J. R. Anderson
This opportunity to react to contemporary work on climate prediction in agriculture is a welcome one for someone who occasionally and mainly youthfully dabbled in the influence of climate in agriculture (e.g. Anderson 1970, 1979, 1981, 1985, 1991; Anderson and Dillon 1988a; Anderson and Hardaker 1973; Anderson and Hazell 1989), who was excited at the prospects for informative predictions (e.g. Byerlee and Anderson 1969, 1982; Anderson and Dillon 1992) but who has long since been far too remote from the action. Accordingly, to jump across the decades of progress, the point of departure taken here is the opening keynote address by Sivakumar (2006), in which the state of the art is succinctly summarized, albeit in a way that emphasizes the possibilities in a guardedly positively manner. Intriguingly, and seemingly properly in the view of this observer, he uses cautious words such as "could help" when charting the situations where climate forecasting efforts are intended to assist farmers and other agricultural managers in their decisions in the face of climatic uncertainty.
Need for the Assessment of the Value of Climate Forecasts
Workshop participants dealt variously with a variety of interrelated phenomena where sometimes use of terms was less cautious, particularly when skipping among "weather, climate, climate change, etc.", where timescales are surely critical but open to opinion or interpretation. The paper of Meinke et al. (2006) was helpful in sorting out these semantics and thus spares such an attempt here, which had it been broached would have been heavily influenced by the recent work of Zillman et al. (2005). Suffice it to say that "forecast" and "prediction" cover many interpretations, such as: categoric vs. probabilistic; concrete/specific vs. descriptive; etc. so it is not too surprising that analysts and users are often talking at cross purposes. Indeed, such a use happens regularly in other related fields such as analysis of "risk", "uncertainty", "variability", "vulnerability" - again, semantic issues for another time and place (e.g. Anderson et al. 1987; Hardaker et al. 2004).
The latter works just cited deal with measuring forecast value from what is usually referred to as a Bayesian perspective. In this view, information as encapsulated in some type of climate forecast has value when it can influence behavior/decisions. Such information usually also has a cost. So, whether it has positive net value is an empirical question that can be posed both before the forecast is issued, and after (ex ante and ex post). Evidence on this question has been sparse in this Workshop, although it would seem that it should be a key item. One might be tempted to ask in the same vein, "Has CLIMAG been worth the investment in and around it?" The answer is not immediately obvious.
To return briefly to the Bayesian approach to forecasting in an uncertain world (in the spirit of Hardaker et al. 2004): prior probabilities attached to possible states of nature represent uncertainty held before a forecast; forecast information is captured in likelihood probabilities; posterior probabilities come from combining these and can serve as the updated weights to use in decision analysis. Such revision cycles can be treated sequentially, i.e. dynamically, in what constitutes an ongoing learning approach. But such an approach needs to be teamed up with models that represent production decisions about inputs and outputs, such as introduced by Msangi et al. (2006), as Eq. 27.1:
where Q is typically multi-enterprise agricultural output, X is conventional inputs (e.g. land, labor, capital, conventional inputs such as fertilizer), Z is unconventional inputs (e.g. infrastructure), K is technical knowledge (e.g. R&D investment), and U is uncontrollable factors (e.g. weather). It is the fact of interaction between the X and the Z variables that gives probabilistic information on the Zs its potential value (e.g. Byerlee and Anderson 1969). Such production models are often estimated pragmatically, almost by definition simplistically and frequently badly but without some such, little can be done to bring climate forecast information explicitly into decision analysis and valuation of worth.
Estimation, whether done via econometrics, programming, or other methods (such as ad hoc simulation models built around crop growth models), is inherently demanding (e.g. Dillon and Anderson 1990): of conceptualization, including dynamics and participatory insights; of data, especially in LDCs; of estimational skills; of optimization skills; and last but not least, of interpretation skills. The work must also encompass modeling of behavioral factors, which adds to the challenge. For instance, representing risk and lifestyle preferences is a non-trivial step, although it seems reasonable to routinely allow for some degree of risk-aversion (Hardaker et al. (2004) argue for modest levels only). The ability of farmers to adjust should also usually be explicitly accounted, so part of the estimational challenge is to model the possible constraints to adjustment in response to emerging information. Farmers and others are all swimming in the stormy seas of risk, with or without formal climate forecasts. Are such forecasts a marginal part of the picture? This is a good question that can be answered only by careful empirical analysis. Needless to say, given the range of phenomena that must be modeled, a wide range of disciplinary skills is necessarily involved in such demanding research work.
Viewing climatic forecasting work as a particular type of research endeavor naturally raises the question of whether investment in it will be blessed with the same sort of typically high returns that have characterized more conventional agricultural research such as that related to crop improvement and productivity enhancement more generally (e.g. Alston et al. 2000). From the Sivakumar (2006) overview and the material presented at this workshop, it seems the evidence is not yet available to reach a solid conclusion on this, especially given the evident scarcity of formal accounting of the costs of climate prediction work. So, in the meantime, it seems analysts need to strive to provide cogent evaluative evidence that can serve in part to deal with the implicit "competition" for funding that arises from mainstream agricultural research. Of course, some of the "conventional" research products that will have potentially high payoffs in responding to climate predictions present particular new evaluation tasks (e.g. appropriately valuing novel short-cycle cultivars that can "escape" or others that can better "endure" some droughts).
Research themes beyond the usual purview of conventional agricultural research are also necessarily involved in understanding the broad context in which climate prediction research takes place. At the risk of stating the obvious, there is clearly a need to better understand the mechanisms that diverse rural communities use for: managing risk, e.g. borrowing finance, selling assets, choosing technologies, etc.; coping with risk, e.g. calling on friends an relatives in times of need; shifting from risk, e.g. migrating, on a temporary or permanent basis, and so on - a field too large to delve into here (but see e.g. Anderson and Roumasset 1996; Anderson 2003). Agro-meteorologists may not have spent much time grappling with rural financial systems, futures markets, etc. but maybe they will have to do so increasingly? Or perhaps they may elect to work in more engaged manners with research workers who do focus on such themes?
Finally, to return to the title of this brief perspective, some policy dimensions pertaining to climate prediction work should be noted, inevitably reflecting efforts past (e.g. Anderson et al. 1987; Anderson and Dillon 1988b) and more recent (e.g. Anderson and Hazell 1997; Anderson 2000, 2003). As climate predictions inherently serve to modify the environment in which farmer choice is made, good policy making should logically be founded on good understanding of farmer risk management more generally, since climate is just one of the risks in that environment. In some countries uncertainty about property rights (especially land) may be of profoundly greater significance than climate outcomes, especially which it comes to on-farm or within-supply-chain investment decisions. Other enabling aspects such as private sector development (PSD) naturally impinge on decisions more generally, in what are increasingly alluded to as investment climate limitations. The world has largely entered an era where novel financial instruments (such as warehouse receipts, forward contracts, etc. largely provided or managed by private suppliers) can be used for more effective risk management, whether risks arise from the natural environment such as climate or from the economic and political environment (e.g. Larson et al. 2004).
One aspect of PSD of direct consequence for climate prediction is the state of development of the insurance industry in areas that are targeted by climate predictors. There are many contemporary developments in this industry that expand the opportunities open to risk managers along the whole chain from plot to plate. In particular, new products based on index insurance are becoming increasingly available, such as rainfall insurance (e.g. Hess 2004; Hess and Syroka 2005). Given the dependence of such instruments on the timely availability of reliable meteorological data, this provides a natural point of intersection between the climate forecasting community and the agricultural risk management community.
Other communities are also relevant as climate forecasters reach out to their diverse clients. Those who plan emergency policy and intervention are members of one such group, and it seems that implementation of improved safety nets is something of a growth industry around the developing world, and is one that needs to be informed by the fruits of research on climate forecasting. Climate policy making is still usually something of an infant industry but is surely one that should be closely linked with climate research. So, the agenda for climate predictors is large, diverse and challenging (including the closer attention to monitoring and evaluation that Sivakumar called for at the outset of this workshop), and would-be predictors are to be enthusiastically assigned every good wish for success as they emerge from a honeymoon phase for what should ultimately be socially valuable contributions to the future of agriculture and humanity on which it depends.
Alston JM, Chan-Kang C, Marra MC, Pardey P, Wyatt TJ (2000) A meta-analysis of rates of return to agricultural R&D: ex pede Herculem? International Food Policy Research Institute, Washington, USA (Research Report 113)
Anderson JR (1970) Rainfall correlations in the pastoral zone of eastern Australia. Austr Meteorol Mag 18:94-101
Anderson JR (1979) Impacts of climatic variability in Australian agriculture: a review. Rev Mark Agr Econ 49:147-177
Anderson JR (1981) Meteorological services in agronomic and economic evaluations of risk. In: WMO (ed) World climate program. WMO, Geneva, Switzerland (Proceedings of the Technical Conference on Climate - Asia and Western Pacific, WMO-No 578, pp 205-209) Anderson JR (1985) Economic impacts. In: Houghton DD (ed) Handbook of Applied Meteorology. Wiley,
New York (Part IV: Societal impacts, Chapter 41, pp 205-209) Anderson JR (1991) A framework for examining the impacts of climatic variability. In: Muchow RC, Bellamy JA (eds) Climatic risk in crop production: models and management for the semiarid tropics and subtropics. CABI, Wallingford, pp 3-17 Anderson JR (2000) Response and risk in rural ecosystems: from models and plots to defined universes.
Agr Ecosyst Environ 82:261-271 Anderson JR (2003) Risk in rural development: challenges for managers and policy makers. Agr Syst
Anderson JR, Dillon JL (1988a) Socioeconomic impacts of climatic variability. In: Parry ML, Carter TR, Konijn NT (eds) The impacts of climatic variations on agriculture. Kluwer, Dordrecht, The Netherlands (vol II: Assessments in semi-arid regions, pp 719-732) Anderson JR, Dillon JL (1988b) Implications for policymakers and planners. In: Parry ML, Carter TR, Konijn NT (eds) The impacts of climatic variations on agriculture. Kluwer, Dordrecht, The Netherlands (vol II: Assessments in semi-arid regions, pp 733-757) Anderson JR, Dillon JL (1992) Risk analysis in dryland farming systems. FAO, Rome, Italy (Farm Systems Management Series 2) Anderson JR, Hardaker JB (1973) Management decisions and drought. In: Lovett JV (ed) The environmental, economic and social significance of drought. Angus and Robertson, Sydney, pp 220-244 Anderson JR, Hazell PBR (eds) (1989) Variability in grain yields: implications for agricultural research and policy in developing countries. Johns Hopkins University Press for IFPRI, Baltimore Anderson JR, Hazell PBR (1997) Risk considerations in agricultural policy-making. In: Huirne RBM, Hardaker JB, Dijkhuizen AA (eds) Risk management strategies in agriculture: state of the art and future perspectives. Mansholt Institute, Wageningen, The Netherlands (EUNITA AIR3-CT-1654, pp 273-284)
Anderson JR, Roumasset JA (1996) Food insecurity and stochastic aspects of poverty. Asian J Agr Econ 2:53-66
Anderson JR, Dillon JL, Hardaker JB (1977) Agricultural decision analysis. Iowa State University Press, Ames, IA, USA
Anderson JR, Hazell PBR, Evans LT (1987) Variability of cereal yields: sources of change and implications for agricultural research and policy. Food Policy 12:199-212 Byerlee DR, Anderson JR (1969) Value of predictors of uncontrolled factors in response functions. Austr J Agr Econ 13:118-127
Byerlee DR, Anderson JR (1982) Risk, utility and the value of information in farmer decision making.
Rev Mark Agr Econ 50:231-246 Dillon JL, Anderson JR (1990) The analysis of response in crop and livestock production, 3rd edn. Pergamon, Oxford
Hardaker JB Huirne RBM, Anderson JR, Lien G (2004) Coping with risk in agriculture, 2nd edn. CABI Publishing, Wallingford
Hess U (2004) Innovative financial services for India, monsoon indexed lending and insurance for smallholders. World Bank, Washington, DC, USA (Agriculture and Rural Development, ARD Working Paper 9)
Hess U, Syroka J (2005) Weather-based insurance in southern Africa: the case of Malawi. World Bank,
Washington, DC, USA (Agriculture and Rural Development Discussion Paper 13) Larson DF, Anderson JR, Varangis P (2004) Policies on managing risk in agricultural markets. World
Bank Res Obser 19:199-230 Meinke H, Nelson R, Stone R, Selvaraju R, Baethgen W (2006) Actionable climate knowledge - from analysis to synthesis. Climate Res 33(1):101-110 Msangi S, Rosegrant MW, You L (2006) Advancing ex-post assessment methods of climate forecast impacts. Climate Res 33(1):67-79 Sivakumar MVK (2006) Climate prediction and agriculture: current status and future challenges. Climate Res 33(1):3-17
Zillman J, McKibbin WJ, Kellow A (2005) Uncertainty and climate change: the challenge for policy. Academy of the Social Sciences in Australia, Canberra (Occasional Paper 2, Policy Paper 3, available at www.ASSA.edu.au)
This page intentionally blank
Was this article helpful?