While the terrain inclination plays a major role in long-distance races for professional cyclists, this aspect is also important for leisure cyclists when planning a tour.Common smartphone applications for cyclists use elevation models to provide an estimate for the expected gradients during route-planning. In this regard, global models such as SRTM (Shuttle Radiometry Topography Mission) are used, since many of the applications aim for worldwide coverage.Compared to a digital elevation model (DEM) created by Airborne Laser Scanning (ALS), usually for areas with much smaller extent, such global models have much coarser resolution and limited accuracy, which in turn lead to problems to correctly estimate the accumulated gradient.Nowadays, many cyclists record their activities using global navigation satellite systems (GNSS) with their smart devices or fitness computers, which store the recorded data for the purpose of training documentation.This work proposes the utilisation of such crowdsensed mobile GNSS data for the modelling of digital street elevation to improve the planning of cycle routes. Since the accuracy of mobile GNSS varies in planimetry and elevation, it is important to critically inspect this data to ensure statistical certainty and integrity.Over 23000 crowdsensed GNSS trajectories, consisting of over 40 million point measurements, recorded in the span of 3 years between 2019 and 2021, were used as input data.We propose various methods to increase the homogeneity and quality of those crowdsensed data, focusing on the elimination of gross errors and the correction of systematic ones. The data were aggregated into a raster-based street elevation model, which combines both the potential for global availability and statistical robustness against outlier.Even if the computed model is limited in extent to a local study area within the city of Vienna, the methods were developed with global applicability in mind.Since the crowdsensed data shows in general a inhomogenous distribution across the study area, naturally focusing on streets and cycle paths, the certainty of the resulting model can only be guaranteed in equal measure.A comparison with reference data in terms of elevation and planimetry showed that our model is able to deliver significantly better road elevation than is possible when sampled from SRTM.With respect to the elevation reference, our model improves average elevation deviation by 22% (from 1.09m to -0.15m) and standard deviation by 36% (from 5.84m to 2.11m). Since the recorded data originate from cyclists, the model mainly focus on cycle paths and therefore covers about 25-30% of the reference road data in the whole study area, derived by OSM (Open Street Maps) street segments.While our model was primarily developed for the improvement of route planning in a mobile cycling application, it might be valuable for location-based services (LBS) or urban planning applications in general.