Main content area

Modelling livelihoods and household resilience to droughts using Bayesian networks

Merritt, Wendy S., Patch, Brendan, Reddy, V. Ratna, Syme, Geoffrey J.
Environment, development and sustainability 2016 v.18 no.2 pp. 315-346
Bayesian theory, analytical methods, drought, groundwater, household surveys, irrigation, landscapes, livelihood, models, monitoring, natural capital, retrospective studies, socioeconomics, watersheds, India
Over the last four decades, the Indian government has been investing heavily in watershed development (WSD) programmes that are intended to improve the livelihoods of rural agrarian communities and maintain or improve natural resource condition. Given the massive investment in WSD in India, and the recent shift from micro-scale programmes (<500 ha) to meso-scale (~5000 ha) clusters, robust methodological frameworks are needed to measure and analyse impacts of interventions across landscapes as well as between and within communities. In this paper, the sustainable livelihoods framework is implemented using Bayesian networks (BNs) to develop models of drought resilience and household livelihoods. Analysis of the natural capital component model provides little evidence that watershed development has influenced household resilience to drought and indicators of natural capital, beyond an increased area of irrigation due to greater access to groundwater. BNs have proved a valuable tool for implementing the sustainable livelihoods framework in a retrospective evaluation of implemented WSD programmes. Many of the challenges of evaluating watershed interventions using BNs are the same as for other analytical approaches. These are reliance on retrospective studies, identification and measurement of relevant indicators and isolating intervention impacts from contemporaneous events. The establishment of core biophysical and socio-economic indicators measured through longitudinal household surveys and monitoring programmes will be critical to the success of BNs as an evaluation tool for meso-scale WSD.