Uncertainty quantification of the Global Rural-Urban Mapping Project over Polish census data

Abstract

The aim of this study is to describe uncertainty of the Global Rural-Urban Mapping Project (GRUMP) data based on Polish population reference grid created by the Central Statistical Office of Poland, using INSPIRE grid coding system. The adopted population data uncertainty analysis methodology combined three different approaches, i.e. simple change detection algorithm to obtain discrepancies at the grid cell level, statistical analytical approach to investigate these discrepancies’ frequency distribution, and GIS approach to analyse spatial pattern of distinguished population difference classes. The results showed significant differences in population count at the grid cell level. The maximum magnitude of GRUMP vs. Polish Reference Grid overestimation equals 4087 people per 1 sq. km, while the underestimation equals 20,086 people per 1 sq. km. Very few grid cell shows no difference in population count, i.e. 1.5% of total grid cell count. GRUMP data overestimates Polish total population by 0.15%, while it underestimates the average population density by 50%. The highest population underestimations were identified in the centers of the cities, while suburban areas were characterised by the large and regular population overestimations within GRUMP dataset. These GRUMP dataset imperfections can be attributed to country-specific administrative divisions and to the varying effectiveness of the urban centers delimitation mapping using the night sky light intensity, including blooming effects as well as not frequently illuminated small settlements.