Positional Accuracy of Online Geocoding Services: Case Study of Bhaktapur District

Abstract

Data is food for Information Systems and location data is basis for GIS services. Geocoding services provide this data by converting Street Addresses like NCIT, Balkumari to corresponding geographic coordinates. These coordinates are then used in data processing to deliver services. Many services (especially location based), FourSquare, Uber, depend on these services for operation. Nepalese market is also increasingly using these services like Tootle, sarathi cab. Till now nobody knows how accurate the result of such services are. There exist multiple places with same name. Some names are not actual but referential. Accuracy of the services depend on the underlying database, the method used, the actual geographic location and also the actual query. Different methods yield different results. The assessment of accuracy and suitability of the geocoding services has not been conducted, yet they are being used extensively. The objective of this study is to compare the positional difference between two Geocoding methods, OpenStreetMap (OSM) Nominatim Service and Google Geocoding Services and compare them with standard government datasets to measure their discrepancy. For reference, settlement data from NGIID was used. Addresses were first geocoded to street level and positional difference in the results were calculated using havensire formula. The discrepancies were categorized into intervals of 100m. Out of 267 location points, 118 result were found in Nominatim, whereas only 86 were found in Google. Average discrepancy for Nominatim result was 175m and for Google it was 1810m. Comparisons show minimal difference in median and minimum values, while there were larger differences in the maximum value. Nominatim delivered comparatively accurate results and found more addresses than google. Google on the other hand gave huge mismatches for some cases. The study found out that the databases are missing in both cases as shown by the no of “not found" cases and that the results from Nominatim are more reliable than that of Google because of its hierarchical matching system and user friendly interface.