Open Source Software developed by our Climate Extremes team
We want to maximize scientific progress and impact, open source is essential to achieve that goal. In this page we share code that we developed to perform advanced analysis on climate data. We aim to have all our code used for publication online, which greatly enhances the reproducibility of our results.
Response Guided Causal Precursor Detection
The idea behind RG-CPD was to create a python package which can process 3-dimensional climate data, such that relationships based on correlation can be tested for causality.
Causal inference metrics have been proven a valuable to go beyond defining a relationship based upon correlation. Autocorrelation, common drivers and indirect drivers are very common in the climate system, and they lead to spurious (significant) correlations. Tigramite has been successfully applied to 1 dimensional time series in climate science (Kretschmer et al. 2016), in order to filter out these spurious correlations using conditional indepence tests (Runge et al. 2017).
Within RG-CPD, the 1-d precursor time series are obtained by creating point-wise correlation maps and subsequently grouping adjacent significantly correlating gridcells together into precursor regions. These precursor regions are then converted into 1-d time series by taking a spatial mean (Kretschmer et al. 2017).
The final step is the same, where the 1-d time series are processed by Tigramite to extract the causal relationships. This requires thorough understanding of the method, see Runge et al. 2017 http://arxiv.org/abs/1702.07007). These 1d time series contain more information since they are spatially aggregated. The 1d time series of different precursor regions are subsequently tested for causality using the Tigramite package.
Kretschmer, M., Coumou, D., Donges, J. F., & Runge, J. (2016). Using Causal Effect Networks to Analyze Different Arctic Drivers of Midlatitude Winter Circulation. Journal of Climate, 29(11), 4069–4081.
Kretschmer, M., Runge, J., & Coumou, D. (2017). Early prediction of extreme stratospheric polar vortex states based on causal precursors. Geophysical Research Letters, 44(16), 8592–8600.
J. Runge et al. (2015): Identifying causal gateways and mediators in complex spatio-temporal systems. Nature Communications, 6, 8502.
J. Runge, S. Flaxman, and D. Sejdinovic (2017): Detecting causal associations in large nonlinear time series datasets.
