Melt pond retrieval based on LinearPolar algorithm using Landsat data

Yuqing Qin,Jie Su,Mingfeng Wang

doi:10.5194/egusphere-egu21-14243

Abstract

&lt;p&gt;The formation and distribution of melt ponds also have an important influence on the Arctic climate. It is necessary to obtain more accurate information of melt ponds on Arctic sea ice by remote sensing. Present large-scale melt pond products, especially melt pond fraction (MPF), still need a lot of verification, and it is a good way to use the very high resolution optical satellite remote sensing data to verify the retrieval MPF of low-resolution melt pond results.&lt;/p&gt;&lt;p&gt;Most MPF algorithm such as Markus (Markus, et al., 2003) and PCA (Rosel et al., 2011) relying on fixed melt pond albedo, LinearPolar algorithm (Wang et. al., 2020) considers that the albedo of melt ponds albedo is variable, it has been proved the retrieval results of this algorithm has a high accuracy of the MPF than that of the previous algorithm based on Sentinel-2 data in Wang et al.&amp;#8217;s work. In this paper, we applied this algorithm to Landsat 8 data. Meanwhile, Sentinel-2 data as well as SVM and ISODATA method are used as the comparison and verification data. The results show that the MPF obtained from Landsat 8 using LinearPolar algorithm is the much more closer to Sentinel-2 than Markus and PCA algorithms, and the correlation coefficients of the two MPF is as high as 0.95. The overall relative error of LinearPolar algorithm is 53.5% and 46.4% lower than Markus and PCA algorithms, respectively. And in the cases without obvious melt ponds, the relative error is reduced more than that with obvious melt ponds. This is because LinearPolar algorithm can identify 100% dark melt ponds and relatively small-scale melt ponds, and the latter contributes more to MPF retrieval.&lt;/p&gt;&lt;p&gt;The application of LinearPolar algorithm on Landsat can cover a wider range than Sentinel and enhance the verification efficiency. Moreover, because of the longer time series of Landsat data than Sentinel data, the long-term variation trend of sea ice in fixed areas can be monitored.&lt;/p&gt;

Highlights

In the cases without obvious melt ponds, the relative error is reduced more than that of those with obvious melt ponds because the LinearPolar algorithm can identify 100% of dark melt ponds and relatively small melt ponds, and the latter contributes more to the reduction in the relative error of melt pond fraction (MPF) retrieval
Classification of S2 interpolated onto the Landsat 8 (L8) grid; (d) the iterative self-organizing data analysis technique (ISODATA) classification of S2; (e) the ISODATA classification of S2 interpolated onto the L8 grid; (f) the MPF of S2 using LinearPolar algorithm; (g) the result of (f) interpolated on the L8 grid; the MPF of L8 using (h) the Markus algorithm; (i) principal component analysis (PCA); and (j) the LinearPolar algorithm
The results of S2 retrieval, support vector machine (SVM) classification, and ISODATA classification are used as verification data to calculate the classification accuracy of the results of L8 using three retrieval algorithms

Summary

Introduction

Melt ponds are some of the most common phenomena on the surface of Arctic sea ice in warm seasons and generally form in late May and refreeze during late August and early September [1]. The distribution and variability of melt ponds have an important influence on the Arctic climate through sea ice albedo feedback [2]. It has been shown that the melt pond fraction (MPF) of the Arctic in spring can effectively predict the September minimum sea ice extent [3–5], which leads to enhancements in the seasonal forecasting ability of the summer sea ice extent. Accurately obtaining the MPF is extremely vital for understanding and predicting Arctic sea ice changes

Objectives

Methods

Results

Discussion

Conclusion