Conference Agenda

Session
A39: Poster Session (2/2)
Time:
Wednesday, 08/July/2020:
5:00pm - 5:45pm

Session Chair: Martin Sudmanns
Session Chair: Hermann Klug
Location: A - Sessions
short, 2-minute pitch presentations by poster authors

Presentations

Linked Open Geodata in Action: Beispiel COVID-19

Florian Thiery1, Timo Homburg2,1

1Research Squirrel Engineers, Deutschland; 2Hochschule Mainz, Deutschland

Das Poster zeigt die Anwendung von LinkedOpenData in der GIS Community mit QGIS am aktuellen Anwendungsbeispiel COVID-19 und ist in zwei Teile aufgeteilt.

Der erste Teil skizziert, wie man mit Konvertierungstools Geodaten nach RDF überführen und zu Geokonzepten interlinken kann - den sogenannten Semantic Upflit. Diese Geodaten können anschließend als Linked Data in einem Triple Store zur Verfügung gestellt werden. Im Kontext der COVID-19 Pandemie sind hier insbesondere Ländergrenzen, sowie innerdeutsche Grenzen wie Bundesländer und Landkreise von Bedeutung. Diese Geodaten wurden mit Hilfe der Schnittstellen der Johns Hopkins University und des Robert Koch-Instituts mit COVID-19 Daten angereichert. Die integrierten Daten können jedoch noch mit weiteren Linked Data Datenquellen angereichert werden. Hierfür zeigen wir ein Beispiel. Hier bietet sich z.B. Wikidata an, wofür wir Beispiele (Einwohnerzahl, etc.) zeigen.

Der zweite teil zeigt zwei Methoden wie diese Geodaten in der Geocommunity verarbeitet werden können. Der erste Weg ist das SPARQLing Unicorn QGIS Plugin welches es ermöglicht ein SPARQL Query direkt an einen Triple Store abzuschicken und die Ergebnisse als Layer in QGIS zu verarbeiten. Das SPARQLing Unicorn QGIS Plugin kann hierbei jedoch auch als Vorbereitung für eine zweite Methode dienen: den SemanticWFS.

Der SemanticWFS ist ein Web Feature Service, welcher Triple Stores als Backends zulässt. Benutzer können mit dem SPARQLing Unicorn QGIS Plugin ein Query an den Triple Store testen und im SemanticWFS das Query als FeatureType konfigurieren. Anschließend kann das Ergebnis des SPARQL Queries in einem der vom SemanticWFS angebotenen Formate zurückgegeben und zum Beispiel in QGIS als WFS eingebunden werden. Dieser Weg ist für Benutzer gedacht, welche über keine Kenntnisse in der SPARQL Query Sprache verfügen und bietet die Möglichkeit über Linked Data angereicherte Daten zielgerichtet als Geowebservice zur Verfügung zu stellen.

Wir stellen somit zwei neue Wege vor, Linked Data in einem GIS zu nutzen und motivieren die Integration und Nutzung von Linked Data an einem konkreten Beispiel.



Automatic Classification of Lineaments. Is there a Best Practice?

Alexandra Jarna1,2, Ola Fredin1,2, Jan Ketil Rod1

1NTNU, Norwegen; 2Geological Survey of Norway

Geological lineaments are a group of linear or curve-linear features on the Earth’s surface which reflects geological structures such as rock fractures, lineaments faults and bedrock contacts, and can easily be recognized in high resolution different DEM derivatives (e.g. slope, hillshade, northness, eastness. Lineaments are important features for our understanding of the geological history, with important societal implications such as rock avalanche and earthquake hazard prediction. In addition, rock fractures and lineaments are important ground water conduits and provide valuable groundwater resources due to increased water permeability in the fractures. The literature concerning (semi-)automatic mapping of lineaments is limited and has been criticized for including misclassifications (false positives) such as roads or agriculture boundaries. However, this can now be minimized by using non-optical data such as potential field geophysical data. Manual mapping using high-quality data (e.g. LiDAR, Panchromatic Landsat ETM+) is common; however, the process is time consuming and subjective. The main goal with this project is to develop a lineament detection algorithm to increase objectivity and repeatability of the mapped products, where experienced structural geologists can verify the classification based on field work and previous mapping.

Recent papers have shown that GEOBIA methods are more promising than pixel-based methods. GEOBIA requires following steps to be able to give a reasonable result: image segmentation, attribution, classification itself and the ability to connect individual objects-segments. In order to achieve this requires developing methods or rule-sets that can repeat human interpretation of datasets in automated or semi-automated ways and lead to the production of classified datasets while reducing the subjectivity, work, time consuming and costs. The rule set parameters are selected based on visual assessment or prior knowledge. Increasingly, machine learning (ML) is also used on image objects by introducing the training samples and classifiers with the use of algorithms (Naive Bayes (NB), k-Nearest Neighbors (kNN), Random Forests (RF), Support Vector Machines (SVM), decision trees (DT), neuro-fuzzy (NF), genetic algorithm (GA)), with variables to learn from machine readable data. Objects are then classified based on their statistical resemblance to the training data.

Given all these tools, this paper discusses the best methods to automatically map bedrock lineaments.



Flood Susceptibility Mapping using per-pixel and object-based approaches for Salzach Basin, Austria

Thimmaiah Gudiyangada Nachappa, Stefan Kienberger, Sansar Raj Meena

Department of Geoinformatics - Z_GIS, University of Salzburg, Austria

Water is an imperative component for nourishing life and has become a significant focus in recent years in terms of inadequacy, disaster and famine. In recent years, the frequency of natural hazards have significantly increased mainly due to deforestation, increase in population and human-made growth. Among the natural hazards, floods are recurring and costliest natural disasters in terms of human lives and financial losses. Flooding is a frequent, severe, widespread and increasing natural hazard. Universally, the incidence of flooding has intensified by 40% in the past couple of decades. Flooding can have a severe impact on the socioeconomic condition of the region where flooding occurs. The object-based image analysis (OBIA) has advanced as a key part of GIScience devoted to the assessment and analysis of satellite imagery. OBIA constructs on image analysis, which is similar to image segmentation and image classification. The image segmentation process is a crucial pre-requisite for classification and further analysis in the so-called geon approach. Geons are spatial units that are homogenous in terms of varying space and time phenomena under policy concern. This geons model incorporates expert knowledge and semi-automatically delineates regions. Austria is at significant risk from natural hazards owing to the location in the alpine range and its climatic environments. Flood susceptibility is a significant step in identifying and mitigating floods in the future and improve flood management measures. The main objective of this study was to combine the pixel-based frequency ratio (FR) approach for weighting with the object-based geons approach for aggregation in the Salzach river basin, Austria. We used ten flood conditioning factors based on the study area like elevation, slope, aspect, NDVI, TWI, SPI, geology, land cover, distance to drainage and distance to roads. The flood inventory data was derived from HORA, and 1861 flood locations were used for the study where 70% of the flood locations were used for training, and 30% was used for validation. The Receiver Operating Characteristic (ROC) curve was drawn, and the Area Under the Curve (AUC) was calculated to obtain the accuracy of the flood susceptibility maps derived from FR and Geons approach. The AUC results indicated that the combined pixel-based and object-based geons model (0.93) yielded better accuracy than the simple pixel-based FR model (0.88). This research introduces the use of geons as an object-based aggregation approach for flood susceptibility mapping and illustrates that geons can be applied for obtaining meaningful susceptible regions. The main advantage of using object-based geons was to have susceptible areas or entities that are independent of any administrative units. These results can be used for the planning and management of areas vulnerable to floods in order to prevent flood-induced damage, and the results may be useful for natural disaster assessment.



3D Geovisualization of Historic and Contemporary Lead Sediment Contamination in Lake Erie

K. Wayne Forsythe1, Danielle E. Mitchell1, Chris H. Marvin2, Debbie A. Burniston2, Stephen J. Swales3, Michael W. MacDonald1

1Ryerson University, Kanada; 2Environment and Climate Change Canada; 3University of Toronto, Kanada

Lead sediment contamination in Lake Erie stems from a long history of natural and synthetic resource production. Historic and contemporary sediment samples were collected by the Canada Centre for Inland Waters in 1971, 1997/1998, and 2014. Each survey is composed of different sampling densities due to the rising cost of sampling procedures and resources. To draw comparisons between sediment surveys over time and space, the kriging interpolation method was used to create continuous data surfaces of lead sediment contamination for each survey. Change detection analyses identified an overall decreasing trend in high levels of lead sediment contamination from 1971 (36.72%) to 1997/1998 (0%) to 2014 (0.71%). Moderate contamination levels remained relatively consistent in surface area coverage between the three survey periods (47.14% in 1971, 51.18% in 1997/1998, and 50.91% in 2014). Sediments with the lowest concentrations of lead contamination increased in surface area throughout Lake Erie coincident with the banning of leaded gasoline in Canada and the United States in the 1990s (12.98% in 1971, 43.85% in 1997/1998, and 38.62% in 2014). Overall lake pollution levels changed from a basin-wide concern of high lead sediment contamination to a localized situation near Cleveland, Ohio. Lake-wide circulation patterns and bathymetric data were added to interpolated contamination surfaces to enhance the understanding of interrelated hydrodynamic processes and geophysical features in the movement of contaminated sediments. Using advanced visualization tools in Esri’s ArcScene, bathymetric data were employed to enhance the geographic context of contamination maps. Highly contaminated sediments appear to follow circulation patterns from the mouth of the Detroit River along the Ohio shoreline, where currents are strongest throughout the lake. The physical barriers to sediment transportation created by the Pelee-Lorraine Sill and the Long Point-Erie Sill can be visualized in three-dimensions. These elevated features between lake basins are easily recognized as impedances to lake currents when circulation directions are draped over the bathymetric model. By using illumination tools and techniques, geovisualizations of lead sediment contamination throughout Lake Erie create a scientific communication tool for a wide audience to use in multiple-criteria decision making for environmental remediation of sediment contamination.



GIS gestützte Analyse zur Unterstützung der Organisation und Planung an Pflichtschulen in Kärnten

Melissa Tischhart, Christina Unterwandling, Sebastian Ertl

FH Kaernten, Österreich

Die Bildungsdirektion in Kärnten, hat es zur Aufgabe, Analysen im Bereich des Schulwesens durchzuführen, die unterschiedliche planerische und organisatorische Tätigkeiten im Bereich der Pflichtschulen abdecken. Um dies zu tun, setzt die Bildungsdirektion ein geographisches Informationssystem (GIS) ein, da viele der dafür verwendeten Daten räumlicher Natur sind (Longley et al., 2015). Einige dieser GIS-gestützten Analysen wurden im Rahmen eines stu-dentischen BSc Projektes durchgeführt und sind in diesem Beitrag dokumentiert.

Literaturrecherche zur GIS-gestützten Planung für Bildung (siehe z.B., Al-Rasheed und El-Gamily, 2013; Mendelsohn, 1996) sowie Gespräche mit der Bildungsdirektion wurden durch-geführt, um benutzerspezifische Anforderungen herauszuarbeiten. Daraus entstand ein An-forderungskatalog. Mit Hilfe von einem Entitiy-Relationship-Modell (ERM) wurden die notwen-digen Daten modelliert und in einem ER Diagramm visualisiert. Das ERM/D dient als Spezifi-kation zur Erstellung einer Geo¬datenbank. Die darauf aufbauende GIS Applikation wurde in ArcGIS 10.6 Umgebung aufgesetzt und Analyse¬methoden identifiziert, welche folgende Auf-gabenbeispiele lösen:

(1) Darstellung der aktuellen Schülerzahlen 2018/19 bzw. der Schülerzahlen im Schul-jahr 2025/26 im Bereich der Volksschulen: Wie wird sich die Schülerzahl an den Volks-schulen in den einzelnen Gemeinden entwickeln? Wo gibt es Volksschulstandorte mit Schülerzahlen <= 30 Schüler; <= 40 Schüler. Wie sieht die durchschnittliche Schüler-zahl pro Volksschulstandort aus?

(2) Analyse und Visualisierung der Altersstruktur der Pflichtschullehrer an Volksschu-len, Neuen Mittelschulen und Polytechnischen Schulen gesamt, pro Schulstandort und in den einzelnen Gemeinden/Bezirken.

(3) Auswirkungen der Geburtenrate auf Schulstandorte: Ist eine mögliche Optimierung von Schulstandorten aufgrund der Geburtenrate denkbar?

Neben dem Konzeptentwurf inkludiert der Lösungsansatz Identifikation jener (nicht-) räumli-chen Analysemethoden, die sich für die o.g. Aufgaben bestens eignen, wie z.B. Statistik, räumliche Statistik oder Trend Analyse (de Smith et al., 2020). Geoverarbeitungsmodelle sind für einzelne Aufgaben erstellt (ModelBuilder), um zugehörige Workflows zu automatisieren und für weitere Nutzer in der Bildungsdirektion verfügbar zu machen (ESRI, 2020). Ergebnis-se sind in Form von thematischen Karten und statistischen Graphen und nach den Vorgaben der Bildungsdirektion geovisualisiert. Sie bieten der Bildungsdirektion Kärnten zusätzliche Informationen, um Optimierungen im Bereich der Pflichtschulen durchzuführen bzw. bilden die Ergebnisse eine Grundlage für eine positive Entwicklung des Schulwesens in Kärnten.

Literaturverzeichnis

Al-Rasheed, K.; El-Gamily, H. I. (2013): GIS as an Efficient Tool to Manage Educational Ser-vices and Infrastructure in Kuwait. In: Journal of Geographic Information Systems (5), S. 75–86.

deSmith, M. J., Goodchild, M. F., & Longley, P. (2020). Geospatial Analysis: A comprehensive guide to Principles, Techniques and Software Tools. Sixth edition. Retrieved from https://www.spatialanalysisonline.com/HTML/index.html

ESRI. (2020). Was ist Geoverarbeitung? [online] https://desktop.arcgis.com/de/arcmap/latest/analyze/main/what-is-geoprocessing.htm Zugriff am 09.05.2020.

Longley, P. A.; Goodchild, M. F.; Maguire, D. J.; Rhind, D. W. (2015): Geographic Information Science and Systems. 4. Aufl. Hoboken: John Wiley & Sons.

Mendelsohn, J. M. (1996): Education planning and management and the use of geographical information systems. Paris: Imprimerie Gauthier-Villars.



Searching for Vienna’s Boundary A

Kira Lappé

University of Vienna, Österreich

Anthropogenic, i.e. human-made sediments as results of the human impact on the Earth’s surface are globally widespread and can be detected almost everywhere – yet, in urban environments, these deposits of anthropogenic origin are especially massive. In Vienna, a city with more than 2,000 years of history, almost 10,000 well core drillings and 70 excavation sites help to throw light on the underground of the city centre. Aim of my UNIGIS MSc Thesis, embedded in the interdisciplinary research project „The Anthropocene Surge“ funded by the Vienna Science and Technology Fund (WWTF ESR 17-040), was the creation of an interpolation of the Boundary A, the boundary between the “natural ground”, untouched by humans, and the anthropogenic deposits. This interpolation was used to produce figures on the current thickness as well as volume of the anthropogenic sediments. By this, the human impact on the underground of a modern, but long-lived city can be visualised and quantified.

A study area of 12.25 m², covering the Inner City and parts of the neighbouring districts, was selected because of its high density of data available and historical importance, being the centre of settlement from Prehistory to Modern times. Statistical analysis of well core logs and archaeological data made figures on the thickness of anthropogenic sediments in the study area available for the first time. Using Universal Kriging, an interpolation of the lower boundary of anthropogenic sediments was achieved, allowing to map the thickness of artificially modified ground (AMG) and to calculate the volume of anthropogenic deposits. The distribution of AMG in the study area is directly reflecting historical and natural reasons relevant to the study area. Comparison with the results of a study conducted on two districts of London showed similar, but somewhat higher accumulations of anthropogenic deposits in Vienna.



SemantiX - A cross-sensor semantic Earth observation data cube to open and leverage essential climate variables with scientists and the public

Martin Sudmanns1, Stefan Wunderle2, Steffen Reichel3, Philipp Hummer4, Andrea Baraldi5, Hannah Augustin1, Luuk van der Meer1, Dirk Tiede1,6

1University of Salzburg, Department of Geoinformatics - Z_GIS, Austria; 2University of Bern, Institute of Geography, Switzerland; 3Spatial Services, Austria; 4Spotteron, Austria; 5Italian Space Agency, Italy; 6UNIGIS Lehrender

Long time series of essential climate variables (ECVs) derived from satellite data are an important contribution to climate research. They are a critical, independent source of information for comparison with climate model results, but they can also be used to directly detect and monitor changes in our environment. The longest European time series (1981-now) of Advanced Very-High-Resolution Radiometer (AVHRR) data is compiled, archived and processed using physical methods and algorithms (e.g., lake surface temperature, snow cover, vegetation dynamics) by the remote Sensing Research Group at University of Bern’s Institute of Geography. Until now, AVHRR data have only been accessible via sequential access, requiring a significant time investment and expert knowledge to find relevant data for analysis. Interested scientists from other disciplines or the broader public currently do not have any access to these methods or data.

SemantiX is a new project (project duration 2020-2022 funded by the Austrian Research Promotion Agency (FFG)) with the goal to complement and expand the existing AVHRR time series using Copernicus Sentinel-3 A/B data and make them accessible using a semantic Earth observation (EO) data cube. A semantic EO data cube can facilitate easier data access for scientists and can be the backend of a smartphone application for providing visualisation targeted to non-expert users. The proposed data cube of AVHRR and Sentinel-3 data as well as derived ECVs will be linked to a mobile citizen science smartphone application giving the public a new, direct and interactive access point to EO data and derived information products. SemantiX is coordinated by the Interfaculty Department of Geoinformatics – Z_GIS at the University of Salzburg with project partners Institute of Geography of the University of Bern, Spatial Services, and Spotteron.

This contribution presents the project idea to continuing climate-relevant AVHRR time series of EO data with Sentinel-3 data for the Austrian-Swiss alpine region, a European region that is currently experiencing serious changes due to climate change that will continue to create challenges well into the future. For the first time, various target groups will have simplified access to these EO data and derived information, including ECVs. Scientists from disciplines unrelated to remote sensing, students (i.e., the next generation of scientists) as well as interested members of the public will have direct access to long EO data time series for a variety of applications and location based access through the mobile citizen science application.



Nachhaltige Siedlungsentwicklung im Rhein-Erft-Kreis: Entwicklung eines Landnutzungsmodell mit QGIS, Postgis und Python auf der Basis von Vektordaten.

Mirko Blinn1, Dr. Sven Lautenbach2, Prof. Dr.-Ing. Theo Kötter1

1Universität Bonn, Deutschland; 2Universität Heidelberg, Deutschland

Das Wachstum der Städte verursacht weltweit vielfältige Probleme. Dazu zählt neben verschiedenen negativen Folgen für Umwelt und Natur, insbesondere der Verlust an landwirtschaftlicher Produktionsfläche. Im Rahmen des vom BMBF geförderten Projektes „Nachwuchs“ wird versucht, durch die Entwicklung von neuen und innovativen Raum- und Siedlungsbildern am Beispiel der sogenannten S.U.N.-Region (Stadt Umland Netzwerk) westlich von Köln diesen Effekten entgegenzuwirken und so für eine nachhaltigere Regionalentwicklung zu sorgen. Ein wichtiger Teilaspekt des Projektes ist die Entwicklung eines Landnutzungsmodells zur Vorhersage möglicher zukünftiger Siedlungsentwicklungen.

Im Rahmen der Modellentwicklung zeigte sich, dass schon vorhandene Landnutzungsmodelle wie beispielweise CLUMondo aufgrund von technischen Restriktionen und einer ursprünglich anderen inhaltlichen Ausrichtung nicht optimal für die Zwecke des Projektes geeignet sind.

Durch den Fortschritt im Bereich des maschinellen Lernens und der inzwischen hohen Verbreitung von Python im wissenschaftlichen Bereich ist es möglich, Landnutzungsmodelle auch auf Basis von Vektordaten zu entwickeln und direkt in QGIS ein zubinden.

Durch die Einbindung in QGIS wird die praktische Anwendung des Modells wesentlich erleichtert sowie die direkte Verwendung der Ergebnisse, beispielweise bei der Vorbereitung von Verfahren zur Neuaufstellung und Fortschreibung von Regionalplänen, möglich.

Zur Verwaltung der benötigten Ausgangsdaten und Ergebnisse bietet sich eine PostGIS- Datenbank gerade zu an.

Dank der zunehmenden Verfügbarkeit freier Geodaten steht je nach simulierter Region ein fast unerschöpflicher Fundus an verwertbaren Daten zur Verfügung. Für die Projektregion westlich von Köln sind dies beispielsweise:

• Alle Daten von Open Geodata NRW

• Die CORINE Land COVER Daten

• Daten aus OpenStreetMap

• Ergebnisse von Isochronenanalysen beispielsweise mit OpenRouteService

Durch die Verwendung ausschließlich freier Software und der hohen Variabilität möglicher Eingangsdaten ist später eine Nutzung des Modells durch dritte oder innerhalb anderer Projekte leicht möglich. Es ist angedacht das Modell und seine Komponenten mittelfristig unter einer freien Lizenz Dritten zur Verfügung zu stellen.

Im Rahmen des Posters wird das dem Modell zugrunde liegende Konzept vorgestellt sowie erste Simulationsergebnisse gezeigt.



Phenological objects: Towards object-based analysis

Hugo Bendini, Anderson Soares, Thales Körting, Leila Fonseca

National Institute for Space Research (INPE), Brasilien

Geographic Object-Based Image Analysis was defined as a sub-discipline of Geographic Information Science that is devoted to developing automated methods to partition remote sensing imagery into meaningful image-objects. This concept represented a paradigm shift on remote sensing, from pixel to objects, and it was successfully employed in different applications. With the launching of new satellites and advances on computer sciences, a new paradigm came up, concerning to satellite image time series (SITS). It allowed to assess dynamics on land surface and many advances were achieved in terms of change detection and multitemporal classification. However, SITS approaches are pixel-based, and despite the capacity to perform precise maps, the quality can be improved with factors that transcend pixel information through the combination of spatial-temporal heterogeneity and autocorrelation. In this context, we introduce Phenobia, a Phenology Object-Based Image Analysis approach and its phenological object. This object can be related to phenoregions, a concept that was already proposed, which are regions where a phenological cycle is homogeneous. However, it has a specific focus and application domain concerning to forest monitoring. Here we present a more generalist model where the phenological object is defined by a domain, which may be native vegetation in a biome, with different physiognomies, or also different agricultural practices or crops. It is also characterized by a period, that defines its phenological variation, that is given by a vegetation index. During this period phenology can be similar between different groups and different domains. This variation is expressed by a set of spectral-temporal attributes that are calculated over a statistic obtained by an aggregation function that encompasses all pixels that have been spatially grouped by a spatial-temporal segmentation. This set of attributes concerns the characteristics of phenology, such as the beginning of green-up. Therefore, they can be used to distinguish semantic groups, as well to derive information on vegetation dynamics or agricultural monitoring. We believe that this concept can be applied for a precise delimitation of targets, which can be fragments of natural vegetation, with applications for example on biodiversity modelling or, in the case of agriculture and pasture, for delimitation of agricultural plots or cattle pickets, to individualize these objects in order to allow object management systems. There's still the need to take into account the potential land change that can occur during the studied period of the event. We believe this can be done with change detection techniques. We proceed a preliminary study case as a proof of concept where we reached interesting results. Our future works consist on assess different spatial-temporal segmentation and change detection algorithms and testing the approach on other domains and more study areas to validate the concept.



Comparison of cycling trip characteristics among various sports tracking apps

Angela Schirck-Matthews1, Hartwig H. Hochmair1, Gernot Paulus2, Dariia Strelnikova2

1University of Florida, Davie, Florida (USA); 2Carinthia University of Applied Sciences, Villach, Carinthia (Austria)

Data from fitness tracker apps have become a prominent source for studying cycling behavior. This type of crowd-sourced data provides larger datasets than were previously attainable by travel surveys and cyclist counts, which allows for the comparison of trip characteristics between geographic regions and the study of temporal trends in bicycle ridership. Researchers acknowledge that different types of biases come with GPS tracking data from fitness apps, such as self-selection bias. Also, the target audience varies among fitness tracker apps, which begs the question as to whether the behavior of cyclists varies among the apps, leading to additional biases. To provide a first insight into this question, this research analyzes trips reported on three fitness tracker apps, Bikemap, Endomondo, and MapMyFitness, for South Florida (Miami-Dade, Broward, and Palm Beach counties) and North Holland. Comparison of trip characteristics is made among the three apps and across both study regions. Results show that cycling behavior observed in the three apps is similar relative to a set of control trips in each region (e.g. fewer primary roads than reference trips observed), but that there are some pronounced differences in trips recorded with the different apps between both regions. For example, Bikemap trips were significantly longer than Endomondo trips in North Holland, whereas the opposite was true in South Florida. This suggests that geographic region plays a role in how trip characteristics recorded on different apps compare to each other, demonstrating the presence of an additional aspect of geographic bias in crowd-sourced cycling data.