How to overcome the ‘maybe’ explanations of ‘why’: explaining causality?

We are proposing a “mini-research challenge” that would explore and gather methods for deriving and/or assigning directional influences, drivers, or causal links, particularly when undertaking remote sensing analysis or attributing climate adaptation. The question of “why” is inevitable when presenting a phenomenon detected or described.

*“Why are we seeing this/ why is this happening?” *“Could this be because of a policy intervention?” *“Could we say that this change we are seeing is linked to anthropogenic climate change? *“Are farmers changing their cropping patterns because…

• they’ve been they are trying to prevent weather
• induced loss on a specific crop,-they’re trying to diversify their production and income,
• they’re neighbor convinced them to, or
• because they learned a new cropping strategy at a farmer training program last season?”

How can we, researcher, enhance the certainty with which we answer this ‘why’? & can we draw causal links among observed phenomena?

The obstacle of scientifically justifying the ‘why’ is not limited to the realm of short time series and climate adaptation studies. Challenges such as distinguishing and accounting (or controlling) for multiple factors driving a specified phenomenon, ‘random’ correlations, detecting the direction of statistical relationships, factors operating at multiple temporal and spatial scales and the formidable work of proving causality, are common across disciples. Within our working group on climate Adaptation in Agricultural Systems, some colleagues are dealing with these questions through statistical methods, some through triangulation with field work and others through forms of data integration. For example, one statistical method recently applied in our group to investigate climatic driver effects on wheat yield was a “causal discovery algorithm,” that adapts a variable selection method, with stricter requirements, to “eliminate spurious correlations,” through specifically addressing auto-correlation, indirect links, common drivers. However, when researcher from Geo.Society group face these challenges, traditional algorithms or methods for causality often cannot be applied for local scale data or when dealing with shorter time-series, as is also the case for applications of the Sentinel sensors (launched in 2014).

We suppose that each field has their own way of dealing with these issues and we would like to learn how other Geo.X members have dealt with or are currently identifying causal links. The “expertise” for this research challenge would be drawn/ asked for in advanced primarily from Geo.X members. And would then workshop applications of these different schemes for explaining influences, drivers and causality into other cases (across Geo.X topics). There would not be a single, streamlined outcome, but rather a “cross-pollination” of methods.

This “mini-research challenge” has highly relevant outcomes to the Autumn School’s topic of science communication, as directional influences, drivers, or causal links all contribute to the ‘why’ andrelevance of research, building blocks for writing a narrative. Commonly the ‘why’ is framed in the introduction, interpreted in results, or hypothesizing in the discussion. In this “mini challenge,” we could then explore how having a science-based justification of the ‘why’ can fortify any storyline, including e.g. a funding proposal or teaching & learning material.

Contact person

Gina Maskell, Roopam Shukla Potsdam Institute for Climate Impact Research