Abstract:

The Global, Self-consistent, Hierarchical, High-resolution Geography (GSHHG) Database, provided by the National Centers for Environmental Information and widely used worldwide, includes global multi-scale coastline data. Although many scientists have utilized this database in their studies, the database itself has not been thoroughly evaluated. The different scales available through the GSHHG database include full resolution, high resolution, intermediate resolution, low resolution, and crude resolution. The Pearl River estuary was selected as our study site. These studies involve the use of the GSHHG database, Landsat MSS from 1978, and Landsat OLI from 2018 to evaluate the differences in coastline length and fractal dimensions among GSHHG datasets collected at five different resolution levels. This study also selected eastern and western sub-regions to compare differences between coastline maps from the GSHHG and Landsat datasets. In addition, the coastline extracted from resampled Landsat OLI data from 2018 (80 m), covering the eastern part of the Pearl River Estuary was compared with the same coastline data extracted from Landsat OLI at 30 m spatial resolution. Coastline lengths according to full resolution, high resolution, intermediate resolution, low resolution, and crude resolution data are 1509.47 km, 1398.12 km, 1212.54 km, 692.00 km, and 326.40 km, respectively. The corresponding fractal dimensions from the GSHHG database were 1.2983, 1.2832, 1.2588, 1.0990, and 1.0262 for full resolution, high resolution, intermediate resolution, low resolution, and crude resolution, respectively. Separate eastern and western segments of the Pearl River Estuary demonstrate that for a given area, full resolution of GSHHG data provides the longest and most detailed rich coastline data relative to the coastline data extracted from the Landsat data and GSHHG database. The length of the coastline extracted from the 1978 Landsat MSS was most similar to the length obtained from high-resolution GSHHG. The data on the morphology of the coastline in the 1978 Landsat MSS were much closer to the full resolution, high resolution, and intermediate resolution data obtained from GSHHG. Compared to the coastline data extracted from 80 m Landsat OLI and 30 m Landsat OLI, a 17.68% length difference was found. Fractal dimension differences were 1.7%. The results show that: (1) The length and fractal dimensions of coastline data produced at five different resolutions are substantially different. Higher-resolution data produce longer coastline lengths and more complex coastline morphologies; (2) Using coastline data extracted from Landsat data in 1978 and 2018 as references, GSHHG coastline dataset was found to be generally consistent with real data shoreline length and morphological characteristics. The morphological characteristics of the GSHHG coastline are closer to the real data from 1978, which does not reflect the current situation of the Pearl River Estuary. The GSHHG database only includes one phase of data, implying that scientists cannot accomplish studies related to coastline change without using additional data. Nevertheless, studies on comparisons between different regions of coastline could select data based on the characteristics of a study area. It is suggested that scientists select coastline data based on coastline length, complexity, and morphology when studying coastline changes in response to situations of interest.

Key words: coastline, Pearl River Estuary, GSHHG, Canny, Fractal Dimension