Spatially weighted supervised classification for remote sensing

A simple approach for incorporating a spatial weighting into a supervised classifier for remote sensing applications is presented. The classifier modifies the feature-space distance-based metric with a spatial weighting. This is facilitated by the use of a non-parametric (k-nearest neighbour, k-NN) classifier in which the spatial location of each pixel in the training data set is known and available for analysis. A remotely sensed image was simulated using a combined Boolean and geostatistical unconditional simulation approach. This simulated image comprised four wavebands and represented three classes: Managed Grassland, Woodland and Rough Grassland. This image was then used to evaluate the spatially weighted classifier. The latter resulted in modest increase in the accuracy of classification over the original k-NN approach. Two spatial distance metrics were evaluated: the non-centred covariance and a simple inverse distance weighting. The inverse distance weighting resulted in the greatest increase in accuracy in this case.     

Self-organised clustering for road extraction in classified imagery

The extraction of road networks from digital imagery is a fundamental image analysis operation. Common problems encountered in automated road extraction include high sensitivity to typical scene clutter in high-resolution imagery, and inefficiency to meaningfully exploit multispectral imagery (MSI). With a ground sample distance (GSD) of less than 2 m per pixel, roads can be broadly described as elongated regions. We propose an approach of elongated region-based analysis for 2D road extraction from high-resolution imagery, which is suitable for MSI, and is insensitive to conventional edge definition. A self-organising road map (SORM) algorithm is presented, inspired from a specialised variation of Kohonen's self-organising map (SOM) neural network algorithm. A spectrally classified high-resolution image is assumed to be the input for our analysis. Our approach proceeds by performing spatial cluster analysis as a mid-level processing technique. This allows us to improve tolerance to road clutter in high-resolution images, and to minimise the effect on road extraction of common classification errors. This approach is designed in consideration of the emerging trend towards high-resolution multispectral sensors. Preliminary results demonstrate robust road extraction ability due to the non-local approach, when presented with noisy input.     

Non-stationary Approaches for Mapping Terrain and Assessing Prediction Uncertainty

It is well known that terrain may vary markedly over small areas and that statistics used to characterise spatial variation in terrain may be valid only over small areas. In geostatistical terminology, a non-stationary approach may be considered more appropriate than a stationary approach. In many applications, local variation is not accounted for sufficiently. This paper assesses potential benefits in using non-stationary geostatistical approaches for interpolation and for the assessment of uncertainty in predictions with implications for sampling design. Two main non-stationary approaches are employed in this paper dealing with (1) change in the mean and (2) change in the variogram across the region of interest. The relevant approaches are (1) kriging with a trend model (KT) using the variogram of residuals from local drift and (2) locally-adaptive variogram KT, both applied to a sampled photogrammetrically derived digital terrain model (DTM). The fractal dimension estimated locally from the double-log variogram is also mapped to illustrate how spatial variation changes across the data set. It is demonstrated that estimation of the variogram of residuals from local drift is worthwhile in this case for the characterisation of spatial variation. In addition, KT is shown to be useful for the assessment of uncertainty in predictions. This is shown to be true even when the sample grid is dense as is usually the case for remotely-sensed data. In addition, both ordinary kriging (OK) and KT are shown to provide more accurate predictions than inverse distance weighted (IDW) interpolation, used for comparative purposes.     

Spatial Concentration in Australian Regional Development,Exogenous Shocks and Regional Demographic Outcomes: a South Australian case study

ABSTRACT

As a tribute to the massive contribution of our friend and colleague Graeme Hugo to the population and settlement geography of Australian rural areas, this paper presents a longitudinal study from his home State. It forms part of a wider study of the long-term demographic relationships between Australia’s rapidly growing regional cities and their surrounding functional regions. Of particular interest is the question of what effect the accelerating concentration of population and economic activity into a given regional city will have for the longer term demographic sustainability of its functional region as a whole. Taking the case of Port Lincoln, regional capital of most of South Australia’s Eyre Peninsula, it examines the nature of change in the functional region over the period 1947–2011, and investigates the forces feeding, and partly counteracting, the population concentration process, informed by concepts of evolutionary economic geography. In particular it traces the demographic impact (particularly differential migration and ageing trends) of exogenous shocks to the region’s essentially primary productive economic base during the period of major change from 1981 to 2011.     

Blind Source Separation for Compositional Time Series

Many geological phenomena are regularly measured over time to follow developments and changes. For many of these phenomena, the absolute values are not of interest, but rather the relative information, which means that the data are compositional time series. Thus, the serial nature and the compositional geometry should be considered when analyzing the data. Multivariate time series are already challenging, especially if they are higher dimensional, and latent variable models are a popular way to deal with this kind of data. Blind source separation techniques are well-established latent factor models for time series, with many variants covering quite different time series models. Here, several such methods and their assumptions are reviewed, and it is shown how they can be applied to high-dimensional compositional time series. Also, a novel blind source separation method is suggested which is quite flexible regarding the assumptions of the latent time series. The methodology is illustrated using simulations and in an application to light absorbance data from water samples taken from a small stream in Lower Austria.

     

Can sediment supply variations create sequences? Insights from stratigraphic forward modelling

Classic sequence stratigraphy suggests depositional sequences can form due to changes in accommodation and due to changes in sediment supply. Accommodation‐dominated sequences are problematic to define rigorously, but are commonly interpreted from outcrop and subsurface data. In contrast, supply‐dominated sequences are much less commonly identified. We employ numerical stratigraphic forward modelling to compare stratal geometries forced by cyclic changes in relative sea level with stratal geometries forced by sediment discharge and water discharge changes. Our quantitative results suggest that both relative sea‐level oscillations and variations in sediment/water discharge ratio are able to form sequence‐bounding unconformities independently, confirming previous qualitative sequences definitions. In some of the experiments, the two types of sequence share several characteristics, such as an absence of coastal‐plain topset deposits and stratal offlap, something typically interpreted as the result of falling relative sea level. However, the stratal geometries differ when variations in amplitude and frequency of relative sea‐level change, sediment/water discharge ratio, transport diffusion coefficient and initial bathymetry are applied. We propose that the supply‐dominated sequences could be recognised in outcrop or in the subsurface if the observations of stratal offlap and the absence of coastal‐plain topset can be made without any strong evidence of relative sea‐level fall (e.g. descending shoreline trajectory). These quantitative results suggest that both supply‐dominated and accommodation‐dominated sequences are likely to occur in the ancient record, as a consequence of multiple, possibly complex, controls.     

Large-scale laboratory experiment on erosion of sand beds by moving circular vertical jets

This study investigates the topographic deformation due to the erosion of a sand bed impinged by a moving submerged turbulent round jet in a large-scale laboratory. The test conditions represent the case of discharges beneath a vessel while operating in water with a limited clearance such as a shallow navigation channel. The jet moves horizontally and discharges water vertically downward towards the bed. The distance between the jet nozzle and the bed equals six times the jet diameter so the jet flow is in the potential core region. The speed of the jet horizontal motion was varied to examine its effect on the scour profile. The characteristic lengths of the scour profile in the asymptotic state were determined by modifying the empirical formulas in Aderibigbe and Rajaratnam [1996. Erosion of loose beds by submerged circular impinging vertical turbulent jets. Journal of Hydraulic Research 34(1), 19–33]. The maximum scour depth, the scour hole radius, and the ridge height were found to be a function of the ratio of the jet exit to jet translation velocities and were modeled using a hyperbolic function. Empirical equations describing the scour profile were developed and the scour profile was found to be self-similar when normalized by appropriate length scales.     

Topographic Rossby waves in the Gulf of Mexico

Observations of topographic Rossby waves (TRW), using moored current meters, bottom pressure gauges, and Lagrangian RAFOS floats, are investigated for the deep basin of the Gulf of Mexico. Recent extensive measurement programs in many parts of the deep gulf, which were inspired by oil and gas industry explorations into ever deeper water, allow more comprehensive analyses of the propagation and dissipation of these deep planetary waves. The Gulf of Mexico circulation can be divided into two layers with the ∼800-1200 m upper layer being dominated by the Loop Current (LC) pulsations and shedding of large (diameters ∼300-400 km) anticyclonic eddies in the east, and the translation of these LC eddies across the basin to the west. These processes spawn smaller eddies of both signs through instabilities, and interactions with topography and other eddies to produce energetic surface layer flows that have a rich spectrum of orbit periods and diameters. In contrast, current variability below 1000 m often has the characteristics of TRWs, with periods ranging from ∼10-100 days and wavelengths of ∼50-200 km, showing almost depth-independent or slightly bottom intensified currents through the weakly stratified lower water column. These fluctuations are largely uncorrelated with simultaneous upper-layer eddy flows. TRWs must be generated through energy transfer from the upper-layer eddies to the lower layer by potential vorticity adjustments to changing depths of the bottom and the interface between the layers. Therefore, the LC and LC eddies are prime candidates as has been suggested by some model studies. Model simulations have also indicated that deep lower-layer eddies may be generated by the LC and LC eddy shedding processes.In the eastern gulf, the highest observed lower-layer kinetic energy was north of the Campeche Bank under the LC in a region that models have identified as having strong baroclinic instabilities. Part of the 60-day TRW signal propagates towards the Sigsbee Escarpment (a steep slope at the base of the northern continental slope), and the rest into the southern part of the eastern basin. Higher energy is observed along the escarpment between 89°W and 92°W than either under the northern part of the LC or further south in the deep basin, because of radiating TRWs from the western side of the LC. In the northern part of the LC, evidence was found in the observations that 20-30-day TRWs were connected with the upper layer through coherent signals of relative vorticity. The ∼90° phase lead of the lower over the upper-layer relative vorticity was consistent with baroclinic instability. Along the Sigsbee Escarpment, the TRWs are refracted and reflected so that little energy reaches the lower continental slope and a substantial mean flow is generated above the steepest part of the escarpment. RAFOS float tracks show that this mean flow continues along the escarpment to the west and into Mexican waters. This seems to be a principal pathway for deepwater parcels to be transported westward. Away from the slope RAFOS floats tend to oscillate in the same general area as if primarily responding to the deep wave field. Little evidence of westward translating lower-layer eddies was found in both the float tracks and the moored currents. In the western gulf, the highest deep energy levels are much less than in the central gulf, and are found seaward of the base of the slope. Otherwise, the situation is similar with TRWs propagating towards the slope, probably generated by the local upper-layer complex eddy field, being reflected and forcing a southward mean flow along the base of the Mexican slope. Amplitudes of the lower-layer fluctuations decay from the northwest corner towards the south.     

Climate change and coral reef bleaching: An ecological assessment of long-term impacts,recovery trends and future outlook

Since the early 1980s, episodes of coral reef bleaching and mortality, due primarily to climate-induced ocean warming, have occurred almost annually in one or more of the world's tropical or subtropical seas. Bleaching is episodic, with the most severe events typically accompanying coupled ocean–atmosphere phenomena, such as the El Niño-Southern Oscillation (ENSO), which result in sustained regional elevations of ocean temperature. Using this extended dataset (25+ years), we review the short- and long-term ecological impacts of coral bleaching on reef ecosystems, and quantitatively synthesize recovery data worldwide. Bleaching episodes have resulted in catastrophic loss of coral cover in some locations, and have changed coral community structure in many others, with a potentially critical influence on the maintenance of biodiversity in the marine tropics. Bleaching has also set the stage for other declines in reef health, such as increases in coral diseases, the breakdown of reef framework by bioeroders, and the loss of critical habitat for associated reef fishes and other biota. Secondary ecological effects, such as the concentration of predators on remnant surviving coral populations, have also accelerated the pace of decline in some areas. Although bleaching severity and recovery have been variable across all spatial scales, some reefs have experienced relatively rapid recovery from severe bleaching impacts. There has been a significant overall recovery of coral cover in the Indian Ocean, where many reefs were devastated by a single large bleaching event in 1998. In contrast, coral cover on western Atlantic reefs has generally continued to decline in response to multiple smaller bleaching events and a diverse set of chronic secondary stressors. No clear trends are apparent in the eastern Pacific, the central-southern-western Pacific or the Arabian Gulf, where some reefs are recovering and others are not. The majority of survivors and new recruits on regenerating and recovering coral reefs have originated from broadcast spawning taxa with a potential for asexual growth, relatively long distance dispersal, successful settlement, rapid growth and a capacity for framework construction. Whether or not affected reefs can continue to function as before will depend on: (1) how much coral cover is lost, and which species are locally extirpated; (2) the ability of remnant and recovering coral communities to adapt or acclimatize to higher temperatures and other climatic factors such as reductions in aragonite saturation state; (3) the changing balance between reef accumulation and bioerosion; and (4) our ability to maintain ecosystem resilience by restoring healthy levels of herbivory, macroalgal cover, and coral recruitment. Bleaching disturbances are likely to become a chronic stress in many reef areas in the coming decades, and coral communities, if they cannot recover quickly enough, are likely to be reduced to their most hardy or adaptable constituents. Some degraded reefs may already be approaching this ecological asymptote, although to date there have not been any global extinctions of individual coral species as a result of bleaching events. Since human populations inhabiting tropical coastal areas derive great value from coral reefs, the degradation of these ecosystems as a result of coral bleaching and its associated impacts is of considerable societal, as well as biological concern. Coral reef conservation strategies now recognize climate change as a principal threat, and are engaged in efforts to allocate conservation activity according to geographic-, taxonomic-, and habitat-specific priorities to maximize coral reef survival. Efforts to forecast and monitor bleaching, involving both remote sensed observations and coupled ocean–atmosphere climate models, are also underway. In addition to these efforts, attempts to minimize and mitigate bleaching impacts on reefs are immediately required. If significant reductions in greenhouse gas emissions can be achieved within the next two to three decades, maximizing coral survivorship during this time may be critical to ensuring healthy reefs can recover in the long term.