The Lake Nakuru (Kenya) faces flooding since 2010, thus interfering with the growth of riparian vegetation, and requiring automated image analysis to monitor the effect of flooding on the Lake Nakuru Riparian Reserve vegetation species. The vegetation specie that was affected by the flooding lake is Acacia Xanthophloea, i.e. a habitat and feed for many animal species within the National Park. In this study, we explore how riparian vegetation classification can ease the detection of degraded Acacia Xanthophloea spp. To overcome the limitations of pixel-based analysis (e.g. salt and pepper effect in the classification map), we have conducted object-based image analysis on satellite imagery (Landsat-5TM and Landsat-8OLI) acquired at different dates: a baseline consisting of an image acquired in 2000 (i.e. before flooding), and several images acquired in 2010, 2012, 2014 and 2016. Such Landsat Earth Observation satellites provide information in the visible, near-infrared and shortwave regions of the electro-magnetic spectrum, which are of utmost importance to assess vegetation vitality. The OBIA methodology employed here consists in several steps: some pre-processing (georeferencing, radiometric corrections, geometric corrections, pan-sharpening, ortho-rectification) is applied on each image ; objects are extracted from each image by eCognition segmentation tool ; spatial and spectral information were combined on riparian reserve vegetation to extracting information about land cover classes ; vegetation was extracted through NDVI thresholding based on visual analysis with eCognition ; thematic re-classification and feature extraction was performed with Arc-GIS ; supervised classification was achieved for each input image using Weka (an open source data mining software) considering several algorithms, including Naïve Bayes classifier, J48 pruned tree, CART, Random Forests and Random Tree; relevant changes were identified as objects that were classified as Acacia Xanthophloea in the baseline, but not anymore in the more recent images. The classification accuracy was evaluated using cross-fold validation and a reference dataset of 232 instances and 8 attributes, and reaches 92% assuming 9 land cover classes (Cynodon/Chloris/Themada grasslands, Olive and Teclea forest, sedges and rashes, Lake Nakuru, Chloris Gayana, Sporobulus Spicatus and Acacia woodland). Furthermore, we were able to identify important features and to define rulesets to be used in a GEOBIA framework. Our study thus shows that: from a thematic point of view, GEOBIA was able to identify Acacia Xanthophloea from Landsat satellite imagery, and comparing the resulting classification maps allows us to achieve monitoring of this specie through time; from a software point of view, it might be necessary to involve several different tools (both proprietary or open source) since there is still some missing functionalities in the existing GEOBIA software solutions.