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.

Global assessment of ocean carbon export by combining satellite observations and food-web models

The export of organic carbon from the surface ocean by sinking particles is an important, yet highly uncertain, component of the global carbon cycle. Here we introduce a mechanistic assessment of the global ocean carbon export using satellite observations, including determinations of net primary production and the slope of the particle size spectrum, to drive a food-web model that estimates the production of sinking zooplankton feces and algal aggregates comprising the sinking particle flux at the base of the euphotic zone. The synthesis of observations and models reveals fundamentally different and ecologically consistent regional-scale patterns in export and export efficiency not found in previous global carbon export assessments. The model reproduces regional-scale particle export field observations and predicts a climatological mean global carbon export from the euphotic zone of ~6 Pg C yr−1. Global export estimates show small variation (typically < 10%) to factor of 2 changes in model parameter values. The model is also robust to the choices of the satellite data products used and enables interannual changes to be quantified. The present synthesis of observations and models provides a path for quantifying the ocean's biological pump.

Ecological controls on biogeochemical fluxes in the western Antarctic Peninsula studied with an inverse foodweb model

Sea ice in the western Antarctic Peninsula (WAP) region is both highly variable and rapidly changing. In the Palmer Station region, the ice season duration has decreased by 92 d since 1978. The sea-ice changes affect ocean stratification and freshwater balance and in turn impact every component of the polar marine ecosystem. Long-term observations from the WAP nearshore and offshore regions show a pattern of chlorophyll (Chl) variability with three to five years of negative Chl anomalies interrupted by one or two years of positive anomalies (high and low Chl regimes). Both field observations and results from an inverse food-web model show that these high and low Chl regimes differed significantly from each other, with high primary productivity and net community production (NCP) and other rates associated with the high Chl years and low rates with low Chl years. Gross primary production rates (GPP) averaged 30 mmolC.m-2.d-1 in the low Chl years and 100 mmolC.m-2.d-1 in the high Chl years. Both large and small phytoplankton were more abundant and more productive in high Chl years than in low Chl years. Similarly, krill were more important as grazers in high Chl years, but did not differ from microzooplankton in high or low Chl years. Microzooplankton did not differ between high and low Chl years. Net community production differed significantly between high and low Chl years, but mobilized a similar proportion of GPP in both high and low Chl years. The composition of the NCP was uniform in high and low Chl years. These results mphasize the importance of microbial components in the WAP plankton system and suggest that food webs dominated by small phytoplankton can have pathways that funnel production into NCP, and likely, export.