Back to results list
Show full item record
Please use this identifier to cite or link to this item:
|Title:||The long-term spatiotemporal dynamics of sand dunes in arid environment studied with optical satellite imagery||Authors:||Saleh, Eslam Ali Hussein||Degree:||Ph.D.||Issue Date:||2022||Abstract:||Dunes are considered the most common landform in terrestrial and extraterrestrial deserts. In several desert areas on Earth, dune and sand sheet instabilities pose a significant threat to transportation networks, water supply routes, urban areas, and cultural sites. Monitoring dune migration in the spatiotemporal domains can, therefore, contribute to a deeper understanding of the underpinning aeolian processes and their interaction with environmental change. Moreover, information on dune migration can be used as an indicator of large-scale trends in windiness over large deserts (e.g., the Sahara). The scarcity of metrological stations in vast deserts hinders drawing a complete picture of dynamic dune behavior. Additionally, scaling measurements of single dunes at a specific time scale to a large spatiotemporal domain are considered a problem. Optical image cross-correlation (OICC) is considered a valuable tool for semi-automatic monitoring of deformation at the Earth's surface (i.e., earthquakes, glaciers, landslides, and dunes) with dense spatiotemporal resolution down to 1/10th of the pixel. However, limited studies using OICC to monitor dune migration by adjacently matching dozens of images over several years do not provide a complete picture of the temporal evolution of dune migration, especially over short timescales, due to sparse temporal coverage. Additionally, the dependency on adjacent matching with low redundancy levels would affect the solution uncertainty and spatial coverage. Previous studies that monitored other moving targets (i.e., landslides and glaciers) employed the inversion of OICC measurements; however, the inversion of time series is in its infancy. Time series inversion by constructing the full network of OICC measurements is considered promising in reducing signal-to-noise ratio (SNR), enhancing spatial coverage, and reducing uncertainty. However, the full network inversion suffers from the following: 1) high computational cost and data overhead, especially for free archives that provide images with a high temporal resolution, 2) the inversion results are affected by low-quality images and temporal decorrelation, and 3) false displacements may occur especially for targets with uniform albedo and apparent topographic fluctuations (e.g., dunes and glaciers). In response to these limitations, this work is strongly motivated by the availability of the free Landsat-8 and Sentinel-2 archives, which offer data with enhanced radiometric, spatial resolutions, co-registration, and orthorectification accuracies. Thus, the main objective of this work is to introduce new quantitative methods for measuring dune velocity using these free archives with high spatiotemporal resolution and reliability based on optical image matching.
The thesis is divided into three main parts: In part one, the improved optical image matching selection and inversion (OPTSI) algorithm is introduced to monitor the temporal evolution of dune movement in the Northwest Sand Sea of North Sinai (NSSS). This method is considered a small baseline subset (SBAS) simulation used extensively in the synthetic interferometric aperture radar (InSAR) domain. The methodological workflow of the OPTSI algorithm includes two main steps: (1) defining the baselines of the optical image matching and weighting these baselines according to their respective contribution to the uncertainty of the matching measurements, and (2) applying selection criteria that limit the baseline thresholds, especially the solar angles disparities while maintaining the connectivity of the network to reduce the oscillation of the singular value decomposition (SVD) inversion. To test the relationship between the baselines and the uncertainties, approximately 576 matching measurements are established over stable targets. The results show that the baselines can be weighted as follows: sun azimuth difference, sun elevation difference, temporal baseline, and spatial baseline. The OPTSI algorithm significantly reduces uncertainty on average up to 25% and improves spatial coverage on average up to 15% with low data overhead, especially for free image archives.
In part two, a broader application of the OPTSI is performed with the merit of fusion between offset maps from two sensors before inversion to condense the temporal sampling to a weekly time scale. Condensing the temporal sampling of time series is considered of great importance for fast-moving targets that require matching images with short temporal span to preserve the surface changes and keep updating about the dynamic behavior; however, short span pairs are more suspectable to geolocation errors. Moreover, the presence of cloud cover, dust storms, or haze would inherently affect the number of images and, consequently the temporal sampling of time series. Therefore, the OPTSI algorithm was extended and applied to monitor the fastest barchan dunes of northern Chad's fastest barchan dunes in the Bodélé Depression (BD). This builds on a previous study by Vermeesch and Leprince (2012) that introduced 26 years of time-series data (1984 to 2010) but did not provide information on the short-term behavior of dune migration and associated wind patterns in this region due to sparse temporal sampling. Furthermore, the adjacent paring criterion of matching images used to produce a time series fails to provide high spatial coverage and low uncertainty. The inversion algorithm first selects the appropriate images by constraining cloud coverage. However, the number of available scenes is primarily limited by cloud cover, especially during the rainy seasons prevalent in tropical regions, resulting in lower temporal sampling. Therefore, the fusion of two or more sensors is feasible for improving temporal sampling resolution and revealing complex deformation patterns at weekly timescales. The results show a recent deceleration occurred in the recent decade in the activity of aeolian features in the Bodélé Depression compared to the previous decade.
In part three, the work is extended to monitor the spatiotemporal variability of dune velocities and corresponding uncertainties for the entire NSSS and demonstrate their determination from the matched measurements. Matched pairs are selected so that the differences in solar angles are small and span at least one year. Such a selection scheme helps reduce shadowing artifacts in the deformation fields and the error budget in converting displacements to annual velocities. The fusion of individual velocities allows the estimation of final velocities for approximately 98.8% of the studied dune areas. The stable regions are used to estimate 95% confidence intervals for the final velocities and extend these calculations to the dune targets. The coherence of the final velocity vectors is then estimated, which is used as an indicator of the homogeneity of migration directions between offset maps. The process of selecting high-quality pairs and then merging individual maps shows high performance in terms of spatial coverage and reliability of the extracted velocities.
In summary, the following points can be concluded: 1) the OPTSI algorithm shows promise for evaluating the temporal evolution of dune movement with low uncertainty, low data burden, and high spatial coverage. More broadly, the algorithm can be applied to dune studies with different environmental conditions and other mobile targets (e.g., glaciers and landslides), 2) the fusion of offset maps from two sensors before the inversion would help condense the temporal sampling up to a weekly time scale, especially for fast-moving targets and regions where the number of images is limited by the presence of haze, dust, or cloud cover, 3) the selection criteria before pairing images helps control artificial deformation and shadowing artifacts, 4) the fusion of groups of individual offset maps helps estimate the average annual rates (AAR) of dune velocities, prevailing migration direction (PMD) and the corresponding uncertainties. The spatial uncertainties of the final velocities after fusion of individual offset maps help investigate the seasonal patterns of dune migration and consequently understand the variability of wind regimes. To sum up, information obtained on migration rates and directions from the matching measurements with large spatiotemporal resolution can be used as a proxy for windiness, especially given the paucity of metrological records in many desert regions and compensate for the scarcity of wind data. Additionally, such information can be used to prioritize field studies, especially in high-risk areas, to implement stabilization procedures.
Hong Kong Polytechnic University -- Dissertations
|Pages:||xx, 340 pages : color illustrations|
|Appears in Collections:||Thesis|
View full-text via https://theses.lib.polyu.edu.hk/handle/200/11986
Citations as of Apr 2, 2023
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.