Monitoring and modelling disturbances to the Niger Delta mangrove forests

The Niger River Delta provides numerous ecosystem services (ES) to local populations and holds a wealth of biodiversity. Nevertheless, they are under threat of degradation and loss mainly due to the population increase and oil and gas extraction activities. Monitoring mangrove vegetation change and understanding the dynamics related with these changes is crucial for the short and longer-term sustainability of the Niger Delta Region (NDR) and its mangrove forests. Over the last two decades, open access remote sensing data, together with technological and algorithmic advancements, have provided the ability to monitor land cover over large areas through space and time. However, the analysis of land cover dynamics over the NDR using freely available optical remote sensing data, such as Landsat, remains challenging due to the gaps in the archive associated with the West African region and the issue of cloud contamination over the wet tropics. This thesis applies state-art-of-the-art remote sensing techniques and integrated modelling approaches to provide reliable information relating to monitoring and modelling of land cover change in the NDR, focusing on its mangrove forests. Spectral-temporal metrics from all available Landsat images were used to accurately map land cover in three time points, using a Random Forests machine learning classification model. The performance of the classification was tested when L-band radar data are added to the Landsat-based metrics. Results showed that Landsat based metrics are sufficient in mapping land cover over the study region with high overall classification accuracies over the three time points (1988, 2000, and 2013) and degraded mangroves were accurately mapped for the first time. Two additional assessments: a change intensity analysis for the entire NDR and, fragmentation analysis focusing on mangrove land cover classes were carried out for the first time ever. The drivers of mangrove degradation were assessed using a Multi-layer Perceptron, Artificial Neutral Networks (MLP-ANN) algorithm. The results reveal that built-up infrastructure variables were the most important drivers of mangrove degradation between 1988 and 2000, whilst oil and gas infrastructure variables were the most important drivers between 2000 and 2013. Results also show that population density was the least important driver of mangrove degradation over the two study periods. Future land cover changes and mangrove degradation were predicted under two business-as-usual scenarios in the short (2026) and longer-term (2038) using a Multi-Layer Perceptron neutral network and Markov chain (MLP-ANN+MC) model. The model’s accuracy was assessed using the highly-accurate land cover classification of 2013. Results show that that mangrove forest and woodlands (lowland and freshwater forests) are demonstrating a net loss, whilst the built-up areas and agriculture are indicating a net increase in both the short and longer-term scenarios. However, degraded mangroves are demonstrating a net increase in the short-term scenario. Interestingly, in the longer-term scenario, more than double the net increase of mangroves degraded in the short-term scenario, are predicted to recover to their healthier state. The thesis results could provide useful information for planning conservation measures for sustainable mangrove forest management of the entire NDR