tion. For example the current scale of the FAO world soil maps at 1:5,000,000 needs finer resolution (FAO, 2000).

Early warning, forecasting systems. Timely forecasts, in­cluding the starting date of the rainy season, average weather conditions over the coming season, conditions within the sea­son that are critical to staple crops and animals, and appropri­ate responses can increase the economic, environmental, and social stability of agricultural systems and associated commu­nities. Advances in atmospheric and ocean sciences, a better understanding of global climate, and investments in moni­toring of the tropical oceans have increased forecasting skill at seasonal to interannual timescales. Early warning systems using seasonal forecasts (such as the FAO Global Informa­tion and Early Warning System) and monitoring of local commodity markets, are increasingly used to predict likely food shortfalls with enough advance warning for effective responses by marketing systems and downstream users.
        Traditional coping mechanisms depend on the ability to anticipate hazard patterns, which are increasingly erratic with the advent of climate change. One option for improv­ing early detection and warning would be to broaden the use of GIS-based methodologies such as those employed by the Conflict Early Warning and Response Network (CEWARN), the Global Public Health Information Network (G-PHIN). Early warning  systems  are  important  because  they help to untangle the multiple but interdependent crises that characterize complex emergencies, particularly in response to climate change. In other words, continuous information gath­ering serves to identify the socioecological ingredients of com­plex crises before they escalate into widespread violence. This means technological systems are also needed. To this end, the added value of technological early warning systems should therefore be judged on their empowerment of local people-centered systems that build on the capacity of disaster-af­fected communities to recover with little external assistance following a disaster. Further applied research is needed on local human adaptability in decentralized settings as well as self-adaptation in dynamic disaster environments.
        Linking early warning to more effective response re­quires a people-centered approach to climate change (UN, 2006). The quest for early warning must be more than just an "exercise in understanding how what is happening over there comes be known by us over here" (Adelman, 1998). Instead, the international community should focus on the real stakeholders and add to their capacity for social resil­ience. On the policy front, the lack of institutionalized early warning systems that survey the localized impact of climate change on ecological and political crises inhibits the formula­tion of evidence-based interventions (Levy and Meier, 2004). Regrettably, little collaboration currently exists between the disaster management and conflict prevention communities despite obvious parallels in risk assessments, monitoring and warning, dissemination and communication, response capability and impact evaluation (Meier, 2007).
        Bringing climate prediction to bear on the needs of ag­riculture requires increasing observational networks in the most vulnerable regions, further improvements in forecast accuracy, integrating seasonal prediction with information at shorter and longer time scales, embedding crop models within climate models, enhanced use of remote sensing,

quantitative evidence of the utility of forecasts for agricul­tural risk management, enhanced stakeholder participation, and commodity trade and storage applications (Giles 2005; Hansen, 2005; Hansen et al., 2006; Doblas-Reyes et al., 2006; Sivakumar, 2006). For seasonal climate forecasts to be an effective adaptation tool, advances in forecasting skills need to be matched with better pathways for dissemination and application, such as by linking forecasts to broader livelihood and development priorities, and by training or­ganizations, such as extension agencies, to facilitate the end users' ability to make effective decisions in response to forecasts (Ziervogel 2004; Garbrecht et al., 2005; Hansen 2005; Vogel and O'Brien, 2006). Substantial investments by national and international agricultural and meteorological services are needed.

Improve crop breeding potential for drought, salinity and heat tolerance. Abiotic stress of agricultural crops is expect­ed to increase in most regions due to warmer temperatures, experienced both as episodic heat waves and mean tempera­ture elevation, prolonged dry spells and drought, excess soil moisture, and salinity linked to higher evapotranspiration rates and salt intrusion. Expected temperature increases of 2-3°C by mid-century could significantly impair productiv­ity of important staple crops of the developing world, such as wheat, and in truly marginal areas, millet. One-third of irrigated agricultural lands worldwide are affected by high salinity, and the area of salt-affected soils is expected to in­crease at a rate of 10% per year (Foolad, 2004). The mag­nitude of these impacts could test our capacity to achieve breakthroughs in germplasm improvement equivalent to the challenge at hand.
         Advances in plant genomics, linked to the Arabidopsis model system, and the integration of genomics with physiol­ogy and conventional plant breeding could lead to the devel­opment of new varieties with enhanced tolerance to drought, heat, and salinity. Emerging genomic tools with future po­tential include whole-genome microarrays, marker-assisted selection using quantitative trait loci, bioinformatics, and microRNAs (Edmeades et al., 2004; Foolad, 2004; Ishitani et al., 2004; White et al., 2004; Denby and Gehring, 2005). Phenological adaptation, e.g., matching crop duration to available season length, is central to successful breeding ef­forts; thus conventional breeding, augmented with genomic tools, is a likely configuration of future plant breeding pro­grams. An example of this would be the integration of phe-notyping (differences in crop germplasm performance under different stress environments) with functional genomic ap­proaches for identifying genes and mechanisms (Edmeades et al., 2004; Ishitani et al., 2004). Improvement in seasonal forecasting and in the use of remote sensing and other obser­vational tools could also be used to further support breed­ing programs, through better characterization of cropping environments.
         Future breakthroughs in understanding how crop plants respond to abiotic stress are very likely, given the scientific resources dedicated to investigating the Arabidopsis thali-ana, a model system used for plant genetics and genomics studies with a small, completely sequenced genome and a short life cycle. For example, progress in genomics related to salt tolerance in Arabidopsis mutants has enhanced un-