Retro Rainfall

Characterizing the tropical hydroclimate response to the 8.2ka Event using data, models, and theory.

For my graduate thesis project, I am combining published paleoclimate datasets with state-of-the-art isotope-enabled global climate models, and using new statistical tools to quantitatively evaluate how, why, and when tropical rainfall patterns responded to the 8.2ka Event, a period of abrupt global climate change that occurred approximately 8,200 years ago. This event was triggered by the release of huge volumes of freshwater from melting glaciers into the North Atlantic Ocean, which disrupted ocean circulation patterns and led to widespread climate changes. Given the parallels between this freshwater forcing during the 8.2ka event and current high northern latitude ice sheet melt, this research sheds light on the potential hydroclimate impacts of large-scale changes in ice sheet mass balance. The tropics are especially sensitive to changes in climate, and changes in tropical rainfall patterns can have significant impacts on ecosystems and human populations in these regions. This is particularly important given that the tropics are home to roughly 40% of Earth’s human population, making it vital to understand the potential hydroclimate impacts of past and future climate change in these areas.

In my project, I am quantitatively reconstructing tropical hydroclimate variability using increasingly complex methods to circumvent the many uncertainties surrounding paleoclimate data. One of the major limitations of using paleoclimate data for global-scale climate reconstructions is the inherent uncertainties and noise associated with proxy records, such as non-climatic factors that can influence proxy values, and the lack of temporal resolution in some records. Additionally, accurately constraining the timing, magnitude, and duration of hydroclimate anomalies based on proxy records alone can be challenging due to the complex interplay of multiple factors that influence these patterns.

A significant challenge in using paleoclimate data is age uncertainty, which arises from a number of sources including errors in the calibration of radiocarbon dates, the selection of appropriate age models, and the calibration of chronologies using statistical techniques. To address these uncertainties, I have developed robust age-depth models for each dataset using Bayesian methods and recalibrated radiometric age data using updated calibration curves. By incorporating these methods to account for age uncertainties, I aim to provide more accurate estimates of the timing of hydroclimate anomalies associated with the 8.2ka Event.

I use the changepoint package in R to detect shifts in the statistical properties of time series data, specifically for identifying changes in hydroclimate patterns during the 8.2 ka Event. For my analyses, I utilize the Pruned Exact Linear Time (PELT) algorithm with penalties imposed by the Modified Bayesian Information Criterion (MBIC). The PELT algorithm is a computationally efficient approach for finding the exact number and locations of change points, while the MBIC penalty helps to reduce overfitting and improve the accuracy of detected change points. To improve the robustness of my results, I generate 1000-member time series ensembles that take into account age and paleodata uncertainties, and I test each ensemble against an isospectral noise model using 100 null hypothesis ensemble members. This approach allows me to better quantify the significance of the detected change points and assess the likelihood that they result from noise or other sources of uncertainty. The use of these methods allows me to quantitatively detect changes in hydroclimate variability associated with the 8.2 ka Event with greater accuracy and precision.