Transferability and Upscaling of Fuzzy Classification for Shoreline Change over 30 Years

Abstract

Local authorities require information on shoreline change for land use decision making. Monitoring shoreline changes is useful for updating shoreline maps used in coastal planning and management. By analysing data over a period of time, where and how fast the coast has changed can be determined. Thereby, we can prevent any development in high risk areas. This study investigated the transferability of a fuzzy classification of shoreline changes and to upscale towards a larger area. Using six sub areas, three strategies were used: (i) Optimizing two FCM (fuzzy c-means) parameters based on the predominant land use/cover of the reference subset, (ii) adopting the class mean and number of classes resulting from the classification of reference subsets to perform FCM on target subsets, and (iii) estimating the optimal level of fuzziness of target subsets. This approach was applied to a series of images to identify shoreline positions in a section of the northern Central Java Province, Indonesia which experienced a severe change of shoreline position over three decades. The extent of shoreline changes was estimated by overlaying shoreline images. Shoreline positions were highlighted to infer the erosion and accretion area along the coast, and the shoreline changes were calculated. From the experimental results, we obtained m (level of fuzziness) values in the range from 1.3 to 1.9 for the seven land use/cover classes that were analysed. Furthermore, for ten images used in this research, we obtained the optimal m = 1.8. For a similar coastal characteristic, this m value can be adopted and the relation between land use/cover and two FCM parameters can shorten the time required to optimise parameters. The proposed method for upscaling and transferring the classification method to a larger, or different, areas is promising showing κ (kappa) values > 0.80. The results also show an agreement of water membership values between the reference and target subsets indicated by κ > 0.82. Over the study period, the area exhibited both erosion and accretion. The erosion was indicated by changes into water and changes from non-water into shoreline were observed for approximately 78 km2. Accretion was due to changes into non-water and changes from water into shoreline for 19.5 km2. Erosion was severe in the eastern section of the study area, whereas the middle section gained land through reclamation activities. These erosion and accretion processes played an active role in the changes of the shoreline. We conclude that the method is applicable to the current study area. The relation between land use/cover classes and the value of FCM parameters produced in this study can be adopted.

A review and meta-analysis of remote sensing data, GIS methods, materials and indices used for monitoring the coastline evolution over the last twenty years

“…Volumetric Changes in beaches' sand studied by Horacio et al (Jesús et al, 2019); Kaliraj et al (2017);Sakthivel et al (2014); Gares et al (Gares Paul et al, 2006). CVI (coastal vulnerability index) is another significant feature that researchers used (Hoque et al, 2019;Joevivek & Saravanan Sakthivel, 2013;Mujabar & Chandrasekar, 2013;Nageswara Rao et al, 2008;Ratna et al, 2018). Tracer stick method has been used by Schwarzer et al (2003).…”

“…Landsat series have been used by most of the researchers such as (Valderrama-Valderrama-Landeros & Flores-de-santiago, 2019); Al-Amin (Hoque et al, 2019;Chao Chen Chao Chen et al, 2019;Mohammed et al, 2019;Yan et al, 2019;Yulianto et al, 2019;Ratna, 2019;Ratna & Bijker, 2019;Karim et al, 2019;Sandeep et al, 2018;Yawo et al, 2018;Behling et al, 2018;Asib et al, 2018;Joevivek & Saravanan Sakthivel, 2018;Do et al, 2018;Zed Abu et al, 2018;Liu & John, 2018;Ratna et al, 2018;Xu Nan, 2018;Kaliraj et al, 2017;Mitra et al, 2017;Qingxiang Liu et al, 2017;Ratna et al, 2017;Luca et al, 2017;Manjulavani et al, 2017;Wenyu & Gong, 2016;Jaime et al, 2016;Vittal & Akshaya, 2015;Usha et al, 2015;Salghuna & Aravind Bharathvaj, 2015;George et al, 2015;Prasita, 2015;Kankara et al, 2015;Gudina Feyisa et al, 2014;Sedar et al, 2014;Yanxia Liu et al, 2013;Joevivek & Saravanan Sakthivel, 2013;Luca et al, 2013;Pardo-Pascual et al, 2012;Bouchahma et al, 2012;…”

“…Senevirathna et al (2018) conducted a survey using the past and recent Quick Bird data derived from GE to investigate the coastal zone trend in Unawatuna beach of Sri Lanka. Ratna et al (2018), used GE images to check Sentinel-2, ASTER and Landsat TM imagery accuracy, a combination of them were used to study the shoreline changes in Indonesia.…”

“…Structure from Motion (SfM) has been used by (Alphonse et al, 2019;Christopher et al, 2015;Elsner Paul et al, 2015;Ian et al, 2016;James et al, 2015;Nino et al, 2015;Papakonstantinou et al, 2016;Pauline et al, 2018). FCM (fuzzy c-means) has been used by (Ratna (2019); Ratna and Wietske (Ratna & Bijker, 2019); Ratna et al (2018); Ratna et al (2017)). K-means algorithm used by Paravolidakis et al (Vasilis et al, 2018); Vandebroek et al (2017); Paravolidakis et al (2016).…”

Benchmarking satellite-derived shoreline mapping algorithms

“…A variety of satellite-based shoreline detection methods are presently available. To extract shoreline observations from satellite imagery, many established SDS algorithms employ different image processing methods, including contouring of a land/water threshold 12,[17][18][19] ; maximum-gradient contouring methods [20][21][22] ; and soft classification techniques [23][24][25] . These methods can also be divided into 'at pixel resolution' and 'sub-pixel resolution,', where 'at pixel resolution' methods tend to create a stair-cased waterline, while sub-pixel methods integrate the information of neighboring pixels to obtain a smoother contour by using, for example, the Marching Squares algorithm 26 .…”