Carefully Choose the Baseline: Lessons Learned from Applying XAI Attribution Methods for Regression Tasks in Geoscience

Abstract

Methods of eXplainable Artificial Intelligence (XAI) are used in geoscientific applications to gain insights into the decision-making strategy of Neural Networks (NNs) highlighting which features in the input contribute the most to a NN prediction. Here, we discuss our “lesson learned” that the task of attributing a prediction to the input does not have a single solution. Instead, the attribution results depend greatly on the considered baseline that the XAI method utilizes; a fact that has been overlooked in the geoscientific literature. The baseline is a reference point to which the prediction is compared so that the prediction can be understood. This baseline can be chosen by the user or is set by construction in the method’s algorithm – often without the user being aware of that choice. We highlight that different baselines can lead to different insights for different science questions and, thus, should be chosen accordingly. To illustrate the impact of the baseline, we use a large ensemble of historical and future climate simulations forced with the SSP3-7.0 scenario and train a fully connected NN to predict the ensemble- and global-mean temperature (i.e., the forced global warming signal) given an annual temperature map from an individual ensemble member. We then use various XAI methods and different baselines to attribute the network predictions to the input. We show that attributions differ substantially when considering different baselines, as they correspond to answering different science questions. We conclude by discussing important implications and considerations about the use of baselines in XAI research.

Using Explainability to Inform Statistical Downscaling Based on Deep Learning Beyond Standard Validation Approaches

“…As a result, different researchers may reach different conclusions when examining the same saliency maps. Moreover, different XAI techniques can yield diverse explanations, as previously examined in (Mamalakis et al, 2022(Mamalakis et al, , 2023. This poses a significant challenge in establishing the reliability of these techniques since, in the absence of a ground truth relationship, determining which technique to trust becomes problematic.…”

The Choice of Baseline Period Influences the Assessments of the Outcomes of Stratospheric Aerosol Injection

“…In the context of climatic changes due to the increase or reduction of anthropogenic emissions, previous studies have analyzed potential impacts (future ones, or avoided ones) in light of a counterfactual that hasn't materialized: such a framing has been used when discussing the beneficial effects of the Montreal Protocols, coining the term "World avoided" (Morgenstern et al, 2008;Wilka et al, 2021) when referring to a future where CFC emissions were not abated. Similarly, for global warming, Mamalakis et al (2023) discussed how the choice of baseline period is fundamental when performing attribution studies and should be chosen carefully depending on the scientific question answered. This work highlights that SAI studies, by adding a novel dimension to the ability to influence global warming impacts, need even more care when explaining how they are defining a certain simulated impact.…”

“…Similarly, for global warming, Mamalakis et al. (2023) discussed how the choice of baseline period is fundamental when performing attribution studies and should be chosen carefully depending on the scientific question answered. This work highlights that SAI studies, by adding a novel dimension to the ability to influence global warming impacts, need even more care when explaining how they are defining a certain simulated impact.…”

Using Explainable Artificial Intelligence to Quantify “Climate Distinguishability” After Stratospheric Aerosol Injection

“…Deep SHAP has been chosen for two reasons: (a) it allows the user to define the baseline for which the attribution is derived (see Mamalakis et al. (2023) on the importance of baselines); and (b) it satisfies the completeness property (Sundararajan et al., 2017), which holds that the attributions add up to the difference between the network output at the current sample and the one at the baseline. For details on the Deep SHAP algorithm, please see Text S2 in Supporting Information .…”