Abstract:

Data quality evaluation is the basis of land cover product selection and the premise of application.Taking Pu 'er City of Yunnan Province as the research area,weadopted the correlation coefficient of area,confusion analysisof classification type,consistency analysis for spatial pattern,and confusion matrix as the indicators for evaluation.Based on the classification results using high spatial resolution airborne images and the sub-meter visual interpretation results,we analyzed the classification accuracy and spatial consistency of 4 land cover products with 10m spatial resolution (Pu'er LC,Esri_Land_Cover_2020_10m,ESA World Cover,Dynamic World) in Pu'er City,Yunnan Province.Ourresults showed that:1) Land types and distribution patterns of 4 land cover products were basically consistent,mainly including woodland,followed by shrub,grassland and cultivated land.The area correlations were relatively high,all the correlation coefficients were greater than 0.8.2) The high spatial consistency area accounted for 64.87% of the Pu 'er city,mostly was woodland;In the low spatial consistency regions,shrub and grassland,woodland,cultivated land,construction land and other land cover types were staggered.3) The absolute accuracy evaluation showed that the overall accuracy of Pu'd LC data was the highest (88.52%),and rest of the three products had the classification accuracy for about 75%.This paper provided a good reference for the quality evaluation and the selection for land cover products.

Key words: land cover, multi-source data, consistency analysis, accuracy evaluation, Pu'er City