The suitability of imagery from the Advanced Very High Resolution Radiometer (AVHRR) on the NOAA series of satellites for the synoptic classification of circulation trends in the European Arctic is ...assessed with reference to data from three climate stations. Simplified synoptic cyclonic classifications are derived from the satellite imagery and tested against climate data. Five classes of frontal system are derived from the tracking of systems over the UK-Scandinavia-Baltic region using 1460 satellite images over two years. An index based on the qualitative interpretation of satellite imagery was related to the reference data. The tracking of the systems in the imagery also facilitates a comparison of travel times across the region and the frequency of occurrence. Frontal systems that remain largely stationary over the Baltic were found to correlate best with precipitation at the reference sites. The paper thus investigates the use of AVHRR imagery for the categorisation of weather patterns towards deriving quantitative relationships between circulation classes and weather elements (such as temperature and precipitation) where, for example, climate data are sparse or where skills required for the interpretation of Height Potential Fields are lacking.
Celotno besedilo
Dostopno za:
BFBNIB, DOBA, IZUM, KILJ, NUK, PILJ, PNG, SAZU, UILJ, UKNU, UL, UM, UPUK
Abstract Small glaciers have short response times to climate change and therefore offer a powerful means of climate monitoring. Glacier responses to climate, or their mass change, may be suggested by ...a change in the Equilibrium Line Altitude (ELA). However, regional climatic reconstructions have repeatedly neglected the importance of local variations in ELAs in preference for regional trends. For small glaciers close to the glaciation level, ignoring the importance of local topographic components in mass balance estimates may lead to erroneous climatic reconstructions. Of 510 small valley and cirque glaciers digitised across northern Scandinavia, 284 were objectively deemed suitable for inferring an ELA. The inferred ELA was derived from the median elevation and several local topographic variables using regression analysis. The glacier elevation, area, slope and aspect parameters were found to be the best predictors of the local ELA. ELA estimations improved from 77% up to 94% accuracy when topographic parameters for every grid-cell within rasters representing glacier surfaces were computed rather than using subjective measurements from topographic maps. Regional ELA trend surfaces, interpolated between the local ELA values, differed in effectively representing the local variability, depending upon the distribution and accuracy of the local ELA values. A second-order polynomial trend surface most accurately represented ELA variations across the study area, within the initial local measurement accuracy of ±100 m. It is concluded that current subjective topographic map-based analyses are unlikely to be sufficiently accurate for predicting the regional ELA of small, sensitive and marginal glaciers, unless CIS-based spatial analyses are made at a reasonable resolution.
Jökulhlaups, or glacier outburst floods, have occurred during the Holocene from the northern margin of the Vatnajökull ice cap, Iceland. Relatively little is known about the origin, magnitude and ...frequency of these jökulhlaups. The volcanic rifting zone of northern Iceland provides a new environment in which to examine jökulhlaups. Jökulhlaup reconstructions have to date omitted 2D hydrodynamic modelling techniques. This research therefore reconstructs jökulhlaups from Kverkfjöll volcano, a discrete source of meltwater from northern Vatnajökull. This research describes a suite of erosional and depositional landforms that distinguish Kverkfjöll jökulhlaup routeways. Some of these; clinker-scoured lava, gorges with walls of pillow and subaerial lava, lava steps, cataract-fill mounds and imbricated boulder clusters and run-ups, are previously undocumented jökulhlaup impacts. These landforms may be diagnostic of volcanic and/or rifting landscape jökulhlaups. Cross-cutting relationships and sedimentary stratigraphy suggest at least three Holocene jökulhlaups from Kverkfjöll. Kverkfjöll jökulhlaups were reconstructed using palaeocompetence, slope-area and 2D hydrodynamic modelling. Jökulhlaups were volcanically triggered, had linearly-rising hydrographs and peak discharges of 50,000-100,000 m3s-1, which attenuated by ~75% within 25km. Flows were highly varied spatially and temporally, and strongly controlled by topography, geology and sediment supply. Frontal flow velocities were ~2ms-1 but as stage increased, mean velocities reached 5-15ms-1. Shear stress and stream power reached 1x104 Nm-2 and 1x105 Wm-2 respectively. Flows were initially hyperconcentrated and subsequently more fluidal, supercritical and highly turbulent. Kverkfjöll jökulhlaups achieved geomorphic work comparable to that generated by the largest known terrestrial floods. Landscaping resulted from topographic confinement, steep channel gradients, high hydraulic roughness and an initially abundant but rapidly depleted supply of volcaniclastic sediment. These controls on, and impacts of, jökulhlaups are important for distinguishing high-magnitude water-sediment inputs to the North Atlantic, for recognising jökulhlaups in the rock record and for flood hazard mitigation in similar landscapes and upon glaciated volcanoes.
In this paper we compare a variety of unsupervised probabilistic models used to represent a data set consisting of textual and image information. We show that those based on latent Dirichlet ...allocation (LDA) out perform traditional mixture models in likelihood comparison. The data set is taken from radiology; a combination of medical images and consultants reports. The task of learning to classify individual tissue, or disease types, requires expert hand labeled data. This is both: expensive to produce and prone to inconsistencies in labeling. Here we present methods that require no hand labeling and also automatically discover sub-types of disease. The learnt models can be used for both prediction and classification of new unseen data.
