UC Santa Cruz

Effective ecological management of agroecosystems for both productivity and sustainability is by design a messy and complex task, producing problems which benefit from highly data-focused analyses. Here, three case studies using quantitative approaches to these problems are presented. First, using results from soil nitrogen monitoring in a long-term organic vegetable/strawberry cropping dataset, an ensemble machine learning model and process model are contrasted and used to reveal key drivers of soil mineral nitrogen asynchrony and loss potentials. Environmental factors, nutrient inputs, and management practices interact to determine the magnitude of nitrogen mineralization, and key combinations of these factors, such as early- or late-season disturbance and irrigation, may increase the risk of generating loss-vulnerable pools. Second, a Bayesian network modeling approach is used to synthesize data across multiple lab and field experiments, using cross-experimental data in an integrative manner, furthering our understanding of treatment dynamics in anaerobic soil disinfestation, an ecological soil pathogen control method, and providing a step forward in recommendations to strawberry growers seeking to optimize implementation in their systems. A strong relationship between carbon inputs and soil temperatures suggests that growers may be able to ease environmental restraints with additional inputs during treatment applications. Finally, an unsupervised cluster analysis is applied to a broad survey of on-farm management practices and approaches to disease control in walnut production, detecting two primary groups of divergent management practices. These groups, broadly characterized by moderate versus high levels of data and technology use, utilize markedly different approaches towards the integration of information and technology into on-farm management decisions.