Reconstructing past hydroclimatic variability using climate-sensitive paleoclimate proxies provides context to our relatively short instrumental climate records and a baseline from which to assess the impacts of human-induced climate change. However, many approaches to reconstructing climate are limited in their ability to address sampling variability inherent in different climate proxies. We iteratively optimise an ensemble of possible reconstruction data series to maximise the Gaussian kernel correlation of Rehfeld et al. (2011) which reconciles differences in the temporal resolution of both the target variable and proxies or covariates. The reconstruction method is evaluated using synthetic data with different degrees of sampling variability and noise. Two examples using paleoclimate proxy records and a third using instrumental rainfall data with missing values are used to demonstrate the utility of the method. While the Gaussian kernel correlation method is relatively computationally expensive, it is shown to be robust under a range of data characteristics and will therefore be valuable in analyses seeking to employ multiple input proxies or covariates.
- Gaussian kernel correlation
- Uneven sampling