Accuracy assessment of Web Map Services using UAV based orthophotos

High-resolution geospatial data is expensive and not easily accessible, and therefore, using a reliable base map for thematic overlay is considered as one of the best practices in GIS. In this regard, the Web Map Service (WMS) has been a game-changer for use by academics and industry, providing a collection of compiled or pre-assembled map imageries containing both vector and raster layers. These are available as third-party plug-ins or readily available layers in GIS software or dynamic web-based platforms. The available raster base maps, which are a collection of aerial imageries provide a high resolution 2D rendition of geographical area serving as a useful background reference dataset. Due to the datasets being a mash-up of different sources, the compiled images may suffer variability in temporal and spatial scale. The information regarding the processing of the data sets and the warranty regarding the accuracy of these datasets is also seldom available, and therefore, the accuracy of these compiled datasets is an open question. The study is aimed at filling the gap in the existing literature regarding the assessment of the accuracy of two most widely used open-source base maps, namely Google and Bing aerial imageries. For this purpose, we have used orthophotos generated via photogrammetric methods acquired through Unmanned Aerial Vehicles (UAVs). The orthophotos were generated over varied topographic regions in the Indian foreland, comprising plateau, hills, and alluvial plains. For further validation, more than 350 additional ground control points (GCPs) were also collected using Real Time Kinematics (RTK) in these areas. Two separate assessments have been made to check the accuracy of the corresponding WMS layers, which varied between 1.6 meters to 4.2 meters in Google and 2.2-5.2 meters for Bing images. We have analyzed the accuracy of the GCPs derived from the RTK survey with respect to the GCPs derived from the orthophotos by calculating the average deviation or RMSE. We have also analyzed the positional accuracy by measuring the respective distances between the point and polygon features obtained through the generated orthophotos and reference WMS layers.