Data science is a powerful tool that supports decision-making in sectors as diverse as healthcare, education, and finance. As the costs of natural disasters continue to rise against the backdrop of a changing climate, data science can play an important role in helping to identify climate-adaptive management pathways for natural resources systems. This dissertation will introduce specific data-driven approaches to identify robust pathways to reduce the impact of climate extremes in agricultural and water resources systems. Several pitfalls of statistical modeling for prediction in the era of climate change are introduced. Specifically, local surface meteorological variables are thermodynamically coupled in a way that impedes our ability to diagnose causality among a suite of meteorological variables in observational analysis of climate stress. Climate change will impact both large-scale dynamical and thermodynamical atmospheric processes, and will lead to non-linear shifts in local climate. A diagnostic-predictive framework is forwarded as a way to build statistical models that are in line with process-based understanding of environmental systems, and provide tailored information to support climate-adaptive management.