The researchers used Geographic Information System and remote sensing techniques to study urban landscape patterns at different scales. They found that the spatial variance of diversity, contagion, and fractal dimension indices was similar at different scales, showing spatial dependence. Smaller scales had stronger spatial dependence, revealing more details. The contribution of spatial autocorrelation to total variance increased with smaller scales. Different landscape indices had different semivariogram models at the same scale, making them incomparable. The ideal scale for studying urban landscape patterns in Shanghai was 1 km. The spatial variance of landscape indices depended on scale, with different scales showing different patterns: complex and irregular at small scales, polycentric at moderate scales, and circle-zoning at large scales.