On the plausibility of climate variable trends

Abstract

<p> </p> <p>Studying long-term changes in physical systems (e.g., ocean, ice, and land water) helps to understand the impact of climate change, natural variability, and human interventions on these systems. Most of the time, our ability to conduct deductive reasoning on climate variable trends (CVTs) is compromised by insufficient data and limitation of deterministic statistical methods, and our reasoning is reduced to inductive (i.e., plausible). Nevertheless, quantifying such plausibility is rarely addressed—instead, traditional binary beliefs (i.e., significant versus insignificant) are held for CVTs.</p> <p>Here, we derive the plausibility of three CVTs (i.e., sea level rise, ice melting, and total water storage change) using satellite data (e.g., altimetry and Gravity Recovery Climate Experiments (GRACE)) between 1993 and 2021. We regressed the satellite data using a non-parametric probabilistic approach (e.g., Gaussian process regression) and derived two probabilistic indices for the trend changes. We probabilistically evaluated the CVTs magnitude; trendiness (i.e., change in trend strength) and stability (i.e., maintaining the progression in the same direction) over a 29-yr period.</p> <p>Our results show the global sea level (GSL) is increasing with 85% plausibility at an average magnitude of 3.1 mm/yr over 29 yr. Pauses and hiatus in the GSL rise were quantified and attributed to El Niño Southern Oscillation. We find acceleration in the regional sea level in six hot spots, where the trend plausibility exceeds 60% beyond the modulation of interannual and decadal climate variability. Results were also obtained for the ice melting in ice sheets and glaciers, and the total water storage over 21 yr.</p> <p>Assessing the trend strength and change goes beyond the simple linear and quadratic regressions and the binary beliefs. Trends evolve over short and long-time scales as a response to the driving processes. Understanding and quantifying how plausible these changes over certain periods improves the physical attribution and overcomes the limitations of data and methods. It is also practical for policymaking and helps to communicate science effectively.</p>