Biodiversity can be studied at four levels: molecular, species, ecosystem and landscape. Previously, many researchers have focused on the analysis of biodiversity at a single level, and few have explored the relationships among the different levels of biodiversity or conducted an integrated analysis of the four levels. The research methodologies are relatively well established at each level of biodiversity, especially for species diversity and landscape diversity. However, few studies have put forward a research method that spans all four levels of biodiversity. Species diversity is always measured using field samples collected in an inventory survey, and landscape diversity is usually calculated using remote sensing techniques, a geographic positioning system, and a geographic information system technical analysis supplemented with field surveys. Additive partitioning of diversity has been used in regional biodiversity analysis to calculate α, β, and γ diversity. The α, β, and γ diversity indices are calculated by the Shannon-Wiener index, Simpson diversity index, and the total diversity at a regional scale. These indices indicate that spatial-scale effects exist in the analysis of four-level diversity. With increasing spatial scale, the values of α diversity and the Shannon-Wiener index increase sharply and then more gradually beyond some threshold, whereas β diversity increases almost constantly. The behavior of spatial-scale effects differs for different biological species. From this viewpoint, an objective assessment of regional biodiversity is not possible based only on species diversity estimated from a field survey in a restricted geographic area. In contrast to the well-established method of analyzing species biodiversity, the method of evaluating ecosystem diversity is still developing because it is based on various biogeographic classification systems that differ among ecological regions and countries, and which are occupied by their own local biological communities. There is an obvious spatial-scale effect on α, β, and γ diversity in analysis of landscape diversity, and the effect may be highly significant at some spatial scales. There are obvious resolution grain effects and spatial-extent effects in landscape diversity analysis. The grain size can reflect the behavior of the dominant landscape class. This may be useful for landscape management. In remote sensing data the grain size and spectral heterogeneity are inherent, thus it is reasonable to analyze landscape diversity based on a variety of remotely sensed images. With regard to relationships among the four levels of biodiversity, habitat diversity has a close relationship with species diversity and, to some extent, rich landscape diversity is always accompanied by high species diversity, attributable to α indices or β indices. At some scales, this relationship may be statistically significant. Scale effects also exist in regional biodiversity in response to human disturbance. To monitor and assess the biodiversity status of a given area, an integrative analysis involving multiple levels of species, ecosystem and landscape diversity is suggested to elucidate the complicated relationships among these levels. Such research will be helpful to explore the coexisting regimes that contribute to regional biodiversity, and to establish a red list of key ecosystems as assessed by ecosystem functions or in other terms in China, as well as to develop regional biodiversity monitoring and assessment methods. We recommend that a multiple-level analysis of biodiversity should be conducted when assessing the biodiversity status of a region.