Project Lead: Masha Vernik, UW Environmental and Forest Sciences

Data Science Lead: Curtis Atkisson

As seasons become less predictable and extreme weather events become more intense and frequent, farmers learn to adapt. Adaptation can mean anything from building new infrastructure that’s out of a flood zone to installing improved irrigation lines; it can mean shifting crop plans or building deeper relationships with customers and fellow farmers. But for some farmers, adaptation is easier than for others.

What influences the challenges farmers face when adapting to climate change? To answer this question, I’m building a model to explain farmers’ ratings of how challenging adapting to climate change has been for them. Specifically, I’ll be looking to see if land access affects organic farmers’ perceived capacity to adapt to climate change, controlling for their level of exposure to climate change. Land access is measured by farm size, land tenure, and perceived challenges accessing land; climate change is measured by heat extremes, drought, precipitation variability, and temperature variability. I chose to explore land access based on 28 interviews I
conducted with organic vegetable farmers in western WA about how they’re experiencing and adapting to climate change, where I repeatedly heard that insecure land tenure can hamper the ability to make necessary investments to adapt to changing conditions.

The farm-level data was collected as part of a national survey of organic farmers by the Organic Farming Research Foundation for the 2022 National Organic Research Agenda, in conjunction with the Organic Seed Alliance for the 2022 State of Organic Seed Report.

With the support of eScience Data Scientist Curtis Atkisson, I have been able to progress towards building a model for organic farmers’ perceived capacity to adapt to climate change. At the beginning of the program, we deliberated on a modeling framework and settled on building a Bayesian structural equation model. Bayesian
models are powerful because they allow you to incorporate your hypotheses (priors) about how the explanatory variables will affect the outcome variable. Before beginning the modeling, we spent time developing metrics for temperature variability, precipitation variability, and extreme heat, which we then used in the model. These metrics were developed from temperature and precipitation data from the ERA 5 – Land dataset, using Google Earth Engine’s Python API. We are now in the process of building the Bayesian model using the R package BRMS.

Image Caption: Photo of a diversified, organic farm in western Washington.