Consistency of the Multiple Remote Sensing-based Land Use and Land Cover Classification Products in the Global Coastal Zones

Land Use and Land Cover (LULC) classification products play an indispensable role in ecosystem assessment, climate change simulation, national geographical condition monitoring, and macro-control policy analysis at the global scale; consistency analysis is the precondition of applying various LULC classification products. This paper assessed the area consistency and spatial consistency of five LULC classification products - MCD12Q1-2010, GlobCover2009, CCI-LC2010, FROM-GLC2010 and GlobeLand30-2010- in the global coastal zones. The five products were compared in terms of the deviation coefficient, correlation coefficient, error matrix, and spatial confusion of LULC types. The main findings are as follows: (1) The spatial patterns of LULC in five products demonstrate relatively strong overall consistency, but can have significant local inconsistency. (2) The five products are qualitatively consistent yet quantitatively inconsistent in classifying the LULC in the global coastal zones ? in terms of structure, water ranks top one, followed by forest and unused land, next are farmland, grassland and shrubland, and lastly wetland and artificial surface, yet the exact area of each LULC type differs among different products. (3) For the correlation coefficient, overall accuracy and Kappa coefficient, MCD12Q1-2010/GlobCover2009 have the minimum values, 0.8814, 67.46% and 0.5748, respectively; while GlobCover2009/CCI-LC2010 have the maximum values, 0.9869, 81.50% and 0.7505, respectively; it is because GlobCover2009 and CCI-LC2010 obtained from the same production organization have the same classification system, while MCD12Q1-2010 is different from GlobCover2009 in terms of the production organization, data source, classification system, and classification method. (4) For the spatial confusion/misclassification between any two different products, grassland, shrubland, and wetland have the highest mix-up ratios, followed by farmland and artificial surface, and lastly forest, unused land, and water; this difference is because forest, unused land, and water have distinctive spectral characteristics and clear spatial textures, while grassland, shrubland, and wetland have similar spectral characteristics and fuzzy spatial distributions. (5) There are 28.81% land area in the global coastal zones with relatively low consistency, i.e., with severe spatial confusion; specifically, the misclassification of farmland, forest, grassland, shrubland, wetland, and unused land has direct influence on the spatial consistency of the five products. This paper is hoped to serve as a reference of selecting data from the five available LULC products for researching coastal zones.