Landscape Pattern Analysis of Impervious Surface Based on Sub-Pixel Model: A Case Study in Nansha, Guangzhou

The original vegetation-based natural landscape has been gradually replaced by the artificial impervious surface because of the rapid expansion of urban space, it's of great significance to study the fraction of a city's impervious surface for quantitatively describing the heterogeneity of natural cover within the urban land and reasonable urban planning. Based on V-I-S (vegetation - impervious surface - soil) model, the 2008 Landsat TM data of Nansha in Guangzhou was unmixed by normalized linear spectral mixture model to get the spatial distribution of the impervious surface fraction, on this basis to do the landscape classification, landscape indices calculation and landscape pattern analysis. The results suggested that the combination of V-I-S model and normalized linear spectral mixture model to unmix sub-pixels has a higher accuracy when compared with the conventional landscape classification with a pixel as the unit and the land cover index calculation; On the whole, the average impervious surface fraction of Nansha was relatively low, indicating the urbanization level was not high, the proportion of each level of urban land existed large difference; From Huangge Town to Nansha Street, Hengli Town and Wanqingsha Town, the level of landscape fragmentation became more weakened, the mean patch area increased gradually, the proportion of high-density urban land use declined; The high-density urban land and the moderate-density urban land had the highest patch density, perimeter area fractal dimension and the level of landscape fragmentation, the shape was the most complex.