Analysis of land use and land cover maps suitability for estimating population density in urban areas - exclusion of the equifinality phenomenon on population maps developed from CLC, UA and OBIA data

Abstract

The series of articles contains a comparison of the possibilities of using for dasymetric estimation of population distribution of spatial information about buildings. The buildings come from three sources characterized by different spatial, thematic and temporal accuracy. These are data from Corine Land Cover (CLC) and Urban Atlas (UA) projects and the result of object classification (OBIA) of RapidEye data. The experiment was carried out in the area of Krakow. Statistical data from 141 city urban units (u. u.) were used.In the first two parts of the cycle, population conversions based on CLC, UA, OBIA and OBIA in combination with UA were presented . In total, 12 maps of Kraków's population were obtained. RMSE and MAPE mean errors were calculated as well as population density for each category of residential development. The results were discussed.In the third part of the cycle, the obtained population maps were analyzed in detail, referring to the Bronowice district (the north-western area of the city) prepared especially by the population. The reference map has been made in high resolution. The methodology of its elaboration has been described in detail. Complementary use of orthophotomap from aerial photographs together with public databases (Geoportal, OpenStreetMap, GoogleStreetView) was presented. The proprietary MMAPE parameter has been proposed. The parameter analyzes the similarity of the reference map of Bronowice with the dasymetric maps. It allows you to statistically describe their credibility and exclude the phenomenon of equivalence.As a result of the conducted research, an erroneous population distribution was detected for the variant OBIA, in which the weights were determined by minimizing the MAPE error. From the remaining experiments, the three best results were obtained by maps using information about urban development from Urban Atlas (MMAPE 100m = 19.3-22.1%). Complementary use of OBIA and UA did not bring any synergy effect -the results were worse than for UA (21.6-24.3%). High errors were noted for OBIA -it is only worth to notice a better result from the binary OBIA method (MMAPE 100m = 22.8%) than the result from the binary CLC method (MMAPE 100m = 24.3%).At this stage of the research, UA data is recommended for the conversion of population. The object classification methods are not a reliable source of data on building types, and such information is necessary for the use of surface-by-weight methods. The use of OBIA is possible only in the binary method and gives results similar to the use of data from CLC.In the fourth part, it is planned to verify the population maps using the Central Statistical Offic (CSO) kilometer network for the whole of Poland, which was made available in 2017. On the basis of multivariate tests and two-stage verification, the authors plan to provide the advantages and disadvantages of the described methods of population conversion and to develop a ranking of the obtained Krakow population maps. AbstraktCykl artykułów zawiera porównanie możliwości wykorzystania...

Analysis of the usefulness land use and land cover maps for estimating the population of urban areas – valorisation of multi-variant population maps based on the GUS kilometre network

“…50m and 100m ("MMAPE 50m and MMAPE 100m, subject to Tobler's condition in u.u.") (Pirowski and Berka, 2019). The second control was aggregative: population maps were converted to a kilometre grid and compared with data published in this resolution by the Central Statistical Office ("RMSE and MAPE, errors in the CSO grid, subject to Tobler's condition in u.u.…”

“…Reference to the reference map of Bronowice (Pirowski and Berka, 2019), characterized by high spatial resolution of population mapping, made it possible to detect the equifinality phenomenon for the OBIA(MAPE) variant. For this population map an incorrect selection of weights was obtained for particular categories of buildings, difficult to catch through the analysis of MAPE and RMSE errors in the u.u.…”

“…-valorisation of twelve developed variants of methods on the basis of a set of independent statistical data; these data were published by the CSO (GUS) at the beginning of 2017 for the whole of Poland in the form of a kilometre grid; these data were not used to calculate the population based on CLC/UA/OBIA; they allow for an objective assessment and the answer which of the generated population maps is the best, i.e. after re-aggregation, this time to the "meshes" of the GUS grid, the lowest deviations are recordedafter re-aggregation, this time to the "meshes" of the GUS grid -in the smallest deviations from the GUS grid; -due to the possibility of the occurrence of the equifinality phenomenon -reference to the analysis made on the basis of a high-resolution photo-interpretation map of the population of the Bronowice district; detailed methodology and results were published in the previous article of the cycle (Pirowski and Berka, 2019); -development of a ranking of 12 analysed variants of population maps; -providing advantages and limitations of the proposed methodology for determining weights in the dasymetric method, in the form of a set of recommendations.…”