Open Access

This report contains two parts: The first part presents an overview on studies concerning the Danube, inland navigation or the impact of climate change on either of those. The second part gives a more detailed analysis of inland navigation on the Danube, partly based on studies presented in part one. Part two covers the current situation along the Danube by covering the topic of bottlenecks and other limitations for shipping along the Danube. Based on these informations, an estimation of the economic impacts of low water periods on inland navigation is made. As a last step, measures to reduce the impact of low water on inland navigations are presented. The report shows, that inland navigation still is an important transport mode, along the Danube as well as in other european regions. Especially in Romania, inland navigation still has a large share of more than 20% and rising in total transport. However, inland navigation depends strongly on good conditions of its infrastructure. These good conditions are limited mainly by two factors: one are the so called bottlenecks. Those are areas with sub-optimal shipping conditions e.g. due to solid rock formations in the river that lead to a reduced water depths. The other factor is the weather (and, on a longer time scale, the climate) which, mostly depending on precipitation and evaporation, can lead to low water levels seasonally. In addition to these two natural factors, laws which e.g. regulate the maximum number of barges allowed. Human build structures like locks limit the size of vessels as well as the speed they can travel with. These limiting factors are identified and located in the first chapter of part two of this report, before the water depth needed by several ship sizes as well as the cargo fleet available along the Danube are presented. One of the targets of this report is to estimate the economic impact of low water periods. All the factors named above as well as the freight prices charged for connections along the Danube are used to each this target in chapter II.4. To estimate the impact of low water periods on the freight prices, a method developed by Jonkeren et al. (2007) for the Rhine is transfered to the Danube. By transfering Jonkeren et al. (2007) method, regression equations for several transport connections along the Danube are identified that give a first estimate for the connection of freight prices and water levels. With the help of these regression equations, an estimation of the total expenses for transport via inland navigation for several years is possible. The yearly and seasonal variability is identified as well as the additional expenses due to water levels below 280cm. But additional expenses are not the only impact of changing water levels on inland navigation. Another is, that while the demand for transport stays at the same level, sometimes the water levels are not sufficient enough to use the full capacity of the fleet. Therefore, the (theoretical) amount of cargo that could not be transported due to low water levels is calculated as well and presented in chapter II.5. Finally, some measures to overcome some of the here named problems of inland navigation due to low water levels are presented. These are separated into two general approches: change the ship or change the river. Both methods have their advantages and disadvantages due to technical as well as regulatory and other factors. The list presented here however is incomplete and only gives a few ideas of how some problems can be overcome. In the end, an individual mix for the different regions along the river and sometimes for the individual companies must be found.

Offline handwriting recognition systems often include a decoding step, that is retrieving the most likely character sequence from the underlying machine learning algorithm. Decoding is sensitive to ranges of weakly predicted characters, caused e.g. by obstructions in the scanned document. We present a new algorithm for robust decoding of handwriting recognizer outputs using character n-grams. Multidimensional hierarchical subsampling artificial neural networks with Long-Short-Term-Memory cells have been successfully applied to offline handwriting recognition. Output activations from such networks, trained with Connectionist Temporal Classification, can be decoded with several different algorithms in order to retrieve the most likely literal string that it represents. We present a new algorithm for decoding the network output while restricting the possible strings to a large lexicon. The index used for this work is an n-gram index with tri-grams used for experimental comparisons. N-grams are extracted from the network output using a backtracking algorithm and each n-gram assigned a mean probability. The decoding result is obtained by intersecting the n-gram hit lists while calculating the total probability for each matched lexicon entry. We conclude with an experimental comparison of different decoding algorithms on a large lexicon.

A method is investigated by which tight bounds on the range of a multivariate rational function over a box can be computed. The approach relies on the expansion of the numerator and denominator polynomials in Bernstein polynomials. Convergence of the bounds to the range with respect to degree elevation of the Bernstein expansion, to the width of the box and to subdivision are proven and the inclusion isotonicity of the related enclosure function is shown.

Vortrag bei der studentischen Konferenz Nachhaltigkeit an der HTWG, Konstanz, 04.05.2016 Vortrag beim Papierzentrum im Murgtal, Gernsbach, 09.05.2016

Ein Rückblick auf das akademische Jahr 2015/16 Berichtszeitraum 1.9.2015 - 31.8.2016