Coarse clastic sediments (boulders) on coastlines have seen a groundswell in geomorphic research interest over recent years, associated in part with the potential of boulder evidence for interpreting characteristics of high-energy wave processes. Yet, the fundamental property of boulder volume is normally difficult to measure accurately owing to complex clast morphology and irregular surface texture. To tackle this problem, this paper concentrates on creating precise, measurable and textured threedimensional (3D) models of coastal boulders without physical contact with the object, based on multi-view image measurement techniques. This method has several advantages over traditional measurements that are inaccurate or alternative solutions using costly techniques such as terrestrial laser scanning. Our methods propose the use of low-cost equipment (digital cameras) that can be used in various coastal environments to easily acquire numerous images of the object of interest. Initial results can be rapidly assessed in the field for immediate quality control. Resulting 3D models, built from overlapping multi-view digital photographs, allow the reconstruction of realistic-looking and textured boulder surfaces. A particular interest in this task is the family of algorithms known as structure from motion (SFM). The work presents analysis of SFM techniques by examining 3D models of boulders observed at a coastal field site on Lu Dao Island in south-eastern Taiwan.

This paper introduces a new metric for tropical cyclone track shape within the tropical South Pacific (TSP) basin, based on measurements of track sinuosity. A sinuosity index (SI) is developed by applying a simple cube-root transformation to original track sinuosity values. Based on the resulting near-normal SI distribution, an ordinal fourcategory (quartile) naming system is then proposed for track-type classification. Track sinuosity patterns are also investigated over the last four decades . Analytical findings suggest that cyclone track sinuosity is an important parameter influencing the potential vulnerability of island archipelagoes to cyclone hazard. Principally, sinuously moving cyclones show some tendency for greater longevity and intensity than straightertracking storms and make up a larger proportion of systems forming in the western tropical South Pacific than those generated farther east. Although no long-term statistical trend can be established, track sinuosity is highly variable through time, implying that the TSP basin and the islands therein will continue to experience large but irregular inter-annual fluctuations in cyclone track morphology.

This study analyses the regional cyclone archive for the tropical South Pacific (160°E-120°W, 0°-25°S) maintained by the designated Regional Specialized Meteorological Centre located at Nadi in the Fiji Islands. The historical cyclone record was examined over 4 decades from the 1969-1970 cyclone season to the 2007-2008 season. Cyclogenesis origins, minimum pressures, durations and track parameters (azimuth and length) of 291 individual storms were investigated. Temporal variability in separate cyclone parameters was highly variable but not necessarily matching on an interannual basis. Anomalous periods of cyclone behaviour can be detected in 1976, 1981, 1983, 1991, 1998, 2001-2002 and 2003. Strong and significant inter-relationships are indicated between storm longevity, track length and minimum sea-level pressure (MSLP) attained, and also between seasonally averaged measures of latitude of cyclone origin and the strength of the Southern Oscillation Index and Multivariate ENSO Index. Yet no overall long-term linear trends were detected in the data, with the exception of MSLP which showed a spurious decreasing trend -a problem already highlighted in other cyclone archives. These findings suggest that the South Pacific cyclone basin and the islands therein will continue to experience strong but irregular interannual fluctuations in cyclone and track characteristics. Such anomalies will remain a much more dominant feature of temporal patterns than possibly evolving changes in long-term average cyclone activity resulting from climate change.

No abstract

In this article, multilayer perceptron (MLP) network models with spatial constraints are proposed for regionalization of geostatistical point data based on multivariate homogeneity measures. The study focuses on non‐stationarity and autocorrelation in spatial data. Supervised MLP machine learning algorithms with spatial constraints have been implemented and tested on a point dataset. MLP spatially weighted classification models and an MLP contiguity‐constrained classification model are developed to conduct spatially constrained regionalization. The proposed methods have been tested with an attribute‐rich point dataset of geological surveys in Ukraine. The experiments show that consideration of the spatial effects, such as the use of spatial attributes and their respective whitening, improve the output of regionalization. It is also shown that spatial sorting used to preserve spatial contiguity leads to improved regionalization performance.