Forecasting the Performance of Agroforestry Systems

Agroforestry has received considerable attention from scientists and development practitioners in recent years. It is recognized as a cornerstone of many traditional agricultural systems, as well as a new option for sustainable land management in currently treeless agricultural landscapes. Agroforestry systems are diverse, but most manifestations supply substantial ecosystem services, including marketable tree products, soil fertility, water cycle regulation, wildlife habitat and carbon sequestration. While these benefits have been well documented for many existing systems, projecting the outcomes of introducing new agroforestry systems, or forecasting system performance under changing environmental or climatic conditions, remains a substantial challenge.

Due to the various interactions between system components, the multiple benefits produced by trees and crops, and the host of environmental, socioeconomic and cultural factors that shape agroforestry systems, mechanistic models of such systems quickly become very complex. They then require a lot of data for site-specific calibration, which presents a challenge for their use in new environmental and climatic domains, especially in data-scarce environments. For supporting decisions on the scaling up of agroforestry technologies, new projection methods are needed that can capture system complexity to an adequate degree, while taking full account of the fact that data on many system variables will virtually always be highly uncertain.

This paper explores what projection methods are needed for supplying decision-makers with useful information on the performance of agroforestry in new places or new climates. Existing methods are discussed in light of these methodological needs. Finally, a participatory approach to performance projection is proposed that captures system dynamics in a holistic manner and makes probabilistic projections about expected system performance. This approach avoids the temptation to take spuriously precise model results at face value, and it is able to make predictions even where data is scarce. It thus provides a rapid and honest assessment option that can quickly supply decision-makers with system performance estimates, offering an opportunity to improve the targeting of agroforestry interventions.