A critical analysis using remote sensing and GIS techniques for spatial distribution and macro-morphological analyses of rockfalls in the Golden Gate Highlands National Park, South Africa
De Lemos, Hugo Jose
A variety of mass movement and depositional geomorphic phenomena in the Golden Gate Highlands National Park have been described in the literature over the past few decades, yet notably, the prominent rockfalls remain understudied. The objective of the MSc study is to undertake detailed mapping of the GGHNP rockfalls, and in so doing ascertain a better understanding of their macro-morphologies, associated spatial-size dynamics and relationships to lithology, using both field-based and desktop image processing techniques. Ground truthing, using the highest possible resolution achievable through differential GPS (DGPS) and field measurements with an accuracy of cm to mm, serves to quantify the accuracy of mapping rockfall phenomena through 0.5 m GSD colour aerial imagery. In addition, field measurements are used as inputs for feature extraction, such that rockfalls associated with the Clarens, Elliot and Molteno Formation Sandstones may be defined using object orientated classification techniques. The dimensions, orientation and absolute coordinates of rockfalls for select representative sites were captured using both field based and desktop techniques. The rockfall coordinates were taken at the midpoint of each measured rock using a DGPS, with an accuracy of ~ 2 cm on the x, y and z axis. Object Based Image Analysis (OBIA) of Clarens and Molteno Fm. rockfalls was performed successfully using ground-truthed rockfall measurements to guide the creation of segmentation and classification rulesets. Multiple linear regression modelling can be used to model rockfall characteristics from ground-truthing with remotely sensed imagery, albeit to a very limited extent. Elliot Formation rockfalls could not be mapped and analysed mainly due to resolution limitations. Uni- and bi-variate statistics show promise in interpreting rockfall distribution and weighting with environmental variables derived from a DEM and geological vector. Point density analyses found that for the entire GGHNP ~2 and ~4 rockfalls are found per Km² of Clarens and Molteno Formation rockfall areas
A dissertation submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of requirements for the degree of Master of Science. 11/30/2013.