Detailed studies on the sediment budget may reveal valuable insights into the successive build-up of the Canterbury Plains and their modification by Holocene fluvialaction connected to major braided rivers. Additionally, they bear implications beyond these fluvial aspects. Palaeoseismological studies claim to have detected signals of major Alpine Fault earthquakes in coastal environments along the eastern seaboard of the South Island (McFadgen and Goff, 2005). This requires high connectivity between the lower reaches of major braided rivers and their mountain catchments to generate immediate significant sediment pulses. It would be contradictory to the above mentioned hypothesis though. Obtaining better control on sediment budgets of braided rivers like the Waimakariri River will finally add significant value to multiple scientific and applied topics like regional resource management. An essential first step of sediment budget studies Is to systematically map the geomorphology, conventionally in the field and/or using remote-sensing applications, to localise, genetically identify, and classify landforms or entire toposequences of the area being investigated. In formerly glaciated mountain environments it is also indispensable to obtain all available chronological information supporting subsequent investigations.
Land cover change information in urban areas supports decision makers in dealing with public policy planning and resource management. Remote sensing has been demonstrated as an efficient and accurate way to monitor land cover change over large extents. The Canterbury Earthquake Sequence (CES) caused massive damage in Christchurch, New Zealand and resulted in significant land cover change over a short time period. This study combined two types of remote sensing data, aerial imagery (RGB) and LiDAR, as the basis for quantifying land cover change in Christchurch between 2011 – 2015, a period corresponding to the five years immediately following the 22 February 2011 earthquake, which was part of the CES. An object based image analysis (OBIA) approach was adopted to classify the aerial imagery and LiDAR data into seven land cover types (bare land, building, grass, shadow, tree and water). The OBIA approach consisted of two steps, image segmentation and object classification. For the first step, this study used multi-level segmentation to better segment objects. For the second step, the random forest (RF) classifier was used to assign a land cover type to each object defined by the segmentation. Overall classification accuracies for 2011 and 2015 were 94.0% and 94.32%, respectively. Based on the classification result, land cover changes between 2011 and 2015 were then analysed. Significant increases were found in road and tree cover, while the land cover types that decreased were bare land, grass, roof, water. To better understand the reasons for those changes, land cover transitions were calculated. Canopy growth, seasonal differences and forest plantation establishment were the main reasons for tree cover increase. Redevelopment after the earthquake was the main reason for road area growth. By comparing the spatial distribution of these transitions, this study also identified Halswell and Wigram as the fastest developing suburbs in Christchurch. These results provided quantitative information for the effects of CES, with respect to land cover change. They allow for a better understanding for the current land cover status of Christchurch. Among those land cover changes, the significant increase in tree cover aroused particularly interest as urban forests benefit citizens via ecosystem services, including health, social, economic, and environmental benefits. Therefore, this study firstly calculated the percentages of tree cover in Christchurch’s fifteen wards in order to provide a general idea of tree cover change in the city extent. Following this, an automatic individual tree detection and crown delineation (ITCD) was undertaken to determine the feasibility of automated tree counting. The accuracies of the proposed approach ranged between 56.47% and 92.11% in thirty different sample plots, with an overall accuracy of 75.60%. Such varied accuracies were later found to be caused by the fixed tree detection window size and misclassifications from the land cover classification that affected the boundary of the CHM. Due to the large variability in accuracy, tree counting was not undertaken city-wide for both time periods. However, directions for further study for ITCD in Christchurch could be exploring ITCD approaches with variable window size or optimizing the classification approach to focus more on producing highly accurate CHMs.
