azam-Ali et al., 2001), which is still cultivated from landraces. Research needs for underutilized crops include germplasm collection, marker assisted breeding, assessments of agronomic charac­teristics and nutritional content, development of improved processing technologies, and market analyses. While these crops cannot replace the major cereals, their improvement could significantly enhance food security options for rural communities confronted with climate change.
        Diversification of agriculture systems is likely to become an important strategy for enhancing the adaptive capacity of agriculture to climate change. Diversification strategies in the near term will need to be flexible, given that the disrup­tive impacts of climate change are projected to be experi­enced more in terms of increased variability, than as mean changes in climate. Therefore, improved skill in predicting how short-term climate phenomena, such as the El Niño Southern Oscillation and the North Atlantic Oscillation, affect seasonal and interannual variability, and the timely dissemination of forecasts will be essential for farmer deci­sions about whether to grow high or low water-consumptive crops and use of drought-tolerant varieties (Adams et al., 2003; Stige et al., 2006).

6.8.1.2 On-farm (low input) options
The knowledge and tools currently available could be better deployed to reduce the vulnerability of rainfed agriculture to seasonal climate variability. For example, poor crop estab­lishment is a significant but solvable constraint in semiarid farming environments (Harris, 2006). Similarly, seasonal dry spells can be bridged using improved rainfall catchment and incremental amounts of fertilizer (Rockström, 2004). By focusing on the "manageable part of climatic variability" (Rockström, 2004), AKST could have a significant positive impact on improving the adaptive capacity of rainfed agri­culture to climate change. It is also important to recognize that risk aversion practices are themselves an adaptation to climate variability, and to understand the functional link­ages between existing coping strategies and future climate change adaptation.
         The greatest period of risk in rainfed agriculture is the uncertainty around the timing of sufficient rainfall for crop sowing. High rainfall variability and poor quality seed leads to slow germination and emergence, causing patchy stands, and multiple and delayed replanting, making poor crop es­tablishment a significant contributor to the productivity gap in semiarid agriculture (Harris, 2006). Emphasis can be put on targeting technologies and practices that reduce the ex-