Fidelity of climate data: Trends, extremes, and nonlinearities in instrumental and tree-growth proxies of climate

Abstract

This work looks at inferences of climate -- and specifically moisture and temperature --, past and present. In the present, asks whether we have enough data to understand extremes in temperature and trends in moisture. When looking at the past, it asks whether the assumptions we have taken in climate reconstructions that use tree rings are valid, and what results when we release or change them. Chapter 1 begins by looking at the implications of incorporating nonlinearities in paleoclimatological reconstructions of drought using tree rings and neural networks. This methodology results in more skillful reconstructions of drought over the majority of the United States. Additionally, the nonlinear methodology results in more stable incidences of drought in the United States Southwest over the last two centuries. Linear methods, on the other hand, might be overestimating drought and underestimating wet periods. Chapter 2 continues with nonlinearities and tree rings but shift towards detecting Liebig's Law of the Minimum in tree ring data over the last five centuries. This chapter finds evidence that Liebig's Law of the Minimum operates at the individual tree level, and that temperature has been to some degree limiting tree growth at the hemispheric and regional level. The implications of this finding are that the sensitivity of tree growth to climate and environmental stressors is state-dependent. Additionally, this chapter raises the possibility of using the intrastand correlation as a proxy to witch which to infer low frequency changes in largescale stressors. Chapter 3 changes gears to the present and future, but continues with temperature. This chapter explores our ability to appropriately model temperature extremes using Generalized Extreme Value (GEV) distributions as a function of data availability. We evaluate the effects of lack of independence, skewness, using synthetic records, and limited data availability in the bias and uncertainty of the parametrization of GEVs. We find a consistently low bias in our estimation of the underlying distribution and in particular for return period temperature maxima. Chapter 4 uses soil moisture observations, rather than modeled indices, to look at soil moisture trends in the United States over the last decade and, depending on the dataset, decades. We find largescale increasing moisture trends over the contiguous United States over the last decade. Additionally, we find agreement between different soil moisture data products -- both satellite and in situ -- and further qualitatively validate the data using greenness indices. Importantly though, the satellite product that combines many satellite records underestimates trends, even if they agree relatively to situ records. Additionally, we find that season with the most agreement between the trends of the distinct types of records is Autumn.