Fabian Lächler Optimization of Regionalized Precipitation with Radar and Rain Gauge Data for Trentino-Alto Adige
Total Page:16
File Type:pdf, Size:1020Kb
Fabian Lächler Optimization of Regionalized Precipitation with Radar and Rain Gauge Data for Trentino-Alto Adige MASTER’S THESIS Submitted to the Institute of Geography, University of Innsbruck In partial fulfilment of requirements for the degree Master Advisors: Univ.-Prof. Dr. Ulrich Strasser Ass. Prof. Dr. Thomas Marke September 2017 Abstract The accurate measurement of rainfall is an important prerequisite for various applications of meteorology, hydrology and their subsections. The reliable detection of precipitation fields is difficult, particularly in mountainous regions. To provide the best possible measurement, rain gauge and precipitation radar data are combined in state-of-the-art weather models. However, these state-of- the-art methods have only been extensively used for a short period of time. For the study area, which is mostly located in the central eastern Alps in the Italian federal state Trentino-Alto Adige, no radar data from Gantkofel mountain had been integrated between 2004 and 2009. Therefore the aim of this master thesis was to develop a state-of-the-art algorithm which calculates retrospective precipitation fields for the years from 2004 to 2009. The algorithm, which is strongly orientated towards the INCA model developed by ZAMG (Haiden et al., 2011), computes a weighted combination of rain gauge and radar data. It takes into account the reduced visibility of radar in mountainous areas. Furthermore the elevation dependence of precipitation is considered. Three evaluation methods were selected to determine the accuracy of the developed model. Firstly, image differencing between the self- developed and the INCA model without the Gantkofel radar was carried out. Secondly, a leave-one- out cross-validation was executed. And thirdly an image differencing between the self-developed algorithm and the state-of-the-art INCA, which includes the Gantkofel radar, was performed for 2014. The results showed that the present algorithm is capable in computing accurate precipitation fields with a significantly improved spatial resolution of 500 meters. However the computed precipitation values are inflated in regions with poor radar accessibility. A further integration of all available radar data would be an advantage to eliminate regions that are affected by topographic shielding. ii Acknowledgements First of all, I would like to thank my advisors Ulrich Strasser and Thomas Marke for supporting me and creating the idea for this master thesis. While working on my thesis, I learned a lot about programming and scientific working. I am also grateful to the whole AHC team for supporting me and providing me with helpful advice. Especially Marcel Siegmann helped me with dozens of code snippets and supported me whenever I needed it. Also Daniel Günther and Florian Hanzer always helped me find a solution to solve problems. Furthermore, I would like to thank Mauro Tollardo for the provision of the radar data and his expertise. In addition I would like to thank Erin Naismith and Dominik Scheuer for proofreading the thesis. I would also like to thank Carina Miggitsch, Francesca Pierri and Bettina Wildauer for being the best flatmates and motivating me all the time. Many thanks also to Nicky de Leeuw for treating my back pain during the work for this thesis. A special thank is dedicated to my parents and to my sister, for helping me in every possible way. They enabled me my studies and motivated me every time anew when there were problems. iii Table of Contents 1. Background of precipitation analysis ...................................................................................... 2 1.1 Precipitation measurement techniques.......................................................................... 2 1.1.1 Precipitation Radar ..................................................................................................... 3 1.1.2 Rain gauge .................................................................................................................. 9 1.1.3 Multi-sensor precipitation estimates ........................................................................ 10 1.2 INCA .............................................................................................................................. 11 2 Data and Methods ................................................................................................................ 13 2.1 Study area ..................................................................................................................... 13 2.2 Data .............................................................................................................................. 15 2.2.1 Radar data ................................................................................................................ 15 2.2.2 Rain gauge data ........................................................................................................ 16 2.3 Approach of combined precipitation Analysis (CPF) ..................................................... 17 2.3.1 Pre-Processing .......................................................................................................... 18 2.3.2 Main precipitation analysis ....................................................................................... 19 2.4 Evaluation methods ...................................................................................................... 23 2.4.1 Differences of CPF – INCA 2008 ................................................................................ 23 2.4.2 Leave-one-out cross-validation................................................................................. 24 2.4.3 Differences of CPF – INCA 2014 ................................................................................ 25 3 Results .................................................................................................................................. 26 3.1 Precipitation results ...................................................................................................... 26 3.2 Results of differences CPF – INCA in 2008 .................................................................... 27 3.3 Results of Leave-one-out cross-validation .................................................................... 29 3.4 Results of differences CPF – INCA in 2014 .................................................................... 34 4 Discussion ............................................................................................................................. 35 4.1 Discussion differences of CPF – INCA in 2008 ............................................................... 35 4.2 Discussion of leave-one-out cross-validation................................................................ 36 4.3 Discussion differences of CPF – INCA in 2014 ............................................................... 38 iv 5 Conclusion and Outlook........................................................................................................ 39 References .................................................................................................................................... 42 v List of Figures Figure 1: Marshall and Palmer drop-size distribution functions (dashed lines) compared with the results of Laws and Parsons (solid lines). (Rinehart, 2010 – based on Marshall and Palmer, 1948) ............. 5 Figure 2: Schematic sketch showing the effects of particle coalescence, melting, and changes in terminal velocity on radar reflectivity through the bright band. Zero height is the melting level. The radar reflectivity Ƞ is given along the x-axis (Rinehart, 2010). .......................................................... 7 Figure 3: Schema of CAPPI and CMAX. (Yoon et al., 2014) ..................................................................... 8 Figure 4: Terrain height with radar site in the study area. For better orientation the main sub regions of the central eastern Alps are labelled. ......................................................................................... 15 Figure 5 shows the increase of rain gauges from 2004 to 2009 ........................................................... 17 Figure 6: Monthly accumulated precipitation for January 2008. The maximum precipitation amounts at the Rieserferner group (north-east) are clearly visible. .............................................................. 26 Figure 7: Monthly accumulated precipitation for June 2008. The highest precipitation values in the Zillertal Alps and the Rieserferner group (north-east) are clearly visible. ....................................... 27 Figure 8: Difference image of CPF and INCA in January 2008. .............................................................. 28 Figure 9: Difference image between CPF and INCA in June 2008. ........................................................ 28 Figure 10: Distribution and altitude of all in the leave-one-out cross-validation used stations. .......... 29 Figure 11: Accumulated precipitation for February 2008. Station Hochserfaus was out of order in February. ......................................................................................................................................... 30 Figure 12: Accumulated precipitation for August 2008. The accumulated precipitation is much stronger than in February. ............................................................................................................................. 30 Figure 13: Station recorded and modelled