Quantifying morphologic changes of a coastal foredune using a low-cost remotely piloted aerial system (RPAS)

Abstract

Mid-latitude sandy coasts are dynamic environments. Monitoring coastal morphodynamics is important for understanding the response of coasts to short-term storm events, for understanding the response of coasts to long-term environmental change, and for managing beach-dune systems. Remotely piloted aerial systems (RPAS) (or drones) present new opportunities for coastal monitoring. This type of platform is inexpensive, efficient, requires minimal expertise, and also provides high resolution aerial imagery. Photogrammetry can be used to derive digital surface models (DSMs) or digital terrain models (DTMs) from RPAS imagery. This thesis assesses the efficacy of low-cost RPAS for describing the morphology and morphodynamics of coastal foredunes. The first objective is to compare DSMs produced by RPAS surveying with DTMs derived using conventional survey methods. Objective two assesses the accuracy and precision of RPAS surveying to quantify morphologic changes of a coastal foredune. The third objective is to examine the influence of vegetation on RPAS-derived DSMs. Comparisons are made between total station, RTK-GPS, terrestrial laser scanner and RPAS surveys conducted on the St. Kilda beach foredune, Dunedin. The surveying methods are compared based on survey efficiency, cost, accuracy of the DTM/DSM, and their sensitivity to atmospheric and environmental limitations. RPAS photogrammetry is used to develop a time series of DSMs, which describe short-term patterns of sedimentation and morphological changes in the lee of this foredune. Vegetation surveys were conducted on the foredune at Mason Bay, Stewart Island, and the areas are classified as uniform and dense, variable, and sparse vegetation, or bare sand. Plots containing each class were surveyed with RPAS and RTK-GPS, to produce a DTM and a DSM that are compared to determine the elevation difference. The RPAS survey was the most efficient method for developing DSMs, even when considering the set-up and data processing time (Objective 1). The RPAS produced the second most precise surface, with a RMSE of 8 cm. The RPAS is more sensitive to environmental and atmospheric conditions; however, this method is very rapid, and undesirable weather conditions can be avoided. The results show there is un-modelled systematic error in the DSM caused by lens distortion, which increases outside the GCP network – areas outside the network were not used for subsequent analysis. Vegetation presence can prevent the derivation of accurate DTMs. The RPAS did not accurately quantify sand deposition due to the presence of vegetation (Objective 2). The sand dampened the vegetation, causing a decrease in elevation in the change model. The sensitivity of the RPAS to vegetation is insignificant in areas with bare or sparse vegetation, or when quantifying large-scale changes (for example, foredune erosion). Vegetation height, vegetation cover/density, GSD, the structural properties of the plant, and the surface spectral properties, were identified as factors causing an elevational offset in the DSM (Objective 3). The elevation of the areas with bare sand were statistically equal in the DTM and DSM, however, the dense, variable and sparsely vegetated areas were statistically different. The elevation difference between the DSM and DTM is the largest in the densely vegetated areas (30 cm). Low-cost RPAS are capable of achieving high-quality morphologic surveys of coastal foredunes. The method affords the advantages of efficiency and flexibility. However, due to the sensitivity of the method to vegetation, low-cost RGB RPAS are more suited to quantifying the morphology of bare sand or sparsely vegetated areas, quantifying large-scale changes, or for long-term morphologic monitoring. Low-cost RPAS are not capable of accurately quantifying small-scale changes in areas with dense vegetation. However, as RPAS platforms develop, it is expected that sensors capable of penetrating vegetation will become more accessible for low-cost platforms.