Spatial visualization of uncertainty
| dc.contributor.advisor | Munch, Zahn | en_ZA |
| dc.contributor.author | Christ, Sven | en_ZA |
| dc.contributor.other | Stellenbosch University. Faculty of Arts and Social Sciences. Dept. of Geography and Environmental Studies. | en_ZA |
| dc.date.accessioned | 2017-02-14T13:05:19Z | |
| dc.date.accessioned | 2017-03-29T11:50:00Z | |
| dc.date.available | 2017-02-14T13:05:19Z | |
| dc.date.available | 2017-03-29T11:50:00Z | |
| dc.date.issued | 2017-03 | |
| dc.description | Thesis (MA)--Stellenbosch University, 2017. | en_ZA |
| dc.description.abstract | ENGLISH ABSTRACT: Geospatial information has become more accessible since the early 2000s. Uncertainty has remained a constant in data, due to various factors, including scale and real world conceptualization. Geospatial products are frequently used to inform decision makers on key decisions, with little understanding of the quality of the data. However, accuracy assessments have improved significantly since the visual screening that was used in the 1950s, now providing statistics such as the Kappa coefficient, root mean square error (RMSE) and the confusion matrix. Two questions thus arise: 1) do those using the data inform themselves about the quality of data; and 2) can visualization of the uncertainty in spatial data aid in the communication of the data quality? This research was achieved in three tasks: 1) evaluate the South African perception on data quality; 2) develop an uncertainty visualization tool; 3) evaluate the uncertainty visualization tool. The first task was achieved through a quantitative survey of people working in the South African geospatial industry. Despite a limited response, the findings indicated that those working with geospatial data do not always seek to verify the quality of the data they are using. It also came to light that most of those who do not verify the quality of their data, would like to have the uncertainty in the data visualized. Task 2 aimed at developing a tool for the visualization of spatial uncertainty (Uview). Uview was based on the findings from Task 1 supplemented by recommendations from literature and other uncertainty visualization tools. The tool was developed for continuous raster datasets only and uses the z-score and modified z-score as its main statistics for visualization. Standard accuracy assessment statistics (global data quality statistics), such as RMSE and mean absolute error (MAE) have also been included in Uview to make it an accuracy assessment and uncertainty visualization tool for continuous raster data. Lastly Task 3, the evaluation of Uview was done using a two-pronged approach. The first part encompassed investigating the usability of the tool. In this phase the visualizations were used to derive relationships between digital elevation models (DEM), uncertainty and a watershed product. It was found that Uview does provide useful information, and watersheds are sensitive to deviations from true value at key locations more than the magnitude of the deviation. When Uview was evaluated by twelve people in the geospatial industry they all agreed that though improvements can be made, as it presents itself currently it is already a useable product that can add value. All respondents agreed that the visualization improves the comprehension of the statistics, and so of uncertainty. | en_ZA |
| dc.description.abstract | AFRIKAANS OPSOMMING: Ruimtelike inligting het sedert die vroeë 2000's meer toeganklik geword. Onsekerheid het 'n konstante in data gebly as gevolg van verskeie faktore, insluitend skaal en werklike wêreld konseptualisering. Ruimtelike produkte word dikwels gebruik om besluitnemers in te lig oor belangrike besluite, met min begrip van die kwaliteit van die data. Tog het akkuraatheid assessering aansienlik verbeter sedert die visuele metodes wat in die 1950's gebruik is, ook met die verskaffing van statistiek soos die Kappa-koëffisiënt, wortel-gemiddelde-kwadraat fout (RMSE) en die verwarringsmatriks. Twee vrae ontstaan dus: 1) neem die gebruikers van die data die tyd om hulself te vergewis met die kwaliteit van data; en 2) kan visualisering van die onsekerheid in ruimtelike data die kommunikasie van die data kwaliteit ondersteun? Hierdie navorsing is behaal in drie take: 1) evalueer die Suid-Afrikaanse persepsie oor data kwaliteit; 2) ontwikkel 'n onsekerheid visualisering hulpmiddel; 3) evalueer die onsekerheid visualisering hulpmiddel. Die eerste taak is behaal deur 'n kwantitatiewe opname van mense wat betrokke is in die ruimtelike inligtingsbedryf in Suid-Afrika. Ten spyte van 'n beperkte reaksie, het die bevindinge aangedui dat diegene wat met ruimtelike data omgaan nie altyd daarna streef om die data gehalte te verifieer nie. Dit het ook aan die lig gekom dat die meeste van diegene wat nie hul data gehalte verifieer nie, wel belangstel in ‘n onsekerheid visualisering van die data. Taak 2 was gemik op die ontwikkeling van 'n instrument vir die visualisering van ruimtelike onsekerheid (Uview). Uview is gebaseer op die bevindinge van Taak 1 aangevul deur aanbevelings vanuit die literatuur en ander onsekerheid visualisering hulpmiddels. Die instrument is ontwikkel vir deurlopende roosterdatastelle en maak gebruik van die z-telling en gemodifiseerde z-telling as belangrikste statistieke vir visualisering. Standaard akkuraatheid assessering statistieke (globale data kwaliteit statistieke), soos RMSE en gemiddelde absolute fout (MAE) is ook ingesluit in Uview om dit 'n akkuraatheid assessering en onsekerheidsvisualisering hulpmiddel vir deurlopende roosterdata te maak. Laastens die evaluering van Uview (Task 3) is gedoen met behulp van 'n tweeledige benadering. Die eerste deel het ondersoek ingestel na die bruikbaarheid van die instrument. In hierdie fase is die visualiserings gebruik om verhoudings tussen digitale elevasie modelle (DEM), onsekerheid en 'n waterskeiding produk af te lei. Daar is bevind dat Uview nuttige inligting verskaf, en waterskeidings is meer sensitief vir afwykings van werklike waardes op belangrike plekke meer as die grootte van die afwyking. Tydens die Uview evaluering deur twaalf mense vanuit die ruimtelike inligtingsbedryf, het almal saamgestem dat hoewel verbeteringe gemaak kan word, die produk soos dit tans daar uitsien alreeds 'n bruikbare produk is wat waarde kan toevoeg. Al die respondente het saamgestem dat die visualisering die begrip van die statistieke verbeter, en so ook van onsekerheid. | afr_ZA |
| dc.format.extent | xiv, 160 pages, illustrations | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/10019.1/100949 | |
| dc.language.iso | en_ZA | en_ZA |
| dc.publisher | Stellenbosch : Stellenbosch University | en_ZA |
| dc.rights.holder | Stellenbosch University | en_ZA |
| dc.subject | Uncertainty visualization | en_ZA |
| dc.subject | Geographic Information Systems | en_ZA |
| dc.subject | Data quality | en_ZA |
| dc.subject | Spatial analysis (Statistics) | en_ZA |
| dc.subject | Altman Z-score | en_ZA |
| dc.subject | UCTD | |
| dc.title | Spatial visualization of uncertainty | en_ZA |
| dc.type | Thesis | en_ZA |