Quantifying the spatio-temporal patterns of dune migration near Minqin Oasis in northwestern China with time series of Landsat-8 and Sentinel-2 observations

Abstract

Investigation of the spatio-temporal patterns of dune migration in a large-scale area with optical remote sensing techniques can help us to better understand aeolian phenomena and mitigate sand-dust disasters. With the rapid growth in data volume, extracting more accurate dune displacement time series and rates from optical observations has become possible; however, the method is not yet fully fledged. To address this issue, we propose an extended algorithm for the mature optical imagery cross-correlation (OICC) technique based on Landsat-8 (L8) and Sentinel-2 (S2) acquisitions. The main innovative points of this algorithm are: 1) the proposed pairing strategy for the OICC processing; 2) the modularized post-processing procedures for noise removal; and 3) the introduction of singular value decomposition (SVD) time-series inversion of the redundant optical observations to quantify the dune migration. To test the effectiveness of this algorithm, it was applied in the study of dune migration near Minqin Oasis in northwestern China, using enriched L8 and S2 images collected between April 2013 and April 2018. Compared with the original OICC results in stable areas, the post-processing and inversion of the proposed algorithm reduce the uncertainty by around 22–35% and 3–5% for L8, 29–48% and 5–12% for S2, respectively. The cross-comparison between the L8- and S2-derived displacement time series shows high consistency, and presents a lower uncertainty than the result of the traditional no-inversion method. Furthermore, the derived displacement rates show spatial patterns that are similar to those of the manually digitized results obtained with historical Google™ Earth (GE) images. These comparisons show the advantage of the proposed algorithm in automatically and accurately quantifying dune migration. Taking into account these measurements, the spatio-temporal evolution patterns of dune migration in the study area were analyzed. From the spatial perspective, the sand dunes move along a northwest-southeast axis with four detected transport pathways. Our research also shows that around 1087.7 km2 of dune fields present an active status. The active sand dunes are currently encroaching on around 155.5 km2 and 4.4 km2 of land each year outside and inside the oasis, respectively, representing a problem of rapid desertification. Temporally, the displacement time series along the dominant migration direction appears as seasonal variations that are seemly consistent with the changes in local atmospheric conditions. The proposed algorithm provides a new perspective to investigate the spatio-temporal evolution of dune migration with medium-resolution L8 and S2 optical datasets.