Spatio-temporal monitoring and analysis of Nigerian Bonny Island coastline dynamics and the physical impacts

Tsatsakis, Michail

4. term

Master of Technology Management (MTM) (Continuing education)

2018

2018-09-07

60

Denne afhandling udvikler en hurtig, lavomkostningsmetode til at overvåge, hvordan kystlinjen på Bonny Island i Nigeria har ændret sig over rum og tid fra 1984 til 2018, og til at vurdere de fysiske effekter af disse ændringer. Den bruger gratis Landsat‑satellitbilleder til at opbygge en tidsserie og klassificerer hver pixel som land eller vand med to teknikker: Random Forest (en maskinlæringsmetode, der kombinerer mange beslutningstræer) og Otsus tærskelmetode (en automatisk regel, der skelner mellem klasser ud fra billedets lysstyrke). Kystlinjen udtrækkes som grænsen mellem land og vand, og der oprettes mange tværgående linjer (transekter) med DSAS‑værktøjet for at beregne rater for erosion (landtab) og akkretion (landvinding). Resultaterne viser, at akkretion dominerer langs ca. 75,5 % af øen, med en gennemsnitlig fremrykning af kystlinjen på 6,81 m pr. år. Raterne for erosion og akkretion fra de to klassifikationsmetoder stemmer stærkt overens. En validering for 1984–2003, der beregnede arealer under erosion og akkretion, indikerer, at Otsus metode havde den højeste nøjagtighed. Begge metoder har fordele og ulemper, så valget bør afhænge af det tilsigtede formål. Studiet bemærker også, at brug af gratis data med grov opløsning ikke nødvendigvis reducerer klassifikationsnøjagtigheden. Detaljerede målinger gives i konklusionerne.

This thesis develops a fast, low‑cost way to monitor how Bonny Island’s coastline in Nigeria changed over space and time from 1984 to 2018, and to assess the physical effects of those changes. It uses free Landsat satellite images to build a time series and classifies each pixel as land or water with two techniques: Random Forest (a machine‑learning method that combines many decision trees) and Otsu’s thresholding (an automatic rule to separate classes based on image brightness). The coastline is then taken as the boundary between land and water, and many cross‑shore lines (transects) are created with the DSAS tool to calculate rates of erosion (land loss) and accretion (land gain). The results show that accretion dominates along about 75.5% of the island, with the shoreline advancing by an average of 6.81 m per year. The erosion and accretion rates from the two classification methods agree strongly. A validation for 1984–2003 that calculated the areas under erosion and accretion indicates that Otsu’s method achieved the highest accuracy. Both methods have advantages and drawbacks, so the choice should depend on the intended use. The study also notes that using free, coarse‑resolution imagery does not necessarily reduce classification accuracy. Detailed measurements are provided in the conclusions.

[This abstract was generated with the help of AI]

bonny island ; coastline extracion ; random forests ; otsu ; remote sensing ; spatio-temporal ; time series analysis ; landsat ; nigeria ; niger delta ; accretion ; erosion ; coastal dynamic ; classification ; gis ; masters ; thesis ; transects ; dsas ; rates ; machine learning

Download
View record in AAU Student Projects