Sensitivity Analysis of Arctic Sea Ice Extent Trends and Statistical Projections Using Satellite Data

Ge Peng,Jessica Matthews,Jason Yu

doi:10.3390/rs10020230

Abstract

An ice-free Arctic summer would have pronounced impacts on global climate, coastal habitats, national security, and the shipping industry. Rapid and accelerated Arctic sea ice loss has placed the reality of an ice-free Arctic summer even closer to the present day. Accurate projection of the first Arctic ice-free summer year is extremely important for business planning and climate change mitigation, but the projection can be affected by many factors. Using an inter-calibrated satellite sea ice product, this article examines the sensitivity of decadal trends of Arctic sea ice extent and statistical projections of the first occurrence of an ice-free Arctic summer. The projection based on the linear trend of the last 20 years of data places the first Arctic ice-free summer year at 2036, 12 years earlier compared to that of the trend over the last 30 years. The results from a sensitivity analysis of six commonly used curve-fitting models show that the projected timings of the first Arctic ice-free summer year tend to be earlier for exponential, Gompertz, quadratic, and linear with lag fittings, and later for linear and log fittings. Projections of the first Arctic ice-free summer year by all six statistical models appear to converge to the 2037 ± 6 timeframe, with a spread of 17 years, and the earliest first ice-free Arctic summer year at 2031.

Highlights

Since satellite-based measurements became available in the late 1970s, a 10–15% per decade decrease in the annual Arctic sea ice extent (SIE) minimum has been recorded, as measured by the area within the 15% sea ice fraction contour (e.g., [1,2,3,4])
Relative to the linear decadal trend of the whole analysis record period of 1979–2015, only slight sensitivity is seen to the varying averaging methods and intervals—about 6.67%/1.04% or less for annual SIE maximum/minimum trends, respectively, relative to those derived from the original daily SIE time series
Beyond the scope of this paper, a robust characterization of the associated uncertainties originating from the retrieval of the remotely sensed observations and carried through the statistical model fitting of SIE to produce related confidence intervals could better describe the prediction capacity of the models presented

Summary

Introduction

Since satellite-based measurements became available in the late 1970s, a 10–15% per decade decrease in the annual Arctic sea ice extent (SIE) minimum has been recorded, as measured by the area within the 15% sea ice fraction contour (e.g., [1,2,3,4]). Using September SIE as the indicator and the ice-free threshold of one million square kilometers, Overland and Wang [5] examined three commonly used approaches for estimating the timing of the first ice-free Arctic summer: (1) extrapolation of sea ice volume data; (2) stochastic projection, assuming several more rapid loss events such as 2007 and 2012; and (3) projection of climate models. These three approaches have roughly predicted the first ice-free Arctic summer year (FIASY) to be 2020 or earlier, 2030 with a 10-year margin of error, and 2040 or later, respectively. They have confirmed that climate models in CMIP5 tend to be conservative in their sea ice projections

Results

Discussion

Conclusion