Evaluation of riverbank erosion based on mangrove boundary changes identification using multi-temporal satellite imagery

Abstract

Evaluating riverbank erosion in mangrove forests is dynamic and challenging because of the complex environment that is exposed to tidal and sedimentation factor. Besides, assessing riverbank erosion in this environment requires a technique that reduces dependency on tidal and sedimentation without affecting the quality of the assessment. Hence, this study evaluated riverbank erosion based on mangrove boundary changes using multi-temporal satellite images comprising Quickbird, WorldView-2 and Pleiades-1B. The first objective of this study is to determine mangrove boundary shifting and its long-term impact towards riverbank features followed by validating the mangrove boundary shifting of satellite imagery with field measurement data, which comprise Real Time Kinematic-Global Positioning System (RTK-GPS). Next, the study assessed the rates of changes o f the riverbank erosion and accretion and the final objective developing a riverbank erosion prediction model. In this study, a change detection technique was used to identify the mangrove boundary changes of Kilim River at different timelines. The extracted mangrove boundary from satellite images for the years 2005, 2012 and 2017 were used to identify changes in the riverbank features such as line shifting, river width, erosion, and accretion. Subsequently, a vector image overlay was used to determine the mangrove boundary shifting for the corresponding years and evaluate the erosion and accretion rates using symmetrical difference and erase tool in ArcGIS software. Sequentially, Root Mean Square Error (RMSE) analysis validated the accuracy of image geo- referencing process while residual analysis was employed to validate the accuracy between satellite imagery and field measurement data comprising RTK-GPS and erosion pin data. Then, line buffering and kernel density analysis were used to develop a riverbank erosion prediction model based on three parameters, namely distance of erosion, area of erosion and direction of shifted mangrove boundary. The initial findings of this study showed that the mangrove boundary changes shifted backwards in the opposite direction from the river and the range of shifting was different according to the intensity o f boat traffic. One of the findings showed that the increasing rates of riverbank erosion ranged from 11302.019 square meters in the first epoch to 15674.721 square meters in the second epoch. Another finding illustrated the riverbank erosion prediction model which displayed several areas such as Sections A, B, I and L which are potentially facing serious riverbank erosion problems in the future in comparison to Sections C, D, E, F, G, H and K. The final finding discussed data validation between Pleiades-1B and GPS-RTK which recorded 0.305 of the r-square value whereas 0.477 was recorded as the r-square value for both Pleiades-1B and the erosion pin. The other validation comprised the second epoch of satellite image (WorldView-2 and Pleiades- 1B) and the erosion pin data which revealed the r-square of 0.9347 and showed the strong relationship between both data. As a conclusion, the findings have shown that the evaluation of the riverbank erosion based on mangrove boundary changes using multi-temporal satellite images is capable o f assisting stakeholders including the Langkawi Development Authority (LADA), Department of Irrigation and Drainage Malaysia (DID) and Marine Department Malaysia to have in-depth understanding of riverbank erosion issue that would enable them to prepare a mitigation plan in the future.