Spatio-temporal analysis of extreme wind velocities for infrastructure design. Dissertation submitted in partial fulfillment of the requirementsfor the Degree of Master of Science in Geospatial Technologies.
Alexys Herleym Rodríguez Avellaneda [email protected] https://github.com/alexyshr
Supervised by: Prof. Dr. Edzer Pebesma Institute for Geoinformatics University of Münster - Germany
Co-supervised by: Prof. Dr. Juan C. Reyes Department of Civil and Environmental Engineering Universidad de los Andes - Colombia
Co-supervised by: Prof. Dr. Sara Ribero Information Management School Universidade Nova de Lisboa - Portugal
Abstract
This research aims to create non-hurricane non-tornadic maps of extreme wind speeds for the mean recurrence intervals MRIs 700, 1700, and 3000 years, covering the Colombian territory. For infrastructure design, these maps are combined with existing hurricane wind speed studies, to be used as input loads due to wind.
For each station with non-thunderstorm wind speeds time histories in the input data, following (Pintar, Simiu, Lombardo, & Levitan, 2015), extreme wind speeds corresponding to each MRI are calculated using a Peaks Over Threshold Poisson Process POT-PP extreme value model, then wind velocities with the same MRI are spatially interpolated to generate continuous maps for the whole study area. The annual exceedance probability for all velocity values in 700, 1700 and 3000 years MRIs output maps are respectively 1/700, 1/1700 and 1/3000.
Regarding input data, not only time series of field measurements from IDEAM methodological stations are used, but also post-processed information coming from the Integrated Surface Database ISD, and ERA5 forecast reanalysis data. This condition demanded a comparison of the different data sources, in order to verify the feasibility in the use of ISD and ERA5, this is downscaling support. The result of the comparison showed little similarity between the different sources, but taking into account that complete and adequate measured data from IDEAM was not available. Before to apply POT-PP, ISD and IDEAM data sources were standardized to meet the requirement of three seconds (3-s) wind gust speed, ten (10) meters anemometer height, and terrain open space condition.
Due to the limitation in the classification of thunderstorm and non-thunderstorm data, it was not possible to take real advantage of POT-PP method, which was limited/restricted from non-homogeneous to homogeneous and from non-stationary to stationary, being equivalent to use the most common POT - generalized Pareto approach. Non-hurricane maps were created for data sources ISD and ERA5, using Kriging as spatial interpolation method. After the integration with previous hurricane studies, the results of ERA5 showed the most reliable final maps, despite limitations in the input data to guarantee downscaling support. ISD final map showed very high wind values, which are unlikely. These shortcomings may be corrected when complete IDEAM data-source and storm data classification are available.
A complete R tool was implemented to solve the whole process, which is based on copyrighted code for de-clustering and thresholding generously given by Dr Adam L. Pintar [email protected] - National Institute of Standards and Technology NIST, U.S Department of Commerce.