Analysis of classification accuracy for pre-filtered multichannel remote sensing data

Noise is one of the main factors degrading the quality of original multichannel remote sensing data and its presence influences classification efficiency, object detection, etc. Thus, pre-filtering is often used to remove noise and improve the solving of final tasks of multichannel remote sensing. Recent studies indicate that a classical model of additive noise is not adequate enough for images formed by modern multichannel sensors operating in visible and infrared bands. However, this fact is often ignored by researchers designing noise removal methods and algorithms. Because of this, we focus on the classification of multichannel remote sensing images in the case of signal-dependent noise present in component images. Three approaches to filtering of multichannel images for the considered noise model are analysed, all based on discrete cosine transform in blocks. The study is carried out not only in terms of conventional efficiency metrics used in filtering (MSE) but also in terms of multichannel data classification accuracy (probability of correct classification, confusion matrix). The proposed classification system combines the pre-processing stage where a DCT-based filter processes the blocks of the multichannel remote sensing image and the classification stage. Two modern classifiers are employed, radial basis function neural network and support vector machines. Simulations are carried out for three-channel image of Landsat TM sensor. Different cases of learning are considered: using noise-free samples of the test multichannel image, the noisy multichannel image and the pre-filtered one. It is shown that the use of the pre-filtered image for training produces better classification in comparison to the case of learning for the noisy image. It is demonstrated that the best results for both groups of quantitative criteria are provided if a proposed 3D discrete cosine transform filter equipped by variance stabilizing transform is applied. The classification results obtained for data pre-filtered in different ways are in agreement for both considered classifiers. Comparison of classifier performance is carried out as well. The radial basis neural network classifier is less sensitive to noise in original images, but after pre-filtering the performance of both classifiers is approximately the same.

Using the Continuous Plankton Recorder to investigate the absence of North Atlantic right whales (Eubalaena glacialis) from the Roseway Basin foraging ground

North Atlantic right whales (Eubalaena glacialis) were absent from Roseway Basin, located off southeastern Nova Scotia, for a 7-year period (1993–1999). The objective of this study was to examine the availability of the right whale's main prey, Calanus finmarchicus, in Roseway Basin during those 7 years to determine if the whales’ absence was due to inadequate prey resources. Since we had no historical data on zooplankton abundances at depth on the Scotian Shelf, near-surface zooplankton abundance data from the Continuous Plankton Recorder were used to infer water-column abundances. In addition, environmental parameters that are often correlated with high zooplankton concentrations were examined. The hypotheses tested were that changes in these parameters would be detectable between three time periods: pre-1993, 1993–1999 and post-1999. Calanus finmarchicus abundance was found to be lowest during 1993–1999, suggesting that right whales were not foraging in Roseway Basin because of the near-absence of their main prey species. Decreased in situ salinity and density proved to be indicators of the changes in circulation in the 1990s that may have affected the advection of C. finmarchicus onto the Scotian Shelf.

Variations in GDGT distributions through the water column in the South East Atlantic Ocean

The TetraEther indeX of 86 carbon atoms (TEX86) temperature proxy is widely used in reconstructions of past sea surface temperature. Most current calibrations are based on surface sediment distributions of the glycerol dialkyl glycerol tetraether lipids (GDGTs) that comprise TEX86 and assume that these GDGTs are exported from the upper mixed layer. However, GDGT export from deeper waters could impact sedimentary GDGT distributions and therefore TEX86 paleothermometry. Here we examine GDGT distributions in suspended particulate matter (SPM) and underlying sediments collected from the Southeast Atlantic Ocean. Our results reveal different GDGT distributions - specifically the ratio between GDGTs bearing 2 vs. 3 cyclopentyl moieties, [2/3] ratios - between surface, subsurface (>50-200 m) and deep water (>200 m) SPM, which suggests the occurrence of in situ (deep) production that is not apparent when considering TEX86. The GDGT distributions in sediments match those of subsurface waters rather than surface waters, suggesting that they have not been preferentially derived from the upper mixed layer; this is consistent with GDGT abundances being highest in shallow subsurface SPM (˜100 to 200 m). It remains unclear what governs the different [2/3] ratios throughout the water column, but it is likely related to a combination of temperature and thaumarchaeotal community structure.

Abundance of phytoplankton, heterotrophic nanoflagellates and bacteria through the water column at time series station L4 in the Western English Channel, 2007-2013

short doi:10/rf8 full doi:10.5285/f014becf-d6d6-3bb9-e044-000b5de50f38

Simple shade plots aid better long-term choices of data pre-treatment in multivariate assemblage studies

Shade plots, simple visual representations of abundance matrices from multivariate species assemblage studies, are shown to be an effective aid in choosing an overall transformation (or other pre-treatment) of quantitative data for long-term use, striking an appropriate balance between dominant and less abundant taxa in ensuing resemblance-based multivariate analyses. Though the exposition is entirely general and applicable to all community studies, detailed illustrations of the comparative power and interpretative possibilities of shade plots are given in the case of two estuarine assemblage studies in south-western Australia: (a) macrobenthos in the upper Swan Estuary over a two-year period covering a highly significant precipitation event for the Perth area; and (b) a wide-scale spatial study of the nearshore fish fauna from five divergent estuaries. The utility of transformations of intermediate severity is again demonstrated and, with greater novelty, the potential importance seen of further mild transformation of all data after differential down-weighting (dispersion weighting) of spatially clumped' or schooled' species. Among the new techniques utilized is a two-way form of the RELATE test, which demonstrates linking of assemblage structure (fish) to continuous environmental variables (water quality), having removed a categorical factor (estuary differences). Re-orderings of sample and species axes in the associated shade plots are seen to provide transparent explanations at the species level for such continuous multivariate patterns.