Monitoring Evolution of Coastline Along Mto Tamamba Delta in Kenya North Coast Using Satellite Images

Abstract

Coastal zones are the epicenter of economic activities in most developing countries in the world hence are bound to change. With advancement in technology monitoring such degradation has become critical subject for scientific research. This study reports a remarkable shoreline change; an aftermath of hydrodynamic effects and human influence on the Kenyan coastline. The sequential accumulation of sediments on Kenya North coast over the past three decades (1990-2021) using Remote sensing and GIS is tracked. Literature on this chronological evolution and its impacts along this coastline is meagre; it is therefore anticipated that the outcome of this study will initiate an elaborate probe into circumstances that contributed to landscape change along this coast. Multispectral Landsat and Sentinel-2 images respectively, display a prograding beach along Mto Tamamba Delta (Kenyan North Coast) a consequence of coastal erosion. These change in coastal morphology poses a threat to marine ecosystems, riparian communities and tourism industry. Prior reports indicate that research work done along this coastline were by field survey. This study is aimed at monitoring coastal erosion and accretion using satellite sensors. To achieve these, coastline data for these years were drawn numerically with the aid of RS and Arc GIS on ENVI 5.3 software. Maximum likelihood and Spectral angular mapper algorithms were used to classify the images due to their threshold input ability that ensures optimum classification results. Time series methodology on classified endmembers was then used to infer the presence of erosion and accretion. Using the image subset technique, areas prone to erosion and accretion within the coastal datum were identified and cut. The patterns of spatiotemporal trend of these endmembers were effectively employed to corroborate these findings. To explain reasons for change, time series analysis and confusion matrix were used. Change in distribution of endmembers over the 31 years’ period was assessed using thematic change technique; an approach not previously applied in this study area. From the classified Mto Tamamba image, the pixel coverage area for each endmember was extracted. The regression analysis from class statistics obtained from Mto Tamamba area for bare land and water showed a correlation with R2 =0.5138 with a P-value ˂ 0.005 on Landsat data. On the other hand, Sentinel-2 classified image a correlation of R2= 0.7857 for the same endmembers. Validation using field campaign data gave an overall accuracy of 84% and kappa coefficient of 0.799 for Landsat while for sentinel-2 an overall accuracy of 88% and Kappa coefficient of 0.849. The fact that validation of resampled classified Landsat image using a higher spatial resolution satellite (Sentinel-2) gave a better result of 92% implies that adoption of higher resolution data as validation process can aid in assessing impact of resolution on classification results.